Upload
others
View
3
Download
0
Embed Size (px)
Citation preview
A TRANSITION FROM TRADITIONAL
MANGROVE REMOTE SENSING TO
RECENT ADVANCES: MAPPING AND
MONITORING
by
Muditha Kumari Heenkenda
(Bachelor of Science (Hons), Master of Geo-information Science)
A thesis submitted in fulfilment of the requirements for the degree of
DOCTOR OF PHILOSOPHY
Research Institute for the Environment and Livelihoods,
Faculty of Engineering, Health, Science and Environment,
Charles Darwin University, Darwin, Australia,
August, 2016
ii
Thesis declaration
I hereby declare that the work herein, now submitted as a thesis for the degree
of Doctor of Philosophy at the Charles Darwin University, is the result of my own
investigations, and all references to ideas and work of other researchers have been
specifically acknowledged.
I hereby certify that the work included in this thesis has not already been
accepted in substance for any degree, and is not being currently submitted in
candidature for any other degree.
I give consent to this copy of my thesis, when deposited in the University
Library, being made available for loan and photocopying online via the University’s
Open Access repository eSpace.
Muditha Kumari Heenkenda
August, 2016
iii
Acknowledgements
First and foremost I would like to thank my supervisors: Dr. Karen Joyce,
Associate Professor Stefan Maier, and Dr. Renee Bartolo for their continuous advice,
time and expertise they contributed to my PhD research. I highly appreciate their
patience, insightful comments and encouragement that contributed substantially to
complete this thesis. Special thanks go to Karen who was always available for advice
and prompt with responses.
My field campaign would not have been successful without the tireless
support of my field assistants: Miguel Tovar Valencia, Olukemi Ronke Alaba, Evi
Warintan Saragih, Silvia Gabrina Tonyes and Dinesh Gunawardena. Sampling inside
mangrove forests is never fun, but these supporters did it consistently without any
complaint. They simply neglected the influence of insects, and a fear of crocodiles.
Lab work could have not been completed without the support and generosity
of Anna Skeleton. I highly appreciate her help throughout the period. I am grateful to
the Director of the Research Institute for the Environment and Livelihoods, Professor
Andrew Campbell for approving funds for field work and publications. Associate
Professor Lindsay Hutley and Dr. Keith McGuinnes; your knowledge and expertise
were greatly appreciated at certain stages of my research. Thanks to Dr. Ian Leiper
and Mrs. Sandra Grant for reviewing parts of my thesis, and being involved in PhD
related discussions.
The Remote Sensing group, Charles Darwin University and my friends
provided advice, assistance and humour. In no particular order they are: Olukemi
iv
Ronke Alaba, Grigorijs Goldbergs, Frederika Rambu Ngana, Miguel Tovar Valencia,
Ian Leiper, David Blondeau-Patissier, and Sofia Oliveira. Thank you so much for
your support. I would like to thank Evi Warintan Saragih and Silvia Gabrina Tonyes
for their encouragement and friendship I received during all ups and downs.
Financial support for this project was provided by Charles Darwin University,
Darwin, Australia. My thanks also go to Ms. Julia Fortune and the Northern Territory
Government for providing data for this study. I am very grateful to the Office of
Research and Innovation staff and the International Office staff for their support
throughout this journey. I highly appreciate the support and generosity of Kristen
Barnes, Katherine Suvvas, and Brooke Williamson who were always there for me.
Thanks to my supervisors, I won the Three Minutes Thesis award at Charles
Darwin University in 2014, and the funds I received from the completion was used to
travel to the University of Calgary, Canada as a visiting student for three months. I
appreciate the generous time, support and encouragement offered by Associate
Professor Greg McDermid and his team at the Department of Geography, University
of Calgary, Canada during the final stages of my thesis write-up.
Last but definitely not least, I would like to thank my dearest husband;
Dinesh for his understanding, patience and support over past four years. It has been
many long hours in front of the computer, days of stress, and days in the field, but
you have made it fly by.
v
Table of Contents
Thesis declaration .................................................................................................................... ii
Acknowledgements ................................................................................................................. iii
Table of Contents ..................................................................................................................... v
List of Figures ........................................................................................................................ xii
List of Tables ......................................................................................................................... xx
Abstract ............................................................................................................................... xxiii
Publications contributing to this thesis ................................................................................ xxv
Statement of contribution of each author ........................................................................... xxvii
Chapter 1 – Introduction .......................................................................................................... 1
1.1 Introduction ........................................................................................................................ 2
1.1.1 Remote sensing for mangrove species mapping ........................................................ 7
1.1.2 Quantifying mangrove biophysical variables .......................................................... 14
1.1.2.1 Individual tree crown delineation ................................................................. 16
1.1.2.2 Estimating Above Ground Biomass ............................................................. 18
1.1.2.3 Estimating Leaf Area Index .......................................................................... 21
1.1.2.4 Estimating canopy chlorophyll content ........................................................ 22
1.1.3 Impacts of development pressure on mangroves ..................................................... 23
1.2 Research Problem ............................................................................................................ 24
1.3 Research aim, objectives and research questions ............................................................. 25
1.4 Outline of the thesis ......................................................................................................... 27
1.5 Chapter synopsis .............................................................................................................. 27
1.5.1 Chapter 1 - Introduction ........................................................................................... 27
vi
1.5.2 Chapter 2 - Mangrove species identification: Comparing WorldView-2 with Aerial
Photographs ...................................................................................................................... 28
1.5.3 Chapter 3 - Mangrove tree crown delineation from high resolution imagery .......... 28
1.5.4 Chapter 4 - Quantifying mangrove chlorophyll from high spatial resolution imagery
.......................................................................................................................................... 28
1.5.5 Chapter 5 - Estimating mangrove biophysical variables using WorldView-2 satellite
data .................................................................................................................................... 29
1.5.6 Chapter 6 - Ecological risk assessment using a Relative Risk Model: A case study of
the Darwin Harbour, Darwin, Australia ............................................................................ 29
1.5.7 Chapter 7 - Conclusions and recommendations ....................................................... 30
1.6 References ........................................................................................................................ 31
Chapter 2 - Mangrove Species Identification: Comparing WorldView-2 with Aerial
Photographs ........................................................................................................................... 46
2.0 Abstract ............................................................................................................................ 47
2.1 Introduction ...................................................................................................................... 48
2.2 Data and Methods ............................................................................................................ 52
2.2.1. Study Area .............................................................................................................. 52
2.2.2. Field Survey ............................................................................................................ 53
2.2.3. Remote Sensing Data and Pre-Processing .............................................................. 54
2.2.4. Image Classification ........................................................................................... 57
2.2.4.1. Separating Mangroves and Non-Mangroves ............................................... 59
2.2.4.2. Mangrove Species Classification ................................................................. 62
2.2.5. Accuracy Assessment ............................................................................................. 65
2.3. Results ............................................................................................................................. 65
2.3.1 Field Survey ............................................................................................................. 65
vii
2.3.2Separating Mangroves and Non-Mangroves ............................................................ 66
2.3.3 Mangrove Species Classification ............................................................................. 68
2.3.4 Accuracy Assessment .............................................................................................. 75
2.4 Discussion ........................................................................................................................ 77
2.4.1 Separating Mangroves and Non-Mangroves ........................................................... 77
2.4.2 Comparison of Mangrove Species Classifications................................................... 80
2.4.3 Accuracy Assessment .............................................................................................. 83
2.5 Conclusions and Recommendations ................................................................................ 86
2.6 Acknowledgments ............................................................................................................ 87
2.7 Author Contributions ....................................................................................................... 88
2.8 Conflicts of Interest .......................................................................................................... 88
2.9 References ........................................................................................................................ 88
Chapter 3 - Mangrove tree crown delineation from high resolution imagery ........................ 96
3.0 Abstract ............................................................................................................................ 97
3.1 Introduction ...................................................................................................................... 98
3.2 Data and methods ........................................................................................................... 102
3.2.1 Study area, remote sensing data, and pre-processing ............................................. 102
3.2.2 Extracting mangrove coverage from images ......................................................... 106
3.2.3 Digital Surface Model (DSM) generation .............................................................. 107
3.2.4 Mangrove tree crown delineation .......................................................................... 108
3.2.4.1 WV2 panchromatic, red, and NIR1 bands (PAN-R-NIR1 method) ...... 108
3.2.4.2 WV2 panchromatic, red, and NIR1 bands and DSM
(PAN-R-NIR1-DSM method) ................................................................................ 109
viii
3.2.4.3 The inverse watershed segmentation (IWS) method with the DSM ...... 110
3.2.5 Accuracy assessment ............................................................................................. 110
3.3 Results ............................................................................................................................ 111
3.3.1 The digital surface model ...................................................................................... 112
3.3.2 Mangrove tree crown delineation .......................................................................... 113
3.3.2.1 WV2 panchromatic, red, and NIR1 bands (PAN-R-NIR1 method) ...... 113
3.3.2.2 WV2 panchromatic, red, and NIR1 bands and DSM (PAN-R-NIR1-
DSM method) ......................................................................................................... 114
3.3.2.3 The inverse watershed segmentation (IWS) method with the DSM ...... 114
3.3.3 Accuracy assessment ............................................................................................. 118
3.4. Discussion ..................................................................................................................... 120
3.5 Conclusions and recommendations ................................................................................ 124
3.6 References ...................................................................................................................... 126
Chapter 4 - Quantifying mangrove chlorophyll from high spatial resolution imagery ........ 135
4.0 Abstract .......................................................................................................................... 136
4.1 Introduction .................................................................................................................... 137
4.2 Materials and methods ................................................................................................... 140
4.2.1 Study area and satellite data ................................................................................... 140
4.2.2 Field sampling........................................................................................................ 146
4.2.2.1 Estimating Leaf Area Index ........................................................................ 147
4.2.2.2 Estimating leaf chlorophyll concentration .................................................. 150
4.2.3 Estimating mangrove canopy chlorophyll content from satellite imagery ............ 151
4.2.3.1 Canopy chlorophyll variation with respect to mangrove species ............... 154
ix
4.2.4 Accuracy assessment ............................................................................................. 154
4.3 Results ............................................................................................................................ 155
4.3.1 Field sampling........................................................................................................ 155
4.3.2 Estimating mangrove canopy chlorophyll content from satellite imagery ............ 155
4.3.2.1 Canopy chlorophyll variation with respect to mangrove species ............... 160
4.3.3 Accuracy assessment ............................................................................................. 161
4.4 Discussion ...................................................................................................................... 162
4.4.1 Estimating mangrove canopy chlorophyll content from satellite imagery ............ 164
4.4.2 Accuracy assessment ............................................................................................. 167
4.5 Conclusions and recommendations ................................................................................ 168
Appendix 4.A ....................................................................................................................... 170
4.6 References ...................................................................................................................... 171
Chapter 5 - Estimating mangrove biophysical variables using WorldView-2 satellite data:
Rapid Creek, Northern Territory, Australia ......................................................................... 181
5.0 Abstract .......................................................................................................................... 182
5.1 Introduction .................................................................................................................... 183
5.2 Data and methods ........................................................................................................... 186
5.2.1 Study area .............................................................................................................. 186
5.2.2 Field sampling, satellite data and predictor variables ............................................ 187
5. 2.2.1. Remotely-Sensed Data and Predictor Variables ....................................... 188
5.2.3.1 Estimating Leaf Area Index ........................................................................ 193
5.2.3.2 Estimating Above Ground Biomass ........................................................... 196
5.2.3 Predicting LAI and AGB ....................................................................................... 197
x
5.2.4 Accuracy assessment ............................................................................................. 199
5.3 Results ............................................................................................................................ 199
5.3.1 Predicting Leaf Area Index and Above Ground Biomass ...................................... 202
5.4 Discussion ...................................................................................................................... 207
5.4.1 Predicting LAI ....................................................................................................... 208
5.4.2 Predicating AGB .................................................................................................... 211
5.5 Conclusions and recommendations ................................................................................ 214
5.6 References ...................................................................................................................... 217
Chapter 6 - Regional Ecological Risk assessment using a Relative Risk Model: A case study
of the Darwin Harbor, Darwin, Australia ............................................................................ 224
6.0 Abstract .......................................................................................................................... 225
6. 1 Introduction ................................................................................................................... 226
6.2 Data and method ............................................................................................................ 229
6.2.1 Study area .............................................................................................................. 229
6.2.2 Data input ............................................................................................................... 231
6.2.3 Problem formulation .............................................................................................. 233
6.2.3.1 Ecological Assessment endpoints ............................................................... 235
6.2.4 Risk analysis .......................................................................................................... 238
6.2.4.1 Relative risk calculation ............................................................................. 240
6.2.5 Risk characterisation .............................................................................................. 241
6.2.6 Uncertainty and sensitivity analysis ....................................................................... 242
6.3 Results ............................................................................................................................ 244
6.3.1 Problem formulation .............................................................................................. 244
xi
6.3.1.1 Habitats and ecological assessment endpoints ........................................... 248
6.3.2 Risk analysis .......................................................................................................... 250
6.3.3 Risk characterisation .............................................................................................. 250
6.3.4 Uncertainty and sensitivity analysis ....................................................................... 253
6.4 Discussion ...................................................................................................................... 256
6.4.1 Uncertainty and sensitivity analysis ....................................................................... 261
6.5 Conclusions and recommendations ................................................................................ 262
6.6 Acknowledgement ......................................................................................................... 265
6.7 References ...................................................................................................................... 265
Chapter 7 – Conclusions and Recommendations ................................................................. 270
Highlights ............................................................................................................................. 271
Summary .............................................................................................................................. 272
7.1 Research question 1: How can high spatial resolution remote sensing data be used to
discriminate mangroves at species level? ............................................................................ 275
7.2 Research question 2: How can 3D canopy dynamics combined with spectral properties of
mangroves be used to map mangrove biophysical variability? ........................................... 276
7.3 Research question 3: How can spatial information be combined to analyse the
environmental stresses on different mangrove sites? ........................................................... 279
7.4 Limitations and recommendations for future research ................................................... 280
7.4.1 Mapping with high spatial resolution image data .................................................. 281
7.4.2 Relative Risk Modelling ........................................................................................ 283
7.5 Contribution to Knowledge ............................................................................................ 284
Appendices ........................................................................................................................... 287
xii
List of Figures
Figure 1.1. The boundary of the Darwin Harbour catchment and its mangrove
coverage; Satellite images © ESRI, DigitalGlobe, and GeoEye. .............. 3
Figure 1.2. Urban and rural activities in the Darwin Harbour estuarine and their
effect on the Darwin Harbour (Symbols courtesy of the Integration and
Application Network, University of Maryland Centre for Environmental
Science; http://ian.umces.edu/symbols/). .................................................. 6
Figure 2.1. The study area located in coastal mangrove forest of Rapid Creek in
Darwin, Northern Territory, Australia; WV2 data © Digital Globe.
(Coordinate system: Universal Transverse Mercator Zone 52 L, WGS84).
................................................................................................................. 53
Figure 2.2. Mangrove species mapping using remotely sensed data. Image
segmentation and initial classification was completed in eCognition
Developer 8.7 software. The support vector machine algorithm was
implemented in the ENVI-IDL environment for species classification. . 59
Figure 2.3. The schematic explanation of image objects at different hierarchical
levels. The first level was at pixel level. At the second level: vegetation
and non-vegetation areas were identified. Mangrove areas were extracted
at the final level. ...................................................................................... 62
xiii
Figure 2.4. Spectral profiles of: (a) all features except mangroves extracted from
WV2 image; (b) mangrove species only extracted from WV2 image; (c)
mangrove species extracted from aerial photographs of spatial resolution
0.14 m; and (d) mangrove species extracted from aerial photographs of
spatial resolution 0.5 m. (AM-Avicennia marina, CT-Ceriops tagal,
BE-Bruguiera exaristata, LR-Lumnitzera racemosa, RS-Rhizophora
stylosa, SA-Sonneratiaalba). .................................................................. 66
Figure 2.5. The locations of field samples available for calibration and validation
purposes, and mangrove coverage extracted from the WV2 image. ....... 68
Figure 2.6. (a) Classification of pan-sharpened WV2 image using blue to red-edge
bands; (b) Classification of pan-sharpened WV2 images using red to
NIR1 bands; (c) Classification of WV2 image using blue to red-edge
bands; (d) Classification of WV2 images using red to NIR1 bands; (e)
Classification of aerial photographs with 0.14 m spatial resolution; and
(f) Classification of aerial photographs with 0.5 m spatial resolution. ... 69
Figure 2.7. The comparison of the WorldView2 image and different classification
approaches: (a) Pan-sharpened WV2 image; (b) Classification results of
PS-WV2-VIS; (c) Classification results of PS-WV2-R/NIR1; (d) Aerial
photograph; (e) Classification results of AP0.14M; (f) Classification
results of AP0.5M; (g) WV2 image; (h) Classification results of WV2-
VIS; (i) Classification results of WV2-R/NIR1. ..................................... 71
Figure 2.8. Extents (%) of mangrove species obtained using six different input
datasets: (a) pan-sharpened WV2 image with bands blue to red-edge; (b)
pan-sharpened WV2 image with bands red to NIR; (c) aerial photographs
xiv
with 0.14 m spatial resolution; (d) aerial photographs with 0.5 m spatial
resolution; (e) WV2 image with bands blue to red-edge; and (f) WV2
image with bands red to NIR. ................................................................. 73
Figure 3.1. The study area is located in Rapid Creek in Darwin, Northern Territory,
Australia. The map shows the validation dataset including both field
surveyed and on screen digitised (three dimensionally) trees. Inset shows
the appearance of mangrove tree crowns on the aerial photograph;
Coordinate system: Universal Transverse Mercator Zone 52 L, WGS84;
Aerial photographs © Northern Territory Government (NTG), Australia,
Copyright 2012 NTG. ........................................................................... 103
Figure 3.2. The digital surface model of the Rapid Creek mangrove forest created
from aerial photographs. Inset “A” shows localized circular patterns due
to the image enhancement using focal statistics function in ArcGIS. ... 113
Figure 3.3. (A) Mangrove tree crowns delineated with the PAN-R-NIR1 method
overlaid with the panchromatic image; (B) Overview of the Rapid Creek
mangrove forest; (C) Mangrove tree crowns delineated with the PAN-R-
NIR1-DSM method overlaid with the panchromatic image; (D)
Mangrove tree crowns delineated with the PAN-R-NIR1-DSM method
overlaid with the DSM; (E) Mangrove tree crowns delineated with the
IWS method overlaid with the panchromatic image and tree tops (black
dots); (F) Mangrove tree crowns (black outlines) delineated with the IWS
method overlaid with the DSM and tree tops (black dots). ................... 115
xv
Figure 3.4. (A) Centroid positions obtained from the PAN-R-NIR1 and the
PAN-R-NIR1-DSM methods overlaid with the panchromatic WV2 image
with low pass filtering applied; (B) Overview of the Rapid Creek
mangrove forest; (C) Centroid positions obtained from the IWS method
overlaid with the DSM. ......................................................................... 117
Figure 4.1. The study area is located in the coastal mangrove forest of Rapid Creek in
Darwin, Northern Territory, Australia; (A) WorldView-2 satellite image
and locations of 29 field sampling plots (5 m x 5 m); WorldView-2
images © DigitalGlobe; (B) Australia, the boundary of the Northern
Territory, and the study area; (C) The Rapid Creek mangrove forest
(study area); Coordinate system: Universal Transverse Mercator
Zone 52 L, WGS84. ............................................................................... 141
Figure 4.2. Workflow for mapping chlorophyll content from satellite data and field
samples. This procedure was repeated for the satellite data with 5 m and
10 m spatial resolution. NDVI = Normalized Difference Vegetation
Index from NIR1 and red bands; NDVI2 = Normalized Difference
Vegetation Index from NIR2 and red bands; NDRE = Normalized
Difference Index from NIR1 and rededge bands; NDRE2 = Normalized
Difference Index from NIR2 and rededge bands; GNDVI: Normalized
Difference Index from NIR1 and green bands; CVI: Chlorophyll
Vegetation Index from NIR1, red and green bands. ............................. 145
xvi
Figure 4.3. Linear correlation coefficients against predictor variables with different
spatial resolutions; Green, Yellow, Red, RedEdge, NIR1 and NIR2:
surface reflectance of green, yellow, red, red-edge and near infrared
bands of WorldView-2 image; NDVI = Normalized Difference
Vegetation Index from NIR1 and red bands; NDVI2 = Normalized
Difference Vegetation Index from NIR2 and red bands; NDRE =
Normalized Difference Index from NIR1 and rededge bands; NDRE2 =
Normalized Difference Index from NIR2 and rededge bands; GNDVI:
Normalized Difference Index from NIR1 and green bands; CVI:
Chlorophyll Vegetation Index from NIR1, red and green bands. ......... 156
Figure 4.4. Ranking the importance of predictor variables according to the mean
squared error (%IncMSE); Green, Yellow, Red, RedEdge, NIR1 and
NIR2: surface reflectance of green, yellow, red, red-edge, NIR1 and
NIR2 bands of the WorldView-2 image; NDVI = Normalized Difference
Vegetation Index from NIR1 and red bands; NDVI2 = Normalized
Difference Vegetation Index from NIR2 and red bands; NDRE =
Normalized Difference Index from NIR1 and rededge bands; NDRE2 =
Normalized Difference Index from NIR2 and rededge bands; GNDVI:
Normalized Difference Index from NIR1 and green bands; CVI:
Chlorophyll Vegetation Index from NIR1, red and green bands. ......... 158
Figure 4.5.Canopy chlorophyll variation (g/m2) in the Rapid Creek mangrove forest;
(A) 2 m spatial resolution; (B) 5 m spatial resolution; and (C) 10 m
spatial resolution. .................................................................................. 159
xvii
Figure 4.6. The canopy chlorophyll distribution with mangrove species;
AM- Avicennia marina; BE - Bruguiera exaristata; CT - Ceriops tagal;
LR - Lumnitzera racemosa; RS - Rhizophora stylosa. ......................... 161
Figure 5.1. Rapid Creek coastal mangrove forest, Darwin, Northern Territory,
Australia. The distribution of field sampling plots (5m x 5m) is shown in
the map (the sizes of magenta squares are not to the scale). Aerial
photographs © Northern Territory Government. .................................. 187
Figure 5. 2. Mangrove trees inside the Rapid Creek forest; (A–C) small,
newly-regenerated mangrove trees; (D,E) large, multi-stemmed
mangrove trees; (F,G) densely-clustered areas. .................................... 200
Figure 5.3. Normal score plots of field estimates of (A) Leaf area index; and (B) Log
transformed above ground biomass. ..................................................... 201
Figure 5.4. Cross-validated predictions for the normalized leaf area index (LAI) and
above ground biomass (AGB): (A) predicted versus field measured
normalized LAI with 2-m spatial resolution predictor variables; (B)
predicted versus field measured normalized LAI with 5-m spatial
resolution predictor variables; (C) predicted versus field measured
normalized LogAGB with 2-m spatial resolution predictor variables; (D)
predicted versus field measured normalized LogAGB with 5-m spatial
resolution predictor variables. (RMSEP, root mean square error of
prediction). ............................................................................................ 203
xviii
Figure 5.5. Predicted maps: (A) leaf area index with a 2-m spatial resolution; (B) leaf
area index with a 5-m spatial resolution; (C) above ground biomass with
a 2-m spatial resolution; and (D) above ground biomass with a 5-m
spatial resolution; using the partial least squares regression algorithm.
The distribution of field sampling plots is shown in the maps (the sizes of
the black squares are not to scale). ....................................................... 206
Figure 6.1. The Darwin Harbor catchment and its mangrove distribution .............. 230
Figure 6.2. The workflow diagram for conducting an ecological risk assessment
using the Relative Risk Model following the guidelines published by the
U.S. Environmental Protection Agency ................................................ 233
Figure 6.3. The conceptual diagram showing estuarine activities and their effect on
mangroves, water bodies and then to the Darwin Harbour (Courtesy of
the Integration and Application Network, University of Maryland Centre
for Environmental Science (ian.umces.edu/symbols/)). ....................... 244
Figure 6.4. The conceptual model for the relative risk modelling in the Darwin
Harbor. They were prepared for all risk regions. Exposure and effect
filters were modified with respect to each risk region. ......................... 249
Figure 6.5. The total relative risk to ecological assets in risk regions of the Darwin
Harbor. They were classified as low, medium or high based on the
Jenks’s optimization (Jenks and Caspall 1971). ................................... 252
xix
Figure 6.6. Total risk scores for the ecological endpoints; (A) Maintenance of
mangrove health; (B) Maintenance of water quality ............................. 253
Figure 6.7. The frequency of all possible outcomes of total risk of ecological assets
in the risk regions: (a) Ludmilla Creek, Mitchell Creek, Mitcket Creek,
Palmerston South, and Pioneer Creek; (b) Blackmore River, Elizabeth
River, Howard River, Hudson Creek, and Kings Creek; (c) Rapid Creek,
Reichardt Creek, Sadgroves Creek, Sandy Creek, West Arm, and Woods
Inlet; (d) Bleesers Creek, Buffalo Creek, Charles Point, and Creek A. 254
Figure 6.8. The rank correlations for sub-regions having the highest total risk scores:
(A) Blackmore River; (B) Elizabeth River; (C) Myrmidon Creek; (D)
Bleesers Creek. ...................................................................................... 256
xx
List of Tables
Table 1.1. The most common remote sensing image processing techniques used for
mapping mangroves ................................................................................ 10
Table 1.2. Four main methods for delineating individual tree crowns ...................... 17
Table 1.3. Four main methods and their advantages and disadvantages for estimating
Above Ground Biomass (AGB) of forests (Komiyama et al. 2008;
Metcalfe et al. 2011; Roy and Ravan 1996). ........................................... 18
Table 2.1. Spectral band information of WV2 image and aerial photographs obtained
from UltraCamD camera [29, 30]. .......................................................... 55
Table 2.2. Number of samples (4 m2 each) used for training and number of sample
points generated for validation for each mangrove species. ................... 68
Table 2.3. Overall accuracy and Kappa statistics obtained for each classification. ... 75
Table 2.4. Comparison of producer’s accuracy and user’s accuracy for each species,
obtained from different remote sensing data sources. (AM-Avicennia
marina, CT-Ceriops tagal, BE-Bruguiera exaristata, LR-Lumnitzera
racemosa, RS-Rhizophora stylosa). ........................................................ 76
Table 3.1. Spectral resolution of WV2 image and aerial photographs (DigitalGlobe,
2011; Northern Territory Government, 2010)....................................... 105
xxi
Table 3.2.The number objects extracted as tree tops using three different methods.
............................................................................................................... 116
Table 3.3. A comparison of accuracies of extracted tree crowns: the possible range of
both OverSegmentation and UnderSegmentation values is [0,1] where
zero defines the perfect segmentation. The values closer to zero for
“closeness index” illustrate the closeness of mangrove tree crowns in two
dimensional space which is defined by OverSegmentation and
UnderSegmentation to validation crowns (Clinton et al., 2010). .......... 119
Table 4.1. Vegetation indices used to establish the relationship between mangrove
field samples and remotely sensed data; green, red, red edge, NIR1 and
NIR2 surface reflectance of WorldView-2 satellite data for
corresponding wavelength regions. ....................................................... 143
Table 4.2. Root mean squared errors (RMSE), correlation coefficients, and
coefficients of determinations calculated with respect to predicted
chlorophyll values and independent validation samples ....................... 162
Table 5. 1 Spectral band information of WorldView-2 images ............................... 189
Table 5. 2. Predictor variables generated from the WorldView-2 multispectral image
for estimating above ground biomass and leaf area index. NIR1, NIR2,
red edge, red and green: surface reflectance of Near Infrared 1, Near
Infrared 2, red-edge, red and green wavelength regions, respectively. . 191
xxii
Table 5. 3. Log10-transformed allometric relationships used for different mangrove
species. The equations are in the form: 𝑙𝑙𝑙𝑙𝑙𝑙10𝐵𝐵𝐵𝐵𝑙𝑙𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵 = 𝐵𝐵0 + 𝐵𝐵1 ∗
𝑙𝑙𝑙𝑙𝑙𝑙10𝐷𝐷𝐵𝐵𝐷𝐷; where DBH is the diameter at breast height; B0 and B1 are
regression coefficients. These equations are specific to the Northern
Australia [37,38], North-eastern Queensland, Australia [39] and Sri
Lanka [19] (biomass in kg and DBH in cm). ........................................ 197
.Table 5. 4. Root mean square errors (RMSE’s) and correlation coefficients (r) for
above ground biomass and leaf area index maps with respect to
validation samples ................................................................................. 207
xxiii
Abstract
Mangroves are dense, spatially heterogeneous forests, and provide a wide
range of goods and services. They protect the shore line from powerful waves, and
provide habitat and nursery grounds for variety of animals. Mangroves are one of the
most threatened habitats in the world, and are disappearing rapidly due to urban,
industrial and commercial activities. These threats lead to an increasing demand for
quantitative assessment for mangrove monitoring. Historically, mapping has been
completed with aerial photos over small areas. Recent advances in remote sensing
technology for data acquisition and processing have opened up a wealth of new
options. The aim of this work was to push the boundary of data processing to extract
metrics relevant to mangrove health mapping, monitoring, and management. This
included mangrove species classification, delineating individual tree crowns, and
continuous variable mapping of plant biomarkers. The first ever study classified
mangrove species at Rapid Creek mangrove forest, Australia was achieved 89%
overall accuracy compared to in-situ observations with a WorldView-2 satellite
image and a support vector machine algorithm. The first successful method was
developed to delineate individual tree crowns using object based image analysis
techniques with 92% overall accuracy. One of the innovative prospects of this study
was development of methods to continuous variable mapping of plant biomarkers:
canopy chlorophyll, above ground biomass, and leaf area index using WorldView-2
images, field measurements, random forest and partial least squares regression
algorithms. A relative risk model was developed to assess ecological risk associated
with the Darwin Harbour mangroves. This model identified localized changes:
cleared, stressed and dead mangroves due to natural or anthropogenic influences. In
summary, this research presents unique applications of advanced remote sensing
xxiv
methods for analysing up-to-date spatial information on the current status of Rapid
Creek mangroves, and they are a baseline for mangrove monitoring and
management.
xxv
Publications contributing to this thesis
Peer reviewed papers;
Chapter 2
Heenkenda, M. K., Joyce, K. E., Maier, S. W., & Bartolo, R., 2014. Mangrove
Species Identification: Comparing WorldView-2 with Aerial Photographs, Remote
Sensing, 6 (7): 6064-6088.
Chapter 3
Heenkenda M. K., Joyce K. E., & Maier S. W., 2015a, Mangrove tree crown
delineation from high resolution imagery, Photogrammetric Engineering and Remote
Sensing, Vol. 81, No. 6, pages: 21-30
Chapter 4
Heenkenda, M. K., Joyce, K. E., Maier, S. W. & de Bruin, S. 2015b, Quantifying
mangrove chlorophyll from high spatial resolution imagery, ISPRS Journal of
Photogrammetry and Remote Sensing, vol. 108, 234-244
http://dx.doi.org/10.1016/j.isprsjprs.2015.08.003.
Chapter 5
Heenkenda M. K., Maier S.W & Joyce K. E., 2016, Estimating mangrove
biophysical variables using WorldView-2 satellite data, Journal of Imaging, 2, 24;
doi:10.3390/jimaging2030024.
xxvi
Chapter 6
Heenkenda, M. K. & Bartolo, R. 2015, Regional Ecological Risk Assessment Using
a Relative Risk Model: A Case Study of the Darwin Harbor, Darwin, Australia,
Human and Ecological Risk Assessment: An International Journal, vol. 22, 401 - 423
doi: 10.1080/10807039.2015.1078225.
Formal permission from the publishers was granted to include this material as
attached in Appendices 1 to 4.
Conference Proceedings;
Heenkenda, M. K., Joyce, K. E. & Maier, S. W. 2014, Comparing digital object
based approaches for mangrove tree crown delineation using WorldView-2 satellite
imagery, South-Eastern European Journal of Earth Observation and Geomatics, vol.
3, 169-172.
xxvii
Statement of contribution of each author
The following table detailing relative contribution by me and my co-authors
refers to the five manuscripts taken directly from this thesis. Four manuscripts were
published in peer reviewed international journals prior to thesis submission. One
manuscript has been submitted to a peer reviewed international journal and accepted
subject to minor revisions.
Manuscript 1: Mangrove Species Identification: Comparing WorldView-2 with
Aerial Photographs - Chapter 2
Contributing author Contribution Tasks Performed
M.K. Heenkenda 90% Study design, field data collection, image and
field data analysis, and manuscript writing
K.E. Joyce
10%
Manuscript review and editing
S.W. Maier Manuscript review and editing
R. Bartolo Manuscript review and editing
Manuscript 2: Mangrove tree crown delineation from high resolution imagery -
Chapter 3
Received 2016 ESRI Award for the Second Best Scientific Paper in Geographic
Information Systems from the American Society of Photogrammetry and Remote
Sensing: The imaging and geospatial information society (ASPRS).
xxviii
Contributing author Contribution Tasks Performed
M.K. Heenkenda 90% Study design, field data collection, image and
field data analysis, and manuscript writing
K.E. Joyce 10%
Manuscript review and editing
S.W. Maier Manuscript review and editing
Manuscript 3: Quantifying mangrove chlorophyll from high spatial resolution
imagery - Chapter 4
Contributing author Contribution Tasks Performed
M.K. Heenkenda 85% Study design, field data collection, image and
field data analysis, and manuscript writing
K.E. Joyce
15%
Manuscript review and editing
S.W. Maier Field survey design, manuscript review and
editing
S. de Bruin Advice on field data processing, manuscript
review and editing
Manuscript 4: Estimating mangrove biophysical variables using WorldView-2
satellite data - Chapter 5
Contributing author Contribution Tasks Performed
M.K. Heenkenda 85% Study design, field data collection, image and
field data analysis, and manuscript writing
K.E. Joyce
15%
Manuscript review and editing
S.W. Maier Field survey design, advice on field data
processing, manuscript review, and editing
xxix
Manuscript 5: Regional Ecological Risk Assessment Using a Relative Risk Model:
A Case Study of the Darwin Harbor, Darwin, Australia - Chapter 6
Contributing author Contribution Tasks Performed
M.K. Heenkenda 90% Relative risk model designing, data collection,
meeting stakeholders, data processing, and
manuscript writing
R. Bartolo 10% Relative risk model design, manuscript review
and editing
1
Chapter 1 – Introduction
2
1.1 Introduction
Mangroves are coastal, salt tolerant woody plants found in sheltered estuaries,
along river banks, deltas and shallow-water lagoons in 124 tropical and subtropical
countries (Food and Agriculture Organisation 2007). Mangroves grow best where
there is low wave energy and sediment deposition of fine particles (Alongi 2002).
There are some exceptional situations where they grow on rocky shores, rootling in
silt-filled depressions (Food and Agriculture Organisation 2007). However, in both
situations, they represent an interface between land and ocean communities. Hence,
mangrove ecosystems are significant for sustaining biodiversity, and providing
various benefits to human activities.
Mangrove ecosystems are significant for a wide range of marine and
terrestrial organisms. Mangroves provide habitats, spawning grounds, nurseries and
nutrients for animals ranging from reptiles to mammals and birds (McKee 1996).
They promote coastal stabilisation, and provide a natural barrier against storms,
cyclones, and tsunamis (Bouillon et al. 2009; Stephen Crooks et al. 2011). Further,
they are known to be highly productive and biochemically active ecosystems
(Suratman 2008). They contribute to the carbon budget as a large store of blue
carbon (Bouillon et al. 2009). Unfortunately, urbanisation, population growth, water
diversion, agriculture, aquaculture, and salt pond construction has led to high rates of
mangrove destruction on a global scale.
Given the range of destructive practices around the world, it has become
critical to create baseline datasets for mangroves. The most recent global data
compilation suggests a current areal mangrove extent of about 152,000 km2
3
(Bouillon et al. 2009; Food and Agriculture Organisation 2007), with Indonesia and
Australia together covering about 30% of this extent. Approximately 35% of the
Australian extent lies along the Northern Territory (NT) coastline.
Figure 1.1. The boundary of the Darwin Harbour catchment and its mangrove
coverage; Satellite images © ESRI, DigitalGlobe, and GeoEye.
4
The largest single expanse of mangroves in the NT is found on the Darwin
Harbour shoreline (Harrison et al. 2009). Although the Darwin harbour mangroves
represent 2% of the national coverage, they comprise 35% of this ecosystem within
the NT (Figure 1.1). On a global scale, Darwin Harbour mangroves contribute
approximately 0.1% of the world’s remaining mangrove extent (Brocklehurst and
Edmeades 1994-1995).
While the Harbour and its catchment encompass a range of distinctive and
special environments that require protection the region is subject to development
pressures and conflicts of use. For example, to accommodate the projected
population growth over next 20 years in the NT, there are many proposed and
ongoing commercial, industrial and residential developments. Recently, the
Australian Government released a white paper on developing the north’s water
resources, infrastructure and workforce for a strong and prosperous economy
(Australian Government 2015). Hence, the growing pressure of rural and urban
development will directly or indirectly impact on Darwin Harbor’s vegetation
species. The growing pressure of development may also increase flooding, velocity
and volume of soil runoff with nutrients, chemicals, sedimentation and pollution. A
key challenge to developing northern Australia is to balance environmental values
with the potential economic and social benefits. In order to do this effectively,
understanding those environmental values and their ecology is critical.
A substantial amount of research has been done in the mangrove forests of
Darwin Harbour to understand the ecology of these ecosystems, and providing
information for effective management. Mangrove monitoring programs have been
conducted since 1997 to obtain baseline data on mangrove productivity and ecology
5
(Water Monitoring Branch, 2005). Further, numerous studies have been undertaken
aiming to understand the ecosystem’s economic and ecological values to set the
balance between mangrove conservation and development (Burford et al. 2008;
Lewis 2002; Metcalfe et al. 2011; Moritz-Zimmermann 2002; NRETAS 2010a,
2010b). However, understanding the enormous complexity of interacting processes
that maintain biodiversity and productivity of mangroves remains a major challenge,
and significant gaps still exist in understanding the likely effects of human
disturbances.
There is no clear approach for integrating the risks due to number of stressors
or likely effects of human disturbances with a numerous possible interactions. Figure
1.2 illustrates multiple stressors due to urban and rural development and their
influence on the coastal environment. Land clearing for urban and rural activities,
agriculture and horticulture, aquaculture, and point sources discharges are major
stressors for estuarine health. For example, the Central Business District of Darwin
was developed after clearing 400 ha of mangroves in 2002 (1.5% of the total
mangrove area) (Northern Territory Government 2002b). The exposure to such
stressors creates a biological degradation of the receiving habitats. At the moment,
impacts from the development pressure of past activities are localised and not well
documented. There is no a proposed method to assess impacts from future
development pressure as well.
6
Figure 1.2. Urban and rural activities in the Darwin Harbour estuarine and their
effect on the Darwin Harbour (Symbols courtesy of the Integration and Application
Network, University of Maryland Centre for Environmental Science;
http://ian.umces.edu/symbols/).
Planners and resource managers have a critical need to create accurate
inventories of mangrove communities and their biophysical characteristics especially
their areal extent, species distribution, and biophysical variability. A detailed survey
of mangroves of Darwin Harbour was completed in 1994 - 1995 based on field
measurements and data obtained from aerial photo interpretations. Yet these data are
limited to a scale of 1: 25000 map scale, and 0.25 hectares spatial resolution
(Brocklehurst and Edmeades 1994-1995). To understand the current situation,
regular map updates are essential. Further, there are no detailed studies explaining
7
biophysical variability (for instance variation of above ground biomass or leaf area
index) of mangroves in this area.
Field survey is essential to identify and locate new mangrove communities
present in a study area and to update existing mangrove maps. The survey should be
planned to represent ranges of mangrove communities, their characteristics and
changes over time within the study area to capture the species and biophysical
variability. Therefore, the process involves a significant time and financial
investment. Mangrove forests in the region are often difficult to access due to the
danger of salt water crocodiles. Further, due to the densely clustered nature of
mangroves and stilt roots, moving within the forest is challenging. Remotely sensed
data provides an alternative approach.
1.1.1 Remote sensing for mangrove species mapping
Aerial photographs and high spatial resolution satellite images are the most
common primary remotely sensed data source for mangrove mapping. Satellite data
with low or medium spatial resolution and laser scanning data are other remote
sensing data sources that can be used for mangrove ecosystem assessment. In the
scientific literature, there are a significant number of studies that involve mangrove
forests, remote sensing data and various image processing algorithms. Table 1.1
shows the most common image processing techniques that have been used for
detecting and delineating mangroves. In summary, these methods are not able to
fully discriminate mangroves species. The majority of mangrove remote sensing
studies use high spatial resolution imagery, mostly with a 5 to 100 m pixel size.
Further, it was clear that the type of data and image processing algorithms used have
8
significant impacts on the accuracy of the resultant mangrove maps. Hence,
determining a suitable data source and robust processing method for mapping
mangroves at species level is in demand.
Image processing logic
and classifiers Mangrove studies Advantages Disadvantages
Manual classifiers
Visual image
interpretation, stereo
digitizing, or manual
boundary tracing
Brocklehurst and Edmeades
(1994-1995); Blasco et al.
(2001); Dahdouh-Guebas et
al. (2002); Sulong et al.
(2002); Verheyden et al.
(2002); Dahdouh-Guebas et
al. (2006); Habshi et al.
(2007)
These methods provide overall
accuracy greater than 90% in
separating mangrove coverages
only.
Methods do not require powerful
computers for image processing.
These methods are based on
image interpretation; hence, the
results are purely dependent on
the analyst, spatial and spectral
resolutions of images.
Methods are time consuming and
costly owing to manual
operations.
Methods unable to differentiate
mangroves at species level with
higher accuracy.
9
Image processing logic
and classifiers Mangrove studies Advantages Disadvantages
Classification logics
Unsupervised
classification,
Supervised
classification,
hybrid-supervised and
unsupervised
classification
Gao (1998); Green et al.
(1998); Lucas et al. (2002);
Nandy and Kushwasha
(2010); Held et al. (2003);
Wang et al. (2004); Ji et al.
(2008); Satyanarayana et al.
(2011); Giri et al. (2011);
Giri et al. (2008); Hossain et
al. (2009)
Depending on the spatial resolution
of the input data, these methods
provide higher accuracies for
mangrove extent mapping as they
allow for computational
classification rather than
interpretation.
Some studies classified individual
species up to about 70% of accuracy
level. Among them, aerial
photographs provided the best
results for mapping individual
species. However, aerial
photographs can be more cost
effective over small areas than
satellite images.
These methods still have a level
of subjectivity depending on the
algorithm used – i.e. can’t
routinely apply and get the same
result each time.
Most common computational
classification errors with these
methods are the omission of
fringe mangroves that are less
than the pixel size resulting in
mixed pixels, and confusion
between mangroves and other
vegetation.
10
Table 1.1. The most common remote sensing image processing techniques used for mapping mangroves
Image processing logic
and classifiers Mangrove studies Advantages Disadvantages
Advanced classifiers
Spectral
transformation,
principle component
analysis (PCA),
vegetation indices,
decision tree algorithm
Green et al. (1998); Held et
al. (2003); Heumann
(2011); Wang et al. (2004);
Vaiphasa et al. (2005);
Vaiphasa et al. (2006);
Wang and Sousa (2009);
Satyanarayana et al. (2011);
Liu et al. (2008)
These methods provided higher
accuracy for separating mangroves
and non-mangroves. For instance,
Liu et al. (2008) achieved greater
than 80% of accuracy for mapping
mangrove coverage. Decision tree
algorithm allows incorporating
digital elevation models and
proximity data for the computational
classification. In a comparison of
classification techniques and data
types, Green et al. (1998) found that
the classification of PCA data
provided significantly better
accuracy than raw images.
These methods also have a level
of subjectivity depending on the
algorithms and data used.
11
12
When applying pixel based classification algorithms there remain several
limitations. The confusion between mangroves, non-mangrove vegetation, urban
areas and even mudflats affects classification accuracies (Gao 1998; Green et al.
1998; Held et al. 2003; Ji et al. 2008; Nandy and Kushwasha 2010). Pixel based
approaches resulted in some confusion between saltmarshes and coastal vegetation
(Kamal and Phinn 2011). Fringe mangrove canopies are commonly less than the
pixel size, resulting in mixed pixels and a consequent over or underestimate of true
mangrove coverage depending on the classification decision (Heumann 2011a).
Although assessing mangrove coverage or mangrove density is possible with high
spatial resolution multispectral imagery (Green et al. 1998; Habshi et al. 2007;
Mitchell et al. 2007), detecting individual mangrove species and estimating canopy
structure remains a challenge.
In recent years new techniques have been developed or adapted to improve
the accuracy of detecting individual species. Some studies integrated a relationship
between mangroves and surrounding environmental gradient. Vaiphasa et al. (2006)
incorporated soil pH data and a Bayesian probabilistic model, and improved the
species mapping accuracy by 12%. However, incorporating soil pH data into the
mapping process was not able to resolve confusion between some mangrove species.
Later, attention was given to test multiple data sources within a rule-based
classification scheme or object based image analysis.
Object based image analysis algorithms typically have a decision-tree that
refines a separation of classes at different levels. Multiresolution segmentation and
class-specific rules having a relationship between image objects at different
13
hierarchical levels improved overall mangrove cover classification accuracy in many
studies (Conchedda et al. 2008; Heumann 2011; Quoc et al. 2013). Hence, an object
based image analysis method was identified as a straightforward, repeatable and
reliable method for classifying mangrove coverage.
Kamal and Phinn (2011) found that multiresolution segmentation
outperformed pixel based approaches (spectral angle mapper and linear spectral
unmixing) in mangrove species mapping. The study obtained more than 75%
classification accuracy for mangrove extent and certain species. A species
classification of SPOT5 satellite data using a decision tree approach and local
knowledge from the region of interest showed accuracies higher than 75% for certain
classes at an object level (Quoc et al. 2013). However, in this study, the mangrove
forest cover fraction per object is affected by image segmentation, and is influenced
by physically visible natural boundaries such as shrimp pond dikes. As tested by
different studies, multiresolution segmentation is a powerful technique; however,
selecting an appropriate scale parameter is crucial for creating meaningful objects.
Similar challenges are faced in other terrestrial environments such as
deciduous forests, tropical rainforests, savannah and grasslands. As a result, more
contextual and probabilistic methods have recently been introduced to terrestrial
vegetation species mapping. For example, sub-pixel classification methods are often
preferred, for example, Markov-random-field based, Artificial Neural Network,
Support Vector Machine, and Random Forests algorithms (Ardila et al. 2011;
Mitchard et al. 2012; Mountrakis et al. 2011; Muad and Foody 2010; Pajares et al.
2010). Knowledge gained in these environments may be applicable for mangroves.
14
1.1.2 Quantifying mangrove biophysical variables
An understanding of mangrove biophysical characteristics is useful for
mangrove management planning, conservation, and change monitoring (Wannasiri et
al. 2013). Hence, estimating increasing or decreasing nutrient levels of mangroves,
and biophysical characteristics would be a significant step towards assessing
variations in mangrove plant status or health. Despite the large number of studies
addressing various aspects of organic matter cycling in mangrove forests in Darwin
Harbour (Alongi 1994; Burford et al. 2008; Larsen and Rudemo 1998; Lu 2006;
McHugh 2004; Metcalfe 2007; Metcalfe et al. 2011; Woodroffe et al. 1988), there is
still a demand for estimating biophysical characteristic variations.
The most important biophysical characteristics that deserve attention are Leaf
Area Index (LAI), Above Ground Biomass (AGB), and canopy nutrient levels. This
is because these characteristics reflect the status of physiological processes of plants
such as photosynthesis, evapotranspiration, carbon cycling, net primary production,
habitat condition and plant health (Flores-de-Santiago et al., 2013; Heumann 2011a).
Tree height, diameter at breast height and crown diameter are considered as the key
physical characteristics for estimating mangrove AGB and LAI (Wannasiri et al.
2013). Nitrogen stress restricts plant productivity and it is highly correlated with
chlorophyll content. Chlorophyll content of any tree reflects the nutritional stresses,
photosynthetic capacity and developmental stages of trees (Felella and Penuelas
1994).
Mangrove forests are generally challenging environments for mapping as a
consequence of architectural heterogeneity such as clustered trees, clumping leaves,
15
and underlying soil, water, and even understory vegetation. Producing precise and
up-to-date spatial information on the current status of mangrove forests is almost
impossible by using traditional field surveys because mangrove swamps are
extremely difficult to access. Estimates of LAI, AGB and canopy chlorophyll are
difficult to obtain in forests owing to the labour required to destructively sample
trees, and the time and money that need to be invested.
As an alternative method for estimating biophysical variations and canopy
nutrients, remotely sensed images are an effective tool. Specifically, these images
record the amount of reflected and absorbed light from different surfaces. As plant
pigments, structure, and water content are all known to affect light interactions in
different wavelengths, it is possible to estimate these properties from the recorded
light signal in an image. Analysis of the images then permits an evaluation of
changes in the environment over time, and preparing decision-making programs for
different time frames. Owing to the availability and low cost of imagery compared to
field measurements, remotely sensed image based measurements are increasingly
used in environmental modelling and ecology for delineating individual tree crowns,
as well as estimating AGB and LAI. The increasing usage of these measurements is
supported by new advances in algorithm development, their high accuracy, reliability
and automation.
16
1.1.2.1 Individual tree crown delineation
Individual tree attributes can be used for forest inventories, assessing forest
regeneration and vegetation changes, and quantifying above ground biomass (AGB)
(Chen et al. 2006). In addition, crown foliage area is related to carbon gain and
water loss (Analuddin et al. 2009), and thus productivity. Therefore, being able to
delineate individual crowns from a crowded canopy is significant for long term
monitoring practises. A review of individual tree or group of trees identification
using remote sensing images identifies four main approaches (Table 1.2).
As described above (Table 1.2), all methods are influenced by image
intensities, sensor characteristics, imaging time, angle, elevation, and homogeneity of
vegetation with reasonable height variations between neighbouring trees. There is no
single method that perfoms equally well for different forest types. They cannot
directly be applied to mangroves due to mangrove forests’ structural complexity or
heterogeneity. Therefore, selecting suitable source data, and appropriately
parameterising an algorithm for mangrove tree crown delineation remain a challenge.
17
Table 1.2. Four main methods for delineating individual tree crowns
Local maxima detection Valley following method
Assumes that tree apexes generate
intensity peaks in images. Assuming a
local maximum is located near or at the
centre of the crown, a local maximum
filter with specific kernel is applied to
the image to detect tree locations
(Pouliot et al. 2002). This is the most
intensively investigated image analysis
approach. However, this method is most
successful with coniferous trees rather
than other forest types (Ardila et al.
2011; Leckie et al. 2005; Lopez 2012).
False detections are observed with
spatial heterogeneous areas or images
with large spectral variations.
Considers the image intensities as
elevation of topographic surface, and
creates a hill pattern aiming to extract
the highest part of the ‘hill’ as a tree
top. Boundaries of tree crowns show
dark values, shady rings or valleys
depending on sun illumination angle
(Gougeon 1995; Leckie et al. 2005).
This method provides best results
when identifying individual trees in
homogeneous landscapes with
reasonable height variations between
neighbouring trees.
Tree modelling and image templates Segmentation method or region
grouping
Constructs geometric attributes of the
trees, and matches to images
considering different viewing
conditions and illuminations (Leckie et
al. 2005). Crown diameter, tree height
and convexity are investigated and often
modelled by means of a generalised
ellipsoid.
Involves identification of groups of
similar neighbouring pixels which
corresponds to objects or part of
objects (Leckie et al. 2005). The
construction of regions of the image is
based on homogeneity of images
characteristics; image brightness,
texture, shape, pixel size etc. This
method is most successful when
identifying individual trees in
homogeneous landscapes (Erikson
2004).
18
1.1.2.2 Estimating Above Ground Biomass
There are four main field survey methods for estimating forest AGB
(Table 1.3) (Bouillon et al. 2008; Komiyama et al. 2008; Metcalfe et al. 2011). They
are labour intensive, and require site and species specific field measurements. To
have a good representation of forest coverage, extensive field sampling is essential
though field sampling is difficult due to muddy soil conditions and heavy weight of
the wood (Komiyama et al. 2008). Because of these methodological problems and
associated uncertainties when extending these conventional methods over a large
area, net primary productivity of mangroves is known to be underestimated (Bouillon
et al. 2008).
Table 1.3. Four main methods and their advantages and disadvantages for estimating
Above Ground Biomass (AGB) of forests (Komiyama et al. 2008; Metcalfe et al.
2011; Roy and Ravan 1996).
Method Description Advantages Disadvantages
Harvest
method
Harvests all trees in a
unit area, and
measure weight of all
woody parts.
Accurate Destructive, unable to
replicate as all trees
must be harvested,
unable to produce in
mature forests due to
heavy woody weight,
time, cost, and labour
intensive
19
Method Description Advantages Disadvantages
Mean-tree
Algorithm
Considers harvest of
average size trees
either for the stand or
within given area
(mean tree
dimensions) and
measure weight of all
woody parts.
Accurate, suitable
for forest
plantations
Only suitable for
forests with
homogeneous tree
sizes,
time, cost, and labour
intensive
Allometry Involves harvesting
individual trees over
a wide range in size
and establishing a
relationship between
measureable tree
dimensions (height of
trees and diameter at
breast height) as
biomass in forest
stands.
Accurate, well
formulated
allometric relations,
once relations are
established they are
repeatable, and the
method becomes
non-destructive
Allometric relations are
site and species
specific,
time, cost, and labour
intensive for generating
relations
Litter fall
estimation
Involves stratified
sampling of litter fall
using litter traps.
Cost effective,
simple calculations
Challenges of sampling
and extrapolating litter
fall rates due to within-
stand heterogeneity,
litter traps must be
elevated above the
highest tide level to
avoid tidal flushing of
litter
20
Remotely sensed data has started addressing the major challenges identified
with these field survey methods by avoiding destructive sampling and reducing time
and cost for field sampling (Table 1.3). Several studies have found strong correlation
between biomass and spectral reflectance values at different wavelength regions
within remotely sensed data (Anaya et al. 2009; Roy and Ravan 1996). Later, the
attention was given to spectral vegetation indices generating from remotely sensed
data rather than spectral reflectance values for estimating AGB (Adam et al. 2010;
Mutanga et al. 2012).
Spectral vegetation indices are mathematical combinations of different
spectral bands mostly in the visible and near infrared regions of the electromagnetic
spectrum. The main purpose of these numerical transformations is to enhance the
information contained in spectral reflectance data for detecting within-field spatial
variability, and to minimize soil, atmospheric and sensor geometry effects (Viña et
al. 2011). These indices provide a simple and convenient approach to extract
information from remotely sensed data. For example, one of the most commonly
used vegetation indices is the Normalised Vegetation Difference Index (NDVI)
derived from red and near infrared wavelength regions. Normalised vegetation
indices derived from different combinations of spectral bands, ratio indices, modified
or soil adjusted vegetation indices are also used for biomass estimation of different
forest types (Huete 1988; Mutanga et al. 2012; Roy and Ravan 1996; Serrano et al.
2000). Further, spectral vegetation indices based approaches are capable of
estimating AGB by reducing major problems associated with field sampling.
21
1.1.2.3 Estimating Leaf Area Index
The ratio of leaf surface area to unit ground surface (Leaf Area Index or LAI)
describes the potential surface area available for leaf gas exchange (Breda 2003).
LAI is an important factor controlling many biological and physical processes of
vegetation such as photosynthesis and respiration. Two main approaches have been
developed for estimating LAI remotely: (1) inversion of canopy radiative transfer
models; and (2) empirical relationships between LAI and spectral vegetation indices
(Viña et al. 2011). Vegetation indices have been more widely used due to their ease
of computation in contrast to the use of radiative transfer models.
Ramsey and Jensen (1996), Green et al. (1997) and Kovacs et al. (2004)
explained the relationship between NDVI and LAI for mangroves in southwest
Florida, British West Indies and Mexico, respectively. Ramsey and Jensen (1996)
described complexity and difficulties in assessing the spectral and structural aspects
of mangroves by analysing canopy reflectance spectra and vegetation indices. The
results were limited by the number and types of sites sampled, and the techniques
used. Green et al. (1997) identified remote sensing as a powerful tool for estimating
LAI achieving 88% of overall accuracy. However, their study area was covered by a
mixture of sparse mangroves and open canopies. These areas received reflectance
values from understorey vegetation as well, resulting in low LAI values. LAI of
degraded mangroves were successfully estimated by Kovacs et al. (2004). Although
the study suggested using its methods for both densely and sparsely distributed
mangroves, it also emphasized using very precise and accurate positioning methods
for sampling and ground controlling. All three studies identified correlating field
observations with remotely sensed information, and determining the regression
22
algorithm as limiting factors to the final results. However, they can be used as guides
when attempting to map mangroves using remote sensing. Non-destructive, remote
determination of LAI replaces time consuming and expensive field measurements.
Further, a large number of spectral vegetation indices have been developed for
estimating LAI. In most cases, these indices have been tested for specific species and
geographic locations only. It is not clear whether they can be applied for different
species with varying structural characteristics and across various geographic
locations.
1.1.2.4 Estimating canopy chlorophyll content
Canopy chlorophyll content is another important biomarker of mangrove
plant status. Leaf pigments, mainly chlorophyll a and chlorophyll b are significant
components responsible for photosynthesis, physiology, and other biological
functions (Zhang et al. 2012). Canopy chlorophyll is also an indicator of crop
nitrogen requirements (Hansen and Schjoerring 2003). Since measuring chlorophyll
in the field is time and cost intensive, remote sensing of chlorophyll content is a
valuable tool (Felella and Penuelas 1994).
Empirical modelling of canopy chlorophyll is based on the statistical
regression between the canopy chlorophyll content and the spectral reflectance
within remotely sensed images. Detecting canopy chlorophyll by remote sensing
techniques has been tested for many terrestrial vegetation types, however, there are a
limited number of studies available for mangrove ecosystems. One study
investigated the spectral responses of different mangrove species under various
conditions such as healthy and degraded (Zhang et al. 2012). They observed
23
significant differences in chlorophyll a and chlorophyll b in different mangrove
species at different wavelength regions. These established models are site, species
and time specific, and cannot readily be applied to other types of vegetation or under
different conditions.
1.1.3 Impacts of development pressure on mangroves
Currently the Darwin Harbour catchment is subject to multiple development
pressures and use conflicts, indicating a major challenge for managing the local
environment (The Darwin Harbour Advisory Committee 2003). The north eastern
part of the Darwin Harbour catchment is already highly developed and native
vegetation and tidal flats have been cleared and drained (Northern Territory
Government 2010). By 2005, approximately 18% of the Harbour catchment had been
cleared of native vegetation for mainly urban and rural residential living,
horticulture, agriculture, infrastructure, defence facilities and manufacturing (Skinner
et al. 2009). A major liquefied natural and petroleum gas project is currently running
in this area. Subsea pipelines that transport liquid hydrocarbons run between onshore
and the Darwin Harbour catchment. For the project duration (2012 to 2017), the
proposed land clearing extent is about 300 ha including 11 ha of mangroves (Carter
2013; INPEX 2013). Further, up to 42 ha of mangroves could experience reduced
health and growth (INPEX 2013). Therefore, a regional plan of management has
been developed for Darwin Harbour and its catchment area, and priority actions are
being implemented to protect its biodiversity.
The first management goal identified by the Darwin Harbour regional plan of
management is to maintain a healthy environment. Hence, the plan encourages
24
ecologically sustainable development through planning and management of future
developments (The Darwin Harbour Advisory Committee 2003). Although, the plan
offers an adaptive and coordinated approach managing and planning within the
harbour and catchment, one of the major challenges is to combine knowledge from
different sources into a consistent framework in order to make practical and sound
management decisions.
To ensure long term sustainability of the Darwin Harbour ecosystem, a
statistically sound, and transparent ecological risk assessment model is essential.
Such a model should enable decision makers to evaluate the likelihood that adverse
ecological effects may occur as a result of exposure to certain stressors in the
catchment. An efficient spatial model will strike the balance between the conflicting
uses and values, and addresses the complexities of planning in the region.
1.2 Research Problem
Maintaining mangrove ecosystem services and a healthy environment is one
of the priority goals of the NT government. This must be in line with many
infrastructure developments along the mangrove dominated shoreline of Darwin
Harbour (Harrison et al. 2009; McHugh 2004; Northern Territory Government
2002a). Although many studies have been done in the mangrove forests of Darwin
Harbour to understand these valuable ecosystems, some knowledge gaps still exist.
In particular, baseline mangrove data need to be updated, in addition to providing an
indication of the species that are vulnerable to clearing, death, or changes to drainage
due to urban and rural developments.
25
Although mangrove species around Darwin Harbour were manually mapped
using field survey and some traditional remote sensing techniques in 1994-1995, the
challenge remains to develop and implement image processing techniques for
mapping, and updating available inventories in an automated manner. Hence, there is
a significant demand for developing an accurate, robust, and repeatable method for
mapping individual mangrove species using remote sensing data. Since there are
major urban and industrial developments projects and associated activities around
Darwin Harbour, ensuring the Darwin harbour ecosystem remains healthy at present
and future levels of development is important. One of the key requirements for
environmental monitoring programs is to combine knowledge from different sources
into a consistent framework that enables stakeholders to make practical and sound
management decisions. Therefore, developing an environmental risk assessment tool
that incorporates information from different sources is essential.
1.3 Research aim, objectives and research questions
This research aims to push the boundary of data processing to extract metrics
relevant to mangrove health mapping, monitoring and management in Darwin
Harbour. The study will also develop a statistically sound ecological risk assessment
model that harmonizes development and conservation around Darwin Harbour to
overcome the challenges associated with mangrove monitoring. To achieve this aim,
the following research questions have been identified, each with a series of
objectives.
26
Question 1: How can high spatial resolution remote sensing data be used to
discriminate mangroves at species level?
Objectives:
I. To identify and map mangroves within a case study site of Darwin Harbour
II. To develop a remote sensing method to distinguish individual mangrove
species
Question 2: How can 3D canopy details combined with spectral properties of
mangroves be used to map mangrove biophysical variability?
Objectives:
I. To extract individual tree crowns or stands within mangrove forest
II. To estimate spatial variability of mangrove canopy chlorophyll
III. To estimate spatial variability of mangrove leaf area index and above ground
biomass
Question 3: How can spatial information be combined to analyse the environmental
stresses on different mangrove sites?
Objectives:
I. To identify human induced environmental stressors on different mangrove
sites
II. To develop a conceptual model of stressors on mangroves, and to identify
risk regions
27
III. To qualitatively analyse the relationships between mangrove health and
human induced environmental stressors using a relative risk model
IV. To quantitatively analyse the uncertainties associated with the relative risk
modelling
1.4 Outline of the thesis
This thesis is composed of five journal manuscripts. Four manuscripts have
already been published in international peer-reviewed journals, and the last one is in
review. The manuscripts are organized into five separate chapters (2 to 6), each of
which stand alone and address one or more research objectives. These chapters are
accompanied by a general introduction (this chapter) and conclusions and
recommendations (chapter 7).
1.5 Chapter synopsis
1.5.1 Chapter 1 - Introduction
The introduction provides the conceptual and theoretical background of the
research reported in the dissertation followed by the research significance. It
discusses the importance for mapping and monitoring mangrove ecosystems, and the
current challenges faced when using traditional survey methods in this environment.
A significant gap in the current state of knowledge regarding mapping mangrove
environments is defined, and the potential for using high resolution remotely sensed
data is demonstrated.
28
1.5.2 Chapter 2 - Mangrove species identification: Comparing WorldView-2
with Aerial Photographs
Mangrove species maps derived from two different layer combinations of
WorldView-2 satellite images are compared with those generated using high
resolution aerial photographs. The successful application of high spatial resolution
remotely sensed data and the support vector machine algorithm for mangrove species
mapping is demonstrated with an accuracy assessment related to field survey data.
With the exception of some formatting to maintain the continuity in the thesis, the
content of this chapter is the same as that of the paper by Heenkenda et al. (2014).
1.5.3 Chapter 3 - Mangrove tree crown delineation from high resolution
imagery
Various tree crown delineation methods are tested with respect to different
remotely sensed data sources. An application of object based image analysis method
for this difficult environment is demonstrated. The accuracy of the proposed methods
is compared with the ground validation measurements. The content of this chapter is
similar to that of the paper by Heenkenda et al. (2015a).
1.5.4 Chapter 4 - Quantifying mangrove chlorophyll from high spatial
resolution imagery
The spatial distribution of mangrove chlorophyll content is mapped with a
combination of remotely sensed data, advanced statistical algorithms and laboratory
analysed field samples. Further, the chapter identifies most important predictor
variables that can be derived from high spatial resolution satellite data for estimating
29
mangrove chlorophyll content in the study area. The accuracy of the proposed
method is analysed based on laboratory measured field samples. This chapter is
based on the paper by Heenkenda et al. (2015b) with some page formatting to
maintain the continuity in the thesis.
1.5.5 Chapter 5 - Estimating mangrove biophysical variables using
WorldView-2 satellite data
The spatial distributions of leaf area index (LAI) and above ground biomass
(AGB) of mangroves are estimated using a combination of field samples and remote
sensing data. A digital cover photography method is implemented to estimate LAI of
mangroves. Site and species specific allometric equations are used for calculating
validation data for AGB. Advanced remotely sensed data processing methods are
then demonstrated for estimating biophysical variations. The content of this chapter
is the same as submitted to the Journal of Ecological Indicators for consideration.
1.5.6 Chapter 6 - Ecological risk assessment using a Relative Risk Model: A case
study of the Darwin Harbour, Darwin, Australia
The ecological risk to the Darwin Harbour mangroves in the context of urban
and rural developments was evaluated using a relative risk model. The model
evaluates the interaction of multiple stressors, habitat and assessment endpoints, and
proves the suitability of this spatial modelling approach for large geographic areas
where multiple human induced stressors are involved. With the exception of some
formatting to maintain the continuity in the thesis, the content of this chapter is exactly
the same that of the paper by Heenkenda and Bartolo (2015).
30
1.5.7 Chapter 7 - Conclusions and recommendations
The concluding chapter specifically details the contribution of this thesis to
the field of mangrove remote sensing. It revisits the main findings and techniques
used throughout each preceding chapter, describing how the approach and results are
significant, new and innovative. Finally, some suggestions to improve results and
ideas for future applications are presented.
31
1.6 References
Adam, E., Mutanga, O. & Rugege, D. 2010, Multispectral and hyperspectral remote
sensing for identification and mapping of wetland vegetation: a review, Wetlands
Ecology and Management, vol. 18, 281-296 10.1007/s11273-009-9169-z.
Alongi, D. M. 1994, Zonation and seasonality of benthic primary production and
community respiration in tropical mangrove forests, Oecologia, vol. 98, 320-327.
Alongi, D. M. 2002, Present state and future of the world’s mangrove forests,
Environmental Conservation, vol. 29, 331-349.
Analuddin, K., Sharma, S., Suwa, R. & Hagihara, A. 2009, Crown Foliage Dynamics
of Mangrove Kandelia obovata in Manko Wetland, Okinawa Island, Japan, Journal
of Oceanography, vol. 65, 121-127.
Anaya, J. A., Chuvieco, E. & Palacios-Orueta, A. 2009, Above ground biomass
assessment in Colombia: A remote sensing approach, Forest Ecology and
Management, vol. 257, 1237-1246.
Ardila, J. P., Tolpekin, V. A., Bijker, W. & Stein, A. 2011, Markov-random-field-
based super-resolution mapping for identification of urban trees in VHR images,
ISPRS Journal of Photogrammetry and Remote Sensing, vol. 66, 762 - 765.
32
Australian Government 2015, Our North, Our Future: White Paper on Developing
Northern Australia, viewed 26.08.2015 2015,
<http://northernaustralia.infrastructure.gov.au/white-paper/index.aspx>.
Blasco, F., Aizpuru, M. & Gers, C. 2001, Depletion of the mangroves of Continental
Asia, Wetlands Ecology and Management, vol. 9, 245-256.
Bouillon, S., Borges, A. V., Castaneda-Moya, E., Diele, K., Dittmar, T., Duke, N. C.,
Kristensen, E., Lee, S. Y., Marchand, C., Middelburg, J. J., Rivera-Monroy, V. H.,
III, T. J. S. & Twilley, R. R. 2008, Mangrove production and carbon sinks: A
revision of global budget estimates, global biogeochemical cycles, vol. 22,
10.1029/2007GB003052.
Bouillon, S., Rivera-Monroy, V. H., Twilley, R. R. & Kairo, J. G. 2009, Mangroves,
in D. Laffoley & G. Grimsditch (eds), The Management of Natural Coastal Carbon
Sinks, IUCN, Gland, Switzerland, p. 53.
Breda, N. J. J. 2003, Ground-based measurements of leaf area index: a review of
methods, instruments and current controversies, Journal of Experimental Botany, vol.
54, 2403-2417 doi: 10.1093/jxb/erg263.
Brocklehurst, P. & Edmeades, B. 1994-1995, Mangrove Survey of Darwin Harbour:
Northern Territory, Technical Memorandum No. 96/9.
33
Burford, M. A., Alongi, D. M., McKinnon, A. D. & Trott, L. A. 2008, Primary
production and nutrients in a tropical macro-tidal estuary, Darwin Harbour,
Australia, Estuarine, Coastal and Shelf Science, vol. 79, 440-448.
Carter, S. 2013, Dredging and Spoil Disposal Management Plan - East Arm, INPEX,
Darwin.
Chen, Q., Baldocchi, D., Gong, P. & Kelly, M. 2006, Isolating Individual Trees in a
Savanna Woodland Using Small Footprint Lidar Data, Photogrammetric Engineering
and Remote Sensing, vol. 72, 923-932.
Conchedda, G., Durieux, L. & Mayaux, P. 2008, An object-based method for
mapping and change analysis in mangrove ecosystems, ISPRS Journal of
Photogrammetry and Remote Sensing, vol. 63, 578-589.
Dahdouh-Guebas, F., Verheyden, A., Kairo, J. G., Jayatissa, L. P. & Koedam, N.
2006, Capacity building in tropical coastal resource monitoring in developing
countries: A re-appreciation of the oldest remote sensing method, International
Journal of Sustainable Development & World Ecology, vol. 13, 62-76.
Dahdouh-Guebas, F., Zetterstrom, T., Ronnback, P., Troell, M., Wickramasinghe, A.
& Koedam, N. 2002, Recent changes in land-use in the Pambala-Chilaw lagoon
complex (Sri Lanka) investigated using remote sensing and GIS: Conservation of
mangroves Vs. development of Shrimp farming, Environmental Development and
Sustainability, vol. 4, 185-200.
34
Erikson, M. 2004, Species classification of individually segmented tree crowns in
high-resolution aerial images using radiometric and morphologic image measures,
Remote Sensing of Environment, vol. 91, 469-477 10.1016/j.rse.2004.04.006.
Felella, I. & Penuelas, J. 1994, The red edge position and shape as indicators of plant
chlorophyll content, biomass and hydric status, International Journal of Remote
Sensing, vol. 15, 1459-1470.
Food and Agriculture Organisation 2007, The World's Mangroves 1980-2005, Food
and Agriculture Organisation of the United Nations, Rome.
Gao, J. 1998, A hybrid method towards accurate mapping of mangroves in a
marginal habitat from SPOT multispectral data, International Journal of Remote
Sensing, vol. 19, 1887-1899.
Giri, C., Ochieng, E., Tieszen, L. L., Zhu, Z., Singh, A., Loveland, T., Masek, J. &
Duke, N. 2011, Status and distribution of mangrove forests of the world using earth
observation satellite data, Global Ecology and Biogeography, vol. 20, 154-159.
Giri, C., Zhu, Z., Tieszen, L. L., Singh, A., Gillette, S. & Kelmelis, J. A. 2008,
Mangrove forest distributions and dynamics (1975-2005) of the tsunami-affected
region of Asia, Journal of Biogeography, vol. 35, 519-528.
Gougeon, F. A. 1995, A crown-following approach to the automatic delineation of
individual tree crowns in high spatial resolution aerial images, Canadian Journal of
Remote Sensing, vol. 21, 274-284.
35
Green, E. P., Cleark, C. D., Mumby, P. J., Edwards, A. J. & Ellis, A. C. 1998,
Remote Sensing techniques for mangrove mapping, International Journal of Remote
Sensing, vol. 19, 935-956.
Green, E. P., Mumby, P. J., Edwards, A. J., Clark, C. D. & Ellis, A. C. 1997,
Estimating leaf area index of mangroves from satellite data, Aquatic Botany, vol. 58,
11-19.
Habshi, A. A., Youssef, T., Aizpuru, M. & Blasco, F. 2007, New mangrove
ecosystem data along the UAE coast using remote sensing, Aquatic Ecosystem
Health & Management, vol. 10, 309-319.
Hansen, P. M. & Schjoerring, J. K. 2003, Reflectance measurement of canopy
biomass and nitrogen status in wheat crops using normalized difference vegetation
indices and partial least squares regression, Remote Sensing of Environment, vol. 86,
542-553.
Harrison, L., McGuire, L., Ward, S., Fisher, A., Pavey, C., Fegan, M. & Lynch, B.
2009, An inventory of sites of international and national significance for biodiversity
values in the Northern Territory, Department of Natural Resources, Environment,
The Arts and Sport, Darwin, NT,
<http://www.nretas.nt.gov.au/__data/assets/pdf_file/0020/13934/06_darwin.pdf>.
Heenkenda, M., Joyce, K., Maier, S. & Bartolo, R. 2014, Mangrove Species
Identification: Comparing WorldView-2 with Aerial Photographs, Remote Sensing,
vol. 6, 6064-6088.
36
Heenkenda, M. K. & Bartolo, R. 2015, Regional Ecological Risk Assessment Using
a Relative Risk Model: A Case Study of the Darwin Harbor, Darwin, Australia,
Human and Ecological Risk Assessment: An International Journal, vol. 22, 401-423
10.1080/10807039.2015.1078225.
Heenkenda, M. K., Joyce, K. E. & Maier, S. W. 2015a, Mangrove Tree Crown
Delineation from High Resolution Imagery, Photogrammetric Engineering and
Remote Sensing, vol. 81, 471-479, doi: 10.14358/PERS.81.6.471.
Heenkenda, M. K., Joyce, K. E., Maier, S. W. & de Bruin, S. 2015b, Quantifying
mangrove chlorophyll from high spatial resolution imagery, ISPRS Journal of
Photogrammetry and Remote Sensing, vol. 108, 234-244
http://dx.doi.org/10.1016/j.isprsjprs.2015.08.003.
Held, A., Ticehurst, C., Lymburner, L. & Williams, N. 2003, High resolution
mapping of tropical mangrove ecosystems using hyperspectral and radar remote
sensing, International Journal of Remote Sensing, vol. 24, 2739-2759.
Heumann, B. W. 2011, An Object-Based Classification of Mangroves Using a
Hybrid Decision Tree-Support Vector Machine Approach, Remote Sensing, vol. 3,
2440-2460 10.3390/rs3112440.
Heumann, B. W. 2011a, Satellite Remote Sensing of mangrove forests: Recent
advances and future opportunities, Progress in Physical Geography, vol. 35, 87-108.
37
Hossain, M. Z., Tripathi, N. K. & Gallardo, W. G. 2009, Land Use Dynamics in a
Marine Protected Area System in Lower Andaman coast of Thailand, 1009-2005,
Journal of Coastal Research, vol. 25, 1082-1095.
Huete, A. R. 1988, A soil-adjusted vegetation index (SAVI), Remote Sensing of
Environment, vol. 25, 295-309 http://dx.doi.org/10.1016/0034-4257 (88) 90106-X.
INPEX 2013, Dredging in Darwin Harbour, viewed 12/01/2015,
<http://www.inpex.com.au/oldsite/sites/default/files/pdfs/120803_dredging%20in%2
0darwin%20harbour.pdf>
Ji, M., Wu, Y., Deng, Z., Fan, H. & Zhang, Z. 2008, Mapping mangroves from high-
resolution multispetral imagery using Beilun Estuary, Guangxi, China as a case
study, in Remote Sensing and Modeling of Ecosystems for Sustainability V, San
Diego, CA, USA vol. 7083.
Kamal, M. & Phinn, S. 2011, Hyperspectral Data for Mangrove Species Mapping: A
Comparison of Pixel-Based and Object-based Approach, Remote Sensing, vol. 3,
2222-2242, 10.3390/rs3102222.
Komiyama, A., Ong, J. E. & Poungparn, S. 2008, Allometry, biomass, and
productivity of mangrove forests: A review, Aquatic Botany, vol. 89, 128-137.
Kovacs, J. M., Flores-Verdugo, F., Wang, J. & Aspden, L. P. 2004, Estimating leaf
area index of a degraded mangrove forest using high spatial resolution satellite data,
Aquatic Botany, vol. 80, 13-22.
38
Larsen, M. & Rudemo, M. 1998, Optimizing templates for finding trees in aerial
photographs, Pattern Recognition Letters, vol. 19, 1153-1162 10.1016/s0167-
8655(98)00092-0.
Leckie, D. G., Gougeon, F. A., Tinis, S., Nelson, T., Burnett, C. N. & Paradine, D.
2005, Automated tree recognition in old growth conifer stands with high resolution
digital imagery, Remote Sensing of Environment, vol. 94, 311-326
10.1016/j.rse.2004.10.011.
Lewis, D. 2002, Community Mangrove Monitoring Program: Impact Assessment of
Coastal Development on Estuarine Mangrove Environments in Darwin Harbour –
Technical Manual and Final Report, 26/2002
Liu, K., Li, X., Shi, X. & Wang, S. 2008, Monitoring mangrove forest changes using
remote sensing and GIS data with decision-tree learning, Wetlands, vol. 28, 336-
346.
Lopez, J. P. A. 2012, Object-based methods for mapping and monitoring of urban
trees with multi-temporal image analysis, Ph.D. thesis, University of Twente.
Lu, D. 2006, The potential and challenge of remote sensing-based biomass
estimation, International Journal of Remote Sensing, vol. 27, 1297-1328.
Lucas, R. M., Ellison, J. C., Mitchell, A., Donnelly, B., Finlayson, M. & Milne, A.
K. 2002, Use of stereo aerial photography for quantifying changes in the extent and
39
height of mangroves in tropical Australia, Wetlands Ecology and Management, vol.
10, 161-175.
McHugh, P. 2004, DLNG Mangrove Monitoring Program: Monitoring of mangrove
community structure/composition, soil condition and mangrove productivity in
Darwin harbour, Department of Infrastructure, Planning and Environment, Land
Monitoring Branch, Darwin.
McKee, K. L. 1996, Mangrove Ecology: A manual for a field course, eds I. C. Feller
& M. Sitnik, Smithsonian Environmental Research Centre, Washington DC.
Metcalfe, K. 2007, The Biological diversity, Recovery from Disturbance and
Rehabilitation of Mangroves in Darwin Harbour, Northern Territory, Doctor of
Philosophy thesis, Charles Darwin University, Darwin.
Metcalfe, K., Franklin, D. C. & McGuinness, K. A. 2011, Mangrove litter fall:
Extrapolation from traps to a large tropical macro-tidal harbour, Estuarine, Coastal
and Shelf Science, vol. 95, 245-252.
Mitchard, E. T. A., Saatchi, S. S., White, L. J. T., Abernethy, K. A., Jeffery, K. J.,
Lewis, S. L., Collins, M., Lefsky, M. A., Leal, M. E., Woodhouse, I. H. & Meir, P.
2012, Mapping tropical forest biomass with radar and spaceborne LiDAR in Lop´e
National Park, Gabon: overcoming problems of high biomass and persistent cloud,
Biogeosciences, vol. 9, 179-191 10.5194/bg-9-179-2012.
40
Mitchell, A. L., Lucas, R. M., Donnelly, B. E., Pfilzner, K., Milne, A. K. &
Finlayson, M. 2007, A new map of mangroves for Kakadu National Park, Northern
Australia, based on stereo aerial photography, Aquatic Conservation: Marine and
Fresh Water Ecosystems, vol. 17, 446-467.
Moritz-Zimmermann, A. 2002, Darwin Harbour Mangrove Monitoring Methodology
Technical Manual, Department of Infrastructure Planning and Environment
Mountrakis, G., Im, J. & Ogole, C. 2011, Support vector machines in remote sensing:
A review, International Journal of Photogrammetry and Remote Sensing, vol. 66,
247-259.
Muad, A. M. & Foody, G. M. 2010, Super-resolution mapping of landscape objects
from coarse spatial resolution imagery, in E. A. Addink & F. M. B. V. Coillie (eds),
ISPRS-GEOBIA 2010: Geographic Object-Based Image Analysis Ghent, Belgium,
vol. XXXVIII-4/C7, 2010.
Mutanga, O., Adam, E. & Cho, M. A. 2012, High density biomass estimation for
wetland vegetation using WorldView-2 imagery and random forest regression
algorithm, International Journal of Applied Earth Observation and Geoinformation,
vol. 18, 399-406 http://dx.doi.org/10.1016/j.jag.2012.03.012.
Nandy, S. & Kushwasha, S. P. S. 2010, Study on the utility of IRS LISS-III data and
the classification techniques for mapping of Sunderban mangroves, Journal of
Coastal Conservation, vol. 15, 123-137 DOI 10.1007/s11852-010-0126-z.
41
Northern Territory Government 2002a, Mangrove management in the Northern
Territory, Department of Infrastructure, Planning and Environment.
Northern Territory Government 2002b, Mangroves,
<http://www.nretas.nt.gov.au/__data/assets/pdf_file/0015/21048/mangroves_17.pdf>
Northern Territory Government 2010, Darwin Harbour - Site of Conservation
Significant, Department of Natural Resources, Environment, The Arts and Sport
NRETAS 2010a, Harbour health and water sampling, Department of Land and
Resource Management, <http://lrm.nt.gov.au/water/aquatic/water_sampling>
NRETAS 2010b, Water Quality Objectives for the Darwin Harbour Region-
Background Document, Department of Natural Resources, Environment, The Arts
and Sport, Aquatic unit, Palmerston, NT, Australia.
Pajares, G., Guijarro, M. & Ribeiro, A. 2010, A Hopfield Neural Network for
combining classifiers applied to textured images, Neural Networks, vol. 23, 144-
153.
Pouliot, D. A., King, D. J., Bell, F. W. & Pitt, D. G. 2002, Automated tree crown
detection and delineation in high-resolution digital camera imagery of coniferous
forest regeneration, Remote Sensing of Environment, vol. 82, 322-334
10.1016/s0034-4257(02)00050-0.
42
Quoc, T. V., Oppelt, N., Leinenkugel, P. & Kuenzer, C. 2013, Remote Sensing in
Mapping Mangrove Ecosystem-An Object-Based Approach, Remote Sensing, vol. 5,
183-201.
Ramsey, E. W. & Jensen, J. R. 1996, Remote Sensing of Mangrove Wetlands:
Relating Canopy Spectra to Site-Specific Data, Photogrammetric Engineering and
Remote Sensing, vol. 62, 939-948.
Roy, P. S. & Ravan, S. A. 1996, Biomass estimation using satellite remote sensing
data-An investigation on possible approaches for natural forest, Journal of
Bioscience, vol. 21, 535-561.
Satyanarayana, B., Mohamad, K. A., Idris, I. F., Husain, M.-L. & Dahdouh-Guebas,
F. 2011, Assessment of mangrove vegetation based on remote sensing and ground-
truth measurements at Tumpat, Kelantan Delta, East Coast of Peninsular Malaysia,
International Journal of Remote Sensing, vol. 32, 1635-1650.
Serrano, L., Filella, I. & Peñuelas, J. 2000, Remote Sensing of Biomass and Yield of
Winter Wheat under Different Nitrogen Supplies, Crop Sci., vol. 40, 723-731
10.2135/cropsci2000.403723x.
Skinner, L., Townsend, S. & Fortune, J. 2009, The Impact of Urban Land-use on
Total Pollutant Loads Entering Darwin Harbour, Northern Territory Government,
Aquatic Health Unit, Department of Natural Resources, Environment, the Arts and
Sport.
43
Stephen Crooks, Dorothée Herr, Jerker Tamelander, Dan Laffoley & Vandever, J.
2011, Mitigating Climate Change through Restoration and Management of Coastal
Wetlands and Near-shore Marine Ecosystems - Challenges and Opportunities,
Environment Department Paper 121, World Bank, Washington, DC.
Sulong, I., Mohd-Lokman, H., Mohd-Tarmizi, K. & Ismail, A. 2002, Mangrove
mapping using LANDSAT imagery and aerial photographs: Kemaman district,
Terengganu, Malaysia, Environmental Development and Sustainability, vol. 4, 135-
152.
Suratman, M. N. 2008, Carbon Sequestration Potential of Mangroves in Southeast
Asia, in F. Bravo, R. Jandl, V. LeMay & K. Gadow (eds), Managing Forest
Ecosystems: The Challenge of Climate Change, Springer Netherlands, vol. 17, pp.
297-315, DOI 10.1007/978-1-4020-8343-3_17, <http://dx.doi.org/10.1007/978-1-
4020-8343-3_17>.
The Darwin Harbour Advisory Committee 2003, Darwin Harbour Regional Plan of
Management, <http://www.lrm.nt.gov.au/water/darwin-harbour/regional>
Vaiphasa, C., Ongsmwang, S., Vaiphasa, T. & Skidmore, A. K. 2005, Tropical
mangrove species discrimination using hyperspectral data: A laboratory study,
Estuarine, Coastal and Shelf Science, vol. 65, 371-379.
Vaiphasa, C., Skidmore, A. K. & Boer, W. F. d. 2006, A post-classifier for mangrove
mapping using ecological data, ISPRS Journal of Photogrammetry and Remote
Sensing, vol. 61, 1-10.
44
Verheyden, A., Dahdouh-Guebas, F., Thomaes, K., Genst, W. d., Hettiarachchi, S. &
Koedam, N. 2002, High-resolution vegetation data for mangrove research as
obtained from aerial photography, Environmental Development and Sustainability,
vol. 4, 113-133.
Viña, A., Gitelson, A. A., Nguy-Robertson, A. L. & Peng, Y. 2011, Comparison of
different vegetation indices for the remote assessment of green leaf area index of
crops, Remote Sensing of Environment, vol. 115, 3468-3478
http://dx.doi.org/10.1016/j.rse.2011.08.010.
Wang, L. & Sousa, W. P. 2009, Distinguishing mangrove species with laboratory
measurements of hyperspectral leaf reflectance, International Journal of Remote
Sensing, vol. 30, 1267-1281.
Wang, L., Sousa, W. P., Gong, P. & Biging, G. S. 2004, Comparison of IKONOS
and QuickBird images for mapping mangrove species on the Caribbean coast of
Panama, Remote Sensing of Environment, vol. 91, 432-440.
Wannasiri, W., Nagai, M., Honda, K., Santitamnont, P. & Miphokasap, P. 2013,
Extraction of Mangrove Biophysical Parameters Using Airborne LiDAR, Remote
Sensing, vol. 5, 1787.
Water Monitoring Branch, 2005, The Health of the Aquatic Environment in the
Darwin Harbour Region, 2004, 5/2005D, <http://www.nretas.nt.gov.au/natural-
resource-management/water/aquatic/ausrivas/darwinharbour>
45
Woodroffe, C. D., Bardsley, K. N., Ward, P. J. & Hanley, J. R. 1988, Production of
Mangrove Litter in a Macrotidal Embayment, Darwin Harbour, N.T., Australia,
Estuarine, Coastal and Shelf Science, vol. 26, 581-598.
Zhang, C., Liu, Y., Kovacs, J. M., Flores-Verdugo, F., Flores-de-Santiago, F. &
Chen, K. 2012, Spectral response to varying levels of leaf pigments collected from a
degraded mangrove forest, Journal of Applied Remote Sensing, vol. 6.
46
Chapter 2 - Mangrove Species Identification:
Comparing WorldView-2 with Aerial
Photographs
This work has been published in the following paper:
Heenkenda, M., Joyce, K., Maier, S. & Bartolo, R. 2014, Mangrove
Species Identification: ComparingWorldView-2 with Aerial Photographs,
Remote Sensing, vol. 6, 6064-6088
Publisher confirmed reproducing the journal article without written
permission since “Remote Sensing” journal is an Open Access journal
(Appendix 1).
47
2.0 Abstract
Remote sensing plays a critical role in mapping and monitoring mangroves.
Aerial photographs and visual image interpretation techniques have historically been
known to be the most common approach for mapping mangroves and species
discrimination. However, with the availability of increased spectral resolution
satellite imagery, and advances in digital image classification algorithms, there is
now a potential to digitally classify mangroves to the species level. This study
compares the accuracy of mangrove species maps derived from two different layer
combinations of WorldView-2 images with those generated using high resolution
aerial photographs captured by an UltraCamD camera over Rapid Creek coastal
mangrove forest, Darwin, Australia. Mangrove and non-mangrove areas were
discriminated using object-based image classification. Mangrove areas were then
further classified into species using a support vector machine algorithm with best-fit
parameters. Overall classification accuracy for the WorldView-2 data within the
visible range was 89%. Kappa statistics provided a strong correlation between the
classification and validation data. In contrast to this accuracy, the error matrix for the
automated classification of aerial photographs indicated less promising results. In
summary, it can be concluded that mangrove species mapping using a support vector
machine algorithm is more successful with WorldView-2 data than with aerial
photographs.
Keywords: mangrove species mapping; object-based image analysis; support vector
machine; WorldView-2; aerial photograph
48
2.1 Introduction
Mangroves are salt-tolerant evergreen forests that create land-ocean interface
ecosystems. They are found in the intertidal zones of marine, coastal or estuarine
ecosystems of 124 tropical and sub-tropical countries and areas [1]. Mangroves are a
significant habitat for sustaining biodiversity and also provide direct and indirect
benefits to human activities.
Despite the increased recognition of their socio-economic benefits to coastal
communities, mangroves are identified as among the most threatened habitats in the
world [2]. Degradation and clearing of mangrove habitats is occurring on a global
scale due to urbanization, population growth, water diversion, aquaculture, and salt-
pond construction [3].
In recent years, numerous studies have been undertaken to further understand
the economic and ecological values of mangrove ecosystems and to provide a means
for effective management of these resources [1, 2, 4–6]. Mangrove forests are often
very difficult to access for the purposes of extensive field sampling, therefore
remotely sensed data have been widely used in mapping, assessing, and monitoring
mangroves [4, 7–10].
According to Adam et al. [11], when using remote sensing techniques for
mapping wetland vegetation, there are two major challenges to be overcome. Firstly,
the accurate demarcation of vegetation community boundaries is difficult, due to the
high spectral and spatial variability of the communities. Secondly, spectral
reflectance values of wetland vegetation are often mixed with that of underlying wet
49
soil and water. That is, underlying wet soil and water will attenuate the signal of the
near-infrared to mid-infrared bands. As a result, the confusion among mangroves,
other vegetation, urban areas and mudflats will decrease map classification accuracy
[8, 12–14]. Consequently, remote sensing data and methods that have been
successfully used for classifying terrestrial vegetation communities cannot be applied
to mangrove studies with the same success.
Using remote sensing to map mangroves to a species level within a study area
presents further challenges. For instance, Green et al. [8] reviewed different
traditional approaches for satellite remote sensing of mangrove forests. After testing
them on different data sources, the study confirmed that the type of data can
influence the final outcome. Heumann [10] further demonstrated the limitations of
mapping mangrove species compositions using high resolution remotely sensed data.
The potential for using hyperspectral remote sensing data for wetland vegetation has
been discussed in numerous studies, however the results are still inconclusive when
considering mangrove species discrimination [10, 11, 13, 15]. Therefore, the
selection of data sources for mangrove mapping should include consideration of ideal
spectral and spatial resolution for the species.
Some laboratory studies using field spectrometers have suggested the ideal
spectral range for mangrove species discrimination [8, 16–18]. In one such study,
Vaiphasa et al. [16] investigated 16 mangrove species and concluded that they are
mostly separable at only a few spectral locations within the 350–2500 nm region.
The study didn’t specifically explore the use of airborne or spaceborne hyperspectral
sensors for mangrove species mapping (such as the best spectral range, number of
bands within that range, and optimal spatial resolution). It has, however, encouraged
50
further full-scale investigations on mangrove species discrimination, which could
involve extensive field and laboratory investigations, necessitating high financial and
time investments. A viable alternative may be to use satellite data with narrow
spectral bands lying within the ideal spectral range for species composition
identification.
The WorldView-2 (WV2) satellite imaging sensor provides such data, and
therefore may have increased potential for accurately mapping the distribution of
mangrove species. Its combination of narrow spectral bands and high spatial
resolution provides benefits over the other freely or commercially available satellite
remote sensing systems albeit at a cost. Therefore, the combination of WV2 data
with advanced image processing techniques will be an added value to wetland
remote sensing.
After determining ideal or optimal image data requirements, the selection of
an appropriate method of processing those data for mapping with maximum
achievable accuracy is critical. Kuenzer et al. [19] provided a detailed review of
mangrove ecosystem remote sensing over the last 20 years, and emphasized the need
for exploitation of new sensor systems and advanced image processing approaches
for mangrove mapping. The most promising results for mangrove mapping can be
found in the study by Heumann [20], who demonstrated high accuracy when
discriminating mangroves from other vegetation using a combination of WV2, and
QuickBird satellite images. However, the accuracy was poor for mangrove mapping
to the species level. Though the advanced technological nature of remotely sensed
images demands solutions for different image-based applications, it can be concluded
51
that little has been adapted to mangrove environments compared to other terrestrial
ecosystems.
An approach that may prove fruitful for mapping mangrove environments is
the Support Vector Machine (SVM) algorithm, a useful tool that minimizes structural
risk or classification errors [20]. SVM is a supervised, non-linear, machine learning
algorithm that produces good classification outcomes for complex and noisy data
with fewer training samples. SVM can be used with high dimensional data as well
[21]. Though this is a relatively new technique for mangrove mapping, it has widely
been applied in other remote sensing application domains with different sensors [22].
For example, Huang et al. [23] compared SVM with three different traditional image
classifiers, and obtained significantly increased land cover classification accuracy
with an SVM classifier.
This study aims to investigate the potential of using high spatial resolution
remote sensing data for discriminating mangroves at a species level. In order to
achieve this aim, three objectives were identified: (a) to identify and extract
mangrove coverage from other vegetation; (b) to apply the SVM algorithm to
distinguish individual mangrove species; and (c) to compare the accuracies of these
techniques when using WV2 and aerial photographs in order to identify the most
accurate combination of data input and image processing technique.
52
2.2 Data and Methods
2.2.1. Study Area
This study focused on a mangrove forest within a small coastal creek system:
Rapid Creek in urban Darwin, Australia (Figure 2.1). It is situated on the north
western coast line of the Northern Territory centered at latitude 12°22′S, and
longitude 130°51′E. This area represents relatively diverse, spatially complex, and
common mangrove communities of the Northern Territory, Australia, and is one of
the more accessible areas in the region for field survey. The aerial extent of the study
area shown in the Figure 1 is approximately 400 hectares.
According to the mangrove classification of Brocklehurst and Edmeades [24],
Avicennia marina (Grey mangroves) and Ceriops tagal (Yellow mangroves) are the
most dominant species for this area, though that study was completed nearly 20 years
ago. Based on the more recent study of Ferwerda et al. [25], there are some other
species such as Bruguiera exaristata (Orange mangroves), Lumnitzera racemosa
(Black mangroves), Rhizophora stylosa (Stilt mangroves), and Aegialitis annulata
(Club mangroves) in the Rapid Creek mangrove forest. However, clear boundaries of
individual mangrove species in this area have not been demarcated by any previous
study.
53
Figure 2.1. The study area located in coastal mangrove forest of Rapid Creek in
Darwin, Northern Territory, Australia; WV2 data © Digital Globe. (Coordinate
system: Universal Transverse Mercator Zone 52 L, WGS84).
2.2.2. Field Survey
Field data were collected during January to April 2013, in the Rapid Creek
mangrove forest. Unfortunately, this did not coincide with the overpass of either
sensor utilized (June 2010). However, the field data can still be considered valid for
calibration and validation purposes, as it is unlikely that any large shifts in species
composition of the site would have been experienced since that time [25]. Further,
54
there was no record of any natural or human disturbances that could have had a
devastating effect on the mangroves during this period. In any case, field sample sites
were located away from edges and transition zones to avoid any errors in
classification due to growth, regeneration, or vegetation decline. After observing the
mangrove zonation patterns visible in the WV2 image, a random sampling pattern
within zones was adopted to identify species.
A sample was defined as a homogenous area of at least 4 m2, which is at least
16 pixels in the WV2 image. Coordinates of these sample polygons and available
species were recorded using a non-differentially corrected Global Positioning System
(GPS). Special attention was given to orient the sample plots in north-south and east-
west directions in order to easier locate them in the images. To overcome GPS
inaccuracies, the plots were located with respect to natural features on the ground as
far as possible. For instance: the distance to water features, roads or edges of mudflats
were recorded. Further, 10 readings were averaged for the final location. Mangrove
species in the field were identified using the Mangrove Plant Identification Tool Kit
(published by Greening Australia Northern Territory) and The Authoritative Guide to
Australia’s Mangroves (published by University of Queensland) [26,27].
2.2.3. Remote Sensing Data and Pre-Processing
A WV2 satellite image was selected as the main image data source for this
study. As the sensor produces images with high spectral and spatial resolution, WV2
is an ideal solution for vegetation and plant studies [28]. The image was acquired on
5 June 2010, with 8 multispectral bands at 2.0m spatial resolution and a
panchromatic band with 0.5m spatial resolution. To compare this work with higher
55
spatial resolution remote sensing data (Table 2.1), true color aerial photographs were
used, which were acquired on 7June 2010 using an UltraCamD large format digital
camera [29].
Table 2.1. Spectral band information of WV2 image and aerial photographs obtained
from UltraCamD camera [29, 30].
Band Spectral Range (nm) Spatial Resolution (m)
WorldView-2
Panchromatic 447–808 0.5
Coastal 396–458 2.0
Blue 442–515 2.0
Green 506–586 2.0
Yellow 584–632 2.0
Red 624–694 2.0
Red-Edge 699–749 2.0
NIR1 765–901 2.0
NIR2 856–1043 2.0
Aerial photographs
Blue 380–600 0.14
Green 480–700 0.14
Red 580–720 0.14
WV2 images were radiometrically corrected according to the sensor
specifications published by DigitalGlobe® [31]. Digital numbers were converted to
at-sensor radiance values, and then to top-of-atmosphere reflectance values. The
56
additive path radiance was removed using the dark pixel subtraction technique in
ENVI 5.0 software. Images were geo-referenced using rational polynomial
coefficients provided with the images, and ground control points extracted from
digital topographic maps of Darwin, Australia.
In order to utilize both the high spatial and spectral resolution options
provided with WV2 panchromatic and multispectral layers, pan-sharpening options
were investigated. Pan-sharpening is defined as a pixel fusion method that increases
the spatial resolution of multi-spectral images [32]. Although there are several
methods available for pan-sharpening, the high pass filter method was selected
because it is known to be one of the best which produces a fused image without
distorting the spectral balance of the original image [33, 34]. Once applied, the
statistical values of the spectral information of the input and output multispectral
products are similar. Palubinskas [35]’s study also proposed the high pass filter
method as a fast, simple and good pan-sharpening method for WV2 images, by
analyzing performances of several image fusion methods. In addition, this method is
known to be one of the best choices when the pixel resolution ratio of higher to lower
is greater than 6:1 [36]. However, special attention must be given when selecting the
filter kernel size as it should reflect the radiometric normalization between the two
images. Chavez et al. [33] stated that twice the pixel resolution ratio is an ideal
solution for the kernel size. This means that a kernel size: 15 × 15 is an optimal
solution for WV2 data. All image radiometric, atmospheric and geometric
corrections must be done prior to pan-sharpening in order to minimize geometric and
radiometric errors.
57
Accordingly, the WV2 multispectral image was then pan-sharpened to 0.5 m
spatial resolution to incorporate the edge information from the high spatial resolution
panchromatic band into the lower spatial resolution multispectral bands using the
high pass filter pan-sharpening method. The Coastal band and the NIR2 band were
not used for further processing because of the limited spectral range of the
panchromatic band (Table 2.1).
The aerial photographs were oriented to ground coordinates following digital
photogrammetric image orientation steps in the Leica Photogrammetric Suite (LPS)
software using image orientation parameters extracted from the camera calibration
report. The coordinates of ground control points were extracted from Australian
geographic reference stations near Darwin, Australia and digital topographic maps of
Darwin [37]. The ortho-photograph of the area was then generated to achieve a
geometrically and topographically corrected image with a resultant resolution of 14 cm
for further studies. Radiometric calibration information was not available.
In order to directly compare with WV2 imagery, another ortho-photo was
created with a pixel size of 0.5m. In order to remove the artifacts that can be created
from resampling, a low pass filter was applied to the raw aerial photographs first.
Ortho-rectification was then completed using the cubic convolution interpolation.
2.2.4. Image Classification
Image classification of WV2 and aerial data was undertaken in two steps.
Firstly, mangroves and non-mangroves were separated using eCognition Developer
8.7 software. Then the classification was refined to discriminate mangroves at a
58
species level using ENVI 5.0 software. The process was applied to two sets of WV2
images: the first WV2 image with a spatial resolution of 2m (without pan-
sharpening) and the other one with a spatial resolution of 0.5 m (with pan-
sharpening), and two sets of ortho-photographs: the first with a spatial resolution of
0.14 m (AP0.14M) and the other one with a spatial resolution of 0.5 m (AP0.5M).
This was done in order to investigate the influence of spatial resolution and the pan-
sharpening effect on classification accuracy. Figure 2.2 shows the overall workflow
diagram for this study, which is described in greater detail below.
59
Figure 2.2. Mangrove species mapping using remotely sensed data. Image
segmentation and initial classification was completed in eCognition Developer 8.7
software. The support vector machine algorithm was implemented in the ENVI-IDL
environment for species classification.
2.2.4.1. Separating Mangroves and Non-Mangroves
To separate mangroves from other features, object-based image analysis (OBIA)
was used. OBIA is based on segmentation, which partitions the image into
meaningful, spatially continuous and spectrally homogeneous objects or pixel groups
[15, 20]. The major challenge is in determining appropriate similarity measures
which discriminate objects from each other. Therefore, the spectral profiles of
identifiable features in the satellite image were analyzed.
Class-specific rules were developed incorporating contextual information from the
WV2 image and relationships between image objects at different hierarchical levels,
to separate mangroves and non-mangroves. The segmentation at level 1 identified
objects that can be grouped to coarse classification structures. All spectral bands
ranging from Blue to NIR1 were used, and weights were assigned as 1 for the
segmentation at each level. The segmentation parameters were selected based on the
pixel size and the expected compactness of resulting objects. Buildings, soil, roads
and mudflats were classified as “Buildings-Roads-Mudflats”, and as the spectral
reflectance ratio of Yellow to NIR1 of this class was less than the mean ratio of
Yellow to NIR1 of objects, the ratio was introduced as a threshold value for the
classification to extract “Buildings-Roads-Mudflats”. The brightness calculated from
reflectance values from Blue to NIR1 bands were analyzed, and the low values (less
than the mean brightness of objects) were used to extract water features. The
60
remaining objects at this level were classified as “Candidate-Mangrove-1” claiming
this class as the parent objects for the next level.
At level 2, specific details (home gardens and other vegetation) within parent
objects were identified (Figure 2.3). Home gardens and other vegetation were
removed considering the objects enclosed by “Buildings-Roads-Mudflats” class.
Further, remaining home gardens and other vegetation were identified using a red
edge normalized difference index (NDVIre) obtained from NIR1 and Red-Edge
bands:
𝑁𝑁𝐷𝐷𝑁𝑁𝑁𝑁𝑟𝑟𝑟𝑟 = �𝑅𝑅𝑁𝑁𝑁𝑁𝑁𝑁1−𝑅𝑅𝑁𝑁𝑅𝑅𝑅𝑅−𝐸𝐸𝑅𝑅𝐸𝐸𝑅𝑅��𝑅𝑅𝑁𝑁𝑁𝑁𝑁𝑁1−𝑅𝑅𝑁𝑁𝑅𝑅𝑅𝑅−𝐸𝐸𝑅𝑅𝐸𝐸𝑅𝑅�
(1)
With 𝑅𝑅𝑁𝑁𝑁𝑁𝑅𝑅1 and 𝑅𝑅𝑅𝑅𝑟𝑟𝑅𝑅−𝐸𝐸𝑅𝑅𝐸𝐸𝑟𝑟 being the reflectance in the NIR1 and Red-Edge
bands respectively [38, 39]. The index was used assuming that it would provide a
good measure of biophysical properties of plants: chlorophyll content and water-
filled cellular structures to separate these classes from others due to the rapid change
in reflectance of vegetation in the Red-Edge region [39].
As the NDVIre of home gardens and other vegetation classes were higher than
the mean NDVIre of objects, the value was introduced as a threshold value for the
classification. The remaining objects from this level were classified as
“Candidate-Mangrove-2”.
At the final level, the normalized difference vegetation index (NDVI)
obtained from NIR1 and Red bands:
61
𝑁𝑁𝐷𝐷𝑁𝑁𝑁𝑁 = (𝑅𝑅𝑁𝑁𝑁𝑁𝑁𝑁1−𝑅𝑅𝑁𝑁𝑅𝑅𝑅𝑅)(𝑅𝑅𝑁𝑁𝑁𝑁𝑁𝑁1−𝑅𝑅𝑁𝑁𝑅𝑅𝑅𝑅) (2)
Where 𝑅𝑅𝑁𝑁𝑁𝑁𝑅𝑅1 and 𝑅𝑅𝑅𝑅𝑟𝑟𝑅𝑅 are the reflectance in the NIR1 and Red bands
respectively, was used to separate “Final-Mangroves”. This vegetation index was
also introduced to the classification as a threshold value since “Final-Mangroves”
class have higher NDVI values than the mean NDVI values of the objects. The
classified objects were closely analyzed for final refinements. Refinements were
done to classify objects by incorporating object geometry and neighborhood
information to the process. For example: the relation to the neighboring borders was
analyzed. The transferability of the rule set was maintained using variables instead of
specific values for class separation. Finally, the outline of the mangrove area was
extracted for further analysis. This method was tested on both sets of WV2 images.
62
Figure 2.3. The schematic explanation of image objects at different hierarchical
levels. The first level was at pixel level. At the second level: vegetation and
non-vegetation areas were identified. Mangrove areas were extracted at the final
level.
The same process was then applied to the two sets of aerial photographs
(AP0.14M and AP0.5M). However, when segmenting AP0.14M, different
segmentation parameters were used at each level due to its higher spatial resolution.
When applying OBIA to aerial photographs, the possibility of developing a
successful rule set for the segmentation is limited due to the broad band width and
limited number of spectral bands of the aerial camera. Further, limitations associated
with radiometric resolution of aerial photographs could affect the accuracies. As a
result, the final mangrove area was manually edited to remove objects of known
home gardens, other vegetation, and grasslands.
2.2.4.2. Mangrove Species Classification
With the increase in mangrove studies and mapping using remote sensing
comes a growing implementation of advanced processing techniques. Although
traditional remote sensing can provide important information for monitoring the
ecosystem, changes, and extent of mangroves, more contextual and probabilistic
methods can be utilized to improve the accuracy of classification, and for
discriminating individual species.
Traditional land cover hard classification is based on the assumption that each
pixel corresponds to a single class [40]. This is not always true. When the
63
instantaneous field of view of the sensor covers more than one class of land cover or
objects, the pixel may have reflectance values from more than one class, and is
defined as a mixed pixel [41]. Mangroves are closed forests which can become very
dense due to their limited habitat range. Mixed pixels therefore have to be expected
in the image. Hence, traditional pixel based image analysis techniques do not fully
exploit the contextual variations in species distribution [10, 11, 20]. A better
alternative is soft classification, which predicts the proportion of each land cover
class within each pixel, resulting in more informative representation of land cover
[40, 42, 43].
One of the soft classification algorithms is the SVM, which locates a best
non-linear boundary in higher dimensional feature space. It works with pixels that
are in the vicinity of classes, while minimizing over-fitting errors of training data
[21]. Hence, a small training set is sufficient to achieve accurate classifications. The
mathematical formation and detailed description of the SVM architecture can be
found in Tso and Mather [21] and Mountrakis et al. [22].
When designing the SVM architecture, careful selection of a kernel function
is important to increase the accuracy of the classification. The position of the
decision boundary always varies with the kernel function and its parameters [21, 22,
44]. Descriptive information about the kernel function and its parameter selection
can be found in Tso and Mather [21].
The extracted mangrove areas were classified using the SVM algorithm. Field
data was collected to represent the five main mangrove species occurring in the study
area: Avicennia marina, Ceriops tagal, Bruguiera exaristata, Lumnitzera racemosa,
64
and Rhizophora stylosa. Sonneratia alba, Excoecaria ovalis, and Aegialitis annulata
have not been used for the classification mainly due to low coverage. In addition,
Sonneratia alba does not exhibit distinctive spectral variation compared to other
species (Figure 2.4b). Field data were then divided into two groups, for training (69
samples) and for validation (47 samples) of the classifiers. The multiclass SVM
classifier developed by Canty [45] was modified and implemented in ENVI
extensions in the IDL environment in order to define the case specific parameters
with an iterative process. This helped to determine the best fitting parameters for this
study. The Radial Basis Function (RBF) was found to be the best kernel function
with a gamma value of 0.09, and penalty parameter of 10.
The SVM algorithm was applied to two sets of band combinations of the
WV2 image. One set uses the visible spectral range, i.e., blue, green, yellow, red and
red-edge bands (WV2-VIS) in order to directly compare it with aerial photographs.
The other set consisted of the red, red-edge and NIR1 bands (WV2-R/NIR1). These
bands were selected based on the research by Wang and Sousa [18]. The study
indicated that the majority of mangrove species which occur in the Rapid Creek area
can be discriminated using the influential wavelengths: 630, 780, 790, 800, 1480,
1530 and 1550 nm spectral bands (which correspond to the red and NIR bands of
WV2). The pan-sharpened sets of images were named as PS-WV2-VIS and PS-
WV2-R/NIR1 for easy reference. The same process was applied to AP0.14M and
AP0.5M. The same training data were used for all four datasets to maintain
consistency between methods.
65
2.2.5. Accuracy Assessment
The accuracy of the mangrove species classification was assessed at the pixel
level using descriptive and analytical statistical techniques. The 560 random
validation points were generated inside the field samples (2 m × 2 m plots).
Therefore, there were a large number of validation points for accuracy assessment for
each species (Table 2.2). The generated maps were visually inspected against field
observations, satellite images and aerial photographs according to Congalton [46]. A
confusion matrix was generated, and users’ and producers’ accuracies, together with
kappa statistics, were investigated for each identified mangrove species.
2.3. Results
2.3.1 Field Survey
There are five mangrove species which are most abundant and can easily be
identified along Rapid Creek: Avicennia marina, Ceriops tagal, Bruguiera
exaristata, Lumnitzera racemosa, and Rhizophora stylosa (Stilt mangroves).
Avicennia marina and Ceriops tagal are the most widely spread species in this area,
while Lumnitzera racemosa covers the majority of the hinterland area. Sonneratia
alba (Apple mangroves) are found at two locations within the site, covering
approximately 20 square meters. Excoecaria ovalis (Milky mangroves) and
Aegialitis annulata (Club mangroves) were also recognized during the field
investigation, though they do not represent significant coverage with in the forest.
66
2.3.2Separating Mangroves and Non-Mangroves
The analysis of spectral profiles within the Rapid Creek coastal area was the
key to introducing class specific rules for OBIA (Figure 2.4).
Figure 2.4. Spectral profiles of: (a) all features except mangroves extracted from
WV2 image; (b) mangrove species only extracted from WV2 image; (c) mangrove
species extracted from aerial photographs of spatial resolution 0.14 m; and (d)
mangrove species extracted from aerial photographs of spatial resolution 0.5 m.
(AM-Avicennia marina, CT-Ceriops tagal, BE-Bruguiera exaristata, LR-Lumnitzera
racemosa, RS-Rhizophora stylosa, SA-Sonneratiaalba).
67
Buildings, soil, mudflats and water showed highly distinctive spectral profiles
compared to other features. The mangrove species are most notably separable within
the range of wavelengths of 478.3nm and 832.9nm, with the exception of Sonneratia
alba, while the spectral profile of Avicennia marina generally is more distinctive
from other species across a broader range of the spectrum.
The locations of field samples are shown in Figure 2.5. The mangrove outline
was successfully extracted from the WV2 image using OBIA. The “Final-
Mangroves” class extracted from the WV2 image (without pan-sharpening) and
aerial images did, however, still include both mangroves and the adjacent home
gardens and other vegetation to be edited manually. The WV2 image was more
useful than the aerial image for extracting the overall mangrove coverage, as less
manual editing was required.
68
Figure 2.5. The locations of field samples available for calibration and validation
purposes, and mangrove coverage extracted from the WV2 image.
Table 2.2 shows the number of sample points that were used for validation for
each species together with the number of samples used for training the classification.
Sample points were generated from47 validation samples shown in Figure 2.5.
Table 2.2. Number of samples (4 m2 each) used for training and number of sample
points generated for validation for each mangrove species.
Species
No. of Samples for
Training
(4 m2 or 16 Pixels Each)
No. of Points for Validating
the Classification
Avicennia marina (AM) 22 216
Bruguiera exaristata (BE) 14 106
Ceriops tagal (CT) 10 80
Lumnitzera racemosa (LR) 12 78
Rhizophora stylosa (RS) 11 96
2.3.3 Mangrove Species Classification
Figure 2.6 shows the derived mangrove species maps. Figure 2.6a was
produced from the PS-WV2-VIS image, and the visual appearance is more closely
related to the dominating species of the area than the other five maps
(Figures 2.6 (b) – (f)). Avicennia marina (AM) and Ceriops tagal (CT) dominate
69% of total mangrove area while Bruguiera exaristata (BE) accounts for only 5%.
69
The majority of the hinterland margin is occupied by Lumnitzera racemosa (LR) or
mixed Lumnitzera racemosa, Bruguiera exaristata and Ceriops tagal. Rhizophora
stylosa (RS) dominates only along the creek and its branches (Figure 2.6a). However,
the visual assessment confirmed that some of the Ceriops tagal has been
misclassified as Bruguiera exaristata (especially on the west of the study area
Figure 2.6a).
Figure 2.6. (a) Classification of pan-sharpened WV2 image using blue to red-edge
bands; (b) Classification of pan-sharpened WV2 images using red to NIR1 bands;
(c) Classification of WV2 image using blue to red-edge bands; (d) Classification of
70
WV2 images using red to NIR1 bands; (e) Classification of aerial photographs with
0.14 m spatial resolution; and (f) Classification of aerial photographs with 0.5 m
spatial resolution.
When testing the same method with the PS-WV2-R/NIR1 image, there was
no significant difference in visual appearance of the classification except for the
classes Rhizophora stylosa and Bruguiera exaristata (Figure 2.6a and b). The
hinterland margin and areas along the water features were successfully classified
with their dominated species. Furthermore, the visual appearance of classification of
Rhizophora stylosa and Ceriops tagal are better than that of PS-WV2-VIS
classification.
Figure 2.7 shows an example of positive detection of Lumnitzera racemosa
(orange colour) at hinterland from pan-sharpened WV2 image classifications.
However, when considering the classification results of WV2 image with 2 m spatial
resolution, the detection of Lumnitzera racemosa and Bruguiera exaristata was
relatively poor (Figures 2.7 h and i).
71
Figure 2.7. The comparison of the WorldView2 image and different classification
approaches: (a) Pan-sharpened WV2 image; (b) Classification results of PS-WV2-
VIS; (c) Classification results of PS-WV2-R/NIR1; (d) Aerial photograph;
(e) Classification results of AP0.14M; (f) Classification results of AP0.5M; (g) WV2
image; (h) Classification results of WV2-VIS; (i) Classification results of WV2-
R/NIR1.
72
Figure 2.6c and d show classifications of the WV2 image without pan-
sharpening. In both instances, most of the Avicennia marina was classified correctly,
especially in the classification of WV2-R/NIR1. In addition, there was no significant
detection of Bruguiera exaristata and Ceriops tagal classes in either classification.
The classification results of WV2-R/NIR1 shows misclassification of Ceriops tagal
and Lumnitzera racemose as Rhizophora stylosa. Further, both results were not able
to capture the mangrove zonation pattern observed in the field.
The accuracy assessment of the PS-WV2-R/NIR1 classification revealed
approximately same overall accuracy as the PS-WV2-VIS classification. Although
the total extent of Rhizophora stylosa and Ceriops tagal was approximately
equivalent to the PS-WV2-VIS image classification, the area covered by Ceriops
tagal was smaller than that class in the PS-WV2-VIS image classification
(Figure2.8a and b). Some of the Ceriops tagal areas may be misclassified as
Rhizophora stylosa and Bruguiera exaristata, because the percentage of the extents
of the other species didn’t change significantly (Figure 2.8). The extents of
Lumnitzera racemosa was almost the same in both instances while the extent of
Avicennia marina was reduced by approximately 2.5 hectares compared to the
PS-WV2-VIS classification.
73
Figure 2.8. Extents (%) of mangrove species obtained using six different input
datasets: (a) pan-sharpened WV2 image with bands blue to red-edge; (b) pan-
sharpened WV2 image with bands red to NIR; (c) aerial photographs with 0.14 m
spatial resolution; (d) aerial photographs with 0.5 m spatial resolution; (e) WV2
image with bands blue to red-edge; and (f) WV2 image with bands red to NIR.
The WV2 image with a spatial resolution of 2 m demonstrated rather different
results than the pan-sharpened image classifications. In both instances, the classified
extents of Bruguiera exaristata were less than 1% of the total mangrove area.
However, the extent of Rhizophora stylosa obtained from the WV2-R/NIR1
classification was similar to the pan-sharpened image classification though the visual
appearance indicates some misclassifications around the mudflats and at the edges of
the mangrove coverage (Figure 2.6d). The classification of WV2-VIS did not show
the Ceriops tagal class, whereas the classification of WV2-R/NIR1 represented 2%
74
of the total area (Figure 2.8e and f). The extent of Avicennia marina class was not
considerably different to other pan-sharpened WV2 and aerial image classifications.
The classification results of using the AP0.14M input had several differences
compared to the WV2 classifications (Figure 2.6). The classes derived from the
aerial photographs were patchier, and the classification did not capture the mangrove
zonation pattern described in previous studies of this area well. Figure2.7 shows one
example of capturing mangrove zonation patterns by different classification
approaches. Lumnitzera racemosa class obtained from the classification of AP0.5M
was patchier than others. The percentage of the extent of Avicennia marina is almost
the same at all classifications. Further, the extent of Rhizophora stylosa from the
classification of aerial images is the same as that of the classification of PS-WV2-
VIS image. The extents of Lumnitzera racemosa was significantly larger using the
technique on aerial photographs, while the extent of Ceriops tagal was half that of
the pan-sharpened WV2 classification (Figure 2.8c and d).
When classifying the AP0.5M aerial photograph, there were no significantly
large differences in the extents of all classes compared to the AP0.14M
classifications (Figure2.8c and d). A slight increase in Rhizophora stylosa and
Avicennia marina extents and a decrease in Bruguiera exaristata extent were noted
in the AP0.5M classification compared to the AP0.14M classification. These classes
exhibited fewer contiguous patterns and highly deviated from reality, thus producing
some misclassification.
75
2.3.4 Accuracy Assessment
The map produced from PS-WV2-VIS best visually represents the zonation
pattern of the different species. This classification also has the highest values for
both overall accuracy (89%) and Kappa statistics (0.86, Table 2.3). Despite the
overall accuracy of the PS-WV2-VIS classification being 2% higher than the PS-
WV2-R/NIR1 classification, the results of the Kappa analysis shows these two
classifications were not significantly different whereas the visual appearance of this
classification is better than some of the classes obtained from PS-WV2-VIS.
Table 2.3. Overall accuracy and Kappa statistics obtained for each classification.
PS-WV2-
VIS
PS-WV2-
R/NIR1
WV2-
VIS
WV2-
R/NIR1 AP0.14M AP0.5M
Overall
accuracy 89% 87% 58% 42% 68% 68%
Kappa 0.86 0.84 0.46 0.25 0.60 0.58
Table 2.3 shows the lowest accuracy assessment figures for the maps
generated from non pan-sharpened WV2 images. The Kappa statistics were not
strong enough to represent the good contingency with validation samples. This is
also supported by the visual appearance of these maps (Figure 2.6).
Visual inspection of the maps produced from the aerial photographs
(AP0.14M and AP0.5M) indicated low quality classification results, especially the
76
classification results of AP0.5M. They were not able to capture most of the variations
visible in the photographs. Further, it can be seen that Ceriops tagal and Avicennia
marina species have mostly been misclassified as Rhizophora stylosa and Bruguiera
exaristata (Figure 2.8c). The descriptive analysis of the classification of AP0.14M
has shown the relatively low accuracy of 68%, with Kappa statistics limited to 0.60
(Table 2.3). The accuracy of the map produced from the AP0.5M was slightly lower
(Kappa equals to 0.58), with an overall accuracy of 68% (Table 2.3).
Table 2.4. Comparison of producer’s accuracy and user’s accuracy for each species,
obtained from different remote sensing data sources. (AM-Avicennia marina,
CT-Ceriops tagal, BE-Bruguiera exaristata, LR-Lumnitzera racemosa, RS-
Rhizophora stylosa).
Image Accuracy (%) AM BE CT LR RS
PS-WV2-VIS Producer’s acc. 98 73 55 100 95
User’s acc. 98 72 84 87 89
PS-WV2-R/NIR1 Producer’s acc. 95 54 70 100 100
User’s acc. 100 83 68 72 81
WV2-VIS Producer’s acc. 98 ** ** 82 28
User’s acc. 99 ** ** 13 19
WV2-R/NIR1 Producer’s acc. 94 ** 2 44 70
User’s acc. 98 ** 2 12 13
AP0.14M Producer’s acc. 83 27 45 73 77
User’s acc. 94 46 35 60 59
AP0.5M Producer’s acc. 91 20 65 79 44
User’s acc. 77 25 70 72 65
** Did not calculate individual accuracies
77
For individual species classifications, Table 2.4 shows low user’s accuracy
for Bruguiera exaristata for the PS-WV2-VIS image classification, compared to
other species. In contrast, Ceriops tagal has a user’s accuracy of 84% with 55%
producer’s accuracy. Lumnitzera racemosa, for example, has a producer’s accuracy
of 100% while the user’s accuracy is 87%. This means that there were no omissions
from this class, but were more inaccurate inclusions providing an over-estimation of
this coverage. In the PS-WV2-R/NIR1 classification the extent of Rhizophora stylosa
was also over-estimated.
The individual classification accuracies of Bruguiera exaristata did not
calculate for WV2 images as there was no sufficient coverage identified by
classification to do so. However, both producer’s and user’s accuracies of Ceriops
tagal was 2% from WV2-R/NIR1 (Table 2.4). Lumnitzera racemosa was highly
over-estimated in the WV2-VIS classification whereas Rhizophora stylosa did in the
WV2-R/NIR1 classification. These inaccurate inclusions that provide over-
estimations were highly evidence in their visual maps (Figure 2.6). The only
successfully classified class was Avicennia marina.
2.4 Discussion
2.4.1 Separating Mangroves and Non-Mangroves
One of the advantages of using WV2 data is the comparatively large number
of spectral bands available within a limited spectral range. This enables more
flexibility in applying a wide range of rules in OBIA. For example, although the
visual appearance of home gardens and other vegetation are the same as mangroves,
78
there is a detectable spectral difference in mangroves within NIR1 and red-edge
regions (Figure 2.4 a and b). In a recent study of mangrove mapping using SPOT-5
satellite images, mudflats within mangrove habitat required manual removal [47]. In
this study, the yellow spectral band was very useful for extracting buildings, soil and
mudflats automatically. The normalized differences calculated from NIR1 and Red-
Edge bands successfully isolated home gardens and other vegetation from
mangroves. However, the possibility of repeating this with aerial photographs was
restricted due to spectral band limitations.
Most of the healthy green grassy areas near the mangrove boundary had
similar spectral profiles to the mangrove species at hinterland (Figure 2.4).
Therefore, the main challenge was to separate mangroves from the healthy green
grass. To achieve this, at different hierarchical levels, contextual information,
geometry and neighbourhood characteristics of objects were used. For example: the
analysis of relation to border to Candidate-Mangrove-1 and Candidate-Mangrove-2
classes, and the normalized difference indices extracted from spectral information of
objects related to their parent objects were used successfully. Having increased the
accuracy of the areal extent of mangroves, the extraction procedure was fully
automated for pan-sharpened WV2 images.
Overall, this approach has effectively discriminated different land cover
classes surrounding a mangrove ecosystem using WV2 images. When using pan-
sharpened images, the whole process was automated, and can be repeated in a robust
manner. There was no manual editing involved. However, when applying OBIA to
the WV2 images with a spatial resolution of 2m, there was some manual editing
involved. The next stage will be to test the transferability of the derived rule set to a
79
different location. However, since the rule sets consist of image variables rather than
set numerical values for class discrimination, they can be tested on other areas easily.
Aerial photographs were visually appealing, and it was easy to visually
identify mangroves, as well as gaps between mangroves. When applying OBIA to
the aerial photographs, the high spatial resolution helped to create small, compact
objects in the OBIA environment and then to discriminate vegetation and non-
vegetation features successfully. However, isolating mangroves from home gardens
and other vegetation types was difficult. The OBIA rule set was amended
considering the limitations of spectral and radiometric resolutions of the data. Even
though the aerial image dataset required manual editing, the problem of having a
heterogeneous mixture of vegetation, mangroves, soil and water can be overcome by
isolating the mangrove coverage before further classification.
When comparing the above results, regardless of spatial resolution, a
relatively large number of spectral bands within the limited spectral range of visible
and NIR would be an ideal solution for mangrove coverage identification. This is
supported by the exploratory spectral separability analysis of WV2 images by
Heumann [20]. His study demonstrated increasing accuracy when discriminating
mangroves from other vegetation using WV2 data and a decision tree classification
algorithm. However, the quantitative analysis of radiometric resolution differences of
these datasets must be considered, as it may play a significant role on the
classification accuracies.
80
2.4.2 Comparison of Mangrove Species Classifications
As described in section 3.2, the differences in classification of species
composition reflects the reliability of using various data sources and advanced
algorithms for classification. The overall accuracy of the PS-WV2-VIS image
classification is very good. The mangrove zonation pattern described by Brocklehurst
and Edmeades [24] for this area has successfully been identified (Figure 2.6a).
Further, these results are supported by the field surveys of Ferwerda et al. [25], who
identified the presence of Rhizophora stylosa closer towards creek banks or tidal
flats, and Lumnitzera racemosa and Ceriops tagal located on the high tidal range.
The classifications in this study detected similar patterns (Figure 2.6).
The Brocklehurst and Edmeades [24] study is the only published mangrove
mapping studying Darwin, Australia, undertaken almost 20 years ago. It does not
demarcate individual species boundaries. Field investigation of the study area
identified many changes that have occurred over this time period. For instance,
Rhizophora stylosa, and Lumnitzera racemosa were not yet documented. However,
since their study was done using manual field survey methods and visual
interpretation of remote sensing images, updating the changes by repeating their data
capturing methods would require significant time and financial investments.
However, the method introduced by this study is repeatable and could be performed
at reasonable time intervals in order to constantly update mangrove coverage maps.
When using SVM classification algorithm, attention must be given to the
parameter selection of SVM architecture. For example: since the performance of
SVM is based on the kernel function used and its parameters, the penalty parameter
81
that works with an optimal boundary selection of the training data has to be
considered carefully [21]. In this study, ten was selected as the ideal penalty
parameter to locate training samples on the correct side of the decision boundary. A
larger penalty parameter of the SVM exhibits over-fitting of training data, thus
reducing classification accuracies.
Another consideration is the benefit of image fusion. Even in this study, when
classifying mangroves without pan-sharpening, individual species accuracies were
low. Over the years, with the development of advanced image/signal processing
techniques, image fusion has become a tool that improves the spatial resolution of
images while preserving spectral information. Many recent studies have indicated
that these algorithms are more sophisticated for improved information extraction
rather solely for visualization. For example: Zhang [48] revised recent studies that
extracted information from pan-sharpened data, and concluded that well pan-
sharpened image could improve the information extraction. Although an
appropriately pan-sharpened image could provide more information for feature
extraction, there is room for further development of pan-sharpening techniques [35,
48].
When comparing these results with global mangrove studies, special attention
must be given to the geographic region. Mangrove ecosystems characteristics are
different from region to region due to soil salinity, ocean current, tidal inundation,
and various geomorphic, edaphic, climatic and biotic factors etc. [6, 10, 49]. Given
these considerations, it will be interesting to see if this technique would be a viable
alternative for the tropical arid or sub-arid mangrove environment as it exhibits
greater structural complexity than this study area. Kamal and Phinn [15] compared
82
pixel-based and object-based image analysis techniques using hyperspectral data for
mangrove identification. They were not able to achieve reasonably high accuracies
for many species using WV2 images. Since their study area lies in Queensland,
Australia, and the mangrove ecosystems around this area are known to be similar to
Northern Territory, Australia [4,24] the results can directly be compared to this
study.
Although the most dominating factors for spectral reflectance variation are
biochemical and biophysical parameters of the plants, the reflectance spectra of
mangroves are mostly combined with those of underlying soil, water and
atmospheric vapour. Therefore, a degradation of classification can be expected,
especially in the regions where water absorption is stronger [11]. For example,
although the band selection of PS-WV2-R/NIR1 lies within the ideal spectral range
for classified species identified by Wang and Sousa [18], the results of some classes
were lower than that of PS-WV2-VIS classification due to the broad NIR band,
which includes a region highly sensitive to the water absorption. The classification of
WV2-R/NIR1 indicated the lowest accuracy, demonstrating the importance of high
spectral resolution in achieving high accuracies.
Most of the traditional approaches for mangrove remote sensing are based on
interpretation of aerial photographs. Heumann [10] summarized 11 mangrove studies
using aerial photographs. Among them, Dahdouh-Guebas et al. [50] successfully
mapped individual species using image attributes extracted from aerial photographs.
In that sense, they used visual interpretation techniques rather than computational
classification. However, in this application, aerial photographs with broad spectral
bands could not delineate the features available in the mixed pixels due to spectral
83
resolution limitations resulting high omission errors. This is also evident from the
spectral profile analysis of this study (Figure 2.4). Most of the species were not able
to be discriminated from each other. Therefore, although the same classifier has been
used, the classification from higher spatial resolution aerial photography is of lower
accuracy.
When comparing the outcome of these six data sets, apart from the spatial and
spectral resolutions, radiometric resolution must be taken into account. A sensor with
high radiometric resolution is more sensitive to capture small differences in
reflectance values. Further, electromagnetic characteristics and signal-to-noise ratio
of sensors can influence the classification accuracies. In this study, it was not
possible to take these radiometric effects into account when resampling aerial
photographs to simulate the WV2 image.
2.4.3 Accuracy Assessment
The visual appearance and the statistical values of the PS-WV2-VIS
classification showed the strongest agreement between generated maps and reference
data, according to Congalton [46]. By contrast, the classifications of WV2-VIS and
WV2-R/NIR1 have the lowest level of agreement between species maps and
validation data. The maps generated from AP0.14M and AP0.50M have a moderate
level of agreement to the reference data, having Kappa statistics between 0.60 and
0.64 [46]. The results of the error matrix analysis of species classification were lower
than the pan-sharpened WV2 image classifications (Table 2.4). Despite the visual
clarity and higher spatial resolution of the aerial photographs, the resultant
classifications did not generate better accuracy results than thepan-sharpened WV2
84
images undergoing the same treatment and process. It can be clearly seen that the
WV2 image with a spatial resolution of 2 m was not a successful alternative in this
context.
The error matrix was examined to make more analytical observations about
individual species. This is a very effective way to describe both errors of inclusion
and exclusion of each species represented in the classification [46, 51]. As explained
by Congalton [51], the user’s accuracy is an indication of whether the pixels
classified on the map actually represent the same species on the ground. The
probability of the reference pixel being correctly classified is the producer’s
accuracy.
Scrutiny of the error matrix reveals that there is confusion in discriminating
Ceriops tagal from Bruguiera exaristata and Rhizophora stylosa (Table 2.4). This is
because of their spectral separability measures within that specific range and is also
due to the patchy distribution pattern. For example: mostly Ceriops tagal is mixed
with Bruguiera exaristata and Avicennia marina in this study area [24, 25]. The
WV2 images with low spatial resolution were not able to spectrally discriminate
species and to identify the complex structural situation.
The error matrix also demonstrated the successful detection of Avicennia
marina from the WV2 images with a low spatial resolution. This is because of the
visually distinctive, smooth and similar spatial pattern of Avicennia marina on the
images (or its distinctive spectral profile). Although most of the reference pixels of
Lumnitzera racemosa and Rhizophora stylosa were correctly classified as their
respective classes, their actual representations on maps were poor. This is evident
85
from the lower user’s accuracy (Lumnitzera racemosa got 82% and Rhizophora
stylosa got 70%) than the producer’s accuracy (Lumnitzera racemosa and
Rhizophora stylosa equals to 13%).
Overall, when investigating classifications of both pan-sharpened WV2
images, Lumnitzera racemosa and Avicennia marina have the highest values for the
producer’s accuracy, indicating that the probability of this species being classified as
another is low. The smooth and similar spatial pattern helps the classification
techniques to detect them accurately.
This is evidence that the combination of high spatial resolution remote
sensing data using a relatively large number of spectral bands within the visible and
NIR region and the SVM non-linear machine learning classification technique, is a
powerful tool for mixed environments such as mangroves. However, spatial
autocorrelation will reduce the accuracy up to a certain level. For instance, the noise
from non-leaf surfaces such as tree branches and background can degrade the results
of spectral separability of mangroves and thus can reduce classification accuracy at
species level.
In both instances, the aerial photographs showed lower classification
accuracies than the pan-sharpened WV2 images. For example: at both instances, it
was not possible to successfully differentiate Bruguiera exaristata and Lumnitzera
racemosa from other species, both having lower producers’ and users’ accuracies.
The user’s accuracy of Bruguiera exaristata, Ceriops tagaland Rhizophora stylosa
indicates high errors of commission, because other species were highly misclassified
as them.
86
2.5 Conclusions and Recommendations
This study compared high spatial resolution aerial photographs with satellite
remote sensing data for the purpose of mangrove species discrimination in two steps.
First, mangroves and non-mangroves were separated using object-based image
analysis method. Then, the mangrove coverage only was classified into species level.
The study demonstrated that a large number of spectral bands with higher spatial
resolution (pan-sharpened WV2 image) were more accurate than broad spectral
bands within the blue, green, and red regions, when discriminating mangroves from
other features in an image. In addition, our findings show that, using a calibrated,
high radiometric resolution sensor such as WV2 allows greater classification
automation with reduced manual editing.
When further classifying down to species level, the highest accuracy (overall
accuracy of 89%) was obtained from the pan-sharpened WV2 image using five
spectral bands within the visible range. The pan-sharpened visible image covered the
same spectral range with the same spatial resolution as the aerial photographs.
Therefore, the higher accuracy of the former compared to the overall accuracy of
68% from resampled aerial photographs is attributed to the increased number of
narrow bands available for analysis, rather than the total wavelength range.
Compared to these results, however, there is no considerable difference between the
mangrove species map obtained from the pan-sharpened Red, Red-Edge and NIR1
bands of WV2 image. Further, this study demonstrated the significant increase in
classification accuracy when using pan-sharpened imagery, on the condition that
spectral and radiometric integrity is maintained using an appropriate algorithm such
as high pass filter pan-sharpening method.
87
This study also demonstrated a unique application of the support vector
machine algorithm for mangrove species mapping. While this advanced image
processing technique has previously been used in other environments, it is
particularly beneficial for mangroves because it efficiently deals with the dense,
heterogeneous nature of mangrove forests.
Although the method used in this study is tested on a mangrove forest with a
small number of species, the obtained results were very impressive. These results
provide a valuable contribution to the mangrove species mapping methodologies. We
would recommend repeating this process on larger study areas with greater species
diversity in order to determine the efficiency and accuracy of the proposed data and
SVM methods. The method could also be tested with blue to NIR1 wavelength bands
of pan-sharpened WV2 on a computationally powerful system. The transferability of
the rule set developed for OBIA can also be tested on a different data set. Further, to
obtain a higher degree of species classification accuracies, a quantitative analysis of
the effects of differences between radiometric resolutions should be investigated.
2.6 Acknowledgments
The support of the Northern Territory Government, Australia in providing
aerial photographs for this study is gratefully acknowledged. Authors also appreciate
Sandra Grant and three anonymous reviewers for constructive advice and editorial
comments.
88
2.7 Author Contributions
Muditha K. Heenkenda designed the research, processed the remote sensing data,
and drafted the manuscript with co-authors providing supervision and mentorship
throughout the process.
2.8 Conflicts of Interest
The authors declare no conflict of interest.
2.9 References
1. Food and Agriculture Organization. The World’s Mangroves 1980–2005;FAO
Forestry Paper 153; Food and Agriculture Organization of the United Nations:
Rome, Italy, 2007.
2. Suratman, M.N. Carbon Sequestration Potential of Mangroves in Southeast
Asia. In Managing Forest Ecosystems: The Challenge of Climate Change;
Bravo, F., Jandl, R., LeMay, V., Gadow, K., Eds.; Springer, 2008; pp. 297–315.
3. Bouillon, S.; Rivera-Monroy, V.H.; Twilley, R.R.; Kairo, J.G.. Mangroves. In
The Management of Natural Coastal Carbon Sinks; Laffoley, D., Grimsditch,
G., Eds.; IUCN: Gland, Switzerland, 2009; p. 53.
4. Komiyama, A.; Ong, J.E.; Poungparn, S. Allometry, biomass, and productivity
of mangrove forests: A review. Aquatic Bot.2008, 89, 128–137.
89
5. Metcalfe, K. The Biological Diversity, Recovery from Disturbance and
Rehabilitation of Mangroves in Darwin Harbour, Northern Territory. In Faculty
of Education, Health & Science; Charles Darwin University: Darwin, NT 0909,
Australia, 2007; pp.1 - 17.
6. Alongi, D.M. Present state and future of the world’s mangrove forests. Environ.
Conserv.2002, 29, 331–349.
7. Blasco, F.; Gauquelin, T.; Rasolofoharinoro, M.; Denis, J.; Aizpuru, M.;
Caldairou, V. Recent Advances in mangrove studies using remote sensing data.
Mar. Freshw. Resour.1998, 49, 287–296.
8. Green, E.P.; Cleark, C.D.; Mumby, P.J.; Edwards, A.J.; Ellis, A.C. Remote
sensing techniques for mangrove mapping. Int. J. Remote Sens.1998, 19, 935–
956.
9. Lucas, R.M.; Ellison, J.C.; Mitchell, A.; Donnelly, B.; Finlayson, M.; Milne,
A.K. Use of stereo aerial photography for quantifying changes in the extent and
height of mangroves in tropical Australia. Wetl. Ecol. Manag.2002, 10, 161–
175.
10. Heumann, B.W. Satellite Remote sensing of mangrove forests: Recent advances
and future opportunities. Progr. Phys. Geogr.2011, 35, 87–108.
90
11. Adam, E.; Mutanga, O.; Rugege, D. Multispectral and hyperspectral remote
sensing for identification and mapping of wetland vegetation: A review. Wetl.
Ecol. Manag.2010, 18, 281–296.
12. Gao, J. A hybrid method toward accurate mapping of mangroves in a marginal
habitat from SPOT multispectral data. Int. J. Remote Sens.1998, 19, 1887–1899.
13. Held, A.; Ticehurst, C.; Lymburner, L.; Williams, N. High resolution mapping
of tropical mangrove ecosystems using hyperspectral and radar remote sensing.
Int. J. Remote Sens.2003, 24, 2739–2759.
14. Nandy, S.;Kushwasha, S.P.S. Study on the utility of IRS LISS-III data and the
classification techniques for mapping of Sunderban mangroves. J. Coast.
Conserv.2010, 15, 123–137.
15. Kamal, M.; Phinn, S. Hyperspectral data for mangrove species mapping: A
comparison of pixel-based and object-based approach. Remote Sens.2011, 3,
2222–2242.
16. Vaiphasa, C.; Ongsmwang, S.; Vaiphasa, T.; Skidmore, A.K. Tropical
mangrove species discrimination using hyperspectral data: A laboratory study.
Estuar. Coast. Shelf Sci.2005, 65, 371–379.
17. Vaiphasa, C.; Skidmore, A.K.; Boer, W.F.D.; Vaiphasa, T. A hyperspectral
band selector for plant species discrimination. ISPRS J. Photogramm. Remote
Sens.2007, 62, 225–235.
91
18. Wang, L.; Sousa, W.P. Distinguishing mangrove species with laboratory
measurements of hyperspectral leaf reflectance. Int. J. Remote Sens.2009, 30,
1267–1281.
19. Kuenzer, C.; Bluemel, A.; Gebhardt, S.; Quoc, T.V.; Dech, S. Remote sensing
of mangrove ecosystems: A review. Remote Sens.2011, 3, 878–928.
20. Heumann, B.W. An object-based classification of mangroves using a hybrid
decision tree—support vector machine approach. Remote Sens.2011, 3, 2440–
2460.
21. Tso, B.; Mather, P.M. Support Vector Machines, In Classification Methods for
Remotely Sensed Data, 2nd ed.; CRC Press: New York, NY, USA, 2009; pp.
125-153.
22. Mountrakis, G.; Jungho, I.; Ogole, C. Support vector machines in remote
sensing: A review. Int. J. Photogramm. Remote Sens.2011, 66, 247–259.
23. Huang, C.; Davis, L.S.; Townshend, J.R.G. An assessment of support vector
machines for land cover classification. Int. J. Remote Sens.2002, 23, 725–749.
24. Brocklehurst, P.;Edmeades, B. The Mangrove Communities of Darwin Harbour:
Northern Territory; Department of Lands, Planning and Environment, Northern
Territory Government: Darwin, NT 0801, Australia, 1996.
92
25. Ferwerda, J.G.; Ketner, P.; McGuinness, K.A. Differences in regeneration
between hurricane damaged and clear-cut mangrove stands 25 years after
clearing.Hydrobiologia2007, 591, 35–45.
26. Wightman, G. Mangrove Plant Identikit for North Australia’s Top End;
Greening Australia: Darwin, NT, Australia, 2006; p. 64.
27. Duke, N.C. Australia’s Mangroves: The Authoritative Guide to Australia’s
Mangrove Plants; University of Queensland: Brisbane, QLD, Australia, 2006; p.
200.
28. DigitalGlobe. The Benefits of the 8 Spectral Bands of WorldView-2;
DigitalGlobe: Longmont, CO 80503, USA, 2009.
29. Northern Territory Government. Northern Territory Digital Data and
Information; Department of Lands, Planning and the Environment: Darwin, NT
0801, Australia, 2010.
30. DigitalGlobe. DigitalGlobe Core Imagery Products Guide; DigitalGlobe:
Longmont, CO 80503, USA, 2011.
31. Updike, T.; Comp, C. Radiometric Use of WorldView-2 Imagery. In Technical
Note; DigitalGlobe: Longmont, CO 80503, USA, 2010; pp. 1-17.
93
32. Amro, I.; Mateos, J.; Vega, M.; Molina, R.; Katsaggelos, A.K. A survey of
classical methods and new trends in pansharpening of multispectral images.
EURASIP J. Adv. Signal Process.2011, doi: 10.1186/1687-6180-2011-79.
33. Chavez, J.P.S.; Sides, S.C.; Anderson, J.A. Comparison of three different
methods to merge multiresolution and multispectral data: Landsat TM and
SPOT panchromatic. J. Photogramm. Eng. Remote Sens.1991, 57, 295–303.
34. Chavez, J.P.S.; Bowell, J.A. Comparison of the spectral information content of
Landsat thematic mapper and SPOT for three different sites in the Phoenix,
Arizona. J. Photogramm. Eng. Remote Sens.1988, 54, 1699–1708.
35. Palubinskas, G. Fast, simple, and good pan-sharpening method. J. Appl. Remote
Sens.2013, doi:10.1117/1.JRS.7.073526.
36. Intergraph Corporation. IMAGINE Workspace: HPF Resolution Merge Manual;
Intergraph: Huntsville, AL 35813, USA, 2013.
37. Australian Government. 2013. Available online: http://www.ga.gov.au/
(accessed on 25 January 2013).
38. Henrich, V.; Krauss, G.; Götze, C.; Sandow, C. Index Database. 2012. Available
online: http://www.indexdatabase.de/db/i-single.php?id=126 (accessed on 29
May 2014).
94
39. Ahamed, T.; Tian, L.; Zhang, Y.; Ting, K.C. A review of remote sensing
methods for biomass feedstock production. Biomass Bioenergy2011, 35, 2455–
2469.
40. Mertens, K.C.; Verbeke, L.P.C.; Ducheyne, E.I.; De Wulf, R.R. Using genetic
algorithms in sub-pixel mapping. Int. J. Remote Sens.2003, 24, 4241.
41. Tatema, A.J.; Lewisa, H.G.; Atkinsonb, P.M.; Nixona, M.S. Super-resolution
land cover pattern prediction using a Hopfield neural network. Remote Sens.
Environ.2002, 79, 1–14.
42. Nguyen, M.Q.; Atkinson, P.M.; Lewis, H.G. Super-resolution Mapping using
Hopfield Neural Network with Panchromatic image. Int. J. Remote Sens.2011,
32, 6149–6176.
43. Tatem, A.J.; Lewis, H.G.; Atkinson, P.M.; Nixon, M.S. Multiple-class land-
cover mapping at the sub-pixel scale using a Hopfield neural network. Int. J.
Appl. Earth Obs. Geoinforma.2001, 3, 184–190.
44. Chapelle, O.; Haffner, P.; Vapnik, V.N. Support vector machines for histogram-
based image classification. Trans. Neural Netw.1999, 10, 1055–1064.
45. Canty, M.J. Supervised Classification: Part 1. In Image Analysis, Classification,
and Change Detection in Remote Sensing with Algorithm for ENVI/IDL, 2nd
ed.; CRC Press: Boca Raton, FL, USA, 2009; pp. 1- 441.
95
46. Congalton, R.G. Accuracy assessment and validation of remotely sensed and
other spatial information. Int. J. Wildl. Fire2001, 10, 321–328.
47. Quoc, T.V.; Oppelt, N.; Leinenkugel, P.; Kuenzer, C. Remote sensing in
mapping mangrove ecosystem-An object-based approach. Remote Sens.2013, 5,
183–201.
48. Zhang, Y.: Pan-Sharpening for Improved Information Extraction. In Advances
in Photogrammmetry, Remote Sensing and Spatial Information Sciences; Taylor
& Francis: London, UK, 2008; pp. 185 - 202.
49. Alongi, D.M. Zonation and seasonality of benthic primary production and
community respiration in tropical mangrove forests.Oecologia1994, 98, 320–
327.
50. Dahdouh-Guebas, F.; Verheyden, A.; Kairo, J.G.; Jayatissa, L.P.; Koedam, N.
Capacity building in tropical coastal resource monitoring in developing
countries: A re-appreciation of the oldest remote sensing method. Int. J. Sustain.
Dev. World Ecol.2006, 13, 62–76.
51. Congalton, R.G. A review of assessing the accuracy of classification of remotely
sensed data. Remote Sens. Environ. 1991, 37, 35–46.
© 2014 by the authors; licensee MDPI, Basel, Switzerland. This article is an open
access article distributed under the terms and conditions of the Creative Commons
Attribution license (http://creativecommons.org/licenses/by/3.0/).
96
Chapter 3 - Mangrove tree crown delineation
from high resolution imagery
This work has been published in the following papers:
Heenkenda, M. K., Joyce, K. E. & Maier, S. W. 2015, Mangrove Tree
Crown Delineation from High Resolution Imagery, Photogrammetric
Engineering and Remote Sensing, vol. 81, 471-479; doi:
10.14358/PERS.81.6.471.
Formal permission from the publisher is attached in Appendix 2.
(Reproduced with permission from the American Society for
Photogrammetry and Remote Sensing: The imaging and geospatial
information society).
AWARD - 2016 ESRI Award for the Second Best Scientific Paper in Geographic
Information Systems from the American Society of Photogrammetry and Remote
Sensing: The imaging and geospatial information society (ASPRS).
97
3.0 Abstract
Mangroves are very dense, spatially heterogeneous, and have limited height
variations between neighbouring trees. Delineating individual tree crowns is thus
very challenging. This study compared methods for isolating mangrove crowns using
object based image analysis. A combination of WorldView-2 imagery, a digital
surface model, a local maximum filtering technique, and a region growing approach
achieved 92% overall accuracy in extracting tree crowns. The more traditionally used
inverse watershed segmentation method showed low accuracy (35%), demonstrating
that this method is better suited to homogeneous forests with reasonable height
variations between trees. The main challenges with each of the methods tested were
the limited height variation between surrounding trees and multiple upward pointing
branches of trees. In summary, mangrove tree crowns can be delineated from
appropriately parameterized object-based algorithms with a combination of high
resolution satellite images and a digital surface model. We recommend partitioning
the imagery into homogeneous species stands for best results.
98
3.1 Introduction
Mangrove forests represent a land-ocean interface ecosystem in tropical and
sub-tropical regions of the world. They provide a wide range of goods and services
that have been acknowledged for decades at local, national, and global levels (Food
and Agriculture Organisation, 2007). Nevertheless, a rapid rate of destruction of
these intertidal forests due to various anthropogenic disturbances over the last few
decades has been recorded (Suratman, 2008). While earlier descriptive or
observational studies focused on community usage of mangroves, more recently
there has been an increase in analytical studies quantifying the diverse value of
mangrove forests (Komiyama et al., 2008; Voa et al., 2012). To quantitatively
evaluate mangrove forests, estimating individual tree growth, biomass, and
productivity has become important.
The expansion of leaf area or number of leaves available for photosynthesis is
an indicator for the biomass production of trees, and thus leaves composing a crown
indicate individual tree growth, senescence, and death (Analuddin et al., 2009). To
determine biophysical parameters of trees (height, diameter at breast height, growth
rate etc.) using remote sensing methods, individual trees must first be isolated and
tree crown boundaries delineated.
Isolating individual trees and extracting tree structure has a significant
contribution to a variety of applications. Individual trees and their structure can be
used for forest inventories, assessing forest regeneration, quantifying above ground
biomass, and assessing vegetation damage (Chen et al., 2006). Therefore, an
99
investigation of mangrove tree crown foliage dynamics would be an added value to
estimate the productivity and health of mangrove forests.
Measuring precise crown foliage dynamics manually in the field is a
challenging task as the irregularities of crown shapes are difficult to capture using
standard field survey equipment. The available field surveying methods for
individual tree attribute extraction are labour intensive and time consuming. As an
alternative, very high resolution satellite and aerial images, as well as laser scanners
might provide a viable option for extracting this information. This is supported by
advances in algorithmic developments towards higher precision, reliability, and
automation. A number of studies have already been conducted to extract individual
tree crowns for various vegetation types using remotely sensed data with varying
degree of success (Blaschke et al., 2011; Chen et al., 2006; Erikson, 2004; Erikson
and Olofsson, 2005; Gougeon and Leckie, 2003; Hirata et al., 2010; Kaartinen et al.,
2012; Larsen et al., 2011; Vauhkonen et al., 2012; Wannasiri et al., 2013; Whiteside
et al., 2011). However, very little information is available for using remote sensing
for delineating mangrove tree crowns.
A review of methods for delineating individual trees or groups of trees in
optical images identified four main approaches: local maxima detection algorithm,
segmentation methods, valley following algorithm, and tree modelling and image
template construction (Lopez, 2012). Kaartinen et al. (2012) tested several other tree
crown extraction methods using airborne laser scanning data. However, in both
instances, the most intensively investigated technique is local maxima detection. It
assumes that tree apexes generate intensity peaks in images, thus local maximum
image brightness values relate to tree top locations (Leckie et al., 2005; Pouliot et al.,
100
2002). As this method does not delineate crown boundaries, it has to be combined
with a segmentation method to identify crowns. A segmentation method involves
identification of groups of similar neighbouring pixels that corresponds to objects or
part of objects. There are many approaches for this: grouping textures, morphological
operators, and joining convex-shaped edges. Hence, construction of regions of the
image is based on homogeneity of image characteristics such as image brightness,
texture, shape, and size etc. Furthermore, in advanced approaches, semantic
information can also be incorporated during segmentation (Eisank et al., 2010).
As an alternative approach for a combination of local maxima detection
algorithm and a region growing, the valley following method considers image
intensities as an analogy for a topographic surface (Gougeon, 1995). It creates a hill
pattern, where the tree tops are actually seen as hill tops. Boundaries of tree crowns
show as dark values, shady rings or ‘valleys’ depending on sun illumination angle
(Gougeon, 1995). The valley following algorithm defines a set of semantic rules to
enclose the valleys, leading to the separation of individual crowns (Gougeon, 1995).
However, this method is more suitable for sparsely distributed forests or plantations
than densely clustered natural forests. It works best when crowns have clear
boundaries, are approximately symmetrical, similar in size, and do not overlap each
other (Bunting and Lucas, 2006). Similarly, the topographic concept of a watershed
can also be applied to the contour detection and image segmentation according to
Beucher and Lantuejoul (1979). And, the topographic concept of a watershed can
also be extended to delineate individual tree crowns. It is another more commonly
applied method for determining tree crowns (Edson and Wing, 2011), and called
Inverse Watershed Segmentation (IWS).
101
The IWS method converts the elevation of the surface model to form the
equivalent of individual hydrologic drainage basins (Edson and Wing, 2011; Swamer
and Houser, 2012; Wannasiri et al., 2013). These individual drainage basins (canopy
basins) represent tree crowns. As the algorithm relies on an inverted surface model,
the accuracy and the resolution of the digital surface model is important. For
example, Edson and Wing (2011) found that the canopy height model with 1 m
resolution of a conifer forest decreased the commission errors caused by upward
facing branches of the same tree when tree tops were 1m apart from each other.
The tree modelling and image template method constructs geometric models
of trees (templates) and matches them to images (Korpela et al., 2007; Wolf and
Heipke, 2007). During this process, shape parameters such as crown diameter, tree
height, and convexity are investigated and often modelled by means of a generalized
ellipsoid. Therefore, the model of the tree which is similar to the visible trees in the
remotely sensed image in terms of viewing and illumination is created as a three-
dimensional projection of a tree crown (Larsen et al., 2011; Larsen and Rudemo,
1998). Hence, the optimal shape and the placement of bounding ellipse in an image
vary with image acquisition geometry (Larsen and Rudemo, 1998).
Although many algorithms have been investigated for individual tree crown
delineation, it is clear that no single algorithm is universally successful in delineating
the crowns of all different vegetation types (Kaartinen et al., 2012; Larsen et al.,
2011; Li et al., 2008). Most methods are based on image intensities, elevation, and
homogeneity of vegetation with reasonable height variations. Further, these methods
are more successful with plantation forests having regular distances between trees
rather than naturally grown forests (Bunting and Lucas, 2006). Unfortunately, such
102
variations are rare in mangrove forests. When considering forest types, although
most already established methods have been successful in coniferous forests, often
they do not work well in deciduous and mixed species forests (Jing et al., 2012).
Deciduous tree crowns are also relatively flat (Jing et al., 2012) and thus resemble
mangrove crowns. Therefore, the methods that have already been developed for
identifying individual trees of terrestrial vegetation may not be suitable for mangrove
environments.
This study evaluated the potential of using high spatial resolution optical
remote sensing data and an object based image analysis (GEOBIA) approach for
outlining mangrove tree crowns by: (a) applying GEOBIA to WorldView-2 (WV2)
data; (b) applying GEOBIA to the combined WV2 and a high resolution digital
surface model (DSM) derived from aerial photography; and (c) applying the ISW
method to the DSM.
3.2 Data and methods
3.2.1 Study area, remote sensing data, and pre-processing
This study focused on a small mangrove forest situated in the Rapid Creek
coastal system near Darwin, Australia (Figure 3.1). The Rapid Creek mangrove
forest covers approximately 60 hectares in extent, and is centred at E/12o22’
S/130o51’. The mangrove forest naturally regenerated after a proposed housing
project was abandoned shortly after a clearing in 1969 (Ferwerda et al., 2007). The
area is comprised of diverse, spatially complex, and common mangrove communities
in the Northern Territory of Australia.
103
Figure 3.1. The study area is located in Rapid Creek in Darwin, Northern Territory,
Australia. The map shows the validation dataset including both field surveyed and on
screen digitised (three dimensionally) trees. Inset shows the appearance of mangrove
tree crowns on the aerial photograph; Coordinate system: Universal Transverse
Mercator Zone 52 L, WGS84; Aerial photographs © Northern Territory Government
(NTG), Australia, Copyright 2012 NTG.
A WV2 satellite image acquired on 5 June, 2010 was used as the main source
of remotely sensed data. The image has 8 multispectral bands with 2.0 m spatial
resolution, and a panchromatic band with 0.5 m spatial resolution. The wavelength
range of the WV2 satellite images vary from 447 nm to 1043 nm (Table 3.1). To
104
generate a digital surface model, true colour aerial photographs were used. The date
of acquisition of these photographs is 7 June, 2010 and the sensor is UltraCamD
large format digital camera (Table 3.1).
The WV2 image was radiometrically, atmospherically, and geometrically
corrected according to the method adopted by Heenkenda et al. (2014). The red and
the NIR1 bands were pan-sharpened to 0.5 m spatial resolution to incorporate edge
information from the high spatial resolution panchromatic band to the lower spatial
resolution multispectral bands. The high pass filter pan-sharpening method was
utilised as it produces a fused image without distorting the spectral balance of the
original WV2 satellite images (Chavez and Bowell, 1988; Chavez et al., 1991;
Heenkenda et al., 2014; Palubinskas, 2013). Since the spectral range of the
panchromatic band does not cover the spectral ranges of both coastal and NIR2
bands, they were not used for further processing.
105
Table 3.1. Spectral resolution of WV2 image and aerial photographs (DigitalGlobe,
2011; Northern Territory Government, 2010)
Band Spectral range (nm) Spatial resolution (m)
WorldView-2 satellite image
Panchromatic 447 – 808 0.5
Coastal 396 – 458
2.0
Blue 442 – 515
Green 506 – 586
Yellow 584 – 632
Red 624 – 694
Red-Edge 699 – 749
NIR1 765 – 901
NIR2 856 – 1043
Aerial photographs from UltraCamD camera
Blue 380-600
0.14 Green 480-700
Red 580-720
106
The image was smoothed with a low-pass filtering process. Low-pass filtering
minimizes some of the natural variability of the data and the noise related to sensor
deficiency. It also helps to reduce difficulties associated with having too many details
within crowns (Gougeon, 1995). The focal statistics (maximum) algorithm in
ArcGIS was applied to the panchromatic, red and NIR1 bands of the WV2 image to
identify the local maximum reflectance values within the neighbourhood (ArcGIS
Resources – ESRI). The specified neighbourhood was defined as circular in shape,
and three pixels in size. This threshold value was selected based on the field
observations, assuming that the average radius of a mangrove tree crown was
approximately 1.5 m. These pre-processed panchromatic, red, and NIR1 bands were
then used for further processing.
3.2.2 Extracting mangrove coverage from images
According to Bunting and Lucas (2006), the first step of any tree crown
delineation is to separate trees to be examined and non-trees of the area (generate a
forest mask). Hence, the mangrove coverage was extracted from the WV2 image,
creating an outline of the mangrove area to be further analysed. Full description of
the methodology is available in Heenkenda et al. (2014), but in brief summary, class
specific rules that incorporated texture, geometry, location information, and
relationships between image objects at different hierarchical levels from the WV2
image were used. This mangrove coverage outline was used to avoid mixing other
surrounding features such as other vegetation types, mudflats within mangroves,
roads and buildings, when isolating individual tree crowns.
107
3.2.3 Digital Surface Model (DSM) generation
Aerial photographs captured by an UltraCamD camera were oriented to
ground coordinates following the digital photogrammetric image orientation steps in
the Pix4DMapper software. The image orientation parameters were extracted from
the UltraCamD camera calibration report. A DSM was then created with 15 cm
resolution. Ground control points obtained from “National Geospatial Reference
System, Australia” (Geoscience Australia) were included to minimize the errors that
can arise from image matching. Once the DSM was created, the mangrove outline
was used to extract the region of interest for further processing.
The DSM was smoothed using a low pass filter to eliminate unexpected,
random height variations (typically noise) and small gaps, and to enhance crown
edges. The focal statistics (maximum) algorithm in ArcGIS was introduced to the
DSM to identify local maximum values within the neighbourhood. This focal
statistics-maximum tool considers each pixel in the raster, and calculates maximum
values with respect to identified neighbourhood (ArcGIS Resources ESRI, 2012).
Therefore, the resulting pixel gets the maximum value of the given neighbourhood.
In this study, a circular neighbourhood (a kernel) was applied to select the maximum
values within a range of 10 pixels (0.15 x 10 m), thus the number of pixels
corresponds the average radius of tree crowns.
108
3.2.4 Mangrove tree crown delineation
Mangrove tree crown extraction was tested on several different layer
combinations using local maxima detection and a region growing algorithm. These
are explained in greater detail below.
3.2.4.1 WV2 panchromatic, red, and NIR1 bands (PAN-R-NIR1 method)
Individual trees were first extracted using a combination of panchromatic,
red, and NIR1 bands of the WV2 image. The red and NIR1 multispectral bands were
selected based on the interaction of light at these wavelengths with vegetation. Red
wavelengths are sensitive to chlorophyll absorption, and the NIR region is useful for
analysing vegetation biomass. The image was initially segmented based on shape
and homogeneity of the WV2 bands using the eCognition software package. Tree
tops were detected by investigating local maxima in the panchromatic and NIR1
bands, assuming they appear brighter than surrounding shaded areas. The resultant
tree tops were 1 pixel in size. An iterative procedure of region growing was
implemented to expand the identified tree tops, and to draw ovals of tree crowns. The
extent of an individual oval should therefore represent the extent of each crown. As a
control parameter for the region growing, the ratio of the NIR1 band of the tree top
object to the neighbouring objects was used assuming that the NIR1 band value is
unique for the entire crown. To remove falsely detected crowns, the mean value of
the Normalised Difference Vegetation Index (NDVI) and the mean standard
deviation of the red band were used. For example: if the NDVI value of the selected
tree crown object is less than the mean NDVI value of objects, and the standard
deviation of the red reflectance is less than the mean standard deviation of the red
109
values of objects, then the object was classified as the false tree crown. This helped
to avoid extending tree crowns into the gaps between mangrove trees. Finally, the
tree crowns were smoothed (reshaped avoiding jagged edges) using the morphology
algorithm available in eCognition.
3.2.4.2 WV2 panchromatic, red, and NIR1 bands and DSM (PAN-R-NIR1-DSM
method)
Mangroves often appear very dense and spatially heterogeneous without
considerable tree height differences between neighbouring trees. However,
underlying height values and variations can be used to extract individual tree crowns
with object based image analysis (GEOBIA). The aerial photography derived DSM
was incorporated into the PAN-R-NIR1 method to investigate the possibility of
further improvements of the shapes of tree crowns using actual height variations.
Tree tops were detected by searching local maxima in the DSM. Because of the
limited height variations of mangrove trees between neighbouring trees, in addition
to their densely clustered nature, spectral information of the panchromatic band of
the WV2 image was also incorporated for tree top detection. As some mangrove
species appear to have multiple tree tops due to the complex nature of their branches
(multiple upward pointing branches), all tree tops that were less than 3 m apart were
identified as false tree tops. This 3 m control distance was chosen in congruence with
the average distance between two trees according to the field observations. After
that, an iterative process of region growing from these tree tops was introduced to
delineate crown boundaries. Two control parameters were used to limit the crown
boundary growing. One was the ratio of the NIR1 band value of the tree top object to
neighbouring objects, and the other one was the height difference between the tree
110
top object and the edge of the crown. It was assumed that the maximum height of
tree crowns was about 2 m. Finally, a morphology analysis was done to re-shape the
isolated crowns.
3.2.4.3 The inverse watershed segmentation (IWS) method with the DSM
To apply the IWS method, the DSM that did not undergo low pass filtering
was used. First the DSM was inverted, which resulted in turning the tree tops upside
down into depressions or ponds. Tree branches and crown boundaries therefore
became watersheds. The local minimum values (pour points) were then detected and
classified as tree tops. Small imperfections (a pixel or set of pixels whose flow
direction is undefined) in any digital surface model are called sinks (ArcGIS
Resources – ESRI, 2012). Once the sinks were identified from the inverted surface
model, they needed to be filled to generate the flow directions. The watersheds were
based on these flow directions and positions of tree tops. Finally, watersheds were
converted to polygons assuming that they represented tree crowns.
3.2.5 Accuracy assessment
The quality of image segmentation is largely based on the quality of the
source data, especially the radiometric, spatial, and spectral resolutions, in addition to
optimal selection of parameter values for the segmentation. A meaningful rule set
then enables adaptation of the segmentation results correctly based on target objects.
However, to identify the best segmentation method and corresponding data source
for mangrove tree crown isolation, a goodness of polygon matching method
described by Clinton et al. (2010) was used. This calculates OverSegmentation,
111
UnderSegmentation, and the closeness index (D), which explains the closeness in the
two dimensional space defined by both OverSegmentation and UnderSegmentation.
OverSegmentation (correctness) and UnderSegmentation (completeness) indicate the
inaccuracies associated with too many or too few segments (Moller et al., 2007;
Zhan et al., 2005). And they are in the range of [0, 1] where zero represents the
perfect segmentation. The closeness (distance) index D is in the range of [0, 21/2].
Validation data were collected by manually digitizing 268 tree crown
boundaries using the Stereo Analyst module in ERDAS IMAGINE software. This
included 56 field surveyed tree locations as well. In order to assess the potential of
mangrove tree crown delineation from high resolution remotely sensed data,
OverSegmentation, UnderSegmentation, and the closeness indices were calculated.
3.3 Results
The Rapid Creek mangrove forest has some areas with homogeneous
mangrove species, and some areas with a more heterogeneous mix. There are five
commonly found mangrove species: Avicennia marina, Ceriops tagal, Bruguiera
exaristata, Lumnitzera racemosa, and Rhizophora stylosa (Heenkenda et al., 2014).
Although Excoecaria ovalis and Aegialitis annulata can also be recognized, their
coverage is relatively limited. Field observations determined that an average distance
between two trees is about 3 m and an average radius of tree crowns is about 1.5 m.
112
3.3.1 The digital surface model
The digital surface model (DSM) created for the Rapid Creek mangrove
forest is shown in Figure 3.2. The horizontal accuracy of the DSM is approximately
0.15 m compared to the ground control points. The vertical accuracy is about 0.2 m
with respect to field measurements of tree heights. The DSM showed localised
circular patterns throughout (mostly corresponding to mangrove tree crowns – Figure
2.2 inset A) highlighted by the application of the focal statistics function.
113
Figure 3.2. The digital surface model of the Rapid Creek mangrove forest created
from aerial photographs. Inset “A” shows localized circular patterns due to the image
enhancement using focal statistics function in ArcGIS.
3.3.2 Mangrove tree crown delineation
3.3.2.1 WV2 panchromatic, red, and NIR1 bands (PAN-R-NIR1 method)
Mangrove tree crowns extracted with the PAN-R-NIR1 method are shown in
Figure 3.3 (A). Some of the polygons represent aggregated trees rather than single
trees. Successful detection of maximum brightness values representing tree tops in
114
the image is evident when polygons are overlaid with the panchromatic image
(Figure 3.4 (A)). However, smooth crown boundaries cannot be perceived. Also,
gaps exist between crown boundaries obtained from the PAN-R-NIR1 method,
though the mangrove forest is very dense by nature.
3.3.2.2 WV2 panchromatic, red, and NIR1 bands and DSM (PAN-R-NIR1-DSM
method)
Figures 3.3 (C) and (D) illustrate the tree crowns delineated using the
PAN-R-NIR1-DSM method overlaid with the panchromatic image and the DSM
respectively. Once the DSM was incorporated, shapes of the canopies were improved
dramatically. When overlaid with the DSM, it can be seen that there are some
clusters of mangrove trees detected rather than individuals (Figure 3.3 (D)). Extents
of the gaps between tree canopies are less obvious than the PAN-R-NIR1 method.
3.3.2.3 The inverse watershed segmentation (IWS) method with the DSM
The IWS method mainly detected clusters of tree canopies with similar
heights, rather than individuals. Figures 3.3 (E) and (F) show the delineated
mangrove tree clusters and initially extracted tree tops overlaid with the
panchromatic image and the DSM respectively. Most of the neighbouring tree tops
were merged together to form a single crown. The DSM showed that there was no
significant range in height differences within tree clusters (Figure 3.3 (F)).
115
Figure 3.3. (A) Mangrove tree crowns delineated with the PAN-R-NIR1 method
overlaid with the panchromatic image; (B) Overview of the Rapid Creek mangrove
forest; (C) Mangrove tree crowns delineated with the PAN-R-NIR1-DSM method
overlaid with the panchromatic image; (D) Mangrove tree crowns delineated with the
116
PAN-R-NIR1-DSM method overlaid with the DSM; (E) Mangrove tree crowns
delineated with the IWS method overlaid with the panchromatic image and tree tops
(black dots); (F) Mangrove tree crowns (black outlines) delineated with the IWS
method overlaid with the DSM and tree tops (black dots).
Table 3.2 shows the number of objects extracted as tree tops using each
method. The number of objects extracted using the PAN-R-NIR1-DSM method
reported the highest number and from the PAN-R-NIR1 method reported the lowest
number. The difference between these two methods was approximately 11,500
objects. The IWS method extracted 8767 more objects than the PAN-R-NIR1
method.
Table 3.2.The number objects extracted as tree tops using three different methods.
The centroid positions of tree crowns obtained for part of the mangrove forest
is shown in Figure 3.4. All the centroid positions of the PAN-R-NIR1 method are
very close (i.e. they occur at the same location) to the corresponding positions from
PAN-R-NIR1-DSM method (Figure 3.4 (A)). This is because the PAN-R-NIR1
method detected the highest brightness values of the images as tree tops, while the
PAN-R-NIR1-DSM method identified both the highest brightness and height values
as tree tops. When creating the watershed from the inverted surface model, most of
Method No. of objects extracted as tree tops
WV2 image only (PAN-R-NIR1) 21,027
WV2 image and DSM (PAN-R-NIR1-DSM) 32,789
Inverse Watershed Segmentation method (IWS) 29,794
117
the tree crowns that had small height differences within their neighbourhood were
clustered together resulting a single tree crown (Figure 3.3 (F)).
Figure 3.4. (A) Centroid positions obtained from the PAN-R-NIR1 and the
PAN-R-NIR1-DSM methods overlaid with the panchromatic WV2 image with low
pass filtering applied; (B) Overview of the Rapid Creek mangrove forest; (C)
Centroid positions obtained from the IWS method overlaid with the DSM.
118
3.3.3 Accuracy assessment
The visual appearance of the results obtained with the PAN-R-NIR1-DSM
method is the one closest to reality. Most of the tree crowns that have reasonable
height variations between neighbouring trees have successfully been detected. The
crown boundaries were smoother than two other methods. Although the tree crowns
obtained with the PAN-R-NIR1 method detected tree tops quite successfully, crown
boundaries had a distinctive jagged shape representing the pixel size of the image
bands. However, this was quite different when considering PAN-R-NIR1-DSM
method. The integration of the high spatial resolution DSM significantly improved
shapes of the crown boundaries. Although the IWS method detected a number of
tree tops closer to the one detected from the PAN-R-NIR1-DSM method (Table 3.2),
it was not possible to delineate the same amount of tree crowns after processing the
surface raster.
The distribution of validation dataset: in-situ surveyed and on screen digitized
tree crowns is shown in Figure 3.1. Table 3.3 shows the calculated statistical values
for different GEOBIA approaches. Tree crowns delineated from the PAN-R-NIR1-
DSM method resulted in the most successful closeness index of 0.11 (relative
accuracy of 92%) followed by the PAN-R-NIR1 combination with a closeness index
of 0.15 (relative accuracy of 89%). There is a slight improvement in tree crown
delineation when incorporating the DSM with optical imagery. OverSegmentation
(correctness) represents the ratio of the area of geographic intersection of reference
objects and tree crowns to the total area of reference objects. Hence, the PAN-R-
NIR1 method got the highest error rate (15%). Completeness (UnderSegmentation)
represents the ratio of the area of geographic intersection of reference objects and
119
tree crowns to the total area of corresponding objects (Clinton et al., 2010).
Completeness of both methods is very good, showing 0.02 for UnderSegmentation.
Table 3.3. A comparison of accuracies of extracted tree crowns: the possible range of
both OverSegmentation and UnderSegmentation values is [0,1] where zero defines
the perfect segmentation. The values closer to zero for “closeness index” illustrate
the closeness of mangrove tree crowns in two dimensional space which is defined by
OverSegmentation and UnderSegmentation to validation crowns (Clinton et al.,
2010).
The visual appearance of results obtained from the IWS method was poor
(Figures 3.3 (E) and (F)), and the quantitative accuracy assessment confirmed this in
Table 3.3. A closeness index value of 0.94 indicated very poor completeness due to
high UnderSegmentation value. The OverSegmentation value was close to zero
indicating that the most objects delineated using the IWS method coincided with
reference tree crowns.
Method OverSegmentation UnderSegmentation Closeness index
Overall relative
accuracy PAN-R-NIR1 0.15 0.02 0.15 89%
PAN-R-
NIR1-DSM
0.11 0.02 0.11 92%
IWS 0.02 0.94 0.94 35%
120
3.4. Discussion
Various approaches have been developed for detecting individual tree crowns
from remotely sensed data. However, most of these were developed with deciduous
or coniferous forests. Trees within these forests have crowns with conical shapes.
These methods are also generally successful in plantation forests where there are
regular distances between trees (Kaartinen et al., 2012; Leckie et al., 2005;
Tiede et al., 2005). According to Vauhkonen et al. (2012), the success of tree crown
delineation algorithms was significantly affected by the forest structure, especially
tree density and clustering. To confirm that, their study investigated six different tree
crown delineation algorithms for different forest types such as boreal forests in
Norway and Sweden, coniferous and broad leaved forests in Germany, and a tropical
pulpwood plantation in Brazil using laser scanning data. Due to the development of
tree crown delineation algorithms for specific forest types and species compositions,
they are less likely to be applied to mangroves. For example: Hirata et al. (2010)
segmented fewer mangrove tree crowns using QuickBird panchromatic satellite
images and a watershed segmentation than the reality. The presence of intermingled
mangrove tree crowns and the structural complexity of mangrove forests make them
unique and densely clustered.
In this study, tree top detection using local maxima, and crown boundaries
identification using a region growing method was tested on different data layer
combinations from high resolution optical images. This approach is more flexible
compared to the inverse watershed method in detecting tree crowns with varying
sizes as it uses the local maxima as the starting point of the growing seed, and the
local minima as the ending point. For example; the PAN-R-NIR1-DSM combination
121
used local maximum values of the DSM and NIR1 band as the starting point of the
seeds, grew them until the ratio of the NIR1 spectral value of the seed object and
neighbouring objects reached to 0.9, and the height difference between the seed
object and the neighbouring objects was up to 2 m. Further, instead of only
investigating the spectral variations of images, incorporating height variations of
canopies was fruitful as the generated DSM identified minor height variations within
canopies (Figure 3.2).
The PAN-R-NIR1 method showed a good visual appearance of mangrove
tree crowns when overlaid on the panchromatic image (Figure 3.3 (A)). This is
supported by the detailed study of Kamal et al. (2014) who confirmed the possibility
of identifying a single mangrove tree crown from remotely sensed images with pixels
smaller than 2 m. However, the number of trees detected using the PAN-R-NIR1
method was less than the PAN-R-NIR1-DSM method, and had relatively large gaps
between demarcated crowns. This was due to the similarities in spectral reflectance
causing some trees to be aggregated and demarcated as one. Therefore, it can be
concluded that the homogeneity within each mangrove species, a percentage of
crown overlap and limited height variations of canopies can affect the accuracy of
delineation. For example: if neighbouring trees were as tall as the seed trees,
brightness variations in the image were unlikely to be seen, and they were therefore
aggregated as one crown (Figure 3.3 (A)). However, conically shaped crowns and
crowns having considerable distance between each other was accurately delineated.
Successful tree crown delineation is strongly influenced by the shape of tree
crowns. Gougeon and Leckie (2006) stated that the detection of tree locations using
local maxima provides good results for medium to dense coniferous stands with
122
conical shape crowns in conjunction with GEOBIA. However, this is not the case
with mangroves. Most mangrove species tend to be rather flat as opposed to conical
in shape (Duke, 2006; Wightman, 2006), and therefore the local maxima approach is
less suitable at some places. The best approach is to divide the area into homogenous
forest stands before applying crown delineation methods. This finding supports that
of Larsen et al. (2011), who evaluated six different individual tree crown delineation
algorithms in varying forest conditions.
When the mangrove tree crowns obtained from the PAN-R-NIR1 DSM
combination were overlaid with the high resolution panchromatic image, the visual
appearance was better than the PAN-R-NIR1 combination (Figures 3.3 (C) and (D)).
The shapes of the crown boundaries were improved due to incorporation of both
image reflectance and height variations for region growing. The number of trees
identified using the PAN-R-NIR1-DSM method was larger than the other methods
(3.2). Since the DSM is sensitive to height variations (spatial resolution of the DSM
was 15 cm), trees with small height variations compared to neighbouring trees can
also be detected. Figure 3.4 (C) shows some examples of low laying mangrove tree
identification. However, there might be some false tree identifications as well. If the
DSM is sensitive to small height variations within canopies, multiple upward
pointing branches of the same tree are identified as different trees. Hence, more than
one tree crown can be counted per tree. Although there was a tree top filtering
technique applied in terms of average distance between trees, it was not possible to
avoid false tree identification completely. The filtering distance was calculated
according to field observations and generalized to the Rapid Creek area. To achieve
higher accuracy, this could be done on a species specific level although that in itself
is a challenge (Heenkenda et al., 2014). Further, there might be some inaccuracies
123
with respect to the DSM generation. For example; the homogeneous forest coverage
within communities and the repetitive pattern of mangroves can significantly impact
on the image matching accuracy (Baltsavias et al., 2008; Gruen, 2012). Introducing
ground control points and manual editing processes cannot avoid such irregularities
completely. One of the alternatives is to use laser scanning data with high point
density to generate canopy height models.
Although the IWS method detected tree tops as successfully as the other two
methods (Figure 3.4 (C) and Table 3.2), once we implemented the processing steps,
the method clustered neighbouring mangrove trees with similar characteristics
together and delineated them as a single tree crown (Figures 3.3 (E) and (F)). The
key input for the IWS algorithm is a flow direction map that shows direction of flow
out of each pixel of the inverted surface raster. If the height irregularities to the
neighbouring pixels are negligible with respect to the introduced control algorithms,
the system cannot identify corresponding flow direction (ArcGIS Resources - ESRI,
2012). Although the study considered the average height difference between tree top
and the crown edge (crown height) as a control parameter, the investigation of the
flow direction map showed the clear aggregation of neighbouring trees with
approximately similar heights. As we understand, the only solution to avoid such a
situation is to deal with homogeneous forest stands separately. However,
Wannasiri et al. (2013) stated that the inversed watershed method performed well in
mangrove forests having low percentage of crown overlap conditioning with a
canopy height model generated from airborne laser scanning data. The main reasons
would be the accuracy of the canopy height model and relatively large crowns
(average diameter is about 3 15 m). Further, the increasing percentage of crown
overlap decreases the extraction accuracy of individual tree parameters from canopy
124
height models. Another drawback of this method is that it involved creating many
intermediate processing files, thus the processing itself is complicated. This method
is a good solution for sparsely distributed, homogeneous forests with reasonable
height variations between trees (Swamer and Houser, 2012) but less suitable for
mangrove forest applications.
A quantitative accuracy assessment provides the measure of a goodness of
polygon matching with respect to validation samples. However, highest evaluation
results do not necessarily involve a good segmentation (Clinton et al., 2010). The
obtained results were validated against 268 tree crowns manually digitized from the
stereo models (Figure 3.1). The PAN-R-NIR1-DSM method showed the best results
having the lowest closeness index value or 92% overall relative accuracy (Table 3.3).
The completeness error of this method was also very low indicating almost all
reference objects were classified as tree crowns. When comparing the numerical
values of the accuracy assessment, there was no significant difference between the
PAN-R-NIR1 and the PAN-R-NIR1-DSM methods. However, the crown shape and
the number of trees identified were improved when incorporating the height
information rather than only using spectral information.
3.5 Conclusions and recommendations
This study tested different data layer combinations for delineating individual
tree crowns in a mangrove forest environment. WV2 imagery was used as the
primary data source, supplemented by an aerial photo derived DSM. The
combination of a WV2 image, a DSM and a GEOBIA algorithm was found to
successfully delineate mangrove tree crowns. The achieved overall relative accuracy
125
was 92%. Minor errors were detected where neighbouring trees with similar
characteristics were aggregated together and conversely, where multiple crowns were
assigned to a single tree (upward pointing branches of the same tree identified as
different trees). The WV2 image with a GEOBIA algorithm also provided good
results compared to the PAN-R-NIR1-DSM method. The statistical values show only
minor differences in accuracies between these two methods though the visual
appearance was dramatically improved when incorporating the DSM. The IWS
method, which has traditionally been successfully used in other environments with
more sparse canopies, did not provide data representative of mangrove tree crowns.
Therefore, it can be concluded that the combination of a WV2 image, a DSM
generated from image matching, and a GEOBIA is a relatively low cost, accurate,
and robust approach, which can be applied to heterogeneous forests like mangroves.
The study also confirmed that it is important to classify the imagery into
homogeneous species stands before applying individual tree detection algorithms.
The validation of the detected locations of trees or tree tops is recommended before
applying region growing algorithm in order to overcome the structural complexity
problems associated with mangroves. We also recommend using orthorectified WV2
images for detecting tree tops to eliminate topographic errors and spatial mismatches.
Further, we suggest testing a combination of WV2 image and airborne LiDAR
derived canopy height model to delineate mangrove tree crowns.
126
Acknowledgement
The support of the Northern Territory Government, Australia in providing aerial
photographs for this study is gratefully acknowledged. Authors appreciate Miguel
Tovar Valencia, Evi Warintan Saragih, Silvia Gabrina Tonyes, Olukemi Ronke
Alaba, and Dinesh Gunawardena for their assistance with field work. Authors would
also like to thank three anonymous reviewers and Ian Leiper for constructive advice
and editorial comments.
3.6 References
Analuddin, K., Sharma, S., Suwa, R. and Hagihara, A., 2009. Crown Foliage
Dynamics of Mangrove Kandelia obovata in Manko Wetland, Okinawa Island,
Japan, Journal of Oceanography, 65: 121-127.
ArcGIS Resources - ESRI. 2012. How Focal Statistics works, URL:
http://resources.arcgis.com/en/help/main/10.1/index.html#//009z000000r7000000,
(last date accessed: June 2013).
Baltsavias, E., Gruen, A., Eisenbeiss, H., Zhang, L. and Waser, L. T., 2008. High-
quality image matching and automated generation of 3D tree models, International
Journal of Remote Sensing, 29 (5): 1243-1259.
Beucher, S. and Lantuejoul, C., 1979. Use of Watersheds in Contour Detection,
Proceedings of the International Workshop on Image Processing: Real-time Edge
127
and Motion detection/estimation, 17-21 September 1979, Rennes, France, pp. 2.1-
2.12.
Blaschke, T., Johansen, K. and Tiede, D., 2011. Object-Based Image Analysis for
Vegetation Mapping and Monitoring, Advances in Environmental Remote Sensing
(Q. Weng, editor), CRC Press, Boca Raton, FL, USA, pp. 241-272.
Bunting, P. and Lucas, R., 2006. The delineation of tree crowns in Australian mixed
species forests using hyperspectral Compact Airborne Spectrographic Imager (CASI)
data, Remote Sensing of Environment, 101 (2): 230-248.
Chavez, J. P. S. and Bowell, J. A., 1988. Comparison of the Spectral Information
Content of Landsat Thematic Mapper and SPOT for Three Different Sites in the
Phoenix, Arizona, Journal of Photogrammetric Engineering and Remote Sensing, 54
(12): 1699-1708.
Chavez, J. P. S., Sides, S. C. and Anderson, J. A., 1991. Comparison of three
Different Methods to Merge Multiresolution and Multispectral Data: Landsat TM
and SPOT Panchromatic, Journal of Photogrammetric Engineering and Remote
Sensing, 57 (3): 295-303.
Chen, Q., Baldocchi, D., Gong, P. and Kelly, M., 2006. Isolating Individual Trees in
a Savanna Woodland Using Small Footprint Lidar Data, Photogrammetric
Engineering and Remote Sensing, 72 (8): 923-932.
128
Clinton, N., Holt, A., Lyan, L. and Gong, P., 2010. An accuracy assessment measure
for object based image segmentation, Proceedings of the International Archives of
the Photogrammetry, Remote Sensing and Spatial Information Sciences, Beijing,
vol.XXXVII Part 4B, pp. 1189-1194.
DigitalGlobe 2011, DigitalGlobe Core Imagery Products Guide, DigitalGlobe,
<http://www.digitalglobe.com/node/1094>.
Duke, N. C., 2006, Australia's Mangroves - the authoritative guide to Australia's
mangrove plants, University of Queensland, University of Queensland, Brisbane,
200 p.
Edson, C. and Wing, M. G., 2011. Airborne Light Detection and Ranging (LiDAR)
for Individual Tree Stem Location, Height, and Biomass Measurements, Remote
Sensing, 3: 2494 2528.
Eisank, C., Drăguţ, L., Götz, J. and Blaschke, T., 2010. Developing a Semantic
Model of Glacial Landforms for Object-based Terrain Classification –The Example
of Glacial Cirques Proceedings of the GEOBIA 2010: Geographic Object-Based
Image Analysis Ghent, Belgium
Erikson, M., 2004. Species classification of individually segmented tree crowns in
high-resolution aerial images using radiometric and morphologic image measures,
Remote Sensing of Environment, 91 (3–4): 469-477.
129
Erikson, M. and Olofsson, K., 2005. Comparison of three individual tree crown
detection methods, Machine Vision and Applications, 16 (4): 258-265.
Ferwerda, J. G., Ketner, P. and McGuinness, K. A., 2007. Differences in
regeneration between hurricane damaged and clear-cut mangrove stands 25 years
after clearing, Hydrobiologia, 591: 35-45.
Food and Agriculture Organisation 2007, The World's Mangroves 1980-2005, Food
and Agriculture Organisation of the United Nations, Rome.
Geoscience Australia. URL: http://www.ga.gov.au/ngrs/, (last date accessed: June
2013).
Gougeon, F. A., 1995. A crown-following approach to the automatic delineation of
individual tree crowns in high spatial resolution aerial images, Canadian Journal of
Remote Sensing, 21(3): 274-284.
Gougeon, F. A. and Leckie, D. G. 2003, Forest information extraction from high
spatial resolution images using an individual tree crown approach, Canadian Forest
Service, Pacific Forestry Centre, Victoria, British Columbia.
Gougeon, F. A. and Leckie, D. G., 2006. The Individual Tree Crown Approach
Applied to Ikonos Images of a Coniferous Plantation Area, Photogrammetric
Engineering & Remote Sensing, 72 (11): 1287 – 1297.
130
Gruen, A., 2012. Development and status of image matching in photogrammetry,
The Photogrammetric Record, 27 (137): 36-57.
Heenkenda, M., Joyce, K., Maier, S. and Bartolo, R., 2014. Mangrove Species
Identification: Comparing WorldView-2 with Aerial Photographs, Remote Sensing,
6 (7): 6064-6088.
Hirata, Y., Tabuchi, R., Patanaponpaiboon, P., Poungparn, S., Yoneda, R. and
Fujioka, Y., 2010. Estimation of aboveground biomass in mangrove forest damaged
by the major Tsunami disaster in 2004 in Thailand using high resolution satellite
data, Proceedings of the International Archives of the Photogrammetry, Remote
Sensing and Spatial Information Science, Kyoto Japan, vol. XXXVIII, pp. 643-646.
Jing, L., Hu, B., Noland, T. and Li, J., 2012. An individual tree crown delineation
method based on multi-scale segmentation of imagery, ISPRS Journal of
Photogrammetry and Remote Sensing, 70 (0): 88-98.
Kaartinen, H., Hyyppä, J., Yu, X., Vastaranta, M., Hyyppä, H., Kukko, A.,
Holopainen, M., Heipke, C., Hirschmugl, M., Morsdorf, F., Næsset, E., Pitkänen, J.,
Popescu, S., Solberg, S., Wolf, B. M. and Wu, J.-C., 2012. An International
Comparison of Individual Tree Detection and Extraction Using Airborne Laser
Scanning, Remote Sensing, 4: 950-974.
Kamal, M., Phinn, S. and Johansen, K., 2014. Characterizing the Spatial Structure of
Mangrove Features for Optimizing Image-Based Mangrove Mapping, Remote
Sensing, 6 (2): 984-1006.
131
Komiyama, A., Ong, J. E. and Poungparn, S., 2008. Allometry, biomass, and
productivity of mangrove forests: A review, Aquatic Botany, 89: 128-137.
Korpela, I., Dahlin, B., Schäfer, H., Bruun, E., Haapaniemi, F., Honkasalo, J.,
Ilvesniemi, S., Kuutti, V., Linkosalmi, M., Mustonen, J., Salo, M., Suomi, O. and
Virtanen, H., 2007. Single Tree forest inventory using LIDAR and aerial images for
3D tree top positioning, species recognition, height and crown width estimation,
Proceedings of the ISPRS Workshop Laser Scanning 2007 and SilviLaser 2007,
Espoo, Finland, vol. XXXVI-3/W52, p. 7.
Larsen, M., Eriksson, M., Descombes, X., Perrin, G., Brandtberg, T. and Gougeon,
F. A., 2011. Comparison of six individual tree crown detection algorithms evaluated
under varying forest conditions, International Journal of Remote Sensing, 32 (20):
5827-5852.
Larsen, M. and Rudemo, M., 1998. Optimizing templates for finding trees in aerial
photographs, Pattern Recognition Letters, 19: 1153-1162.
Leckie, D. G., Gougeon, F. A., Tinis, S., Nelson, T., Burnett, C. N. and Paradine, D.,
2005. Automated tree recognition in old growth conifer stands with high resolution
digital imagery, Remote Sensing of Environment, 94 (3): 311-326.
Li, Z., Hayward, R., Zhang, J. and Liu, Y., 2008. Individual Tree Crown Delineation
Techniques for Vegetation Management in Power Line Corridor, Proceedings of the
10th International Conference on Digital Image Computing: Techniques and
Applications, 1-3 December 2008, Canberra, Australia.
132
Lopez, J. P. A., 2012. Object-based methods for mapping and monitoring of urban
trees with multitemporal image analysis, Ph.D. dissertation, University of Twente,
176 p.
Moller, M., Lymburner, L. and Volk, M., 2007. The comparison index: A tool for
assessing the accuracy of image segmentation, International Journal of Applied Earth
Observation and Geoinformation, 9:311-321.
Northern Territory Government 2010, Northern Territory Digital Data and
Information, Department of Lands, Planning and the Environment, June, 2010.
Palubinskas, G., 2013. Fast, simple, and good pan-sharpening method, Journal of
Applied Remote Sensing, 7: 073526-1-12.
Pouliot, D. A., King, D. J., Bell, F. W. and Pitt, D. G., 2002. Automated tree crown
detection and delineation in high-resolution digital camera imagery of coniferous
forest regeneration, Remote Sensing of Environment, 82 (2–3): 322-334.
Suratman, M. N., 2008. Carbon Sequestration Potential of Mangroves in Southeast
Asia, Managing Forest Ecosystems: The Challenge of Climate Change (F. Bravo, R.
Jandl, V. LeMay and K. Gadow editors), Springer Netherlands, pp. 297-315.
Swamer, M. and Houser, P., 2012. Extraction of Tree Crowns and Heights using
LiDAR, URL: http://web.pdx.edu/~jduh/courses/geog493f12/Student_projects.htm,
(last date accessed: 3 October 2013).
133
Tiede, D., Hochleitner, G. and Blaschke, T., 2005. A full GIS-based workflow for
tree identification and tree crown delineation using laser scanning, Proceedings of
the International Archives of the Photogrammetry, Remote Sensing and Spatial
Information Science, vol.XXXVI (3/W24).
Vauhkonen, J., Ene, L., Gupta, S., Heinzel, J., Holmgren, J., Pitkänen, J., Solberg, S.,
Wang, Y., Weinacker, H., Hauglin, K. M., Lien, V., Packalén, P., Gobakken, T.,
Koch, B., Næsset, E., Tokola, T. and Maltamo, M., 2012. Comparative testing of
single-tree detection algorithms under different types of forest, Forestry, 85 (1): 27-
40.
Voa, Q. T., Kuenzerb, C., Voc, Q. M., Moderd, F. and Oppelte, N., 2012. Review of
valuation methods for mangrove ecosystem services, Ecological Indicators, 23: 431 -
446.
Wannasiri, W., Nagai, M., Honda, K., Santitamnont, P. and Miphokasap, P., 2013.
Extraction of Mangrove Biophysical Parameters Using Airborne LiDAR, Remote
Sensing, 5: 1787-1808.
Whiteside, T., Boggs, G. and Maier, S. W., 2011. Extraction of tree crowns from
high resolution imagery over Eucalypt dominant tropical savannas, Photogrammetric
Engineering and Remote Sensing, 77: 813-824.
Wightman, G., 2006, Mangrove Plant Identikit for North Australia's Top End,
Greening Australia NT, Darwin, Department of Natural Resources, Environment and
the Arts, 64p
134
Wolf, B. and Heipke, C., 2007. Automatic extraction and delineation of single trees
from remote sensing data, Machine Vision and Applications, 18: 317-330.
Zhan, Q., Molenaar, M., Klaus Tempfli and Wenzhong Shi., 2005. Quality
assessment for geo spatial objects derived from remotely sensed data, International
Journal of Remote Sensing, 26 (14): 2953-2974.
135
Chapter 4 - Quantifying mangrove chlorophyll
from high spatial resolution imagery
Graphical abstract
This work has been published in the following paper:
Heenkenda, M. K., Joyce, K. E., Maier, S. W., & de Bruin, S. 2015,
Quantifying mangrove chlorophyll from high spatial resolution imagery,
ISPRS Journal of Photogrammetry and Remote Sensing, vol. 108,
234-244, http://dx.doi.org/10.1016/j.isprsjprs.2015.08.003.
Formal permission from the publisher is attached in Appendix 3.
136
4.0 Abstract
Lower than expected chlorophyll concentration of a plant can directly limit
photosynthetic activity, and resultant primary production. Low chlorophyll
concentration may also indicate plant physiological stress. Compared to other
terrestrial vegetation, mangrove chlorophyll variations are poorly understood. This
study quantifies the spatial distribution of mangrove canopy chlorophyll variation
using remotely sensed data and field samples over the Rapid Creek mangrove forest
in Darwin, Australia. Mangrove leaf samples were collected and analysed for
chlorophyll content in the laboratory. Once the leaf area index (LAI) of sampled
trees was estimated using the digital cover photography method, the canopy
chlorophyll contents were calculated. Then, the nonlinear random forests regression
algorithm was used to describe the relationship between canopy chlorophyll content
and remotely sensed data (WorldView-2 satellite image bands and their spectral
transformations), and to estimate the spatial distribution of canopy chlorophyll
variation. The imagery was evaluated at full 2 m spatial resolution, as well as at
decreased resampled resolutions of 5 m and 10 m. The root mean squared errors with
validation samples were 0.82, 0.64 and 0.65 g/m2 for maps at 2 m, 5 m and 10 m
spatial resolution respectively. The correlation coefficient was analysed for the
relationship between measured and predicted chlorophyll values. The highest
correlation: 0.71 was observed at 5 m spatial resolution (R2=0.5). We therefore
concluded that estimating mangrove chlorophyll content from remotely sensed data
is possible using red, red-edge, NIR1 and NIR2 bands and their spectral
transformations as predictors at 5 m spatial resolution.
137
Keywords: Mangrove, chlorophyll, WorldView-2, satellite imagery, remote sensing
random forests
4.1 Introduction
Mangroves are one of the most productive and biochemically active
ecosystems (Suratman, 2008). Their dense root system reduces coastal erosion, and
protects the coastline from flood, waves and storms. These roots also filter and trap
pollutants, thereby decreasing coastal pollution. Mangrove forests serve as a nursery
area for shrimps, fish and crustaceans. On the global scale, high population pressure
in coastal areas has converted many mangrove forests into infrastructure, salt and
rice production, and aquaculture (Food and Agriculture Organisation, 2007). If these
coastal activities are unsustainably planned and managed, the result will be a large
scale mangrove degradation or deforestation. Therefore, the maintenance of
mangrove ecosystems is important.
Maintenance of mangrove ecosystems requires better knowledge of their
physiological processes such as photosynthesis, net primary production and plant
health (Flores-de-Santiago et al., 2013). The status of these physiological processes
may reflect by the nutrients in vegetation foliage (Filella and Penuelas, 1994). For
example, plant stress may affect the plant pigment system, and thus the
photosynthesis. More specifically, the chlorophyll content of any foliage is correlated
with nitrogen levels, and hence photosynthesis, and developmental stages (Filella
and Penuelas, 1994; Haboudane et al., 2002; Wu et al., 2008). Therefore, assessing
mangrove chlorophyll content is an important tool for ecosystem management.
138
Chlorophyll variations can occur at a variety of scales from individual trees to
communities, or even broader regions. Variations are affected by soil type, soil
nutrients, topography, and daily nutrient intake received from other sources
(Williams, 2012). Compared to other ecosystems, understanding mangrove
biochemical variations especially chlorophyll content is relatively limited (Flores-de-
Santiago et al., 2013; Williams, 2012; Zhang et al., 2012).
Quantifying spatial variability of mangrove chlorophyll from field
observations is time consuming and costly. An intensive sampling scheme is needed
throughout the area of interest to capture fine scale spatial variability. As fine scale
sampling is often prohibitive over large areas, predictive models are created to
extrapolate the information to unsampled regions. Remote sensing is integral in this
process.
Remote sensing has been used for decades to measure the chlorophyll content
of various natural plant communities (Atzberger et al., 2010; Boegh et al., 2012;
Clevers and Kooistra, 2012; Filella and Penuelas, 1994; Joyce and Phinn, 2003). For
example, Wu et al. (2008) compared the performances of several vegetation indices
derived from hyperspectral data in estimating chlorophyll content, and introduced
four new vegetation indices that are highly correlated with canopy chlorophyll.
Gitelson et al. (2003) investigated the spectral characteritics of the relationship
between reflectance and chlorophyll content of maple, chestnut, wild vine and beech
leaves, and developed a technique for non-destructive chlorophyll estimation. These
studies commonly related the absorption and reflectance of light in different
wavelengths with the presence or absence of photosynthetic pigments. Further,
developing statistical relationships between the chemical content extracted from leaf
139
samples and light reflectance in specific wavelengths using airborne or satellite data
allows modelling the spatial variations of chlorophyll over a large area.
Chlorophyll absorbs light strongly in the blue and red wavelength regions for
the purpose of photosynthesis. (Kokaly et al., 2009) described the broad wavelength
region of 400-700 nm as the most active region for leaf pigments or chlorophyll.
Filella and Penuelas (1994) identified the red edge region (wavelengths of 680-750
nm) as one of the best remote sensing descriptors of chlorophyll concentration. Hunt
et al. (2013) developed a triangular greenness index (TGI) considering spectral
reflectance at wavelengths: 480, 550, and 670 nm, and confirmed TGI as the best
spectral index for low-cost chlorophyll mapping. However, Wu et al. (2008) stated
that the blue region should not be used to estimate chlorophyll content due to its
overlapping absorption features with carotenoids. Therefore, the chlorophyll
prediction indices that provide higher accuracies are mainly based on the reflectance
around the 550 nm or 680 nm to 750 nm regions. In summary, there are numerous
options for revealing canopy pigment concentration from remotely sensed data.
The challenge is to identify the spectral bands of remote sensing data and
their transformations with the highest predictive power to model the spatial variation
of mangrove chlorophyll. Statistical relationships between field samples and
corresponding reflectance values provide numerical estimates for analysing the best
predictor. For example, Flores-de-Santiago et al. (2013) introduced Vog1 index
stating that the ratio of reflectance at 740 nm to that of at 720 nm as the best
predictor for the south end of the Urias system mangrove forest in Mexico. Zhang et
al. (2012) identified the red-edge position as the best predictor for the degraded
mangroves of Mexican Pacific during the dry season.
140
The aim of this study was to model the spatial distribution of mangrove
canopy chlorophyll content using remotely sensed data and field samples. Further,
the study analysed the optimal combination of predictor variables and spatial
resolution for chlorophyll mapping with the random forests regression algorithm.
4.2 Materials and methods
4.2.1 Study area and satellite data
This study focused on the Rapid Creek mangrove forest in Darwin, Northern
Territory, Australia (12o 22’43”S, 130o 51’55”E) (Figure 4.1). The extent of the
mangrove forest is about 3.8 hectares. Avicennia marina, Ceriops tagal, Bruguiera
exaristata, Lumnitzera racemosa, and Rhizophora stylosa are the most common
mangrove species in this forest (Heenkenda et al., 2014). There is relatively limited
coverage of the additional species Excoecaria ovalis and Aegialitis annulata.
A WorldView-2 (WV2) 2.0 m spatial resolution, multispectral satellite image
was selected as the remote sensing data source for this study. The image was
acquired on 26th July, 2013, with eight multispectral bands. The image was
radiometrically corrected with the sensor specifications published by DigitalGlobe®
(Updike, 2010). Digital numbers were converted to at-sensor radiance values, and
then to top-of-atmosphere reflectance values. The additive path radiance was
removed using the dark pixel subtraction technique in ENVI 5.0 software. Finally,
the image was geo-referenced using rational polynomial coefficients provided with
the image, and ground control points extracted from digital topographic maps of
Darwin, Australia (Heenkenda et al., 2014). To avoid the confusion between
141
mangroves and non mangroves, mangrove areas were extracted using the object-
based image analysis method described in Heenkenda et al. (2014). We used
contextual information, geometry and neighborhood characteristics of objects at
different hierarchical levels to separate mangrove coverage only.
Figure 4.1. The study area is located in the coastal mangrove forest of Rapid Creek in
Darwin, Northern Territory, Australia; (A) WorldView-2 satellite image and
locations of 29 field sampling plots (5 m x 5 m); WorldView-2 images ©
DigitalGlobe; (B) Australia, the boundary of the Northern Territory, and the study
area; (C) The Rapid Creek mangrove forest (study area); Coordinate system:
Universal Transverse Mercator Zone 52 L, WGS84.
142
The WV2 multispectral bands were resampled to 5 m and 10 m spatial
resolution using the cubic convolution resampling method. This was done to simulate
remote sensing images from other satellite missions that provide multispectral
images within the same spectral region such as RapidEye and SPOT 5. Green (506
nm - 586 nm), red (624 – 694 nm), red edge (699 nm – 749 nm), NIR1 (765 nm –
901 nm) and NIR2 (856 nm – 1043 nm) bands of WV2 image and number of band
ratios and indices were calculated with the image data to test their linear relationship
with mangrove chlorophyll content at three different spatial resolutions (Table 4.1).
The calculated vegetation indices and selected WV2 bands were stacked together to
form a single image with 12 bands for each spatial resolution.
To identify individual mangrove tree crowns, we used a stereo model of the
study area generated from an overlapping pair of WV2 panchromatic images
acquired on the same date. To prepare the stereo model, the image pair was oriented
to ground coordinates following digital photogrammetric image orientation steps in
the Leica Photogrammetric Suite (LPS) in ERDAS IMAGINE software. The ground
references were obtained from the Rational Polynomial Coefficients (RPC’s)
calculated during the image acquisition, and ground control points extracted from the
digital topographic map of Darwin, Australia. The accuracy of the stereo model was
assessed with respect to the ground control points.
143
Table 4.1. Vegetation indices used to establish the relationship between mangrove
field samples and remotely sensed data; green, red, red edge, NIR1 and NIR2
surface reflectance of WorldView-2 satellite data for corresponding wavelength
regions.
Vegetation index Band relationship Source
Normalized Difference
Vegetation Index (NDVI)
(𝑁𝑁𝑁𝑁𝑅𝑅1 − 𝑟𝑟𝑟𝑟𝑟𝑟) (𝑁𝑁𝑁𝑁𝑅𝑅1 + 𝑟𝑟𝑟𝑟𝑟𝑟)⁄ Rouse et al. (1974) and Ahamed et al. (2011)
Normalized Difference
Vegetation Index (NDVI2)
(𝑁𝑁𝑁𝑁𝑅𝑅2 − 𝑟𝑟𝑟𝑟𝑟𝑟) (𝑁𝑁𝑁𝑁𝑅𝑅2 + 𝑟𝑟𝑟𝑟𝑟𝑟)⁄ Mutanga et al. (2012)
Normalized Difference Red
Edge index (NDRE)
(𝑁𝑁𝑁𝑁𝑅𝑅1 − 𝑟𝑟𝑟𝑟𝑟𝑟 𝑟𝑟𝑟𝑟𝑙𝑙𝑟𝑟) (𝑁𝑁𝑁𝑁𝑅𝑅1 + 𝑟𝑟𝑟𝑟𝑟𝑟 𝑟𝑟𝑟𝑟𝑙𝑙𝑟𝑟)⁄ Barnes et al. (2000) and Ahamed et al. (2011)
Normalized Difference Red
Edge index (NDRE2)
(𝑁𝑁𝑁𝑁𝑅𝑅2 − 𝑟𝑟𝑟𝑟𝑟𝑟 𝑟𝑟𝑟𝑟𝑙𝑙𝑟𝑟) (𝑁𝑁𝑁𝑁𝑅𝑅2 + 𝑟𝑟𝑟𝑟𝑟𝑟 𝑟𝑟𝑟𝑟𝑙𝑙𝑟𝑟)⁄ Mutanga et al. (2012)
Green Normalized Difference
Vegetation Index (GNDVI)
(𝑁𝑁𝑁𝑁𝑅𝑅1 − 𝑙𝑙𝑟𝑟𝑟𝑟𝑟𝑟𝑔𝑔) (𝑁𝑁𝑁𝑁𝑅𝑅1 + 𝑙𝑙𝑟𝑟𝑟𝑟𝑟𝑟𝑔𝑔)⁄ Gitelson et al. (1996) and Ahamed et al. (2011)
Chlorophyll Vegetation Index
(CVI)
(𝑁𝑁𝑁𝑁𝑅𝑅1 ∗ 𝑟𝑟𝑟𝑟𝑟𝑟) (𝑙𝑙𝑟𝑟𝑟𝑟𝑟𝑟𝑔𝑔2)⁄ Vincini et al. (2007) and Vincini et al. (2008)
144
Figure 4.2 explains the overall workflow for mapping chlorophyll content
from the satellite data. The chlorophyll content per square unit of leaf area was
calculated using field samples. The leaf area index of each tree was calculated using
vertical photographs. They were used to estimate the canopy chlorophyll content.
Then, with a suspected non-linear relationship between chlorophyll and the spectral
reflectance, the random forests multiple regression algorithm was selected to
establish the relationship between samples and predictor variables. The modelled
relationship was then applied to the satellite imagery to obtain the chlorophyll map.
The method is discussed in greater details below.
145
Figure 4.2. Workflow for mapping chlorophyll content from satellite data and field
samples. This procedure was repeated for the satellite data with 5 m and 10 m spatial
resolution. NDVI = Normalized Difference Vegetation Index from NIR1 and red
bands; NDVI2 = Normalized Difference Vegetation Index from NIR2 and red bands;
NDRE = Normalized Difference Index from NIR1 and rededge bands; NDRE2 =
Normalized Difference Index from NIR2 and rededge bands; GNDVI: Normalized
Difference Index from NIR1 and green bands; CVI: Chlorophyll Vegetation Index
from NIR1, red and green bands.
146
4.2.2 Field sampling
Field sampling was conducted from 21st to 25th July, 2013 in the Rapid Creek
mangrove forest. This period coincided with the date of a satellite image overpass to
reduce discrepancies between sample data and predictor information. A sampling
pattern was determined considering the accessibility, density of mangrove forest,
mangrove greenness, and species variation. However, when selecting sampling
locations, attention was given to avoid areas close to water features due to the danger
of salt water crocodiles that inhabit the region. A total of 29 plots were selected with
an extent of 5 m x 5 m each. Within each plot, at least five trees that can easily be
identified on the image were selected for leaf collection. However, there were some
very dense mangrove plots where we could not select five trees due to their
overlapping canopies. From each tree (n=84), five leaves were collected from
different branches that were exposed to direct sunlight. The leaves were immediately
wrapped in an aluminium foil and placed in airtight and appropriately labelled plastic
bags in a cooler bag to transport to the laboratory. The position of each tree was
recorded using a course acquisition global positioning system (GPS). Further,
auxiliary details such as distances to roads and water features that can easily be
identified in the image were collected to aid locating these trees on the image.
Upward looking photographs of the canopy (at a zero zenith angle), were
taken using a Panasonic Lumix DMC-FT2 compact digital camera to calculate the
leaf area index (LAI) of each tree. Two spirit levels were used to minimize the
camera tilt. The camera was set to an aperture priority-automatic exposure mode.
Sixteen evenly spaced photographs were taken to cover each plot area. The number
of trees within the plot was counted in the field.
147
4.2.2.1 Estimating Leaf Area Index
The digital cover photography method developed by Macfarlane et al. (2007)
was selected to calculate LAI. This is a fast and reliable method for estimating LAI
within densely clustered mangroves. Vertical photographs were used to estimate the
large gaps between mangrove trees, small gaps within mangrove trees, the proportion
of the ground area covered by the vertical projection of foliage and branches, the
crown cover, the crown porosity, and the woody-to-total area ratio of each sampling
plot. A detailed description of the method to estimate LAI using digital photographs
can be found in Macfarlane et al. (2007) and Pekin and Macfarlane (2009).
All digital photographs were manually checked for overlapping mangrove
trees. An object based image analysis (OBIA) method was introduced to classify the
background (sky), mangrove leaves, and branches/stems using eCognition software.
Areas with high blue reflectance values were classified as the sky, considering the
contrast between vegetation and the background of the photographs. The ratio
between the reflectance in the blue and the red bands were further considered to
separate the background and the vegetation. Vegetation was further classified into
leaf area and branches/stems. To identify branches/stems from vegetation, the
greenness calculated from photographs were used as Eq. (1) (Liu and Pattey, 2010).
𝐺𝐺𝑟𝑟𝑟𝑟𝑟𝑟𝑔𝑔𝑔𝑔𝑟𝑟𝐵𝐵𝐵𝐵 = 2 ∗ 𝐺𝐺 − 𝐵𝐵 − 𝑅𝑅 (1)
Where G, B, and R represent the reflectance (intensity levels) recorded with the
green, blue and red bands of the digital camera.
148
The background of the vertical photograph was then classified into large gaps
between trees (gL), and small gaps within trees (gT) using size and relations to
neighboring features. For example, if the area of the gap was less than 50 pixels and
surrounded by vegetation, it was classified as a small gap within a tree. The total
number of pixels for large gaps and small gaps were calculated (see Appendix 4 A
for the detailed description of this classification).
The fraction of the foliage cover (ff) as the proportion of the ground area
covered by the vertical projection of foliage and branches (Macfarlane et al., 2007;
Walker and Tunstall, 1981), and the crown cover (fc) were calculated as per Eqs. (2)
and (3).
𝑓𝑓𝑐𝑐 = 1 − 𝑙𝑙𝐿𝐿 ∑𝑃𝑃𝐵𝐵𝑃𝑃𝑟𝑟𝑙𝑙𝐵𝐵⁄ (2)
𝑓𝑓𝑓𝑓 = 1 − 𝑙𝑙𝑇𝑇 ∑𝑃𝑃𝐵𝐵𝑃𝑃𝑟𝑟𝑙𝑙𝐵𝐵⁄ (3)
Where gL is the total number of pixels of large gaps and gT is the total number
of pixels, fc is the crown cover, ff is the fraction of the foliage cover, and ∑𝑃𝑃𝐵𝐵𝑃𝑃𝑟𝑟𝑙𝑙𝐵𝐵is
the total number of pixels of the photograph.
The crown porosity (ɸ) is the proportion of the ground area covered by the
vertical projection of foliage and branches within the perimeter of the crowns of
individual plants. It was calculated from Eq. (4).
ɸ = 1 − 𝑓𝑓𝑓𝑓 𝑓𝑓𝑐𝑐⁄ (4)
149
Where ff is the fraction of the foliage coverage and fc is the crown cover.
The effective plant area index (Lt) that includes the contribution from woody
elements to the total plant cover was estimated using the modified version of the
Beer-Lambert law as specified in Eq. (5).
𝐿𝐿𝑡𝑡 = −𝑓𝑓𝑐𝑐 ∗ (𝑙𝑙𝑔𝑔(∅) 𝑘𝑘⁄ ) (5)
Where fc is the crown cove, ɸ is the crown porosity and k is the canopy
extinction coefficient.
The canopy extinction coefficient (k) that defines an angle and spatial
arrangement of leaves was taken as 0.5 for this study. Different vegetation types
inherit different k values. For example, Clough et al. (1997) estimated k value of
mangroves as 0.56 considering direct measurements of mangrove leaf angles and
measurement of beam transmittance (DEMON instrument). The value of 0.525 can
also be considered as an appropriate value for mangrove stands (Green and Clark,
1997; Green et al., 1997). However, according to Perera et al. (2013), 0.5 is a good
average value for mangroves when there is no already estimated one.
Then we calculated the woody-to-total area ratio (α) and the actual leaf area
index (LAI) from Eqs. (6) and (7) (Chen, 1996; Chen et al., 1997).
𝛼𝛼 = ∑(𝐴𝐴𝑟𝑟𝑟𝑟𝐴𝐴𝐵𝐵)∑(𝐴𝐴𝑟𝑟𝑟𝑟𝐴𝐴𝑉𝑉) (6)
𝐿𝐿𝐿𝐿𝑁𝑁 = 𝐿𝐿𝑡𝑡 ∗ (1 − 𝛼𝛼) (7)
150
Where ∑(𝐿𝐿𝑟𝑟𝑟𝑟𝐵𝐵𝐵𝐵) is the total area of branches except leaves, ∑(𝐿𝐿𝑟𝑟𝑟𝑟𝐵𝐵𝑉𝑉) is the
total area of vegetation including branches, and Lt is the effective plant area index.
The final results represented an actual leaf area index of individual plots or a
cluster of trees. This was then divided by the number of trees to get a value per tree.
4.2.2.2 Estimating leaf chlorophyll concentration
A 10.4 mm diameter core was extracted from a mangrove leaf using a
sharpened metal tube, and ground with a small amount of 80% acetone (Wellburn,
1994). The slurry was put into a centrifuge tube, and the total amount made to a
volume of 5 ml. The samples were centrifuged for five minutes at 2500 rpm.
Approximately 1-2 ml of the supernatant was pipetted into a quartz cell for
spectrophotometric analysis. The absorbance of the solution was recorded at 663 nm
and 646 nm, and Chlorophyll a and b concentrations (mg/l) were calculated based on
the Eqns. (1) and (2) (Wellburn, 1994). The summation of these two represents the
total chlorophyll concentration (Arnon, 1949; Mackinney, 1938, 1941; Wellburn,
1994). To compare this with spectral reflectance of satellite images as an areal
measure, chlorophyll contents per square unit leaf area were calculated (g/m2). This
method was repeated for five leaves per tree, and the average value was considered
as the chlorophyll content per square unit leaf area of the particular tree. Finally, the
canopy chlorophyll concentration of each canopy was estimated multiplying the
chlorophyll content per square unit leaf area of the particular tree with the
corresponding LAI.
151
Chlorophyll a (Ca) = 12.25A663 – 2.79A646 (8)
Chlorophyll b (Cb) = 21.5A646 – 5.1A663 (9)
Where “A” is the absorbance at the wavelength identified by the subscript (in nm)
4.2.3 Estimating mangrove canopy chlorophyll content from satellite imagery
The random forests regression model is a nonparametric algorithm for
nonlinear multiple regression (Breiman and Cutler, 2001). The model calculates a
response variable (chlorophyll) by growing an ensemble of different decision trees,
and putting each pixel to be modelled down each of the decision trees. The response
is determined by evaluating the responses from all of these trees. In regression, a
resulting value for a pixel is the mean of all of the predictions of the random forests
(Breiman and Cutler, 2001). See Breiman (2001), Breiman and Cutler (2001),
Vincenzi et al. (2011) and Horning (2010) for a more detailed description of random
forests.
To identify mangrove trees on the stereo model of the study area, GPS
locations as well as additional information that were collected during field sampling
were used. Crown outlines of these trees were three dimensionally digitized on-
screen, and were overlaid with the predictors to calculate the corresponding predictor
variable values. Once overlaid, the pixels those have more than 70% of their extent
inside the tree crown were selected, and their corresponding predictor variable values
were noted. This identified pixels for calibration (n=57) and validation (n=22)
purposes.
152
Variables of interest in an experiment (those that are measured or observed)
are called response or dependent variables. Other variables in the experiment that
affect the response, and can be set or measured by the experimenter are called
predictor or independent variables. In this study, response variables are canopy
chlorophyll, and predictor variables are remotely sensed data derived vegetation
indices and WV2 individual bands.
To generate the relationship between field samples and these selected
predictor variable values (Table 4.1), the random forests regression was completed
using the “randomForest” package in R software (Breiman et al., 2014; Horning,
2011; Liaw and Wiener, 2002). After several iterations of fitting the random forests
model with different parameters, the optimal solutions were identified, such as the
number of random trees to be generated, and the number of variables that should be
used at each split when generating random trees.
When growing individual trees, the random forests algorithm uses
approximately two thirds of samples for fitting trees (model calibration)(Breiman,
2001; Vincenzi et al., 2011). The remaining samples are called “out-of-bag” (OOB)
data, and they are used to internally validate the model or to determine the error
drawn at each step (at each terminal node). In regression, the prediction error in the
OOB data is recorded as a mean of squared residuals (MSE). We recorded the MSE
value after parameter optimization as this indicated the predictive ability of the
model.
The importance of each predictor variable to the regression model was then
evaluated. The “randomForest” package provides two measures for variable
153
importance. One is based upon the mean decrease of accuracy in predictions on the
OOB data when the given variable is excluded from the model. First, the prediction
error in the OOB data for each tree is calculated (MSE). Then the same is calculated
after permuting each predictor variable on the OOB data. The difference between
these two instances are then averaged over all trees, and normalized by the standard
deviation of the differences (Liaw and Wiener, 2002).The other variable importance
measure is the total decrease in node impurity that resulted from splits over the
variable, averaged over all trees. Impurity is measured by the residual sum of squares
(Breiman, 2001; Horning, 2010). However, due to impurity bias for selecting split
variables, the variable importance measure can also be biased (Gromping, 2009).
In summary, the importance measure shows how much the mean squared
error (or impurity) changes when the variable is randomly permuted. For example, if
we randomly permute a variable that do not contribute to the prediction, the
prediction may not improve, and a small changes of the mean squared error (and
impurity) can only be noted. Therefore, high mean squared error indicates more
important variable (Breiman, 2001; Breiman and Cutler, 2001).
In this study, the decreasing mean squared error as the measure of variable
importance was used. This measure has also been adapted as the state-of-the-art
approach by various studies (Gromping, 2009). We also used the “varSelRF”
package in R software (Diaz-Uriarte, 2014) to find out the optimal number of
predictor variables that can be used to obtain the predicted map of canopy
chlorophyll at different spatial resolutions. This package uses the OOB error
minimization criteria to eliminate the least important variables.
154
Once the random forests model was calibrated and internally validated, it was
applied to the selected predictors to create a map of mangrove canopy chlorophyll
over the study area. For this study, we used a “raster” package from the R software
(Hijmans and van Etten, 2012) to predict the generated random forests model (with
all trees in the forest) using training data.
4.2.3.1 Canopy chlorophyll variation with respect to mangrove species
Once the chlorophyll map with the highest accuracy was identified, predicted
chlorophyll values for each mangrove species were assessed. Random points were
generated (n=500) using ArcGIS software over the mangrove species map produced
by Heenkenda et al. (2014). Corresponding canopy chlorophyll values were assigned
to these points, and the chlorophyll variation with respect to mangrove species was
investigated.
4.2.4 Accuracy assessment
The mean of squared residuals (OOB error) was analyzed to estimate the
predictive performance of the random forests model. The validation dataset was used
to assess the accuracy of the predicted map. The root mean squared error (RMSE)
and a correlation between predicted chlorophyll values and measured chlorophyll
values were recorded. The coefficient of determination (R2) was calculated to check
how close the data to the fitted regression line. The process was repeated for the data
with 5 m and 10 m spatial resolutions.
155
4.3 Results
4.3.1 Field sampling
The canopy chlorophyll values retrieved from field samples (n=84) varied
from 0.06 to 3.85 g/m2. The mean chlorophyll value was 1.32 g/m2, and the variance
was 0.78 g/m2. The LAI for individual trees ranged from 0.39 to 15.62. The average
LAI was 4.0.
4.3.2 Estimating mangrove canopy chlorophyll content from satellite imagery
Correlation coefficients were used to explore the relationship between
chlorophyll content and satellite data. A weak linear correlation between field
samples and satellite data was found (Figure 4.3). There is variability in correlation
not only between different predictor variables, but also between different spatial
resolutions. In some cases linear trend correlations were even positive at one spatial
resolution, while negative at a different resolution. For instance, although CVI index
had positive correlation for 2 m spatial resolution data, it had negative correlation for
both 5 m and 10 m satellite data (Figure 4.3). Hence, a small variation of any point
may change the trend line from positive to negative or negative to positive.
156
When considering 2 m spatial resolution, the NIR1 band, CVI and NDRE
indices showed the highest positive linear correlation with field samples (0.12, 0.14
and 0.26 respectively) while NIR2 showed the lowest correlation (0.02). Green,
yellow, red and red-edge bands revealed a negative linear correlation with field
samples (Figure 4.3).
Figure 4.3. Linear correlation coefficients against predictor variables with different
spatial resolutions; Green, Yellow, Red, RedEdge, NIR1 and NIR2: surface
reflectance of green, yellow, red, red-edge and near infrared bands of WorldView-2
image; NDVI = Normalized Difference Vegetation Index from NIR1 and red bands;
NDVI2 = Normalized Difference Vegetation Index from NIR2 and red bands; NDRE
= Normalized Difference Index from NIR1 and rededge bands; NDRE2 =
Normalized Difference Index from NIR2 and rededge bands; GNDVI: Normalized
Difference Index from NIR1 and green bands; CVI: Chlorophyll Vegetation Index
from NIR1, red and green bands.
157
From predictor variables with 5 m spatial resolution, NDRE2 index showed
the highest positive correlation while NIR2 band showed the lowest correlation
(0.01). Correlations obtained for green, yellow, red-edge, NIR1 bands and CVI,
NDVI and GNDVI indices of 10 m spatial resolution decreased with increasing
chlorophyll content, displaying a negative linear correlation. The red-edge band and
CVI index obtained the highest negative linear correlation (-0.20 and -0.21
respectively) while NDVI2 and NDRE2 had the highest positive linear correlation
(0.21 and 0.25 respectively). The correlation coefficient with NDRE was lower than
that of 2 m and 5 m resolutions. Therefore, it is difficult to interpret a strong linear
correlation between field samples and predictor variables at these spatial resolutions.
The most suitable parameters for the random forests model were selected
based on the OOB errors. The high number of random trees (n=2000) produced an
optimal solution having the lowest OOB error. OOB errors were 0.42, 0.54 and
0.55 g/m2 for 2 m, 5 m and 10 m resolutions respectively. The optimal number of
variables tried at each split node is four. However, compared to the variance of
canopy chlorophyll samples, these OOB errors are high, indicating relatively low
predictive performances of the models.
It was found that three variables are sufficient for predicting canopy
chlorophyll at 2 m spatial resolution, while four variables are required at 5 m and 10
m spatial resolution. Figure 4.4 shows the importance of each predictor variable
when estimating the response variable (canopy chlorophyll). The NIR1 band,
GNDVI and NDRE2 indices were considered the most important for 2 m spatial
resolution. However, they had relatively low importance at 5 m and 10 m spatial
resolution. The NDRE2, NDVI, and NDVI2 indices and the yellow band were the
158
four most important predictor variables when the spatial resolution was 5 m. Yellow
and green bands, NDRE2 and NDVI2 indices were the four most important
predictors for 10 m spatial resolution. In all instances, the NDRE index showed the
lowest importance (Figure 4.4).
Figure 4.4. Ranking the importance of predictor variables according to the mean
squared error (%IncMSE); Green, Yellow, Red, RedEdge, NIR1 and NIR2: surface
reflectance of green, yellow, red, red-edge, NIR1 and NIR2 bands of the
WorldView-2 image; NDVI = Normalized Difference Vegetation Index from NIR1
and red bands; NDVI2 = Normalized Difference Vegetation Index from NIR2 and
red bands; NDRE = Normalized Difference Index from NIR1 and rededge bands;
NDRE2 = Normalized Difference Index from NIR2 and rededge bands; GNDVI:
Normalized Difference Index from NIR1 and green bands; CVI: Chlorophyll
Vegetation Index from NIR1, red and green bands.
159
The canopy chlorophyll variation in Rapid Creek mangrove forest varies from
0.36 g/m2 to 2.79 g/m2 at three spatial resolutions (Figure 4.5). The lowest predicted
value for the map with 2 m spatial resolution was 0.43 g/m2 while more similar
values were estimated at 5 m and 10 m resolutions (0.36 g/m2 and 0.37 g/m2). The
highest chlorophyll values were shown along the creek (and its branches) at all
resolutions. Lowest chlorophyll values were always close to the edge of the
mangrove forest in the landward direction.
Figure 4.5.Canopy chlorophyll variation (g/m2) in the Rapid Creek mangrove forest;
(A) 2 m spatial resolution; (B) 5 m spatial resolution; and (C) 10 m spatial resolution.
160
4.3.2.1 Canopy chlorophyll variation with respect to mangrove species
As shown in Figure 4.6, Avicennia marina and Rhizophora stylosa had the
highest mean canopy chlorophyll value (1.6 g/m2). However, their canopy
chlorophyll distribution ranges are different. Chlorophyll content of Avicennia
marina ranged from 0.76 to 2.74 g/m2 while that of Rhizophora stylosa is 0.82 to
2.60 g/m2. However, the median canopy chlorophyll content of the Avicennia marina
was1.4 g/m2and, 50% of the dataset was within the range of 1.4 g/m2 to 2.7 g/m2.
Lumnitzara racamosa had the lowest mean canopy chlorophyll value (1.2 g/m2).The
median canopy chlorophyll content of the Lumnitzara racamosa was1.2 g/m2and,
50% of the dataset was within the range of 1.1 g/m2 to 1.3 g/m2 (Figure 4.6). The
distribution range of Bruguiera exaristata is 0.66 to 2.72 g/m2. Ceriops tagal and
Bruguiera exaristata had approximately equal mean values (1.24g/m2 and 1.25 g/m2
respectively) and distribution patterns (Figure 4.6). 75% of their canopy chlorophyll
content values were less than 1.3 g/m2.
161
Figure 4.6. The canopy chlorophyll distribution with mangrove species;
AM- Avicennia marina; BE - Bruguiera exaristata; CT - Ceriops tagal;
LR - Lumnitzera racemosa; RS - Rhizophora stylosa.
4.3.3 Accuracy assessment
The RMSE values, correlation coefficients (r) between predicted maps and
individual validation samples, and R2 are shown in Table 4.2. The 5 m spatial
resolution map showed the highest correlation coefficient (0.71) and R2equals to 0.5
indicating a strong linear relationship between predicted chlorophyll values and
measured chlorophyll samples. At 2 m spatial resolution, relatively low
correspondence between predicted values and measured values was observed. The
map with 10 m spatial resolution also showed a strong correlation with the validation
data.
(g/m
2
)
162
Table 4.2. Root mean squared errors (RMSE), correlation coefficients, and
coefficients of determinations calculated with respect to predicted chlorophyll values
and independent validation samples
4.4 Discussion
Mangroves those are prone to natural or anthropogenic disturbances (stressed
mangroves) experience decreases in leaf chlorophyll concentrations, incident light
absorbance, and overall productivity (Filella and Penuelas, 1994; Flores-de-Santiago
et al., 2013). Although a useful measure of these effects is quantifying nitrogen
stress, this is difficult to measure using remote sensing techniques. As leaf nitrogen
content is highly correlated with its chlorophyll content, chlorophyll can be used as a
surrogate measure. As such, we have demonstrated a method that estimates
mangrove canopy chlorophyll from remotely sensed data.
The LAI measurements obtained from the digital cover photography method
are at a successful level of accuracy. Although there is no recorded previous study in
the region to compare with, results are in line with few other available mangrove
studies around the world. For example, leaf area indices of Avicennia marina
Spatial
resolution
Root Mean Squared
Error (g/m2)
Correlation
coefficient (r)
Coefficient of
determination (R2)
2 m 0.82 0.28 0.08
5 m 0.64 0.71 0.50
10 m 0.65 0.66 0.44
163
plantations in Thailand varied from 0.5 to 5.0 (Laongmanee et al., 2013). A remotely
sensed data derived mangrove leaf area indices of British West Indies varied from
0.8 to 7.0 having 3.96 as the mean (Green et al., 1997). It can therefore be concluded
that the LAI values of sampled trees are at an acceptable level.
Leaf samples exhibited low chlorophyll concentration per square meter
compared to few available studies around the world. For instance, Flores-de-Santiago
et al. (2013) estimated mangrove chlorophyll following the same laboratory
procedure, and obtained the variation from 0.5 to 6.0 g/m2 for both dry and rainy
seasons. On the other hand, some studies used a chlorophyll meter to obtain canopy
chlorophyll values. A lack of concurrent laboratory testing for validation or
mangrove based chlorophyll calculation formulae led these studies to obtain low
chlorophyll concentration per square meter. However, this is not the situation with
this study. Since we used a laboratory based leaf chlorophyll analysis method, the
low chlorophyll concentration may represent some other external factors.
The Rapid Creek catchment is part of the larger Darwin Harbor catchment
including the largest urban area in Darwin, Australia, and therefore, carries relatively
higher pollutant load (Skinner et al., 2009). According to the report published by the
Northern Territory government, the Darwin Harbor catchments’ most soils are
relatively infertile with high potential acidity levels in the mangrove soil (Hill and
Edmeades, 2008). As mangrove nutrients or material exchange through tidal
inundation and flood water is a well-known phenomenon (Adame and Lovelock,
2011) the resultant nutrient deficiencies may affect to the Rapid Creek mangrove
chlorophyll concentration.
164
4.4.1 Estimating mangrove canopy chlorophyll content from satellite imagery
Numerous studies have examined the relationship between spectral
reflectance across the electromagnetic spectrum and leaf/canopy chlorophyll
concentration of various forest types (Hunt et al., 2013). Many studies have shown a
strong linear relationship between spectral reflectance and field samples. However,
this was not observed in this study despite testing various spectral bands and their
spectral transformations at different resolutions (Figure 4.3). A major obstacle to
radiometrically characterize mangroves is the extremely dynamic area where
mangroves grow. The resultant pixels are comprised of vegetation, soil and water
(Blasco et al., 1998; Kuenzer et al., 2011). In this study, special attention was given
to select sampling trees which we could easily identify on the image. Although these
trees had a less percentage of canopy overlapping with neighboring trees, the
resultant image pixels have more tendencies to mix with surrounding soil or mud.
Further, some sampling areas of the Rapid Creek mangrove forest covers densely
clustered, matured trees, and some areas cover with newly regenerated small
mangrove trees. The age factor of trees also affects the chlorophyll concentration of
canopies. Accounting all these reasons, we were not able to see a strong linear
relationship between spectral indices and canopy chlorophyll content as other
terrestrial vegetation exhibits.
We found no correlation or a poor negative linear relationship between
red-edge band and sample data (Figure 4.3), although Filella and Penuelas (1994)
identified the red edge location to evaluate plant chlorophyll. Some mangrove studies
showed a positive linear correlation with red edge position. For instance, during a dry
season, there was a positive correlation between red edge position and the
165
chlorophyll content for degraded mangroves in in Sinaloa, Mexico (Zhang et al.,
2012). Healthy mangroves of the same area exhibited a high positive correlation
between one of the spectral transformations of red edge position (the ratio of the
reflectance at 740 nm to that of 720 nm -Vog1 index) (Flores-de-Santiago et al.,
2013). Instead, in this study, the variable importance measures identified the
red-edge band or its spectral transformations as an important predictor variable at
three spatial resolutions indicating a strong non-linear relationship (Figure 4.4). For
example, NDRE2 derived from NIR2 and red-edge bands was considered as an
important variable at all three instances.
The red, green and yellow bands also got priorities (Figure 4.4). The yellow
band of WV2 image is designed to investigate the yellowness of vegetation. Further,
it is well known that NIR1, NIR2, and green bands are not only important for
extracting information, but they also minimize the spectral contamination of soil,
topography, and viewing angle of the sensor (Hunt et al., 2013). Reflectance in the
near infrared region is strongly influenced by the walls of the spongy mesophyll cells
of leaves, and therefore healthier leaves exhibit stronger reflectance in the near
infrared region (Zhang et al., 2012). Therefore, red, red-edge, NIR1 and NIR2 bands
of the WorldView-2 satellite image, and their derived normalized difference indices
were identified as the best predictor variables.
In this study, since we did not observe a strong linear relationship between
field samples and predictor variables, we used random forests nonlinear, multiple
regression algorithms. In this application, the random forests model provided a
moderate statistical means to quantify mangrove chlorophyll, while having numerous
advantages over other statistical regression methods. The model can support a large
166
number of variables with a relatively small number of training samples. Most
importantly, non-linear random forests regression does not over fit the response
variable (Breiman, 2001; Breiman and Cutler, 2001). The main disadvantage of the
application of this regression model is that it is not possible to predict or extrapolate
the values beyond the range in the training dataset. Therefore, the training dataset
(samples) must be selected carefully to represent the full range of potential
variations.
The canopy chlorophyll content appears to be correlated with proximity to
water features. For example, along the water features of the Rapid Creek mangrove
forest, maps show the highest chlorophyll values. These areas experience high daily
inundation rates, and consequently the trees are healthy and tall. Leaf chlorophyll is
highly dependent on soil nutrient, daily water intake and the exposure to direct
sunlight (Blasco et al., 1998). The landward margin of this mangrove forest receives
a relatively less daily intake of water, and therefore the soil salinity is high. Mostly
these areas are dominated by Lumnitzara racamosa. As shown in Figure 4.6,
Lumnitzara racamosa exhibited the lowest canopy chlorophyll content over the area.
Rhizophora stylosa dominated along water features, and showed relatively high
chlorophyll values. The spatial distribution patterns of all output maps are consistent
(low values closer to the landward margin and high values closer to the water
features) although the accuracy estimates are at low or moderate levels. Our resultant
maps cannot be used to determine chlorophyll values of specific trees but can be
used to investigate the spatial variations throughout the study area at a smaller scale.
167
4.4.2 Accuracy assessment
The map with 5 m spatial resolution showed the lowest RMSE value
(0.64 g/m2) compared to the validation data, and the highest correlation between
predicted and measured values (0.71) and the highest R2 value (0.5). Compared to 10
m spatial resolution map, there is only a slight quantitative difference. However,
pixel values of 10 m might be an integrated measure of individual trees, their
understory, soil and shadows. On the other hand, remotely sensed data with 2 m
spatial resolution had pixel sizes smaller than the size of an individual tree crowns
and a group of pixels characterizes individual trees.
It can be concluded that satellite images with 5 m spatial resolution would be
a suitable solution for chlorophyll mapping in this area. This is supported in part by
the study of Kamal et al. (2014). Although their results are limited to Moreton Bay,
Australia, they confirmed that a pixel size larger than 4 m is suitable for mapping
mangrove vegetation formation, communities and larger mangrove features.
According to their findings, pixel sizes between 0.5 m to 2 m would be an ideal
solution for perceiving structural variations within mangrove canopies (average tree
crowns, foliage clumping, and canopy gaps). However, these results can significantly
vary with characteristics of the mangrove forest and its available species. For
example, the average diameter of the tree crowns and the distance between trees
would make an enormous difference to the end results, increasing or decreasing the
level of information that can be extracted from corresponding pixels. According to
our results, WV2 spectral bands of red, red-edge, NIR1 and NIR2, and their spectral
transformations of 5 m spatial resolution would be the best input data source for
mangrove canopy chlorophyll estimation from the random forests algorithm for the
168
Rapid Creek mangrove forest. As an alternative solution, the RapidEye satellite
images with red-edge and near infrared band (or SPOT5 images) can be used.
Owing to moderate accuracies obtained from this study, a number of errors
were identified throughout the mapping process. Positions of field samples are the
most significant source of uncertainty. The geometric accuracy of the WV2 satellite
image was 0.25 m (half a pixel) compared to the ground control points established in
Darwin, Australia. The positional accuracy of the GPS receiver is about 5-10 m.
Although we used field measured, additional information to identify sampled trees
on images, at the end, field samples may not sufficient representatives of the image
pixels.
For future studies, we would recommend a transect method for field sampling
for more successful results. Two ends of transects should be established outside the
mangrove forest with a highly accurate positioning system. Then, all other
measurements that based on the established transect lines will be with greater
accuracy.
4.5 Conclusions and recommendations
This is the first study that mapped mangrove canopy chlorophyll variation
over the Rapid Creek mangrove forest, combining remotely sensed data, laboratory
measured samples, and random forests regression algorithm. Although we could find
a few studies that analysed the radiometric characteristics of mangrove leaf pigments
and their relation to various vegetation indices derived from satellite data, it is
difficult to find a recorded study for mapping a spatial distribution of mangrove
169
chlorophyll around the world as well. Hence, this study confirmed integrating
remotely sensed data and field samples for mapping mangrove canopy chlorophyll
content over a large area. The random forests regression algorithm showed moderate
results for this application, while there is a weak linear relationship between field
samples and the satellite data.
The best qualitative and quantitative results were obtained from the
chlorophyll maps with 5 m spatial resolution. The red, red-edge, NIR1 and NIR2
bands of the WorldView-2 satellite image, and their derived normalized difference
indices were identified as the best predictor variables. The chlorophyll map with
spatial resolution of 10 m also provided a similar accuracy level. The lowest
accuracy was obtained from the map with 2 m spatial resolution. As an alternative
solution, we would recommend using the RapidEye satellite images, including
red-edge band and near infrared bands.
Although we used field measured additional information to identify sampled
trees on images, it seems that the field samples are not sufficient representatives of
the image pixels. For future studies, we would recommend a transect method for
field sampling with greater positional accuracy. Two ends of these transects should
be established outside the mangrove forest with a highly accurate positioning system,
and the other measurements must be based on these transect lines. Further, the
authors would recommend testing this method on a different mangrove forest to see
the transferability of the method used.
170
Appendix 4.A
171
4.6 References
Adame, M. F. and Lovelock, C. E., 2011. Carbon and nutrient exchange of mangrove
forests with the coastal ocean, Hydrobiologia, 663: 23-50.
Ahamed, T., Tian, L., Zhang, Y. and Ting, K. C., 2011. A review of remote sensing
methods for biomass feedstock production, Biomass and Bioenergy, 35 (7): 2455-
2469.
Alongi, D. M., 1994. Zonation and seasonality of benthic primary production and
community respiration in tropical mangrove forests, Oecologia, 98: 320-327.
Alongi, D. M., 2002. Present state and future of the world’s mangrove forests,
Environmental Conservation, 29 (3): 331-349.
Arnon, D. I., 1949. Copper enzymes in isolated chloroplasts. Polyphenoloxidase in
Beta vulgaris Journal of Plant Physiology, 24 (1): 1-15.
Atzberger, C., Guérif, M., Baret, F. and Werner, W., 2010. Comparative analysis of
three chemometric techniques for the spectroradiometric assessment of canopy
chlorophyll content in winter wheat, Computers and Electronics in Agriculture, 73
(2): 165-173.
Axelsson, C., 2011. Predicting Mangrove Leaf Chemical Content from
Hyperspectral Remote Sensing using Advanced Regression Techniques, Master of
Science dissertation, University of Southampton (UK), Lund University (Sweden),
172
University of Warsaw (Poland), University of Twente, Faculty ITC (The
Netherlands), 62 p.
Barnes, E. M., Clarke, T. R., Richards, S. E., Colaizzi, P. D., Haberland, J.,
Kostrzewski, M., Waller, P., Choi, C., Riley, E., Thompson, T., Lascano, R. J., Li.,
H. and Moran, M. S., 2000. Coincident Detection of Crop Water Stress, Nitrogen
Status and Canopy Density Using Ground-Based Multispectral Data, Proceedings of
the Fifth International Conference on Precision Agriculture, 16-19 July 2000,
Bloomington, MN, USA.
Blasco, F., Gauquelin, T., Rasolofoharinoro, M., J.Denis, M.Aizpuru and Caldairou,
V., 1998. Recent Advances in mangrove studies using remote sensing data, Marine
Freshwater Resources, 49: 287-296.
Boegh, E., Houborg, R., Bienkowski, J. F., Braban, C. F., Dalgaard, T., van Dijk, N.,
Dragosits, U., Holmes, E., Magliulo, V., Schelde, K., Di Tommasi, P., Vitale, L.,
Theobald, M. R., Cellier, P. and Sutton, M., 2012. Remote sensing of LAI,
chlorophyll and leaf nitrogen pools of crop and grasslands in five European
landscapes, J Biogeosciences Discussions, 8: 10149-10205.
Breiman, L., 2001. Random Forests, Machine Learning, 45: 5-32.
Breiman, L. and Cutler, A., 2001. Random Forests, URL:
https://www.stat.berkeley.edu/~breiman/RandomForests/cc_home.htm, (last date
accessed: 10-07-2013).
173
Breiman, L., Cutler, A., Liaw, A. and Wiener, M., 2014. Breiman and Cutler's
random forests for classification and regression-Package 'randomForest', URL:
http://stat-www.berkeley.edu/users/breiman/RandomForests, (last date accessed: 12-
07-2013).
Chen, J. M., 1996. Optically-based methods for measuring seasonal variation of leaf
area index in boreal conifer stands, Agricultural and Forest Meteorology, 80: 135-
163.
Chen, J. M., Rich, P. M., Gower, S. T., Norman, J. M. and Plummer, S., 1997. Leaf
area index of boreal forests: Theory, techniques, and measurements, Journal of
Geophysical Research, 102 (D24): 29429-29443.
Clevers, J. G. P. W. and Kooistra, L., 2012. Using Hyperspectral Remote Sensing
Data for Retrieving Canopy Chlorophyll and Nitrogen Content, IEEE Journal of
selected topics in applied earth observation and remote sensing, 5 (2).
Clough, B. F., Ong, J. E. and Gong, W. K., 1997. Estimating leaf area index and
photosynthetic production in canopies of the mangrove Rhizophora apiculata,
Marine Ecology Progress Series, 159: 285-292.
Croft, H., Chen, J. M., Zhang, Y., Simic, A., Noland, T. L., Nesbitt, N. and Arabian,
J., 2015. Evaluating leaf chlorophyll content prediction from multispectral remote
sensing data within a physically-based modelling framework, ISPRS Journal of
Photogrammetry and Remote Sensing, 102 (0): 85-95.
174
Diaz-Uriarte, R., 2014. varSelRF: Variable Selection using Random Forests, R
News.
Filella, I. and Penuelas, J., 1994. The red edge position and shape as indicators of
plant chlorophyll content, biomass and hydric status, International Journal of Remote
Sensing, 15 (7): 1459-1470.
Flores-de-Santiago, F., Kovacs, J. M. and Flores-Verdugo, F., 2013. The influence of
seasonality in estimating mangrove leaf chlorophyll-a content from hyperspectral
data, Wetlands, Ecology and Management, 21 (3): 193-207.
Food and Agriculture Organisation 2007, The World's Mangroves 1980-2005, Food
and Agriculture Organisation of the United Nations, Rome.
Gitelson, A. A., Kaufman, Y. J. and Merzlyak, M. N., 1996. Use of a green channel
in remote sensing of global vegetation from EOS-MODIS, remote Sensing of
Environment, 58 (3): 289-298.
Gitelson, A. A., Gritz, Y. and Merzlyak, M. N., 2003. Relationship between leaf
chlorophyll content and spectral reflectance and algorithms for non-destructive
chlorophyll assessment in higher plant leaves, Journal of Plant Physiology, 160: 271-
282
Green, E. and Clark, C., 1997. Assessing Mangrove Leaf Area Index and Canopy
Closure, Remote Sensing Handbook for Tropical Coastal Management (extracts) (A.
J. Edwards editor, UNESCO, Department for International Development.
175
Green, E. P., Mumby, P. J., Edwards, A. J., Clark, C. D. and Ellis, A. C., 1997.
Estimating leaf area index of mangrove from satellite data, Aquatic Botany, 58:
11-19.
Gromping, U., 2009. Variable Importance Assessment in Regression: Linear
Regression versus Random Forest, The American Statistician, 63 (4): 308-319.
Haboudane, D., Miller, J. R., Tremblay, N., Zarco-Tejada, P. J. and Dextraze, L.,
2002. Integrated narrow-band vegetation indices for prediction of crop chlorophyll
content for application to precision agriculture, remote Sensing of Environment, 81:
416-426.
Heenkenda, M., Joyce, K., Maier, S. and Bartolo, R., 2014. Mangrove Species
Identification: Comparing WorldView-2 with Aerial Photographs, Remote Sensing,
6 (7): 6064-6088.
Hijmans, R. J. and van Etten, J. 2012, raster: Geographic analysis and modeling with
raster data, R package version 2.0-12, http://CRAN.R-project.org/package=raster.
Hill, J. V. and Edmeades, B. F. J. 2008, Acid Sulfate Soils of the Darwin Region,
Land and Water Division, Department of Natural Resources, Environment the Arts
and Sport, Darwin.
Horning, N., 2010. Random Forests: An algorithm for image classification and
generation of continuous fields data sets, Proceedings of the GeoInformatics for
Spatial-Infrastructure Development in Earth and Allied Sciences (GIS-IDEAS) 9-12-
176
2010, Hanoi, Vietnam, http://wgrass.media.osaka-
cu.ac.jp/gisideas10/viewpaper.php?id=342.
Horning, N., 2011. Training Guide for Using Random Forest to Classify Satellite
Images-v7, American Museum of Natural History, (last date accessed.
Hunt, E. R., Doraiswamy, P. C., McMurtrey, J. E., Daughtry, C. S. T., Perry, E. M.
and Akhmedov, B., 2013. A visible band index for remote sensing leaf chlorophyll
content at the canopy scale, International Journal of Applied Earth Observation and
Geoinformation, 21: 103-112.
Joyce, K. E. and Phinn, S. R., 2003. Hyperspectral analysis of chlorophyll content
and photosynthetic capacity of coral reef substrates, Limnology and Oceanography,
48 (Part 1 and 2): 489-496.
Kamal, M., Phinn, S. and Johansen, K., 2014. Characterizing the Spatial Structure of
Mangrove Features for Optimizing Image-Based Mangrove Mapping, Remote
Sensing, 6 (2): 984-1006.
Kokaly, R. F., Asner, G. P., Ollinger, S. V., Martin, M. E. and Wessman, C. A.,
2009. Characterizing canopy biochemistry from imaging spectroscopy and its
application to ecosystem studies, remote Sensing of Environment, 113: S78-S91.
Kuenzer, C., Bluemel, A., Gebhardt, S., Quoc, T. V. and Dech, S., 2011. Remote
Sensing of Mangrove Ecosystems: A Review, Remote Sensing, 3: 878-928.
177
Laongmanee, W., Vaiphasa, C. and Laongmanee, P., 2013. Assessment of Spatial
Resolution in Estimating Leaf Area Index from Satellite Images: A Case Study with
Avicennia marina Plantations in Thailand, International Journal of Geomatics, 9 (3):
69-77.
Liaw, A. and Wiener, M., 2002. Classification and Regression by randomForest, R
News, 2 (3): 18-22.
Liu, J. and Pattey, E., 2010. Retrieval of leaf area index from top-of-canopy digital
photography over agricultural crops, Agricultural and Forest Meteorology, 150 (11):
1485-1490.
Macfarlane, C., Hoffman, M., Eamus, D., Kerp, N., Higginson, S., McMurtrie, R.
and Adams, M., 2007. Estimation of leaf area index in eucalypt forest using digital
photography, Agricultural and Forest Meteorology, 143 (3–4): 176-188.
Mackinney, G., 1938. Some Absorption Spectra of Leaf Extracts, Journal of Plant
Physiology, 13 (1): 123-140.
Mackinney, G., 1941. Absorption of light by chlorophyll solutions, Journal of
Biological Chemistry, 140: 315-322.
Mutanga, O., Adam, E. and Cho, M. A., 2012. High density biomass estimation for
wetland vegetation using WorldView-2 imagery and random forest regression
algorithm, International Journal of Applied Earth Observation and Geoinformation,
18 (0): 399-406.
178
Pekin, B. and Macfarlane, C., 2009. Measurement of Crown Cover and Leaf Area
Index Using Digital Cover Photography and Its Application to Remote Sensing,
Remote Sensing, 1 (4): 1298-1320.
Perera, K. A. R. S., Amarasinghe, M. D. and Somaratna, S., 2013. Vegetation
structure and species distribution of mangroves along a soil salinity gradient in a
micro tidal estuary on the north-western coast of Sri Lanka, American Journal of
Marine Science, 1 (1): 7-15.
Rouse, J. W., Haas, J., R. H., S. J. A. and Deering, D. W., 1974. Monitoring
vegetation systems in the Great Plains with ERTS, Proceedings of the Third ERTS-1
Symposium, NASA, Washington, DC.
Skinner, L., Townsend, S. and Fortune, J. 2009, The Impact of Urban Land-use on
Total Pollutant Loads Entering Darwin Harbour, Department of Natural Resources,
Environment, the Arts and Sport, Darwin.
Suratman, M. N., 2008. Carbon Sequestration Potential of Mangroves in Southeast
Asia, Managing Forest Ecosystems: The Challenge of Climate Change (F. Bravo, R.
Jandl, V. LeMay and K. Gadow editors), Springer Netherlands, pp. 297-315.
Updike, T.and Comp, C. 2010, Radiometric Use of WorldView-2 Imagery,
Technical Note,
http://www.digitalglobe.com/sites/default/files/Radiometric_Use_of_WorldView-
2_Imagery%20(1).pdf.
179
Vincenzi, S., Zucchetta, M., Franzoi, P., Pellizzato, M., Pranovi, F., de Leo, G. A.
and Torricelli, P., 2011. Application of a Random Forest algorithm to predict spatial
distribution of the potential yield of Ruditapes philippinarum in the Venice lagoon,
Italy, Ecological Modelling, 222 (8): 1471-1478.
Vincini, M., Frazzi, E. and D'Alessio, P., 2007. Comparison of narrow-band and
broad-band vegetation indexes for canopy chlorophyll density estimation in sugar
beet, Proceedings of the 6th European Conference on Precision Agriculture '07,
Wageningen, The Netherlands, pp. 189-196.
Vincini, M., Frazzi, E. and D'Alessio, P., 2008. A broad-band leaf chlorophyll
vegetation index at the canopy scale, Precision Agriculture, 9: 303-319.
Walker, J. and Tunstall, B. R. 1981, Field estimation of foliage cover in Australian
woody vegetation, CSIRO Institute of Biological Resources, Canberra.
Wellburn, A. R., 1994. The Spectral Determination of Chlorophyll a and b, as well
as Total Carotenoids Using Various Solvents with Spectrometers of Different
Resolution, Journal of Plant Physiology, 144: 307-313.
Williams, G. J., 2012. Estimating Chlorophyll Content in Mangrove Forest Using a
Neighbourhood based Inversion Approach, Master of Science dissertation, Lund
University (Sweden), University of Twente, Faculty ITC (The Netherlands)
180
Wu, C., Niu, Z., Tang, Q. and Huang, W., 2008. Estimating chlorophyll content from
hyperspectral vegetation indices: Modelling and validation, Agricultural and Forest
Meteorology, 148 (8–9): 1230-1241.
Zhang, C., Liu, Y., Kovacs, J. M., Flores-Verdugo, F., Flores-de-Santiago, F. and
Chen, K., 2012. Spectral response to varying levels of leaf pigments collected from a
degraded mangrove forest, Journal of Applied Remote Sensing, 6.
181
Chapter 5 - Estimating mangrove biophysical
variables using WorldView-2 satellite data: Rapid
Creek, Northern Territory, Australia
Graphical abstract
This work has been published in the following paper:
Heenkenda M. K., Maier S.W., & Joyce K. E. 2016, Estimating mangrove
biophysical variables using WorldView-2 satellite data: Rapid Creek,
Northern Territory, Australia, Journal of Imaging, 2, 24;
doi:10.3390/jimaging2030024
Publisher confirmed reproducing the journal article without written
permission since “Journal of Imaging” is an Open Access journal.
182
5.0 Abstract
Mangroves are one of the most productive coastal communities in the world.
Although we acknowledge the significance of ecosystems, mangroves are under
natural and anthropogenic pressures at various scales. Therefore, understanding
biophysical variations of mangrove forests is important. An extensive field survey is
impossible within mangroves. WorldView-2 multi-spectral images having a 2-m
spatial resolution were used to quantify above ground biomass (AGB) and leaf area
index (LAI) in the Rapid Creek mangroves, Darwin, Australia. Field measurements,
vegetation indices derived from WorldView-2 images and a partial least squares
regression algorithm were incorporated to produce LAI and AGB maps. LAI maps
with 2-m and 5-m spatial resolutions showed root mean square errors (RMSEs) of
0.75 and 0.78, respectively, compared to validation samples. Correlation coefficients
between field samples and predicted maps were 0.7 and 0.8, respectively. RMSEs
obtained for AGB maps were 2.2 kg/m2 and 2.0 kg/m2 for a 2-m and a 5-m spatial
resolution, and the correlation coefficients were 0.4 and 0.8, respectively. We would
suggest implementing transects method for field sampling and establishing end
points of these transects with a highly accurate positioning system. The study
demonstrated the possibility of assessing biophysical variations of mangroves using
remotely-sensed data.
Keywords: Mangrove, above ground biomass, leaf area index, WorldView-2, partial
least square regression
183
5.1 Introduction
Mangrove forests are a dominant feature of many tropical and subtropical
coastlines. They have a variety of growth forms, including intertidal trees, palms and
shrubs that often grow in dense stands [1]. Although mangroves form valuable
ecosystems along sheltered coastal environments, on the global scale, they are
disappearing at an alarming rate [2]. For instance, by 2000, the worldwide mangrove
extent has fallen below 15 million ha, down from 19.8 million ha in 1980 [3]. The
world has thus lost five million ha of mangroves over twenty years, or 25% of the
extent found in 1980. The main reasons for rapid mangrove destruction and land
clearings are urbanization, population growth, water diversion, aquaculture,
agriculture and salt pond construction.
Land clearings throughout catchments and in urbanized areas can cause an
increased volume of water entering watercourses, carrying substances, such as
topsoil, chemicals, rubbish and nutrients. These substances deposit on sediments in
which mangroves grow. The increased amount of water influences the rate of erosion
or deposition of sediments, causing a significant problem for the health of aquatic
habitats. Therefore, when sustainable developments are progressing through land
clearing, it is necessary to ensure effective mangrove conservation.
To set the balance between mangrove conservation and developments, one of
the vital roles of monitoring and ultimately managing mangroves is to create an
accurate and up-to-date quantitative analysis of their baseline health parameters.
However, assessing mangrove health is not a straightforward task. This is due to the
complex structure of forests, their biophysical variations and the interaction between
184
mangroves, soil, water and salinity. Therefore, indirect measures that correlate with
the levels of vegetation stresses, such as above ground biomass (AGB), canopy
nutrient levels, particularly canopy nitrogen level, and leaf area index (LAI), can
eventually be considered.
Most of the conventional methods that have been developed for estimating
AGB, LAI and canopy nutrient levels have limitations when extended over space and
time. For example, estimating biomass using the allometric method is based on
measureable canopy dynamics, such as tree height and diameter at breast height
(DBH). Due to the within-stand heterogeneity of canopies, labour-intensive, site- and
species-specific field measurements are crucial [4–6]. However, field sampling
within mangrove forests is challenging. Many mangrove species have complex aerial
root systems, which make sampling difficult. Furthermore, mangroves are dense and
rather difficult to walk through. Remotely-sensed data addresses the major
challenges (especially field sampling) identified with already developed conventional
methods that estimate AGB and LAI. Most remote sensing-based approaches are
capable of estimating plant biophysical characteristics by reducing the shortcomings
of field observations.
There are numerous studies to estimate the biophysical characteristics of
vegetation using remotely-sensed data. The key issue is to correlate the intensity of
electromagnetic energy absorbed or reflected by the plant (spectral reflectance) with
ground measurements of biophysical variables. This spectral reflectance is either
measured in situ (using spectrometers) or via airborne or spaceborne sensors. Several
studies have found strong correlations between forest biomass or LAI and spectral
reflectance values at different wavelengths [7,8]. Eckert [9] and Ahamed et al. [10]
185
summarized different studies that analysed the levels of vegetation greenness in
terms of vegetation indices derived from remotely-sensed data for estimating
biomass. For example, Anaya et al. [7] and Satyanarayana et al. [11] used the
normalized difference vegetation index (NDVI) calculated from red and near infrared
wavelengths of remotely-sensed data for estimating biomass. However, compared to
other terrestrial vegetation types, little has been adapted to mangroves.
Compared to other terrestrial ecosystems, a few studies associated with field
samples and remotely-sensed data for producing thematic maps over mangrove
forests can be found [11–16]. Clough et al. [17], Green and Clark [18] and Green et
al. [14] established a relationship between vegetation indices derived from remotely-
sensed data and in situ LAI samples for mangroves.
Kamal et al. [15] studied the spatial resolution of satellite images, spectral
vegetation indices and different mapping approaches for LAI estimation at Moreton
Bay, Australia, and Karimunjawa Island, Indonesia. The study confirmed that the
LAI estimation accuracy using remotely-sensed data was site specific and varied
across pixel sizes and image segmentation scales. Since there is no recorded study
related to the Rapid Creek mangrove forest in Darwin, Australia, the demand still
exists for estimating the LAI of mangroves by integrating in situ samples, remotely-
sensed data and advanced statistical regression algorithms. To the same extent, we
could find only a few studies that mapped AGB of mangroves over a large area. For
example, recently, Zhu et al. [16] retrieved mangrove AGB from field data and
WorldView-2 (WV2) satellite images. This study tested Sonneratia apetala and
Kandelia candel mangrove species and a limited number of vegetation indices for
AGB mapping over the study site. The study confirmed the importance of the red
186
edge band of WV2 satellite image and associated vegetation indices for AGB
estimation over other spectral regions.
Komiyama et al. [19], Komiyama et al. [5], Fu and Wu [4] and Perera and
Amarasinghe [20] derived only a relationship between mangrove canopy dynamics,
AGB and vegetation indices. Therefore, testing the possibility of mapping AGB of
mangroves over a large area from remotely-sensed data is still required.
This study aimed to quantify the spatial distribution of LAI and AGB of
mangroves using field measurements and WV2 satellite images. To achieve this aim,
the study compared two different spatial resolutions of WV2 data (original spatial
resolution of multispectral bands (2 m) with resampled multispectral bands (5 m))
and partial least squares regression algorithm for mapping LAI and AGB variations
over a large area.
5.2 Data and methods
5.2.1 Study area
The Rapid Creek mangrove forest in Darwin, Northern Territory, Australia
was selected as the test site for this study (Figure 5.1). This area is dominated by five
different mangrove species - Avicennia marina, Ceriops tagal, Bruguiera exaristata,
Lumnitzera racemosa, and Rhizophora stylosa [20]. Other species do not represent
significant coverage to be considered separately.
187
Figure 5.1. Rapid Creek coastal mangrove forest, Darwin, Northern Territory,
Australia. The distribution of field sampling plots (5m x 5m) is shown in the map
(the sizes of magenta squares are not to the scale). Aerial photographs © Northern
Territory Government.
5.2.2 Field sampling, satellite data and predictor variables
A total of 29 plots, 5 m × 5 m in extent, were identified in the field. These
plots were selected with trees having similar characteristics (species, age, height and
DBH) to represent calculated values accurately. Each of them was oriented in a
north-south and east-west direction in order to locate them easily in the satellite
image. Inside each plot, five trees were selected to take measurements based on their
188
ability to be identified in satellite images; that is isolated and without clumping to
neighbouring tree crowns. A sampling pattern was determined considering the
accessibility, density of mangrove forest, mangrove greenness and species variation.
Attention was also given to avoid areas close to water features due to the danger of
salt water crocodiles that inhabit the region.
Tree height, diameter at breast height and species were recorded. Species
were identified based on the guidelines provided by Duke [22] and Wightman [23].
The positions of each the tree and the four corners of each plot were recorded using
the non-differentially-corrected Global Positioning System (GPS). Apart from the
GPS measurements, we collected some additional measurements to support
positioning field plots in the image. These measurements especially refer to distances
to closest roads, water features and other features that can easily be identified in the
image.
5. 2.2.1. Remotely-Sensed Data and Predictor Variables
An overlapping pair of WV2 satellite images acquired on 26 July 2013 was
used as the remotely-sensed data for this study. The main reason for having an
overlapping pair of images was to prepare a stereo model of the area to manually
extract individual tree crowns. The satellite image capture coincided with the field
data acquisition. The satellite is in a nearly circular, Sun-synchronous orbit with a
period of 100.2 min. The spatial resolution of multispectral bands was 2.0 m, and the
panchromatic band was 0.5 m. Table 1 shows the detailed spectral band description
of WV2 images.
189
Table 5. 1 Spectral band information of WorldView-2 images
Band Spectral Range (nm) Spatial Resolution
(m)
Panchromatic 447–808 0.5
Coastal 396–458
2
Blue 442–515
Green 506–586
Yellow 584–632
Red 624–694
Red-edge 699–749
NIR1 765–901
NIR2 856–1043
WV2 images were radiometrically corrected according to the method
described by Heenkenda et al. [21]. First, digital numbers were converted to
at-sensor-radiance values, and then, they were converted to top-of-atmosphere
reflectance values. The dark pixel subtraction technique in ENVI 5.0 software was
applied to remove the additive path radiance. Finally, images were geo-referenced
using rational polynomial coefficients provided with the images and ground control
points extracted from digital topographic maps of Darwin, Australia. The
georeferenced surface reflectance values were used for further analysis.
Non-mangrove areas of images were masked as per Heenkenda et al. [21].
Class-specific rules were developed based on the contextual information from the
WV2 images, geometry and neighbourhood characteristics of objects at different
hierarchical levels to separate mangrove coverage only (see Heenkenda et al. [21] for
a detailed description).
190
Plant growth is correlated with the expansion of leaf canopy and the
increasing weights of woody elements. Hence, the AGB and LAI of mangroves can
relate to vegetation indices derived from remotely-sensed data. To take the maximum
advantage of the relatively narrow spectral bands of WV2 multispectral images, this
study selected nine vegetation indices and the green, yellow, red, red-edge, NIR1 and
NIR2 bands of the WV2 image to predict LAI and AGB over the study area
(Table 5.2). These narrowband greenness indices use spectral reflectance values in
the red, red-edge, NIR1 and NIR2 regions, and they produce a measure of the
photosynthetic characteristics of vegetation. Hence, narrowband greenness indices are
more suitable measures of the biophysical variations and vigour of green vegetation
than broadband greenness vegetation indices. Vegetation indices shown in Table 5.2
were calculated for predicting AGB and LAI from WV2 data. Then, spectral bands
of the WV2 satellite image and vegetation indices were stacked together to form one
image with multiple layers (15 layers) for further processing.
The WV2 multispectral bands were resampled to a 5-m spatial resolution
using the cubic convolution resampling method. This was done to simulate remote
sensing images from other satellite missions that provide multispectral images within
the same spectral region, such as RapidEye. The number of band ratios and
vegetation indices was calculated as explained in Table 2. The green (506 nm–586
nm), red (624–694 nm), red-edge (699 nm–749 nm), NIR1 (765 nm–901 nm) and
NIR2 (856 nm–1043 nm) bands of the WV2 image and the calculated band ratios
and indices were stacked together to form a single image with 15 bands of a 5-m
spatial resolution for further processing.
191
Table 5. 2. Predictor variables generated from the WorldView-2 multispectral image
for estimating above ground biomass and leaf area index. NIR1, NIR2, red edge, red
and green: surface reflectance of Near Infrared 1, Near Infrared 2, red-edge, red and
green wavelength regions, respectively.
Vegetation Index Band Relationship Source
Normalized Difference
Vegetation Index
(NDVI)
(NIR1 − red) (NIR1 + red)⁄ Rouse et al. [24] and
Ahamed et al. [10]
Normalized Difference
Red Edge index
(NDRE)
(NIR1 − Red Edge) (NIR1 + Red ⁄ Ahamed et al. [10] and
Barnes et al. [25]
Green Normalized
Difference Vegetation
Index (GNDVI)
(NIR1 − green) (NIR1 + green)⁄
Ahamed et al. [10], Li et
al. [26] and Gitelson et al.
[27]
Green Normalized
Difference Vegetation
Index 2 (GNDVI2)
(NIR2 − green) (NIR2 + green)⁄ Mutanga et al. [28]
Normalized Difference
Vegetation Index 2
(NDVI2)
(NIR2 − red) (NIR2 + red)⁄ Mutanga et al. [28]
Normalized Difference
Red Edge index 2
(NDRE2)
(NIR2 − Red Edge) (NIR2 + Red ⁄ Mutanga et al. [28]
Renormalized
Vegetation Index
(RDVI)
(NIR1 − red) √NIR1 + red⁄ Li et al. [26]
Ratio Vegetation Index
(RVI) NIR1 red⁄ Li et al. [26]
Modified Soil Adjusted
Vegetation Index
(MSAVI)
(1 + 0.5)(NIR1 − red) (NIR1 + r⁄ Qi et al. [29]
192
An overlapping panchromatic image pair was oriented to ground coordinates
following the digital photogrammetric image orientation steps in the Leica
Photogrammetric Suite (LPS)/ERDAS IMAGINE software to obtain a stereo model
of the area. The rational polynomial coefficients (RPC’s) calculated during the image
acquisition and ground control points extracted from the digital topographic map of
Darwin, Australia (scale of 1: 10,000), were used as references. Further, the quality
of the stereo model was assessed by comparison with ground control points. Then,
outlines of the mangrove field plots that were assessed in the field were digitised
using a stereo model. To identify the four corners of field plots in the image, the GPS
locations, as well as the indirect measurements that were collected during field
sampling were used. To calculate the corresponding predictor variable values of each
plot, all pixels within a field plot polygon were considered. However, pixels that
represented less than 70% in extent within a plot were discarded from
further calculation.
The study selected the digital cover photography method developed by
Macfarlane et al. [30] for calculating the LAI of field sampled mangrove trees. Since
mangroves are densely clustered with muddy soil underneath the canopy cover, the
digital cover photography method is more reliable than plant canopy analysers or
area meters (LAI-2200C, LI-3100C, LI-COR 6400, etc.). Vertical photographs of the
canopy, pointing the lens of the camera upwards, were taken using a Panasonic
Lumix DMC-FT2 compact digital camera. Two levelling bubbles were attached to
the camera to ensure that photographs were taken without a tilt. The camera was set
to an aperture priority-automatic exposure mode. Due to the densely-clustered nature
of mangroves, it is impossible to isolate one tree and take photographs that represent
193
its canopy area. Therefore, 16 evenly-spaced photographs were taken within each
plot. The number of trees within the plot was counted.
5.2.3.1 Estimating Leaf Area Index
According to Macfarlane et al. [30], vertical photographs were used to
estimate the large gaps between mangrove trees, small gaps within mangrove trees,
the proportion of the ground area covered by the vertical projection of foliage and
branches, the crown cover, the crown porosity and the woody-to-total area ratio of
each sampling plot. A detailed description of the method to estimate LAI using
digital photographs can be found in Macfarlane et al. [30], Pekin and Macfarlane
[31] and Heenkenda et al. [32].
All digital photographs were classified using eCognition software for further
calculations. By considering the contrast between vegetation and the background (the
sky) of the photographs, areas with higher blue reflectance values were classified as
sky. The ratio between the reflectance of blue and red bands was further considered
to separate the background and the vegetation. The background was further classified
into large gaps (gL) between tree crowns and total gaps (gT), considering the relation
to neighbouring features. For example, if the area of the gap is less than 50 pixels
and it is surrounded by vegetation, it is classified as a small gap within a tree.
Finally, a total number of pixels for large gaps and small gaps was summed to obtain
gT (see Heenkenda et al. [32] for the detailed description).
194
Vegetation was further classified into leaf area and branches/stems. The
greenness values were used to identify branches/stems from vegetation. Hence, the
greenness was calculated from photographs using Equation (1) [33].
𝐺𝐺𝑟𝑟𝑟𝑟𝑟𝑟𝑔𝑔𝑔𝑔𝑟𝑟𝐵𝐵𝐵𝐵 = 2 ∗ 𝐺𝐺 − 𝐵𝐵 − 𝑅𝑅 (1)
Where G, B, and R represent the reflectance (intensity levels) recorded with the
green, blue and red bands of the digital camera.
The fraction of foliage cover (ff) that is the proportion of the ground area
covered by the vertical projection of foliage and branches [30, 31, 34], and the crown
cover (fc) were calculated as Eqs. (2) and (3).
𝑓𝑓𝑐𝑐 = 1 − 𝑙𝑙𝐿𝐿 ∑𝑃𝑃𝐵𝐵𝑃𝑃𝑟𝑟𝑙𝑙𝐵𝐵⁄ (2)
𝑓𝑓𝑓𝑓 = 1 − 𝑙𝑙𝑇𝑇 ∑𝑃𝑃𝐵𝐵𝑃𝑃𝑟𝑟𝑙𝑙𝐵𝐵⁄ (3)
Where gL is the total number of pixels of large gaps (gaps between tree crowns) and
gT is the total number of pixels of gaps, fc is the crown cover, ff is the fraction of the
foliage cover, and ∑𝑃𝑃𝐵𝐵𝑃𝑃𝑟𝑟𝑙𝑙𝐵𝐵 is the total number of pixels of the photograph.
Considering these results, the crown porosity (ф) which is the proportion of
the ground area covered by the vertical projection of foliage and branches within the
perimeter of the crowns of individual plants was calculated from the Eq. (4)
ɸ = 1 − 𝑓𝑓𝑓𝑓𝑓𝑓𝑐𝑐
(4)
Where ff is the fraction of the foliage coverage and fc is the crown cover.
195
The effective plant area index (Lt) includes the contribution from woody
elements to the total plant cover. Hence, it provides an overestimation for leaf area
index [29]. The effective plant area index was estimated using the modified version
of the Beer-Lambert law as specified in Eq. (5).
𝐿𝐿𝑡𝑡 = −𝑓𝑓𝑐𝑐 ∗ln (ɸ)𝑘𝑘
(5)
Where fc is the crown cover; ɸ is the crown porosity; and k is the canopy extinction
coefficient.
As Perera et al. [35] suggested, the canopy extinction coefficient (k) was
taken as 0.5 for this study. This represents the average value of already published k
values for mangroves around the world. The factor k mainly depends on stand
structure and canopy architecture; therefore, different vegetation types have different
values. For example, Breda [36] reviewed k values for forest stands with broad
leaves and found a range of values from 0.42 to 0.58. Macfarlane et al. [30] assumed
k =0.5 for eucalypt forests. Finally, we calculated the woody-to-total area ratio (α)
and the actual leaf area index (LAI) from Equations (6) and (7) [37,38].
𝛼𝛼 = ∑(𝐴𝐴𝑟𝑟𝑟𝑟𝐴𝐴𝐵𝐵)∑(𝐴𝐴𝑟𝑟𝑟𝑟𝐴𝐴𝑉𝑉) (6)
𝐿𝐿𝐿𝐿𝑁𝑁 = 𝐿𝐿𝑡𝑡 ∗ (1 − 𝛼𝛼) (7)
Where ∑(𝐿𝐿𝑟𝑟𝑟𝑟𝐵𝐵𝐵𝐵) is the total area of branches except leaves; ∑(𝐿𝐿𝑟𝑟𝑟𝑟𝐵𝐵𝑉𝑉) is the total
area of vegetation including branches; and Lt is the effective plant area index.
196
The final results represented an actual LAI of individual plots or a cluster of
trees. This was then divided by the number of trees to get a value per tree.
5.2.3.2 Estimating Above Ground Biomass
The study selected the allometric method for estimating AGB from field
measurements. The basic theory of the allometric relationship is that the growth rate
of one part of the organism is proportional to that of another [5], and therefore, it
depends on measurable canopy dynamics, such as tree height and DBH. Once the
regression relationship between canopy dynamics is established, the regression
equation estimates the standing biomass of a tree. However, allometric relationships
for mangroves are species and site specific [5]. Therefore, special attention was
given to use allometric equations developed for mangrove species in Northern
Australia or Southeast Asia. Since there was no equation derived for Sonneratia alba
and Excoecaria agallocha var. ovalis, the common equation developed by Bai [39]
for mangroves in the Northern Territory, Australia, was used (Table 5.3).
The logarithmic transformation of DBH values was calculated using
Microsoft Excel software. Equations from Table 5.3 were used to calculate Log10
(Biomass). Finally, the calculated above ground biomass values were converted to
AGB per square meter (unit = kg/m2) considering the extent of each plot.
197
Table 5. 3. Log10-transformed allometric relationships used for different mangrove
species. The equations are in the form: 𝑙𝑙𝑙𝑙𝑙𝑙10(𝐵𝐵𝐵𝐵𝑙𝑙𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵) = 𝐵𝐵0 + 𝐵𝐵1 ∗ 𝑙𝑙𝑙𝑙𝑙𝑙10(𝐷𝐷𝐵𝐵𝐷𝐷);
where DBH is the diameter at breast height; B0 and B1 are regression coefficients.
These equations are specific to the Northern Australia [37,38], North-eastern
Queensland, Australia [39] and Sri Lanka [19] (biomass in kg and DBH in cm).
Mangrove Species B0 B1 Study
Avicennia marina −0.511 2.113 Comley and McGuinness [40]
Bruguiera exaristata −0.643 2.141 Comley and McGuinness [40]
Ceriops tagal −0.7247 2.3379 Clough and Scott [41]
Lumnitzera racemosa 1.788 2.529 Perera and Amarasinghe [20]
Rhizophora stylosa −0.696 2.465 Comley and McGuinness [40]
Sonneratia alba −0.634 2.248 Bai [39]
Excoecaria agallocha var.
ovalis −0.634 2.248 Bai [39]
5.2.3 Predicting LAI and AGB
Multivariate regression methods are some of the most-widely used methods
for estimating plant biophysical and biochemical variables, in particular for
developing empirical models of variables of interest based on satellite image data.
Among them, partial least squares regression (PLSR) receives much popularity, as it
has been designed to outperform the problems of collinearity of prediction variables
and over-fitting with relatively few samples. Hence, PLSR is suitable when the
number of predictor variables is large and they are highly collinear [26,42,43]. This
198
study selected the PLSR algorithm to model the relationship between field samples
and satellite data. The “pls” package in R software [44] was used for the calculation.
In its simplest form, PLSR specifies a regression model by projecting the
predictor variables and response variables to a new space. That is, data are first
transformed into a different and non-orthogonal basis similar to a principal
component analysis, and the most important partial least square components are
considered for building a regression model. Hence, the PLSR procedure extracts
successive linear combinations of partial least square factors or components. Once
the optimal number of partial least square components is selected, the relationship
can be predicted over a large area. A detailed description of the PLSR algorithm is
available in Mevik and Wehrens [44], Hastie et al. [45] and Wang et al. [43].
The PLSR algorithm performs best with normalized data without outliers
[45]. Therefore, the LAI sample data were normalized with the mean equal to zero
and the variance equal to one and analysed for outliers. The mangrove field plot
polygons were considered to extract predictor variable values from all pixels within
plots, as described previously for a 2-m spatial resolution. Finally, these sampled
pixels were randomly divided into two sets: training (70% of sample data) and
validation (30% of sample data). By using the “pls” package, the optimal number of
components for the prediction or predictive abilities of the model was assessed
considering the prediction root mean square error (RMSEP). To obtain RMSEP, the
model was internally cross-validated using the leave-one-out cross-validation method
with training data. Once the optimal number of components was selected, the model
was used to predict LAI over the study area using the “raster” package in R software
[46]. Finally, we de-normalized the raster maps considering the mean and variance of
199
the original field samples to obtain real LAI values. A low pass filter (3 × 3 kernel)
was applied to smooth data by reducing local variation and removing noise. This
process was repeated to analyse the spatial variation of LAI over the Rapid Creek
mangrove forest with predictor variables of a 5-m spatial resolution.
Field samples of log-transformed AGB values (allometric equations directly
calculated log-transformed AGB) were analysed for the normality and outliers. Then,
the above process was repeated to predict AGB over the study area. The sampling
plot polygons were considered to extract predictor variable values from all pixels
within plots.
5.2.4 Accuracy assessment
The accuracies of models were internally assessed using a leave-one-out
cross-validation method with training samples at all instances. However, the
accuracy of the predicted LAI and AGB maps was assessed using validation data.
Approximately one third of field samples of LAI and AGB separately was
considered as the validation data (randomly selecting 30% of the sampling pixels
within field plots). Root mean squared errors (RMSE’s) and correlation coefficients
(r) between predicted values and field measurements were recorded.
5.3 Results
The locations of field plots are shown in Figure 5.1. Three field plots at the
northeast corner of the study area have low biomass values. Their DBH values vary
from 1.5 cm to 3.5 cm, and tree heights vary from 1.4 m to 3.6 m. Figure 5.2A–C
200
shows mangrove trees in these areas. They are newly-regenerated, small trees. Some
areas are covered with large, wide-spread, multi-stemmed mangrove trees
(Figure 5.2D,E). Their DBH values vary from 10.0 cm to 19.0 cm, and heights range
from 4.0 m to 8.0 m. There are some areas especially near water features covered by
densely-clustered mangroves with different heights (Figure 5.2F,G).
Figure 5. 2. Mangrove trees inside the Rapid Creek forest; (A–C) small,
newly-regenerated mangrove trees; (D,E) large, multi-stemmed mangrove trees;
(F,G) densely-clustered areas.
(A)
(D) (C)
(B)
(G)
(F) (E)
201
The normal score plot of field samples is shown in Figure 5.3. The normal
score represents alternative values to data points within a dataset that would be
expected from a normal distribution. Points on the reference line are closer to
normality, and horizontal departures indicate departures from normality. Hence, few
sampled mangrove trees were removed from further calculations, as they exhibited
non-normality with extremely low or high values (Figure 5.3A,B). The normality
was measured with respect to the coefficient of determination of regression (R2). The
R2 of LAI equals 0.96, and the log-transformed AGB showed 0.9. Once outliers were
removed, the remaining field samples were normally distributed.
Figure 5.3. Normal score plots of field estimates of (A) Leaf area index; and (B) Log
transformed above ground biomass.
(B
(A
202
5.3.1 Predicting Leaf Area Index and Above Ground Biomass
A number of studies indicated PLSR as a powerful tool to extract spectral
signatures and to create reliable models of LAI [25]. The performance of the PLSR
model: a “goodness of fit” is represented by a root mean squared error of prediction
(RMSEP). The RMSEP of the PLSR for LAI prediction was 0.69 with four
components when considering a 2-m spatial resolution. This value confirmed a good
relation between predictor variables and LAI field samples. Therefore, four
components with all predictor variables were used for final mapping. The RMSEP of
a 5-m spatial resolution was 0.78 with six components. Six components with all
predictor variables were used for final mapping. If the prediction model is perfect,
the RMSEP and RMSE (the root mean squared error for training and validation data)
should be very similar. However, the RMSE values of LAI map with a 2-m and a
5-m spatial resolution were 0.75 and 0.78, respectively (Table 5.4). Although the
slight difference with respect to a 2-m spatial resolution shows model overfitting, the
model performance for 5-m spatial resolution data was good.
Figure 5.4 A–D shows the cross-validated predictions with selected
components versus measured values (or field samples). Most of the points
(normalized LAI and AGB values) slightly deviate from the aspect ratio = 1 line.
However, there is no indication of a curvature or other anomalies. The validation
results (RMSEP) are shown in Figure 5.4.
203
Figure 5.4. Cross-validated predictions for the normalized leaf area index (LAI) and
above ground biomass (AGB): (A) predicted versus field measured normalized LAI
with 2-m spatial resolution predictor variables; (B) predicted versus field measured
normalized LAI with 5-m spatial resolution predictor variables; (C) predicted versus
field measured normalized LogAGB with 2-m spatial resolution predictor variables;
(D) predicted versus field measured normalized LogAGB with 5-m spatial resolution
predictor variables. (RMSEP, root mean square error of prediction).
The predicted LAI map with a 2-m spatial resolution is shown in Figure 5.5A.
The highest LAI value is 13.0, and the lowest one is 0.2. The mean LAI value is 3.9
204
with a standard deviation of 1.1. Approximately 70% of data ranged from 0.8 to 4.5.
Although LAI values are normally distributed, the pattern has a high dispersion. The
majority of data values are between 0.8 and 4.5. LAI values are high where closer to
edges of water features. The upper right side of the map (Figure 5.5A) shows low
LAI values, and this area is dominated by relatively small recently re-generated
mangrove plants.
The predicted LAI map with a 5-m spatial resolution is shown in Figure 5.5B.
The highest LAI value is 13.3, and the lowest one is 1.1. The mean LAI value is 4.2
with a standard deviation of 0.8. More than 70% of LAI values are between 3.2 and
4.8 showing a normal distribution. However, LAI values are high where closer to
edges of water features. The upper right side of the map (Figure 5.5B) shows low
LAI values, as shown in the map with a 5-m spatial resolution (Figure 5.5B).
When comparing Figure 5.5A,B at this scale, it can be seen that results
obtained from 2-m spatial resolution images provide finer LAI spatial variations than
results from 5-m spatial resolution images. However, at a larger scale, Figure 5.5A
shows scattered LAI variation patterns. These patterns do not correctly represent the
spatial variations of mangrove trees around those areas and, thus, the spatial
distribution of LAI of the area.
The visual appearances of the AGB maps are in-line with the field
observations. Most of the areas having low AGB values are dominated by relatively
small and young mangrove trees. For instance, three field plots at the northeast
corner of the study area (Figure 5.1) have low biomass values. The average measured
DBH in this area was 1.8 cm, and the average height of mangrove trees was 1.6 m.
205
Mangrove trees are dense and tall along the water features, and these areas showed
high AGB values.
When considering the performance of the PLSR model for AGB, the RMSEP
for a 2-m spatial resolution was 0.89 kg/m2; however, once predicted AGB over the
study area, the RMSE was 2.2 kg/m2. The correlation coefficient obtained with
respect to validation samples was 0.4 (Table 5.4). Hence, it can be concluded that the
model was over fitted. The RMSEP for a 5-m spatial resolution was 1.7 kg/m2, and
the RMSE with respect to validation data was 2.0 kg/m2 with strong linear
correlation between predicted and sampled AGB values.
The predicted AGB values with a 2-m spatial resolution ranged from 0.12
kg/m2 to 425.5 kg/m2 with a mean value of 22.5 kg/m2 and a standard deviation of
8.1 kg/m2 (Figure 5.5C). AGB values with a 5-m spatial resolution ranged from 0.12
kg/m2 to 423.2 kg/m2 with mean value of 18.4 kg/m2 and a standard deviation of 7.2
kg/m2 (Figure 5.5D). At both instances, the AGB values showed a skewed
distribution rather than a normal distribution. The majority of the data (more than
70%) were in between 30 kg/m2 and 267 kg/m2. The main reason should be the
extremely large, multi-stemmed Avicenna marina and Rhizophora stylosa mangrove
trees in this forest. They were spread over a large area without secondary forest
underneath. The highest measured, as well as highest predicted AGB values
represent these areas (423.2 kg/m2 and 425.5 kg/m2)
206
Figure 5.5. Predicted maps: (A) leaf area index with a 2-m spatial resolution; (B) leaf
area index with a 5-m spatial resolution; (C) above ground biomass with a 2-m
spatial resolution; and (D) above ground biomass with a 5-m spatial resolution; using
the partial least squares regression algorithm. The distribution of field sampling plots
is shown in the maps (the sizes of the black squares are not
to scale).
207
5.3.2 Accuracy assessment
The RMSE values with respect to the validation samples and correlation
coefficients between predicted values and validation samples were recorded
(Table 5.4). We used randomly-selected, approximately one third of field samples as
the validation samples for both instances. Although the AGB map with a 2-m
resolution showed low linear correlation between predicted values with validation
samples, the LAI map showed a good correlation. AGB and LAI maps with a 5-m
spatial resolution showed a strong linear correlation between predicted values and
validation samples.
.Table 5. 4. Root mean square errors (RMSE’s) and correlation coefficients (r) for
above ground biomass and leaf area index maps with respect to validation samples
Biophysical Variable RMSE Correlation Coefficient
Spatial resolution 2 m 5 m 2 m 5 m
Above ground biomass
(AGB) 2.2 kg/m2 2.0 kg/m2 0.4 0.8
Leaf area index (LAI) 0.75 0.78 0.7 0.8
5.4 Discussion
The major advantage of assessing AGB and LAI from remotely-sensed data is
the estimation of AGB and LAI over large areas without having extensive field
campaigns. This is also a solution for many logistical and practical problems arising
with field efforts. For example, access to interior extremely-dense mangrove patches
208
is extremely difficult. In addition, the methods are non-destructive, relatively fast and
economical with less labour force.
5.4.1 Predicting LAI
Mangrove canopies are densely clustered with overlaps. Extended root
systems and muddy soil underneath the canopy cover make it difficult for sampling.
Although the widely-accepted method for LAI estimation is the simulation of canopy
light profiles using conventional instruments, such as portable plant canopy analysers
or area meters, the digital cover photography method is more reliable than
conventional instruments for mangrove forests. The digital cover photography
approach is independent from the radiation condition of the forest and can be used
for an extensive sampling. This method was later tested for types of cameras, digital
file compression, image size and ISO equivalences, and little or no effect on
estimating LAI was found [31]. Lui and Pattey [33] also recommended using the
digital cover photography method by comparing its results with LAI estimation from
conventional equipment.
The mean value of the predicted LAI map with a 2-m spatial resolution (3.9)
is slightly lower than previously-recorded mangrove studies around the world. For
instance, Clough et al. [17] estimated the LAI of Rhizophora apiculata mangrove
forest in Malaysia using three different methods and obtained mean LAI values: 4.9,
4.4 and 5.1. LAI of Avicennia marina plantations in Thailand varied from 0.5 to 5.0.
For the homogeneous mangrove stands at Moreton Bay, Australia, LAI values
ranged from 0.26 to 3.23 with 1.97 as the mean value [15]. Kamal et al. [15] also
assessed heterogeneous mangrove stands at Karimunjawa Island, Indonesia, and
209
obtained LAI values from 0.88–5.33 (mean value = 2.98). A remote sensing study on
LAI in the British West Indies showed a range from 0.8–7.0 with 3.96 as the mean.
Statistics from the predicted LAI map of this study with a 5-m spatial resolution (4.2)
was as close as previous mangrove studies. However, comparing results from this
study with other studies is not a perfect approach to make a conclusion regarding the
spatial distribution of LAI. They are site and species specific. When looking at cross-
validated predictions, most of the predicted LAI values are lower than field sample
values (Figure 5.4A,B). One consideration is the canopy extinction coefficient (k)
used to calculate Lt. Although the used value (0.5) represents the average of already
published k values for mangroves around the world, this might not be the correct
value for this area. Further, LAI values vary with the spatial resolutions of remote
sensing images used for data processing. Kamal et al. [15] recently confirmed that
the accuracy of LAI estimation was site specific and depends on the pixel size of the
remotely-sensed images, and their study even indicated two different LAI
distribution patterns for homogeneous and heterogeneous mangrove forests.
The performance of the PLSR model: a “goodness of fit” is represented by a
root mean squared error of prediction (RMSEP). The accuracy of the predicted maps
was assessed using the root mean squared error (RMSE) compared to the validation
samples. If the prediction model is perfect, RMSEP and RMSE (the root mean
squared error for training and validation data) should be very similar. However, in
this study, values related to a 2-m spatial resolution are slightly different, indicating a
model overfitting. The best model performances were shown when using a 5-m
spatial resolution. According to Zheng and Moskal [47], the accuracy of LAI
estimation depends on two main reasons: overlapping and clumping between leaves
within canopies due to the non-random distribution of foliage and light obstruction
210
from canopy branches, trunks and stems. The first reason relates to this study, as
well, due to mangrove canopy overlapping. The latter will not be a problem because
of using vertical photographs rather than simulating light transmission. Another
reason for obtaining low accuracies with a 2-m spatial resolution than a 5-m spatial
resolution would be the spatial mismatch between field samples and predictor
variables due to positional errors.
To establish the relationship between field samples and predictor variables, a
corresponding predictor pixel value for each field sample should correctly be
extracted. Hence, the positional accuracy of field samples and the pixel size is
important. Laongmanee et al. [48] assessed the optimal spatial resolution for
estimating the LAI of Avicennia marina plantations in Thailand. The best results
were produced by satellite images with a 10-m spatial resolution. Although there is
still room to confirm this finding, Green and Clark [18] suggested analysing the
information of satellite data within a 5 m × 5 m block of pixels regardless of the
spatial resolution for LAI estimation. Green and Clark [18] also argued that although
high resolution satellite data provide significantly greater levels of accuracy, they
were not capable of fixing positional errors. Hence, it is clear that the satellite data
with a 2-m spatial resolution are not suitable to establish this match-minimizing
effect from the positioning errors. According to this study, although 2-m or 5-m
spatial resolutions can interchangeably be used for mapping LAI in the Rapid Creek
mangrove forest, we would recommend testing a 5 m × 5 m block of pixels
regardless of the spatial resolution for LAI estimation, as suggested by Green and
Clark [18]. Further, we would recommend using the transect method for field
sampling to minimize positional errors. Two ends of transects can be established
211
outside the mangrove forest with DGPS (differential GPS). Then, all other
measurements can be based on the established transect line with greater accuracy.
5.4.2 Predicating AGB
We used allometric equations for AGB calculation. Allometric relationships
are highly species specific and less site specific [5,19]. Although special attention
was given to using allometric relationships that were specifically developed for
Northern Territory, Australia, mangrove species, we used some equations that were
developed for Northern Queensland, Australia, and Sri Lanka. In this study, it was
not possible to use the common equation proposed by Komiyama et al. [19], as it
requires the vegetation density, which we did not measure in the first place.
Therefore, for the mangrove species for which we could not find already developed
allometric equations, we used a generic equation developed by Bai [39] for the
Northern Territory, Australia, mangroves.
Some of the sampled Avicennia marina and Rhizophora stylosa species are
multi-stemmed mangrove trees. As explained by Clough et al. [49], when there are
multi-stemmed mangrove trees, each branch should be treated as a discrete tree, and
the dry weight of the common butt should be included to obtain a robust AGB of the
tree. However, this discrepancy is important if the consideration is to calculate the
AGB of different woody parts of trees only, rather than total AGB [49]. Hence, in
this study, we treated each branch of multi-stemmed mangrove trees as a separate
tree. The dry weight of the common butt was neglected.
212
The predicted AGB maps are approximately in-line with visual field
observations and AGB calculations. Along water features, mangroves exhibit high
AGB values. Mangroves are known to have low productivity where closer to the
landward margin [35]. These areas experience relatively less tidal inundation
frequency and shorter duration together with a minimal freshwater influence [50].
The low productivity is also connected with higher salinity and less nutrients, and
this is strongly evidenced in the Rapid Creek mangrove forest. Above ground
biomass varies with the age of the tree, temperature, solar radiation and oxygen, as
well [51]. In this study area, there is no doubt that all trees would experience the
same temperature, solar radiation and oxygen, but the age of trees, salinity and
nutrients levels are different. As some parts of the Rapid Creek mangrove forest are
regenerated forests after clearings and natural disasters, there are trees with different
ages. Especially closer to the northeast corner of the forest, trees are younger than the
rest of the forest and small, and they showed low AGB values.
When considering the performance of the PLSR model for predicting AGB
with a 2-m spatial resolution, it was found that the model was over fitted (Figure 4).
The high spectral variation and shadows caused by canopy may create difficulty in
developing PLSR model. However, the model performance for a 5-m spatial
resolution was at an acceptable level. There was a strong linear correlation between
predicted and field sample AGB values. The correlation coefficient is relatively high
compared to the very limited number of mangrove studies that are available to-date
without indicating any model overfitting. As we suspected earlier when estimating
LAI from satellite data, one reason for obtaining low accuracies with a 2-m spatial
resolution than a 5-m spatial resolution would be the spatial mismatch between field
samples and predictor variables due to positional errors. When estimating AGB from
213
satellite data, a 5-m spatial resolution would be the most suitable approach against a
2-m spatial resolution.
Although there are several studies investigating the relationship between the
physical parameters of mangrove trees and AGB and developing allometric
equations, there is a limited number of recorded studies mapping the spatial
distribution of AGB over large areas around the world. For instance, in Qi’ao Island,
Guangdong Province, China, predicted AGB of mature Kandelia candel mangroves
ranged from 15.51 kg/m2 to 40.66 kg/m2 with the average value of 24.77 kg/m2, and
artificially-restored Sonneratia apetala mangroves ranged from 3.4 kg/m2 to 23.42
kg/m2 with the average value of 11.38 kg/m2 [16]. However, in this study, we did not
find Kandelia candel and Sonneratia apetala mangroves, and thus, this limits the
comparison possibilities.
Apart from the statistical analysis of model accuracies, the authors are
confident about the LAI and AGB model performances with respect to a 5-m spatial
resolution at the northeast corner of the study area, southern areas and areas along
water features. For instance, as shown in Figure 5.2A–C, northeast areas are
dominated by regenerated small mangrove trees. Southern areas have relatively tall
and dense mangroves. There is no recorded natural or anthropogenic disasters related
to these areas, and mangrove trees are relatively mature. However, the model
performances along the western edge of the mangrove forest are confusing. These
edges are dominated by Lumnitzera racemosa mangrove trees and are tall and mature
with relatively small DBH values (around 5 cm). On the other hand, these trees
receive less amounts of water, thus nutrients, compared to the other areas due to
ground height variations.
214
The canopy height model of the vegetation has a direct and increasingly well-
understood relationship to above ground biomass, and therefore, one of the best
predictors for biomass would be a canopy height model [8]. Simard et al. [12] also
used elevation data derived from the Shuttle Radar Topography Mission (SRTM) for
estimating mangrove heights and above ground biomass. They used airborne laser
scanning data and extensive field samples to calibrate these elevation data. We
created an accurate digital surface model of the area using a pair of WV2
panchromatic images. However, due to the lack of elevation data with reliable
accuracy, we were not able to create a canopy height model of the area. Hence,
although we identified the possibility of integrating the canopy height model as a
predictor variable, we did not use it.
In this study, we did not consider very small gaps (less than 2 m) between
mangrove trees due to processing difficulties with the spatial resolutions with which
we dealt. Rather than masking out large non-mangrove areas from satellite images,
we would suggest identifying gaps between mangrove trees and excluding them from
further analysis. This would remove the noise of predicted images. Additionally, the
complexity of species composition, stand structures and the densely-clustered nature
of mangrove forests might add some errors, forming mixed pixels of remotely-sensed
data.
5.5 Conclusions and recommendations
One of the key issues associated with effective planning and management of
mangrove forests is up-to-date information about the ecosystem. To collect
information, extensive field sampling is difficult and time consuming. Compared to
215
other terrestrial ecosystems, a few studies that are associated with field samples and
remotely-sensed data for producing thematic maps over mangrove forests can be
found. This study investigated mapping above ground biomass (AGB) and leaf area
index (LAI) from WorldView-2 satellite images and field samples. Site- and species-
specific allometric relationships were used for calculating above ground biomass for
sampled trees. The leaf area index was obtained from the digital cover photography
method. Hence, the field sampling process was fast and economical.
The relationships between both biophysical variables, LAI and AGB, with
predictor variables (2-m spatial resolution) were established separately using the
partial least squares regression algorithm. Once these relationships were established,
they were used to predict response variables over the study area. The accuracies of
predicted maps were analysed compared to validation samples. The process was
repeated for the predictor variables with a 5-m spatial resolution. The LAI map with
a 2-m spatial resolution showed a root mean square error of 0.75, and the map with a
5-m spatial resolution showed a root mean square error of 0.78 compared to the
validation samples. The correlation coefficients between field samples and predicted
maps were 0.7 and 0.8, respectively. Root mean square errors obtained for AGB
maps were 2.2 kg/m2 and 2.0 kg/m2 for a 2-m and a 5-m spatial resolution, and the
correlation coefficients were 0.4 and 0.8, respectively.
The satellite data with higher spatial resolution should be associated with
accurately-positioned field samples. Therefore, we would suggest implementing the
transects method for field sampling and establishing end points of these transects
outside the mangrove forest with a highly accurate positioning system, such as a
216
differential GPS. Further, we would recommend analysing a 5 m × 5 m block of
pixels regardless of the spatial resolution for LAI and AGB estimation.
In conclusion, the study demonstrated the possibility of assessing the
biophysical variations of mangroves using WorldView-2 satellite data. This would
lead to better mangrove conservation and management of the Rapid Creek area.
Acknowledgments: The authors thank Miguel Tovar Valencia, Evi Warintan
Saragih, Silvia Gabrina Tonyes, Olukemi Ronke Alaba and Dinesh Gunawardena for
their assistance in the field surveying.
Author Contributions: Muditha K. Heenkenda designed the research, processed the
remote sensing data, and drafted the manuscript with co-authors providing
supervision and mentorship throughout the process.
Conflicts of Interest: The authors declare no conflict of interest.
217
5.6 References
1. Food and Agriculture Organisation of the United Nations (FAO). The World’s
Mangroves 1980–2005; Food and Agriculture Organisation of the United
Nations: Rome, Italy, 2007.
2. Bouillon, S.; Rivera-Monroy, V.H.; Twilley, R.R.; Kairo, J.G. Mangroves. In
The Management of Natural Coastal Carbon Sinks; Laffoley, D., Grimsditch,
G., Eds.; International Union for Conservation of Nature and Natural Resources
(IUCN): Gland, Switzerland, 2009.
3 Wilkie, M.L.; Fortuna, S.; Forestry Department; FAO. Status and Trends in
Mangrove Area Extent Worldwide; FAO: Rome, Italy, 2003. Available online:
www.fao.org/docrep/007/j1533e/J1533E02.htm (accessed on 17.08.2016).
4. Fu, W.; Wu, Y. Estimation of aboveground biomass of different mangrove trees
based on canopy diameter and tree height. Procedia Environ. Sci. 2011, 10,
2189–2194.
5. Komiyama, A.; Ong, J.E.; Poungparn, S. Allometry, biomass, and productivity
of mangrove forests: A review. Aquat. Bot. 2008, 89, 128–137.
6. Metcalfe, K.; Franklin, D.C.; McGuinness, K.A. Mangrove litter fall:
Extrapolation from traps to a large tropical macrotidal harbour. Estuar. Coast.
Shelf Sci. 2011, 95, 245–252.
7. Anaya, J.A.; Chuvieco, E.; Palacios-Orueta, A. Aboveground biomass
assessment in Colombia: A remote sensing approach. For. Ecol. Manag. 2009,
257, 1237–1246.
8. Roy, P.S.; Ravan, S.A. Biomass estimation using satellite remote sensing data—
An investigation on possible approaches for natural forest. J. Biosci. 1996, 21,
535–561.
218
9. Eckert, S. Improved forest biomass and carbon estimations using texture
measures from WorldView-2 satellite data. Remote Sens. 2012, 4, 810–829.
10. Ahamed, T.; Tian, L.; Zhang, Y.; Ting, K.C. A review of remote sensing
methods for biomass feedstock production. Biomass Bioenergy 2011, 35, 2455–
2469.
11. Satyanarayana, B.; Mohamad, K.A.; Idris, I.F.; Husain, M.-L.; Dahdouh-
Guebas, F. Assessment of mangrove vegetation based on remote sensing and
ground-truth measurements at Tumpat, Kelantan delta, east coast of Peninsular
Malaysia. Int. J. Remote Sens. 2011, 32, 1635–1650.
12. Simard, M.; Rivera-Monroy, V.H.; Mancera-Pineda, J.E.; Castañeda-Moya, E.;
Twilley, R.R. A systematic method for 3d mapping of mangrove forests based
on shuttle radar topography mission elevation data, ICEsat/GLAS waveforms
and field data: Application to Ciénaga Grande de Santa Marta, Colombia.
Remote Sens. Environ. 2008, 112, 2131–2144.
13. Mitchard, E.T.A.; Saatchi, S.S.; White, L.J.T.; Abernethy, K.A.; Jeffery, K.J.;
Lewis, S.L.; Collins, M.; Lefsky, M.A.; Leal, M.E.; Woodhouse, I.H.; et al.
Mapping tropical forest biomass with radar and spaceborne LiDAR in Lope
National Park, Gabon: Overcoming problems of high biomass and persistent
cloud. Biogeosciences 2012, 9, 179–191.
14. Green, E.P.; Mumby, P.J.; Edwards, A.J.; Clark, C.D.; Ellis, A.C. Estimating
leaf area index of mangrove from satellite data. Aquat. Bot. 1997, 58, 11–19.
15. Kamal, M.; Phinn, S.; Johansen, K. Assessment of multi-resolution image data
for mangrove leaf area index mapping. Remote Sens. Environ. 2016, 176, 242–
254.
219
16. Zhu, Y.; Liu, K.; Liu, L.; Wang, S.; Liu, H. Retrieval of mangrove aboveground
biomass at the individual species level with WorldView-2 images. Remote Sens.
2015, 7, 12192–12214.
17. Clough, B.F.; Ong, J.E.; Gong, W.K. Estimating leaf area index and
photosynthetic production in canopies of the mangrove Rhizophora apiculata.
Mar. Ecol. Prog. Ser. 1997, 159, 285–292.
18. Green, E.; Clark, C. Assessing mangrove leaf area index and canopy closure. In
Remote Sensing Handbook for Tropical Coastal Management (Extracts);
Edwards, A.J., Ed.; UNESCO: Paris, France, 2000.
19. Komiyama, A.; Poungparn, S.; Kato, S. Common allometric equations for
estimating the tree weight of mangroves. J. Trop. Ecol. 2005, 21, 471–477.
20. Perera, K.A.R.S.; Amarasinghe, M.D. Carbon partitioning and allometric
relashionships between stem diameter and total organic carbon (TOC) in plant
components of Bruguiera gymnorrhiza (L.) Lamk. And Lumnitzera racemosa
willd. in a Microtidal Basin Estuary in Sri Lanka. Int. J. Mar. Sci. 2013, 3, 72–
78.
21. Heenkenda, M.K.; Joyce, K.E.; Maier, S.W.; Bartolo, R. Mangrove species
identification: Comparing WorldView-2 with aerial photographs. Remote Sens.
2014, 6, 6064–6088.
22. Duke, N.C. Australia’s Mangroves—The Authoritative Guide to Australia’s
Mangrove Plants; University of Queensland: Brisbane, Australia, 2006; p. 200.
23. Wightman, G. Mangrove Plant Identikit for North Australia’s Top End;
Greening Australia: Darwin, Australia, 2006; p. 64.
220
24. Rouse, J.W., Jr.; Haas, R.H.; Schell, J.A.; Deering, D.W. Monitoring vegetation
systems in the great plains with Erts. In Third Earth Resources Technology
Satellite -1 Symposium; NASA: Washington, DC, USA, 1974;pp. 309–317.
25. Barnes, E.M.; Clarke, T.R.; Richards, S.E.; Colaizzi, P.D.; Haberland, J.;
Kostrzewski, M.; Waller, P.; Choi, C.; Riley, E.; Thompson, T.; et al. Coincident
detection of crop water stress, nitrogen status and canopy density using ground-
based multispectral data. In Proceedings of the Fifth International Conference on
Precision Agriculture, Bloomington, MN, USA, 16–19 July 2000.
26. Li, X.; Zhang, Y.; Bao, Y.; Luo, J.; Jin, X.; Xu, X.; Song, X.; Yang, G.
Exploring the best hyperspectral features for lai estimation using partial least
squares regression. Remote Sens. 2014, 6, 6221–6241.
27. Gitelson, A.A.; Kaufman, Y.J.; Merzlyak, M.N. Use of a green channel in
remote sensing of global vegetation from EOS-MODIS. Remote Sens. Environ.
1996, 58, 289–298.
28. Mutanga, O.; Adam, E.; Cho, M.A. High density biomass estimation for wetland
vegetation using WorldView-2 imagery and random forest regression algorithm.
Int. J. Appl. Earth Obs. Geoinf. 2012, 18,
399–406.
29. Qi, J.; Chehbouni, A.; Huete, A.R.; Kerr, Y.H.; Sorooshian, S. A modified soil
adjusted vegetation index. Remote Sens. Environ. 1994, 48, 119–126.
30. Macfarlane, C.; Hoffman, M.; Eamus, D.; Kerp, N.; Higginson, S.; McMurtrie,
R.; Adams, M. Estimation of leaf area index in eucalypt forest using digital
photography. Agric. For. Meteorol. 2007, 143, 176–188.
221
31 Pekin, B.; Macfarlane, C. Measurement of crown cover and leaf area index using
digital cover photography and its application to remote sensing. Remote Sens.
2009, 1, 1298–1320.
32. Heenkenda, M.K.; Joyce, K.E.; Maier, S.W.; Bruin, S.D. Quantifying mangrove
chlorophyll from high spatial resolution imagery. ISPRS J. Photogramm. Remote
Sens. 2015, 108, 234–244.
33. Lui, J.; Pattey, E. Retrieval of leaf area index from top-of-canopy digital
photogrpahy over agricultural crops. Agric. For. Meteorol. 2010, 150, 1485–
1490.
34. Walker, J.; Tunstall, B.R. Field Estimation of Foliage cover in Australian
Woody Vegetation; CSIRO Institute of Biological Resources: Canberra,
Australia, 1981; p. 18.
35. Perera, K.A.R.S.; Amarasinghe, M.D.; Somaratna, S. Vegetation structure and
species distribution of mangroves along a soil salinity gradient in a micro tidal
estuary on the North-Werstern coast of Sri Lanka. Am. J. Mar. Sci. 2013, 1, 7–
15.
36. Breda, N.J.J. Ground-based measurements of leaf area index: A review of
methods, instruments and current controversies. J. Exp. Bot. 2003, 54, 2403–
2417.
37. Chen, J.M. Optically-based methods for measuring seasonal variation of leaf
area index in boreal conifer stands. Agric. For. Meteorol. 1996, 80, 135–163.
38. Chen, J.M.; Rich, P.M.; Gower, S.T.; Norman, J.M.; Plummer, S. Leaf area
index of boreal forests: Theory, techniques, and measurements. J. Geophys. Res.
1997, 102, 29429–29443.
222
39. Bai, L. The Colour of Mud: Blue Carbon Storage in Darwin Harbour; Charles
Darwin University: Darwin, Australia, 2012.
40. Comley, B.W.T.; McGuinness, K.A. Above-and below-ground biomass, and
allometry, of four common Northern Australian mangroves. Aust. J. Bot. 2005,
53, 431–436.
41. Clough, B.F.; Scott, K. Allometric relationships for etimating above-ground
biomass in six mangroves species. For. Ecol. Manag. 1989, 27, 117–127.
42. Carrascal, L.M.; Galvan, I.; Gordo, O. Partial least squares regression as an
alternative to current regression methods used in ecology. Oikos 2009, 118, 681–
690.
43. Wang, F.; Huang, J.; Lou, Z. A comparison of three methods for estimating leaf
area index of paddy rice from optimal hyperspectral bands. Precis. Agric. 2011,
12, 439–447.
44. Mevik, B.; Wehrens, R. The pls pckage: Principal component and partial least
squares regression in R.
J. Stat. Softw. 2007, 18, 1–24.
45. Hastie, T.; Tibshirani, R.; Friedman, J. Linear methods for regression. In The
Elements of Statistical Learning: Data Mining, Inference, and Prediction;
Springer: New York, NY, USA, 2001; p. 533.
46. Hijmans, R.J.; van Etten J. Raster: Geographic Analysis and Modeling. R
Package Version 2.2–31; 2014.
http://CRAN.R-project.org/package=raster
47. Zheng, G.; Moskal, L.M. Retrieving leaf area index (LAI) using remote sensing:
Theories, methods and sensors. Sensors 2009, 9, 2719–2745.
223
48. Laongmanee, W.; Vaiphasa, C.; Laongmanee, P. Assessment of spatial
resolution in estimating leaf area index from satellite images: A case study with
Avicennia marina plantations in Thaliland. Int. J. Geomat. 2013, 9, 69–77.
49. Clough, B.F.; Dixon, P.; Dalhaus, O. Allometric relationships for estimating
biomass in multi-stemmed mangrove trees. Aust. J. Bot. 1997, 45, 1023–1031.
50. Clough, B.F.; Andrews, T.J. Some ecophysiological aspects of primary
production by mangroves in North Queensland. Wetlands 1981, 1, 6–7.
51. Alongi, D.M. Present state and future of the world’s mangrove forests. Environ.
Conserv. 2002, 29, 331–349.
© 2016 by the authors. Submitted for possible open access
publication under the
terms and conditions of the Creative Commons Attribution (CC-BY) license
(http://creativecommons.org/licenses/by/4.0/).
224
Chapter 6 - Regional Ecological Risk assessment
using a Relative Risk Model: A case study of the
Darwin Harbor, Darwin, Australia
This work has been published in the following paper:
Heenkenda, M. K. & Bartolo, R. 2015, Regional Ecological Risk
Assessment Using a Relative Risk Model: A Case Study of the Darwin
Harbor, Darwin, Australia, Human and Ecological Risk Assessment: An
International Journal, vol. 22, 401 – 423,
doi: 10.1080/10807039.2015.1078225.
Formal permission from the publisher is attached in Appendix 4.
225
6.0 Abstract
There are many proposed and ongoing commercial, industrial and residential
developments within the Darwin Harbor catchment in Northern Australia, to
accommodate the projected population growth over the next 20 years. Hence, it is
necessary to ensure the balance between these developments and ecosystem
conservation. We evaluated ecological risk for the Darwin Harbor using a relative
risk model (RRM). The catchment was divided into 22 risk regions based on small
catchment boundaries and their homogeneity. Through the RRM, we ranked and
summed the stressors and habitats within regions. The interaction between stressors
and habitats were modelled through exposure and effect filters. The ecological
assessment endpoints were maintenance of the mangrove health and the maintenance
of water quality. The risk regions: Myrmidon Creek, Blackmore River, Bleesers
Creek and Elizabeth River showed the highest total relative risk for ecological assets.
These risk regions had a high percentage cover of industrial, commercial, residential
areas, diffuse entry points, and climate change effects. The Creek A, Sandy Creek,
West Arm and the Pioneer Creek were the risk regions with lowest total relative risk
scores. The RRM is a robust application that is suitable for a large geographic area
where multiple stressors are of concern.
Indexing terms: Ecological risk assessment, Relative Risk Model, mangrove, water
quality, Darwin Harbor
226
6. 1 Introduction
Darwin Harbor, located in the Northern Territory, Australia is a large
indented embayment with three main arms: East Arm, Middle Arm and West Arm.
The Harbor is globally significant as a macro tidal environment with a tidal range of
about 7 to 8 meters, and monsoonal climate with uniformly high temperatures and
solar radiation (Metcalfe 2007).The Darwin Harbor shoreline is covered by the most
floristically diverse, and one of the largest mangrove habitats in the Northern
Territory (Lee 2003). These mangroves comprise 36 species out of the 50 species
regarded as mangroves worldwide, and cover approximately 20, 400 hectares
(Metcalfe 2007).
There are major development projects and associated activities around
Darwin Harbor such as significant plans for population growth, industrial expansion,
resource processing infrastructure, and consequent increases in shipping and
dredging. These developments pose risks for the natural, cultural, aesthetic and
recreational values of the Harbor, and need to be managed carefully. Currently the
Darwin Harbor catchment is subject to multiple pressures and conflict of use
resulting in a major challenge for the government, industry and the community in
managing the Harbor environment (The Darwin Harbor Advisory Committee 2003).
The catchment encompasses a range of distinctive and special environments. Hence,
it is important for the Harbor and its catchment to remain biologically productive,
and for the water quality and habitats to be protected.
The Northern Territory government undertakes water quality, and ecosystem
monitoring and testing throughout Darwin Harbor region. One of the key
227
requirements for environmental monitoring programs in Darwin Harbor is the ability
to combine knowledge from different sources into a consistent environmental
management framework that enables stakeholders to make practical and sound
management decisions. Due to the current situation, the natural resources of Darwin
Harbor region are experiencing increasing pressure from the growing human
population, and subsequent rural and industrial developments. For example, although
water resources of Darwin Harbor region are considered to be in good condition,
there are areas of concern where sewage outfalls and urban stormwater can impact on
the quality of the receiving water. Further, the assimilative capacity of the Harbor
and waterways to receive pollutants from point and non-point sources, other
understanding about key ecological processes and their vulnerabilities are poorly
known (Northern Territory Government 2010). Hence, the decision makers are
interested in evaluating the likelihood that adverse ecological impacts may occur as a
result of exposure to multiple stressors in the region.
The Northern Territory government and other agencies that are responsible
for industrial developments around Darwin Harbor have initiated their own
individual environmental risk assessment programs rather than a single combined
approach. One example is the dredging program introduced by a liquefied natural
and petroleum gas project (INPEX). The key part of the project is managing
environmental impacts in order to preserve the health of Darwin harbor. Hence, its
main objectives are to monitor and respond to impact on environmentally sensitive
receptors from offshore dredging, and to understand likely impacts of sedimentation
and dredging activities (INPEX 2013). However, there is no way to combine
sedimentation effects from other industrial developments to this assessment. To our
knowledge, there is no integrated regional ecological risk assessment approach for
228
policy makers, community organisations, managers and investors to support more
robust debate, and more transparent decision making regarding rural and urban
developments for Darwin Harbor. Hence, to maintain the sustainability of the Darwin
Harbor ecosystem, a statistically sound, transparent, repeatable ecological risk
assessment approach is essential. This study is a significant contribution to the
regional ecological risk assessment of Darwin Harbor.
The traditional ecological risk assessment methods comprise of evaluating the
interaction of three components: stressor, receptor and response (Landis and Wiegers
1997). This is fairly simple at a single contaminated site with only one stressor
involved. To expand this to a regional scale, the scale, complexity of the structure,
and the regional spatial components such as sources, habitats and impacts to the
assessment endpoints must be considered (Landis and Wiegers 1997). The relative
risk model (RRM) evaluates multiple, dissimilar stressors simultaneously and
comparatively at a regional scale (Landis and Wiegers 1997; Wiegers et al. 1998).
Although this method was introduced to assess the risk of chemical stressors in the
beginning, it has successfully been introduced to analyse biological or ecological
stressors (Colnar and Landis 2007; Moraes et al. 2002; O'Brien and Wepener 2012;
Yu et al. 2015). For example, recently Bartolo et al. (2012) assessed the ecological
risk of Australia’s tropical rivers, and identified the most significant threats to the
ecological assets of the northern tropical rivers.
The objective of this study was to evaluate the relative risk to ecological
assets of the Darwin Harbor catchment. Therefore, the application is; (a) to identify
the sources and habitats in different locations of the Darwin Harbor catchment; (b) to
rank their importance in each location; (c) to integrate them to predict the relative
229
risk levels; and (d) to assess the uncertainties that may propagate through the
modelling process.
6.2 Data and method
6.2.1 Study area
The Darwin Harbor’s management area encompasses 3227 km2, and extends
from Charles Point in the west to Gunn Point in the east. This includes the estuarine
areas and tributaries of Woods Inlet, West Arm, Middle Arm, East Arm, Howard
River, Elizabeth River and the land that drains into these waterways. Hence, the
catchments include the cities of Darwin, Palmerston, Cox Peninsula, and new cities
of rural areas (Figure 6.1). The population was about 120900 in 2006, however, the
predicted population by the Australian Bureau of Statistics as of 2021 is 184500
(Skinner et al. 2009).
230
Figure 6.1. The Darwin Harbor catchment and its mangrove distribution
1 Blackmore River 7 Elizabeth River 13 Micket Creek 19 Sadgroves Creek
2 Bleesers Creek 8 Howard River 14 Myrmidon Creek 20 Sandy Creek
3 Buffalo Creek 9 Hudson Creek 15 Palmerston South 21 West Arm
4 Charles Point 10Kings Creek 16 Pioneer Creek 22 Woods Inlet
5 Creek A (Middle Arm) 11Ludmilla Creek 17 Rapid Creek
6 Darwin Central
Business District (CBD) 12Mitchell Creek 18 Reichardt Creek
231
6.2.2 Data input
The digital topographic data layers of land use and land cover (scale 1:20,
000, produced in 2002) were used as a preliminary data source for this study. The
urban and rural residential areas, and transport and communication features were
updated using Google Earth satellite image acquired on 28th March, 2014. The digital
10K topographic data series was used to extract water features, for example streams,
rivers and dams, in the region (Geoscience Australia 2002). Since rivers and streams
are line features, we introduced buffer zones of 5 m and 2.5 m for rivers and streams
respectively to form polygon features. These buffer distances were selected
considering the average width of water features. The “waterbodies” data set was
prepared combining streams, rivers, dams and other water features in the area. The
digital elevation model produced from the Shuttle Radar Topography Mission
(SRTM), with a spatial resolution of 30m, was used to analyse sub-catchment
vulnerability to sea level rise. The recent reports published by the oil company -
INPEX for the Darwin LNG (liquefied natural gas) plant include the statistics and
spatial distribution of areas that are going to be affected by increasing sedimentation
based on the onshore dredging (van Senden et al. 2013). The statistics of nutrient
loads and other suspended solids were obtained from a report published by the
Northern Territory government analysing the pollutant load entering to Darwin
Harbor due to urban and rural land-use (Skinner et al. 2009). Unfortunately, we were
not able to find point source data for the Ludmilla Creek catchment. The Darwin
Harbor sub-catchment boundaries were adopted from the Australian Collaborative
Land Use Mapping program (V5), and were used as initial data for risk regions.
232
Regional ecological risk assessment methods evaluate the interaction of three
different regional spatial components, which include: sources that release stressors;
the habitat where the receptors reside; and impacts to the assessment endpoints
(Landis and Wiegers 1997). Stressors can be derived from diverse sources, including
available topographic data, reports, and expert knowledge. Receptors are often linked
to habitats. The interaction between these stressors and habitats is explained in terms
of ranked exposure values. The impact to the assessment endpoint is explained
through the ranked effects. According to Landis and Wiegers (1997), at a regional
level, the ranking of exposures and effects must be done considering the spatial and
temporal heterogeneity of them.
The method used in this study was similar to Landis and Wiegers (1997).
First, the method identified sources and their reflecting stressors, habitats, and
ecological assessment endpoints. Then the area was divided into sub-areas to be used
as risk regions based on their homogeneity. In this study, we considered sub-
catchments and their homogeneity to define risk regions. The stressors and habitats
were ranked for each risk region. Finally, an ecological risk for each risk region was
calculated by quantitatively determining the interaction of stressors and habitats.
According to U.S. Environmental Protection Agency, there are three main
phases for ecological risk assessment: problem formulation; risk analysis; and risk
characterization (USEPA (US Environmental Protection Agency) 1998). Each phase
is divided into sub phases as shown by Figure 6.2. They are discussed in greater
detail in the proceeding sections.
233
Figure 6.2. The workflow diagram for conducting an ecological risk assessment
using the Relative Risk Model following the guidelines published by the U.S.
Environmental Protection Agency
6.2.3 Problem formulation
The problem formulation phase is a process that identifies the nature of the
problem through existing data. The foundation for the problem formation is based on
234
how well we understand the system we are investigating. We reviewed available
literature, including reports published by the Northern Territory government
analysing different aspects of health of the Darwin Harbor, the Darwin Harbor
regional management plans, and consulted experts in the field (Bartolo et al. 2012;
Brocklehurst and Edmeades 1994-1995; Darwin Harbor Regional Plan of
Management 2003; Metcalfe 2007; NRETAS 2010a; 2010b; van Senden et al. 2013;
Skinner et al. 2009) for preparing a conceptual diagram to understand the current
situation of the catchment. Varieties of anthropogenic factors have resulted in
development pressure on Darwin Harbor catchment. Based on the information
collected and the conceptual diagram, we identified the major sources of this
pressure, and categorized them into different stressors. For example, the source
“Agriculture” was categorized into the following stressors: cropping; horticulture;
and aquaculture. The corresponding ecosystem response variables for stressors were
identified accordingly. For example, one of the ecosystem response variables for
cropping is influx of nutrients and chemicals.
The habitat defines where the receptors reside at a regional scale (Landis and
Wiegers 1997). The habitats that were directly related to the health of the Darwin
Harbor, and that had spatial data available were selected for this study. Further, we
referred to the values identified by various experts in the field, the percentage of land
cover, vulnerability to urban pressure, and significance of ecosystems in the region to
decide on habitats to be considered for this study. Sources and threats were also
defined based on the prepared conceptual diagram. Special attention was given to the
availability of spatial data.
235
At this stage, we examined the integration and usage of available information
on sources, stressors and their characteristics, exposure opportunities, the ecosystem
components potentially at risk, and ecological effects. Then, a conceptual diagram
was prepared combining all information to summarise our understanding of the
Darwin Harbor ecosystem. This aided in finalising the sources, stressors and habitats
and their interactions that were included in the RRM.
6.2.3.1 Ecological Assessment endpoints
According to the USEPA (US Environmental Protection Agency) (1998),
management goals are desired characteristics of ecological values that need to be
protected. There are five key management goals identified by the Darwin Harbour
regional plan of management (The Darwin Harbour Advisory Committee 2003). The
first goal is to maintain a healthy environment. The outcomes to be achieved through
this goal are; (a) improved understanding and knowledge of the region’s
environment; (b) protection and enhancement of freshwater, estuarine and marine
water quality; and (c) protection of the health and functioning of ecosystems and
conservation of biodiversity. The third goal which is linked to this first goal is to
encourage ecologically sustainable development through planning and management
of future developments (The Darwin Harbour Advisory Committee 2003). The other
three management goals are mainly related to the conflicting uses of the Darwin
Harbor region rather than the ecological values. Hence, we considered the above
explained first and third management goals and their expected outcomes only to
decide on ecological assessment endpoints.
236
Assessment endpoints are expressions of an environmental value that is to be
protected (USEPA (US Environmental Protection Agency) 1998). Ecological related
endpoints often help to sustain the natural structure, function, and biodiversity of an
ecosystem. Hence, deciding on an assessment endpoint is the process of defining a
goal for a risk assessment through an ecological entity and its attributes. However,
there are two main steps involved when selecting ecological endpoints (USEPA (US
Environmental Protection Agency) 1998). The first step is to select ecological values
that are suitable for assessment endpoints. Therefore, the ecological relevance,
vulnerability to potential stressors, and relevance to management goals were
considered (Obery and Landis 2002; USEPA (US Environmental Protection Agency)
1998). The other step is to identify the specific ecological entity and the
characteristics of these entities that required a protection. In that sense, the ecological
assessment endpoints reflect management goals, and the ecosystem they represent.
The Darwin Harbor is a small estuary covered by one of the largest and most
diverse areas of mangroves in the Northern Territory. The Harbor supports a range of
estuarine, freshwater, and terrestrial environments. Although the catchment mostly
comprises vacant Crown land, it also contains a mix of commercial, industrial and
residential areas including Darwin, Palmerston and Cox Peninsula cities. There has
been some excellent research and mangrove monitoring in Darwin Harbor to balance
the ongoing industrial developments and the conservation. However, a significant
knowledge gaps remain. For example, the report; “The mangrove forest of Darwin
Harbour” by Moritz-Zimmermann (2002) stresses the need to study the likely effects
of human disturbances to mangroves. The recommendations made by McHugh
(2004), McGuinnes (2003), NTG (2005) and the Darwin Harbour Regional Plan of
237
Management (2003) also explicitly explain the need of an integrated approach for
mangrove monitoring.
Most of the catchment, about 80%, is undeveloped and comprises savannah
woodlands and forests. They are not in pristine condition due to high fire frequency
and access tracks (Skinner et al. 2009). Extensive growth of invasive weeds (gamba
and mission grass) during the wet season increases the fire frequency and intensity
during the dry season. Further, freshwater lagoons, grassy swamps, paperbark
woodlands, floodplains and rainforest cover approximately 6% of the catchment
(Skinner et al. 2009). Land clearing and urbanization alter overland flow paths of
runoff, reduce the volume of water that infiltrates to groundwater, and the time
runoff takes to enter the rivers and creeks. There are about 1150 hectares of
nationally significant, mostly dry rainforest in the Harbor catchment. They appear as
small patches around the margin of the tidal flats, and most patches are less than 10
hectares in extent (Northern Territory Government 2010). However, there are some
large areas of rainforests or vine-thicket habitat at Blaydin Point, Wickham Point,
Flagstaff Hill and Kings Table. Fortunately, these forest patches are mostly
surrounded and protected by mangroves. The influences from the development
pressure are minimal, and forest patches are in pristine condition (Northern Territory
Government 2010).
The Northern Territory government initiated numerous water quality studies
in Darwin Harbor (Skinner et al. 2009; Smith and Haese 2009). They reported the
contribution from urban and rural runoff and point sources. It was shown that land
use influences the water quality and flow volume that enter the Darwin Harbor, and
consequently changes the water quality of the Darwin Harbor. The Harbor itself is a
238
home for a diverse range of marine species: salt water crocodiles, dugongs, marine
turtles, and a variety of fish. However, the marine invertebrate fauna of the Harbor
region is poorly understood and is still being described. Hence, there is a potential
for severe impacts on coastal water quality and overall ecological health of the
Harbor due to the increasing population and land developments unless the water
quality of rivers and streams entering the Harbor is not well managed.
The ecological endpoints were selected subsequent to this information and
expert opinion about the study area. Further, the characteristics that need to be
protected for these endpoints were identified.
6.2.4 Risk analysis
Risk analysis is the second phase of the RRM. Here, two primary components
of the risk: exposure and effects, and their relationship between each other, and
ecosystem characteristics were analysed (Figure 6.2). First, the area can be divided
into more homogeneous regions: risk regions for further analysis. Risk regions can
be derived based on the topography, habitat or any other criteria that helps to form
homogeneous regions. For example, Yu et al. (2015) sub-divided their study area
into risk regions based on morphological, ecological, environmental and
administrative characteristics. O'Brien and Wepener (2012) introduced a combination
of management objectives, source information, and habitat data as criteria for
forming risk regions in South Africa. Bartolo et al. (2012) identified sub-regions of
Northern Tropical Rivers, Australia based on river basins (catchments). Colnar and
Landis (2007) and Hayes and Landis (2004) used watershed and bathymetric
boundaries for subdivision of the area, while Moraes et al. (2002) cosidered the
239
watersheds and landuse of the area. Catchment area limits, boundaries of
environmentally protected areas, territory boundaries, habitat and source types, the
flow of tributaries into Mountain river/contours, incremental gradient of human
activities of the lower reaches of the catchment are few other criteria used by
different studies (Moraes et al. 2002; Obery and Landis 2002; Walker et al. 2001;
Wiegers et al. 1998).
Most of Darwin Harbor’s catchment is undeveloped and mainly comprises
mangroves, savanna woodlands, urban and rural land uses. We divided the area into
different regions considering the boundaries of catchments (drainage boundaries),
homogeneity, and data availability for further modelling. Some small catchments
were merged with neighbouring catchments that have similar topographical
characteristics and land use forming homogeneous sub-regions (Figure 6.1).
The relationships and the probabilities of interaction between risk
components are described by numerical weighting factors called filters (Colnar and
Landis 2007; Yu et al. 2015). We identified two filters: a conceptual model pathway
from source to habitat as an exposure filter, and a conceptual model pathway from
habitat to endpoint as an effect filter. As described by Colnar and Landis (2007) and
Wiegers et al. (1998), exposure filters were developed considering two major issues:
(a) Will the source release the stressor; (b) Will the stressor then occur and persist in
the habitat. The effect filters were developed considering whether the assessment
endpoint occur in and utilize the habitat, and whether results from interaction with
the stressor are positive (undesirable) or negative (beneficial). If the pathway is
completed, we assigned a value of “one” to the corresponding filter, otherwise the
240
value is “zero”. We did not apply weights for exposure and effect filters in this
study.
The threats and habitats were ranked based on two-point scale within the
range of 0-6, and 0, 2, 4, and 6 represented no, low, moderate and high respectively
(Bartolo et al. 2012; Yu et al. 2015). The percentage cover of each threat and habitat
within sub-regions were considered for ranking. Then, they were classified according
to the Jenk’s optimization method (Jenks and Caspall 1971) in ArcGIS software.
This method minimizes squared deviations of class means or considers a goodness of
variance fit. The optimization is achieved when the quantity goodness of variance fit
is maximized. Hence, the classification is based on the natural grouping that is
inherited in the data. For example, when there is a greater possibility of occurrence
of one specific threat in one risk region compared to other risk regions, the rank
assigned was as six. The rank was equal to zero, when it is very unlikely to occur.
The point source discharge data were sub-divided into the following four
different stressors: nutrients (nitrogen and phosphorus); metals (aluminium, arsenic,
cadmium, chromium, copper, nickel, lead and zinc); Total Suspended Solids (TSS);
and Volatile Suspended Solids (VSS). They were ranked based on the percentage of
the number of kilograms per risk region (with respect to the areal extent of the
corresponding sub-region).
6.2.4.1 Relative risk calculation
The quantitative analysis of risks of each region depends on the interaction of
all components such as stressors, habitats, exposures and effects. Hence, the
241
magnitudes of the total threats in the risk region, the sum of the potential threat
exposure in the region, and the total risk to the ecological assessment endpoints were
calculated as explained by Landis and Wiegers (1997), Moraes et al. (2002) and
Bartolo et al. (2012). The following formulas were used for the calculation. For
example, to calculate the sum of stressors in each risk region, Eqn. (1) was used.
𝑆𝑆𝑆𝑆𝐵𝐵 𝑙𝑙𝑓𝑓 𝐵𝐵𝑠𝑠𝑟𝑟𝑟𝑟𝐵𝐵𝐵𝐵𝑙𝑙𝑟𝑟𝐵𝐵 𝐵𝐵𝑔𝑔 𝑟𝑟𝐵𝐵𝐵𝐵𝑘𝑘 𝑟𝑟𝑟𝑟𝑙𝑙𝐵𝐵𝑙𝑙𝑔𝑔 = ∑(𝑆𝑆𝑠𝑠𝑟𝑟𝑟𝑟𝐵𝐵𝐵𝐵𝑙𝑙𝑟𝑟𝐵𝐵) (1)
When the stressor has potential to impact habitat, the exposure filter received
a value of “one”, and the sum of potential stressor exposure in the region was
calculated as Eqn. (2).
𝑆𝑆𝑆𝑆𝐵𝐵 𝑙𝑙𝑓𝑓 𝑝𝑝𝑙𝑙𝑠𝑠𝑟𝑟𝑔𝑔𝑠𝑠𝐵𝐵𝐵𝐵𝑙𝑙 𝐵𝐵𝑠𝑠𝑟𝑟𝑟𝑟𝐵𝐵𝐵𝐵𝑙𝑙𝑟𝑟𝐵𝐵 𝑟𝑟𝑃𝑃𝑝𝑝𝑙𝑙𝐵𝐵𝑆𝑆𝑟𝑟𝑟𝑟 𝐵𝐵𝑔𝑔 𝑟𝑟𝐵𝐵𝐵𝐵𝑘𝑘 𝑟𝑟𝑟𝑟𝑙𝑙𝐵𝐵𝑙𝑙𝑔𝑔 = ∑(𝑆𝑆𝑠𝑠𝑟𝑟𝑟𝑟𝐵𝐵𝐵𝐵𝑙𝑙𝑟𝑟 ∗ ℎ𝐵𝐵𝑎𝑎𝐵𝐵𝑠𝑠𝐵𝐵𝑠𝑠)
(2)
The total risk to the ecological assessment endpoints in each risk regions was
calculated as Eqn. (3);
𝑇𝑇ℎ𝑟𝑟 𝑠𝑠𝑙𝑙𝑠𝑠𝐵𝐵𝑙𝑙 𝑟𝑟𝐵𝐵𝐵𝐵𝑘𝑘 𝑠𝑠𝑙𝑙 𝑠𝑠ℎ𝑟𝑟 𝑟𝑟𝑒𝑒𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝐵𝐵𝑒𝑒𝐵𝐵𝑙𝑙 𝐵𝐵𝐵𝐵𝐵𝐵𝑟𝑟𝐵𝐵𝐵𝐵𝐵𝐵𝑟𝑟𝑔𝑔𝑠𝑠 𝑟𝑟𝑔𝑔𝑟𝑟𝑝𝑝𝑙𝑙𝐵𝐵𝑔𝑔𝑠𝑠𝐵𝐵 𝐵𝐵𝑔𝑔 𝑟𝑟𝐵𝐵𝑒𝑒ℎ 𝑟𝑟𝐵𝐵𝐵𝐵𝑘𝑘 𝑟𝑟𝑟𝑟𝑙𝑙𝐵𝐵𝑙𝑙𝑔𝑔 =
∑(𝑇𝑇𝑙𝑙𝑠𝑠𝐵𝐵𝑙𝑙 𝑟𝑟𝐵𝐵𝐵𝐵𝑘𝑘 𝑠𝑠𝑙𝑙 𝑟𝑟𝑒𝑒𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝐵𝐵𝑒𝑒𝐵𝐵𝑙𝑙 𝐵𝐵𝐵𝐵𝐵𝐵𝑟𝑟𝐵𝐵𝐵𝐵𝐵𝐵𝑟𝑟𝑔𝑔𝑠𝑠 𝑟𝑟𝑔𝑔𝑟𝑟𝑝𝑝𝑙𝑙𝐵𝐵𝑔𝑔𝑠𝑠) (3)
6.2.5 Risk characterisation
This is the final phase of the risk assessment (Figure 6.2). The results from
the relative risk calculation indicate the regional perspective of risks in the Harbor
242
catchment. The total risk in each risk region, risk in each habitat, and the total risk to
assessment endpoints were used to derive site specific information.
6.2.6 Uncertainty and sensitivity analysis
As risk predictions in the RRM are point estimates that derive from data and
subjective expert knowledge, they are uncertain. There are three main uncertainty
categories that can propagate through the RRM. Model structure uncertainty is
propagated because key elements are not included in the conceptual model due to a
lack of knowledge, data availability and over simplification. Assigning an input
value to the model is considered as another source of uncertainty. This specifically
related to ranking stressors, habitats with a lack of knowledge of interactions
between model components, simplification and assumptions. The quality of the data,
their availability and currency can also be categorised as data uncertainties. The
subjective judgements of experts contribute some errors.
Three uncertainty levels (ranks) were used (low, moderate and high). For
example, when the data are site specific and high quality, the uncertainty level was
categorised as low. When there were no site specific data (or no data), and the quality
was poor, outdated, and if we made one assumption due to less knowledge about
data, was ranked as medium uncertainty. For example, since residential data of this
study is outdated, we updated them with Google Earth images, however, all new
buildings near residential areas were assumed as houses, and were ranked as medium
uncertainty. If there were no site specific data (or no data), and the quality was poor,
outdated, and more than one assumption made due to less knowledge about data,
they were ranked as high uncertainty. The discrete probability distributions were
243
introduced to each component as explained by Yu et al. (2015), and Hayes and
Landis (2004). For example, for the RRM model components with high uncertainty,
a probability of 0.6 was assigned. The adjacent ranks received a probability of 0.2
each. For the model components with medium uncertainty, the probability was 0.8
with two adjacent ranks having the probability of 0.1 each. No distributions were
assigned to the components with low uncertainty.
The RRM produced point estimates for risks based on ranks and filters from
imperfect data. The Monte Carlo simulation was applied to determine the entire
range of possible outcomes and likelihood of achieving each of the model
components using Crystal Ball® software. Here, we introduced already defined
uncertainty levels of each component and ran the simulations for 1000 iterations. We
repeated these simulations different times and noted that there were no differences
after 1000 iterations.
The sensitivity analysis was also undertaken using the Crystal Ball® software.
This analysis explains the influence of uncertainties of sources to the model
sensitivity or parameter uncertainty (Colnar and Landis 2007). The model parameters
were classified according to the correlation coefficients that were generated through
the sensitivity analysis with respect to their contribution to the prediction of
uncertainty. Hence, a parameter that has high impact on uncertainty within the model
should obtain high correlation (Colnar and Landis 2007).
244
6.3 Results
6.3.1 Problem formulation
We referred to available literature and expert knowledge to identify sources,
stressors and their characteristics, exposure opportunities, the ecosystem components
potentially at risk, and ecological effects in the Darwin Harbor catchment. The
resultant conceptual diagram shows the interaction between urban or rural activities
and their effect on coastal environment (Figure 6.3).
Figure 6.3. The conceptual diagram showing estuarine activities and their effect on
mangroves, water bodies and then to the Darwin Harbour (Courtesy of the
Integration and Application Network, University of Maryland Centre for
Environmental Science (ian.umces.edu/symbols/)).
245
The major sources of threats for estuarine areas are land clearing for urban
and rural developments, land clearing for industrial and communication
developments, agriculture, point source discharges and climate change. They will
directly or indirectly impact on species and habitat destruction, increased flooding,
increasing (velocity and volume) soil runoff with nutrients and chemicals, increasing
sedimentation and pollution. For example, vegetation serves to slow the flow of
water and increase its infiltration into soil and groundwater storages. This enables the
water to flow more gradually into watercourses over a longer period of time.
However, because of extensive land clearing throughout most catchments and,
especially in more urbanized areas, the increased volume and velocity of water can
enter watercourses during flood. When it floods, runoff can carry substances such as
topsoil, chemicals, rubbish, nutrients, and oil and grease from roads. This polluted
water can be a significant problem for the health of mangroves. This may lead to the
degradation of mangroves unless these stressors are not addressed appropriately.
Since runoff from the Winnellie industrial area flows into mangroves of the Darwin
Harbor, heavy metal and nutrient concentration have been compared with other areas
which are not affected by urban runoff (Askwel 2008). Marginally higher
concentrations of heavy metals have been observed in the Winnellie mangrove
sediments compared to other sites. However, the conditions of mangroves are poorly
studied over time compared with such chemical pollutions.
The accumulation of sediment can greatly be increased by upstream
disturbances such as vegetation clearing and urbanisation. Although mangroves
colonise sedimentary environments, excessive sediment deposits can reduce growth
or even kill mangroves (McGuinnes 2003). Complete burial of root structure may
disturb gas exchange, killing root tissue and trees (McHugh 2004). Since mangroves
246
are susceptible to pollution, pollutants from urban areas and agrochemicals results in
short term and long term effects on mangroves (Askwel 2008) because mangrove
environments are a trap for rubbish accumulation. A Darwin Harbor mangrove
rubbish survey found the three most common types of rubbish are aluminium cans,
plastic bottles and plastic bags (Lewis 2002). They were thought to be sourced from
storm water drains, tides and direct littering at the sites, and their rate of
decomposition is indefinite or quite high (Lewis 2002).
Currently, there is a major liquefied natural and petroleum gas project around
Darwin Harbor being run by INPEX. Subsea pipelines which transport liquid
hydrocarbons run between the gas fields and onshore at the Middle Arm (Metcalfe
2007). For the entire project, the proposed land clearing extent is about 300 ha
including mangroves. Parallel to this gas project, major industrial developments in
the Middle Arm are being discussed. According to a report published by INPEX at
different stages of the project, there is a high possibility of increasing sedimentation
over certain areas of the Darwin Harbor catchment (van Senden et al. 2013).
The point source discharges can either be hot water outflows (storm water
drainages) or treated sewage discharges. A hot water outflow from storm drainage
reduces mangrove leaf area and increases defoliation (McGuinnes 2003). Increased
water temperature shows mortality of seeds. Also at low levels, nutrients entering
waterways can lead to increasing productivity by promoting the growth of even algae
and plants. One of the consequences of urbanization is discharging pre-treated
sewage to mangrove swamps as efficient, low cost and natural wastewater treatment
(Lewis 2002). Nutrient pollution in the mangroves can have various effects.
Basically, it can enhance tree growth and productivity. On the other hand,
247
excessively high nutrient concentration causes excessive algal growth which can
hinder oxygen exchange and mangrove seedling (McGuinnes 2003). Since there are
waste water treatment plants around Darwin Harbor (Darwin Central Business
District, Palmerston South, Buffalo Creek and Ludmilla Creek), increasing
sedimentation and accumulation nutrients as well as pollutants to the mangroves can
be expected.
Numerous water quality studies of the Darwin Harbor determined the
contribution from diffuse sources. The main nutrient load contributor is sewage
effluent and the effect is more significant at a local scale, such as a tidal creek level
where point source nutrients are discharges rather than at a whole-of-Harbor scale
(Skinner et al. 2009).
Climate change predictions indicate that sea level will increase by one metre
during the course of a 20 year period. This will highly affect mangroves with species
and habitat destruction showing a backward movement of zonation pattern.
We categorized each source into different stressors (Figure 6.4). For example,
the stressors related to the land clearing for urban and rural developments were
identified as residential, transportation, utilities and recreational activities. Utilities
are defined as electricity generation and transmission, gas treatment, storage and
transmission. The source, land clearing for industrial and communication
developments, was divided into three different stressors that included manufacturing
and industrial, dredging, and services. Dredging reflects the areas that are going to be
affected once the ongoing gas project has completed the onshore dredging. Services
include the areas that are allocated to the provision of commercial or public services
248
substantially influencing the natural environment. The agriculture runoff from
croplands and horticulture lands are increasing levels of nutrients and turbidity in
streams and in mangroves. Available topographic data allowed us to categorise the
source agriculture into three different areas: cropping, horticulture and aquaculture.
6.3.1.1 Habitats and ecological assessment endpoints
Since unlocking the many possibilities for increasing infrastructure
development along the mangrove dominated shoreline of Darwin Harbor, ensuring
the maintenance of ecosystem services and a healthy environment is one of the
priority goals of the Government (Harrison et al. 2009; Lee 2003). Although many
studies have been undertaken in the mangrove forests of Darwin Harbor to
understand this valuable ecosystem, some knowledge gaps still exist. The
maintenance of mangrove health (sustainable ecosystem) and maintenance of water
quality (aquatic biodiversity) were selected as assessment endpoints for this study.
The attention was given to the management goals of the Northern Territory
government (The Darwin Harbor Advisory Committee 2003), the ecological
relevance to the stakeholders of the region, and vulnerability to potential stressors.
Most of the areas of the Darwin Harbor catchment are vacant Crown land
(Harrison et al. 2009). Although there are a number of habitat types present in the
catchment, mangroves, streams, rivers and other water bodies are the most available
terrestrial and aquatic habitats that highly influence to the health of the Darwin
Harbor. Hence, we selected mangroves and water bodies as habitats of the RRM.
Change in species structure, change in habitat for flora and fauna, increase in
sedimentation, and influx of nutrients were considered as effects of stressors in
249
mangrove habitat. For water bodies, we accounted for change in waterways, change
in ground water level, influx of nutrients, influx of chemicals, and increase in
sedimentation.
The exposure to the stressor creates a possibility of a biological degradation
on habitats. For example, the land clearing stressor indicates a reduction of mangrove
coverage, change in species structure, reduction of habitat for flora and fauna. Point
source discharges exposures an influx of nutrients. As explained in Figure 6.4,
exposure and effect filters were identified for mangroves and water features.
Figure 6.4. The conceptual model for the relative risk modelling in the Darwin
Harbor. They were prepared for all risk regions. Exposure and effect filters were
modified with respect to each risk region.
250
6.3.2 Risk analysis
Based on sub-catchment boundaries, we created 22 risk regions from an
initial 42 sub-catchements of the Harbor. For example: Howard River risk region
includes six sub-catchments; Blackmore River and Elizabeth River risk regions were
formed merging seven and eight of the Darwin Harbor’s sub-catchments
respectively. Howard River, Elizabeth River and Blackmore River risk regions are
the largest sub-regions including major river systems. According to Skinner et al.
(2009), they are classified as rural type land uses. Further, Sadgroves Creek, Hudson
Creek and Reichardt Creek risk regions comprise with the largest area of light
industrial land use. Hudson Creek is subjected to an industrial expansion of the
nearby East Arm Port. Rapid Creek, Buffalo Creek, Ludmilla Creek and the Darwin
Central Business Districtrisk regions include extensive urban areas.
6.3.3 Risk characterisation
The total relative risk was calculated for the 22 risk regions. Myrmidon
Creek, Blackmore River, Bleesers Creek and Elizabeth River risk regions showed the
highest total relative risk of ecological assets in the risk region. The Creek A, Sandy
Creek, West Arm and the Pioneer Creek were the risk regions with the lowest total
relative risk to ecological assets. The model did not calculate the total relative risk to
the Darwin Central Business District risk region as its mangrove habitat rank is zero.
The Bleesers Creek catchment is adjoining to the Winnellie industrial area, and
therefore, the percentage of metal, TSS, and sedimentation are high compared to
other catchments. Myrmidon Creek is closer to Palmerston city, has influence from
residential and industrial areas, and contains a sewage plant and diffuse entry point.
251
Blackmore River and Elizabeth River catchments have high impact from agriculture
and point sources.
1 Blackmore River 7 Elizabeth River 13 Micket Creek 19 Sadgroves Creek
2 Bleesers Creek 8 Howard River 14 Myrmidon Creek 20Sandy Creek
3 Buffalo Creek 9 Hudson Creek 15 Palmerston South 21 West Arm
4 Charles Point 10 Kings Creek 16 Pioneer Creek 22Woods Inlet
5 Creek A (Middle Arm) 11 Ludmilla Creek 17 Rapid Creek
6 Darwin Central
Business District (CBD) 12 Mitchell Creek 18 Reichardt Creek
252
Figure 6.5. The total relative risk to ecological assets in risk regions of the Darwin
Harbor. They were classified as low, medium or high based on the Jenks’s
optimization (Jenks and Caspall 1971).
We investigated the total relative risk to each assessment endpoint separately.
The highest risks related to the maintaining mangrove health were found in
Sadgroves Creek, Reichardt Creek, Blackmore River, Myrmidon Creek, Bleesers
Creek and Elizabeth River (Figure 6.6 (A)). Mangroves of Creek A, Sandy Creek,
West Arm and Pioneer Creek are in a pristine condition with low risks. Although we
did not include influence from point source discharges to the Ludmilla Creek risk
region, it showed moderate total risk to the ecological assets. If the point source
discharges were included, the total relative risk of Ludmilla Creek risk region would
be high.
The total risk values obtained for the maintenance of water quality correspond
to the risk values for the maintenance of mangrove health (Figures 6.6 (A) and (B)).
Although the risk scores for both maintenance of mangrove health and maintenance
of water quality of Buffalo Creek catchment are moderate, this will be high in the
near future due to the expanding Lyons and Muirhead residential developments. To
the same extent, the Palmerston South and Mitchell Creek catchments will
experience higher risk scores due to Palmerston’s new urban footprints: Bellamack
and Roseberry suburbs.
253
Figure 6.6. Total risk scores for the ecological endpoints; (A) Maintenance of
mangrove health; (B) Maintenance of water quality
6.3.4 Uncertainty and sensitivity analysis
The simulations of total risk to each ecological endpoint were modelled. The
frequency of the variation of total risk values was shown in Figure 6.7. As shown in
Figure 6.7 (a), the simulated total risk scores of the Pioneer Creek, Micket’s Creek,
Creek A, Sandy Creek and the West Arm are characterised by distributions with data
1 Blackmore River 7 Elizabeth River 13 Micket Creek 19 Sadgroves Creek
2 Bleesers Creek 8 Howard River 14 Myrmidon Creek 20 Sandy Creek
3 Buffalo Creek 9 Hudson Creek 15 Palmerston South 21 West Arm
4 Charles Point 10 Kings Creek 16 Pioneer Creek 22 Woods Inlet
5 Creek A (Middle Arm) 11 Ludmilla Creek 17 Rapid Creek
6 Darwin Central Business
District (CBD) 12 Mitchell Creek 18 Reichardt Creek
254
spikes. They indicated less uncertainty in source-habitat-endpoint pathways. The
Rapid Creek and the Woods Inlet distributions indicated moderate uncertainty of the
components (Figure 6.7 (b)). However, Reichardt Creek, Bleesers Creek, Myrmidon
Creek, Palmerston South, Elizabeth River and Blackmore River showed the most
widely spread distributions. This indicated higher uncertainties of model
components.
Figure 6.7. The frequency of all possible outcomes of total risk of ecological assets
in the risk regions: (a) Ludmilla Creek, Mitchell Creek, Mitcket Creek, Palmerston
South, and Pioneer Creek; (b) Blackmore River, Elizabeth River, Howard River,
Hudson Creek, and Kings Creek; (c) Rapid Creek, Reichardt Creek, Sadgroves
Creek, Sandy Creek, West Arm, and Woods Inlet; (d) Bleesers Creek, Buffalo Creek,
Charles Point, and Creek A.
255
The sensitivity analysis calculated rank correlations for all model components
(i.e. risk regions, sources, habitats, and assessment endpoints). The rank correlation
indicates whether the uncertainty of any model component has an effect on the
uncertainty of the forecast. Figure 6.8 shows the rank correlation coefficients of the
risk regions having the highest total risk scores. These graphs show the relative
magnitude of influence for each variable. Assumptions made for water bodies within
the Elizabeth River, Bleesers Creek, and Blackmore River risk regions have a
dominant effect on the uncertainty of the forecast (Figures 6.8 (A), (B) and (D)). In
two instances (Blackmore River and Elizabeth River), it accounted nearly 80% of the
variation in forecast values, and can consider as the most important assumption in the
model. Treated sewage/nutrients of the Elizabeth River contributed least to the
forecast variance. It means that this assumption has a small effect; it could be
ignored or even could be eliminated from the analysis. The influence from the habitat
mangroves was relatively higher than other components.
256
Figure 6.8. The rank correlations for sub-regions having the highest total risk scores:
(A) Blackmore River; (B) Elizabeth River; (C) Myrmidon Creek; (D) Bleesers
Creek.
6.4 Discussion
Ecological risk assessment is an important tool for supporting and improving
environmental decision-making. The RRM was used as such a tool for assessing the
impact of potential sources on habitats within the Darwin Harbor catchment in the
context of ongoing developments. A conceptual model integrating existing
information and expert opinion focused on describing source-habitat-endpoint
pathways was developed as an important foundation in the RRM process. This
257
approach to conceptual model development has limitations, and we suggest
organising a series of stakeholder meetings to gather information rather than informal
meetings for future work. This would facilitate stakeholders to share their own
perspectives, come to agreement on what the conceptual model should include, and
to avoid subjective opinions.
Accurate and current spatial information is required for aquatic habitats, point
source discharges and land clearing. For example, we did not obtain point source
data for the Ludmilla Creek risk region which encompasses an abandoned treated
sewage discharge point. The Mitchell Creek risk region does not include the
influence from the recently populated Bellamack suburb. The current stressors and
effects from the LNG project in the Harbor need to be incorporated, especially those
effects due to the increase in sedimentation.
Although the results of the RRM applied here are representative of current
knowledge of stressors and effects on mangroves, we would recommend applying
the model to individual sub-catchments (42) rather than the merged sub-catchments
that resulted in the 22 risk regions. For example, the nutrient and total pollutant loads
of the Howard River, Blackmore River and the Elizabeth River are higher than other
risk regions due to their larger areal extent (Skinner et al. 2009). However, to
minimize such effects in this study, we calculated the percentage of kilograms per
each point source values with respect to the areal extent of each risk region.
It is clear that catchments with a high percentage cover of industrial and
commercial land use, and residential areas received the highest risk to each
ecological endpoint. Mangroves close to development and population centres
258
suffered some clearing and localized damages from contaminants. These catchments
received higher ranks for residential, nutrients, metal, and VSS than undeveloped
areas. For example, Elizabeth River risk region has the highest total risk to the
ecological endpoints, and dredging, horticulture, cropping and aquaculture obtained
relatively high ranks. Mangroves and waterbodies also ranked high. The Creek A
obtained the lowest total risk to the ecological endpoints. Currently, although there is
high impact from sea level rise and sedimentation stressors to this risk region, other
stressors have minimal or no impact. Further, mangrove habitat was also ranked high
due to impacts from sedimentation and sea level rise. The prediction of climate
change effects has a major role in risk to mangroves. Due to the elevation variation
of the catchments, some areas will suffer from climate change effects or sea level
rise. Most of the catchments where the total risk of ecological assets is low have less
influence from land clearing for industries, diffuse entries and climate change.
This model provides a robust framework for assessing the regional ecological
risk and the benefit of Monte Carlo analysis in describing uncertainty in a RRM. The
outcome of this study should not be used for decision making; rather it can be used
as a framework for future studies. This is a useful approach for integrating all
components: multiple assessment endpoints, threats and habitats together, and
addressing the knowledge gap. According to the Northern Territory’s long-term
development plan, parts of Mitchel Creek, Palmerston South, Elizabeth River and
Blackmore River will become residential and light industrial areas to accommodate
the increasing population. These catchments already have high or moderate levels of
risks. For example, Palmerston South risk region already accounted for high risks for
various stressors namely: residential, utilities, sedimentation, sea level rise, waste
water/metals, and stormwater drainage/suspended solids. Hence, this is an important
259
issue to be considered when making decisions. The development plan is being
proposed for the Middle Arm of the Darwin Harbor (the City of Weddell) especially
for the Pioneer Creek catchment. Currently, the Pioneer Creek catchment
experiences low total risk of the ecological assets compared to others. For example, it
has low impact from stressors: land clearing for residential and transport and
communication, sedimentation and point sources discharges and surface runoff, and
there is no influence from other sources. The proposed developments will increase
the land clearing for residential and commercial activities, and subsequently the total
relative risk of the region. To the same extent, Skinner et al. (2009) predicted a
significant increase of pollutant load as a result of these developments. Therefore,
this is another important issue to be considered by the decision makers. Instead, the
risk regions with moderate total risks, for example, Sandy Creek and Kings Creek
have more possibilities for developments. Currently, these risk regions have minimal
or no influence from the majority of sources such as agriculture, aquaculture, climate
change, and land clearing for residential or commercial developments. However,
although the RRM can provide useful information about the ecological risk of
corresponding areas, it is difficult to select one risk region over another for locating
developments without a proper methodological approach.
The Integrated Water Resources Management (IWRM) approach promotes
the coordinated development and management of water, land and related resources,
in order to maximize the resultant economic and social welfare in an equitable
manner without compromising the sustainability of vital ecosystems. This has been
introduced by the Global Water Partnership (GWP) (Global Water Partnership 1996).
According to IWRM, stakeholder participation, river basin planning, pollution
control, and monitoring are few activities that constitute water management. This is a
260
similar approach as to what has been undertaken through the “Darwin Harbour
Strategy” by the Northern Territory Government.
The “Darwin Harbour Strategy” is a guide for sustainable development of the
Darwin Harbor region, and it supports the integrated management of the Darwin
Harbor region’s diverse environmental, social, cultural and economic values and
uses. However, the Northern Territory's water resources are considered to be under
relatively little pressure due to a comparatively small population base and low
intensity of land use in the catchment. This situation is changing due to a predicted
increasing population in the near future and therefore the Northern Territory is
increasingly becoming subject to higher demands, so integrated assessment
approaches are recommended. Hence, we would also suggest continuing the already
introduced “Darwin Harbour Strategy” approach to have coordinated development
and management of water, land and related resources in the region.
In summary, increasing population and consequent land development has a
potential for more severe impact on water quality and ecological health of the
Darwin Harbor catchment. For example, the potential increase of pollutant load and
nutrient load from the medium and long-term plans for the Harbor catchment
developments has already been predicted. Land clearing, residential and industrial
activities will increase. Therefore, we demonstrated the successful application of the
RRM to assess these risks. The model is repeatable and can use to predict the
ecological risk in the near future with various scenarios.
261
6.4.1 Uncertainty and sensitivity analysis
Although the Darwin Harbour catchment is a rapidly developing area,
Blackmore River, Howard River and Elizabeth River catchments are classified as
rural land uses compared with catchments closer to the Darwin Central Business
District which are classified as urban areas. These rural areas are poor in data
availability and data quality. Available topographic data are outdated; more than 10
years old. For example, through the sensitivity analysis, the waterbodies data
contributed to the majority of uncertainty for the Elizabeth River, Blackmore River,
and Bleesers Creek catchments as there is less spatial data available for waterbodies
in these catchments (Figure 6.8). The error contribution from mangroves to the risk
analysis in Blackmore River, Elizabeth River, Myrmidon Creek and Bleesers Creek
regions are also higher than other regions. The mangrove coverage spatial data set is
almost 20 years old, and it does not include already cleared and newly regenerated
areas exhibiting higher data uncertainty. For instance, the mangrove clearing in the
Middle Arm due to the activities of the natural gas project is not included in the
analysis. Therefore, the main limitation related to this application is data input. The
outdated or lack of data impacts the analysis. Although we updated newly formed
residential areas using Google Earth, there may be some errors associated with this
interpretation due to difficulties in determining industrial and residential land use
from the satellite imagery without field verification (that is, both cover types appear
similar in the imagery) .
The introduction of subjective expert opinion may add some gross errors to
the model. The ranks, exposure and effect filters are subjective. This subjectivity can
change the model outputs. A further issue related to the conceptual model is that the
262
model can over-simplify the real world situation. The assigning of ranks in the
uncertainty analysis is subjective. This is also based on expert opinion and existing
information. Hence, we would recommend organising a series of stakeholder
meetings and gathering information from various users, and collecting current
information before implementing the RRM. If uncertainties are minimised, the RRM
can be used by decision-makers and land managers to prepare action plans to reduce
threats from urban development to the coastal environment and mangroves.
6.5 Conclusions and recommendations
There are many ongoing and proposed industrial, commercial and residential
development projects in the Darwin Harbor catchment, and subsequently the natural
resources of Darwin Harbor region are experiencing an increasing development
pressure. As it is necessary to ensure the balance between these developments and
conservation, appropriate tools that combine knowledge from different sources into a
consistent environmental management framework are needed. One of the key
requirements for environmental monitoring programs in Darwin Harbor is an
integrated approach for ecological risk assessment that enables stakeholders to make
practical and sound management decisions. Therefore, in this study, we introduced
the application of a RRM to assess the ecological risk of the Darwin Harbor, and this
application is a significant contribution to the current knowledge in this context.
The application of the RRM involves three major phases: problem formation;
risk analysis; and risk characterization. We investigated existing knowledge and
consulted expert opinion in the problem formation phase. Then, a conceptual model
263
was created combining all information. Spatial data from topographic databases and
statistics relating to various Darwin Harbor monitoring projects were collected.
In the second phase of the analysis, we identified the ecological endpoints and
corresponding model components including sources, stressors and habitats. The
management goals of Darwin Harbor were considered in determining the ecological
assessment endpoints which were the maintenance of mangrove health and the
maintenance of water quality. The large Darwin Harbor catchment boundary and
boundaries of sub-catchments were adapted from the Australian Collaborative Land
Use Mapping program (V5). However, homogeneous small catchments were merged
and formed 22 risk regions for further analysis. We created a RRM to rank and sum
each source of stressors, and habitat in each risk region. Once model components
were ranked according to the percentage coverage of sub-regions (with respect to
individual areal extent), the pathway of sources-habitats-endpoints was created.
Risk characterisation is the last phase of the modelling process. The
uncertainty of each model component was investigated and classified into low,
medium and high, and the discrete probability distribution was introduced. The
Monte Carle simulation was used to obtain all possible values of risk scores. We
analysed the total relative risk of the ecological assets in risk regions. They were
classified into three classes according to the Jenk’s optimization.
The sub regions: Myrmidon Creek, Blackmore River, Bleesers Creek and
Elizabeth River showed the highest total relative risk to ecological assets in the risk
regions. Creek A, Sandy Creek, West Arm and Pioneer Creek showed the lowest
total relative risk of ecological assets. The study did not calculate the total relative
264
risk for the Darwin Central Business District region. These results indicated that risk
regions with a high percentage cover of industrial, commercial, residential areas,
diffuse entry points, and climate change effects contributed to the highest risk to each
ecological endpoint.
The relative risk model is a robust application that is suitable for a large
geographic area where multiple stressors are of concern. Data uncertainty is the
major limitation associated with the model. For example, one of the selected habitats
to be analysed is mangroves and spatial data on mangrove coverage is almost 20
years old. It does not include either recently cleared or regenerated mangrove areas.
We would not recommend using these results for decision making due to the various
uncertainties, however, this can be used as a framework for future studies. We
further would recommend organising a stakeholder meeting at the first phase of the
model for data collection. This will reduce the effect of subjective opinions. We
would also suggest continuing the already introduced “Darwin Harbour Strategy”
approach to have coordinated development and management of water, land and
related resources in the region.
265
6.6 Acknowledgement
6.7 References
Askwel R. 2008. Ichthys gas field development: Referral of proposed action-Browse
Basin and Blaydin Point, Darwin, DEV-EXT-RP-0053. INPEX Browse, Ltd.
Bartolo RE, van Dam RA, Bayliss P 2012. Regional Ecological Risk Assessment for
Australia's Tropical Rivers: Application of the Relative Risk Model. Human and
Ecological Risk Assessment 18: 16-46.
Brocklehurst P, Edmeades B. 1994-1995. Mangrove Survey of Darwin Harbour:
Northern Territory, Technical Memorandum No. 96/9.
Colnar AM, Landis WG 2007. Conceptual Model Development for Invasive Species
and a Regional Risk Assessment Case Study: The European Green Crab, Carcinus
maenas, at Cherry Point, Washington, USA. Human and Ecological Risk Assessment
13: 120-55.
Darwin Harbour Regional Plan of Management 2003, Terrestrial and freshwater
environments of the Darwin Harbour catchment, in Darwin Harbour Public
Presentations
Geoscience Australia. 2002. GEODATA TOPO 10M 2002 (Double precision),
ANZLIC unique identifier: ANZCW0703005262. Geoscience Australia.
266
Global Water Partnership 1996, What is IWRM?, viewed 08/07/2015,
<http://www.gwp.org/The-Challenge/What-is-IWRM/>.
Harrison L, McGuire L, Ward S, et al. 2009. An inventory of sites of international
and national significance for biodiversity values in the Northern Territory,
Department of Natural Resources, Environment, The Arts and Sport, Darwin, NT,
http://www.nretas.nt.gov.au/__data/assets/pdf_file/0020/13934/06_darwin.pdf.
Hayes EH, Landis WG 2004. Regional Ecological Risk Assessment of a Near Shore
Marine Environment: Cherry Point, WA. Human and Ecological Risk Assessment
10: 299-325.
INPEX 2013, Dredging in Darwin Harbour, viewed 12/01/2015,
<http://www.inpex.com.au/oldsite/sites/default/files/pdfs/120803_dredging%20in%2
0darwin%20harbour.pdf>.
Jenks, G. F. & Caspall, F. C. 1971, Error on Choroplethic Maps: Definition,
Measurement, Reduction, Annals of the Association of American Geographers, vol.
61, 217-244.
Landis WG, Wiegers JA 1997. Design Considerations and a Suggested Approach for
Regional and Comparative Ecological Risk Assessment. Human and Ecological Risk
Assessment 3: 287-97.
Lee GP. 2003. Mangroves in the Northern Territory, 25/2003D, Department of
Infrastructure, Planning and Environment, Darwin, Darwin.
267
Lewis D. 2002. Community Mangrove Monitoring Program: Impact Assessment of
Coastal Development on Estuarine Mangrove Environments in Darwin Harbour –
Technical Manual and Final Report, 26/2002
McGuinnes KA. 2003. The mangrove forests of Darwin Harbour: a review of
research on the flora and invertebrate fauna, Department of Industry, Planning and
Environment, Darwin Harbour Region: current knowledge and future needs.
McHugh P. 2004. DLNG Mangrove Monitoring Program: Monitoring of mangrove
community structure/composition, soil condition and mangrove productivity in
Darwin harbour, Department of Infrastructure, Planning and Environment, Land
Monitoring Branch, Darwin.
Metcalfe K 2007, The Biological diversity, Recovery from Disturbance and
Rehabilitation of Mangroves in Darwin Harbour, Northern Territory, Doctor of
Philosophy thesis, Charles Darwin University, Darwin.
Moraes, Landis WG, Molander S 2002. Regional risk assessment of a Brazilian rain
forest reserve. Human and Ecological Risk Assessment 8: 1779-803.
Moritz-Zimmermann A. 2002. Darwin Harbour Mangrove Monitoring Methodology
Technical Manual- Land Monitoring Series No. 3, Report No. 25/2002, Department
of Infrastructure Planning and Environment, Palmerston, NT, Australia.
268
NRETAS.2010a. Harbour health and water sampling, Department of Land and
Resource Management, Available at
http://lrm.nt.gov.au/water/aquatic/water_sampling.
NRETAS. 2010b. Water Quality Objectives for the Darwin Harbour Region-
Background Document, Department of Natural Resources, Environment, The Arts
and Sport, Aquatic unit, Palmerston, NT, Australia.
Northern Territory Government 2005. Mangrove Monitoring, Department of Land
Resources Management, Darwin.
Northern Territory Government 2010, Water Quality Objectives for the Darwin
Harbour Region - Background Document, Department of Natural Resources,
Environment, The Arts and Sport, Aquatic unit, Darwin.
O'Brien GC, Wepener V 2012. Regional-scale risk assessment methodology using
the Relative Risk Model (RRM) for surface freshwater aquatic ecosystems in South
Africa. Water SA 38: 153-66.
Obery AM, Landis WG 2002. A Regional Multiple Stressor Risk Assessment of the
Codorus Creek Watershed Applying the Relative Risk Model. Human and Ecological
Risk Assessment 8: 405-28.
vanSenden D, Hennecke W, Buckee J, et al. 2013. Mangrove Community Health
Remote Sensing Baseline Report: Ichthys Nearshore Environmental Monitoring
Program.
269
Skinner L, Townsend S, Fortune J. 2009. The Impact of Urban Land-use on Total
Pollutant Loads Entering Darwin Harbour, Northern Territory Government, Aquatic
Health Unit, Department of Natural Resources, Environment, the Arts and Sport.
Smith J, Haese RR 2009. The role of sediments in nutrient cycling in the tidal creeks
of Darwin Harbour. AUSGEO 95.
The Darwin Harbour Advisory Committee. 2003. Darwin Harbour Regional Plan of
Management, Available at <http://www.lrm.nt.gov.au/water/darwin-
harbour/regional>.
USEPA (US Environmental Protection Agency). 1998. Guidlines for Ecological Risk
Assessment, US Environmental Protection Agency: Risk Assessment Forum,
Washinton DC, USA.
Walker R, Landis W, Brown P 2001. Developing A Regional Ecological Risk
Assessment: A Case Study of a Tasmanian Agricultural Catchment. Human and
Ecological Risk Assessment 7: 417-39.
Wiegers JK, Feder HM, Mortensen LS 1998.A Regional Multiple-Stressor Rank-
Based Ecological Risk Assessment for the Fjord of Port Valdez, Alaska. Human and
Ecological Risk Assessment 4: 1125-73.
Yu, W., Zhang, L., Ricci, P. F., Chen, B. & Huang, H. 2015, Coastal ecological risk
assessment in regional scale: Application of the relative risk model to Xiamen Bay,
China, Ocean & Coastal Management, vol. 108, 131-139.
270
Chapter 7 – Conclusions and Recommendations
271
Highlights
• Mangrove species within the Rapid Creek mangrove forest were classified
using remotely sensed images. When discriminating mangroves from other
features in the study area, a large number of spectral bands with higher spatial
resolution were more accurate than broad spectral bands within the blue, green,
and red regions. A significant increase in classification accuracy was achieved
when using pan-sharpened imagery of five spectral bands within the visible
range. This approach can be used repeatedly not only to map new mangroves
but also to update existing maps.
• Mangrove tree crowns were successfully delineated using a combination of
WorldView-2 imagery, a digital surface model, and a GEOBIA algorithm.
Minor errors were detected where neighbouring trees with similar
characteristics were aggregated together and when upward pointing branches
of the same tree identified as different trees. However, these results are
invaluable especially to identify dead mangrove trees, to calculate number of
trees newly re-generated and to quantify biomass of each trees.
• The spatial variability of mangrove canopy chlorophyll was mapped using the
red, red-edge, NIR1 and NIR2 bands of the WorldView-2 satellite image, and
their derived normalized difference indices. The field samples were not
sufficient representatives of the image pixels due to inaccuracies of GPS
measurements and exhibited some errors. This is the first step towards
analysing mangrove canopy health and the approach can be improved
introducing more precise calibration such as very precisely measured tree
272
locations and very high resolution remotely sensed images preferably from
Unmanned Aerial Vehicles.
• The spatial variability of Leaf Area Index and Above Ground Biomass of
mangroves were also mapped using high resolution satellite images and digital
photographs. The results identified where the most healthy and unhealthy
mangrove forests were located.
• The ecological risk associated with Darwin Harbour mangroves was assessed
in the context of rural and urban development using a relative risk modelling
approach. The catchment was divided into 22 risk regions based on small
catchment boundaries and their homogeneity. The interaction between stressors
and habitats were modelled through exposure and effect filters. For instance,
the risk regions: Myrmidon Creek, Blackmore River, Bleesers Creek and
Elizabeth River showed the highest total relative risk for ecological assets.
These risk regions had a high percentage cover of industrial, commercial,
residential areas, diffuse entry points, and climate change effects. Therefore,
the tool can be served as an instrument that ensures the balance between rural
and urban developments and the mangrove conservation in the future.
Summary
Mangroves are one of the most globally important, salt tolerant woody plants
that are subjected to numerous natural and anthropogenic disturbances. Fast and
reliable methods to map and monitor mangrove ecosystems are vital for their
protection and management. However, detailed mangrove maps at the community or
273
species level are challenging to produce. Field observations are time consuming,
expensive, and difficult due to densely clustered nature of trees and muddy soil.
Therefore, remote sensing is the only way to obtain spatially extensive information
over the large areas that they cover. At the scale of an image pixel, mangrove
ecosystems are heterogeneous. Commonly used per-pixel analyses cannot adequately
describe the variations in their biophysical properties. Advanced digital image
analysis algorithms suitable for classifying mangrove species and quantifying
biophysical variations at sub-pixel level were tested. The Rapid Creek mangrove
forest, Darwin, Australia was selected as the test site.
The most recent study related to the Rapid Creek mangrove forest was found
in between 1994 and 1995. Since individual mangrove species have varying
tolerances to the period, frequency, and depth of inundation, precise and up-to-date
spatial information on the current status of this area is a prerequisite for sustainably
conserving and monitoring the ecosystem. The study discriminated mangrove and
non-mangrove areas using object based image classification. Integrating mangrove-
environment relationships was exploited in order to improve the mapping accuracy.
Mangrove areas were then further classified into species using a support vector
machine algorithm. Attention was also given to delineate individual tree crowns for
assessing and monitoring mangrove species.
Attributes of individual mangrove trees may help to analyse their biophysical
variations, growth patterns, and regeneration after natural or anthropogenic
disturbances. An object based image analysis method was tested for different
remotely sensed data sources for individual tree crown delineation. The combination
of high resolution satellite image, a digital surface model of the area, and object
274
based image analysis methods gave promising results. Further, the study analysed
biophysical variations of mangrove trees. According to the scientific literature, most
important plant biophysical properties are canopy chlorophyll, above ground biomass
and leaf area indices. Hence, this study developed fast, reliable, and low cost
methods for estimating these biomarkers from remotely sensed data. Developed
methods were validated using in-situ field measurements. Overall the results of this
study reveal the potential of using remotely sensed data for deriving spatial
information on current status of mangroves. However, to build resilience into
mangrove conservation plans, decision makers need to identify and protect
mangroves that are prone to disturbances. The study developed a spatial model that
assesses the ecological risk to the mangroves.
The traditional ecological risk assessment methods evaluate the interaction of
stressor, receptor and response components of the systems. When expanding these
methods to regional scale, the scale, complexity of the structure, and the regional
spatial components must be incorporated. Hence, this study introduced a relative risk
model that evaluates multiple, dissimilar stressors simultaneously and comparatively
at a regional scale.
The aim of this dissertation was to push the boundary of remote sensing data
processing to extract metrics relevant to mangrove health mapping, monitoring and
management in Darwin Harbour. The three main research questions and their
subsequent objectives are the core of the thesis. Their significance is discussed in the
following sections. The dissertation presents not only new outcomes for mapping
mangrove species and their biophysical variations but also an ecological risk
assessment tool to monitor mangrove ecosystems around Darwin Harbour.
275
7.1 Research question 1: How can high spatial resolution remote sensing data be
used to discriminate mangroves at species level?
The question was answered in two steps. First, the study separated mangroves
and non-mangroves using object based image analysis method. Then, the mangrove
coverage only was classified into species level. To my knowledge, Chapter 2 is the
first study of its kind to comprehensively compare the application of high resolution
satellite imagery with aerial photographs for mangrove mapping.
The study demonstrated that a large number of spectral bands with higher
spatial resolution (pan-sharpened WorldView-2 (WV2) image) were more accurate
than broad spectral bands (aerial photographs) within the blue, green, and red
regions, when discriminating mangroves from other features in an image. The spatial
relationship between mangroves and their surrounding environment (contextual
information) help to improve the mapping accuracy. When further classifying down
to species level, the highest accuracy (overall accuracy of 89%) was obtained from
the pan-sharpened WV2 image using five spectral bands within the visible range.
The pan-sharpened visible image covered the same spectral range with the same
spatial resolution as the aerial photographs. Therefore, the higher accuracy of the
former compared to the overall accuracy of 68% from resampled aerial photographs
is attributed to the increased number of narrow bands available for analysis, rather
than the total wavelength range. Compared to these results, however, there is no
considerable difference between the mangrove species map obtained from the pan-
sharpened Red, Red-Edge and NIR1 bands of WV2 image. Further, this study
demonstrated the significant increase in classification accuracy when using pan-
sharpened imagery, under the condition that spectral and radiometric integrity is
276
maintained using an appropriate algorithm such as high pass filter pan-sharpening
method.
This study also demonstrated a unique application of the support vector
machine algorithm for mangrove species mapping. While this advanced image
processing technique has previously been used in other environments, it is
particularly beneficial for mangroves because it efficiently deals with the dense,
heterogeneous nature of mangrove forests. However, spatial autocorrelation will
reduce the classification accuracy up to a certain level. The noise from non-leaf
surfaces such as tree branches and background can degrade the results of spectral
separability of mangroves.
7.2 Research question 2: How can 3D canopy dynamics combined with spectral
properties of mangroves be used to map mangrove biophysical variability?
The research question was answered in three steps. First, Chapter 3 tested
different remote sensing data sources and object based image analysis algorithms for
delineating mangrove individual tree crowns. A number of studies have already been
conducted to extract individual tree crowns for various vegetation types using
remotely sensed data with varying degree of success. Very little information is
available for mangrove forests, and already developed methods are more successful
with plantation forests having regular distances between trees rather than naturally
grown forests. Plantation forests show more similarities in size due to the similar age
of individuals as well. Unfortunately, similar sizes and regular distances between
mangroves trees are rare, and those methods cannot readily be applied to mangroves.
Therefore, Chapter 3 provided a significant contribution to this context. Following on
277
from this work, Chapter 4 mapped spatial variations of mangrove canopy
chlorophyll. There are a few studies that analysed mangrove biophysical variables
and their relation to various vegetation indices derived from satellite data, however,
to my knowledge, it is difficult to find a recorded study for mapping a spatial
distribution of mangrove chlorophyll. Chapter 5 evaluated Above Ground Biomass
(AGB) and Leaf Area Index (LAI) values over the Rapid Creek mangrove forest.
Although the digital cover photography method has already been tested with a few
homogeneous forest types, this is the first study that used this method to estimate
LAI over a heterogeneous forest like mangroves. Calculated LAI values were used as
in-situ calibration and validation data. The random forests regression algorithm and
partial least squares regression algorithm showed promising results when mapping
spatial variation of these biomarkers, especially when there is a weak linear
relationship between field samples and the satellite data.
In this study, to delineate individual tree crowns, WV2 imagery was used as
the primary data source, supplemented by an aerial photo derived Digital Surface
Model (DSM). The combination of a WV2 image, a DSM and an object based image
analysis was found to successfully delineate mangrove tree crowns. The achieved
overall relative accuracy was 92%. Minor errors were detected where neighbouring
trees with similar characteristics were aggregated together and conversely, where
multiple crowns were assigned to multiple trees (upward pointing branches of the
same tree identified as different trees). The visual appearances of tree crowns were
dramatically improved when incorporating the DSM. The inverse watershed method,
which has been successfully used in other forest environments with sparser canopies,
did not provide data representative of mangrove tree crowns. The combination of a
WV2 image, a DSM, and an object based image analysis approach is a relatively low
278
cost, accurate, and robust approach, which can be applied to heterogeneous forests
like mangroves.
To measure mangrove chlorophyll content, mangrove leaf samples were
collected and analysed in the laboratory. Then, canopy chlorophyll content was
calculated using LAI of sampled trees. The nonlinear random forests regression
algorithm was used to describe the relationship between canopy chlorophyll content
and remotely sensed data, and to estimate the spatial distribution of canopy
chlorophyll variation. The imagery was evaluated at full 2 m spatial resolution, as
well as at reduced resolutions of 5 m and 10 m. The root mean squared errors with
validation samples were 0.82, 0.64 and 0.65 g/m2 for maps at 2 m, 5 m and 10 m
spatial resolution respectively. The correlation coefficient was analysed for the
relationship between measured and predicted chlorophyll values. The highest
correlation (0.71) was observed at 5 m spatial resolution. Hence, this study
confirmed integrating remotely sensed data and field samples for mapping mangrove
canopy chlorophyll content over a large area. The random forests regression
algorithm showed moderate results for this application, while there is a weak linear
relationship between field samples and the satellite data.
Chapter 5 describes quantification of spatial variations of AGB and LAI in
the study site. Field measurements, vegetation indices derived from WV2 images,
and a partial least square regression algorithm were used to derive LAI maps and
AGB maps. The highest accuracies for LAI and AGB were obtained at 5 m spatial
resolution. The correlation coefficients between field samples and predicted maps
were 0.8 in both instances. Further, there was a strong linear correlation (0.7)
between LAI and AGB as found for other forest types. Hence, the study
279
demonstrated the possibility of assessing standing characteristics of mangroves using
remotely sensed data.
7.3 Research question 3: How can spatial information be combined to analyse
the environmental stresses on different mangrove sites?
This research question was answered in the Chapter 6. Regional ecological
risk assessment methods evaluate the interaction of three different regional spatial
components, which include: sources that release stressors; the habitat where the
receptors reside; and impacts to the assessment endpoints. This is fairly simple at a
single contaminated site with one stressor. However, the relative risk model (RRM)
evaluates multiple, dissimilar stressors simultaneously and comparatively at a
regional scale. The application of the RRM involves three major phases: problem
formulation, risk analysis, and risk characterization. During the problem formulation
phase, existing knowledge about the region was investigated. Experts in this context
were consulted. Then, a conceptual model was created combining all information.
Spatial data from topographic databases and statistics relating to various Darwin
Harbor monitoring projects were collected. In the second phase of the analysis,
ecological endpoints and corresponding model components including sources,
stressors and habitats were identified. The management goals of Darwin Harbor were
considered in determining the ecological assessment endpoints which were the
maintenance of mangrove health and the maintenance of water quality. The large
Darwin Harbor catchment boundary and boundaries of sub catchments were adapted
from the Australian Collaborative Land Use Mapping program (V5). However,
homogeneous small catchments were merged and formed 22 risk regions for further
analysis. Once model components were ranked according to the percentage coverage
280
of sub-regions (with respect to individual areal extent), the pathway of sources
habitats endpoints was created. Risk characterisation was the last phase of the
modelling process. The uncertainty of each model component was investigated and
classified into low, medium and high, and the discrete probability distribution was
introduced. The Monte Carle simulation was used to obtain all possible values of risk
scores. The total relative risk of the ecological assets in risk regions were analysed
and classified according to the Jenk’s optimization theory. Risk regions with a high
percentage cover of industrial, commercial, residential areas, diffuse entry points,
and climate change effects contributed to the highest risk to each ecological
endpoint.
In conclusion, successful development of advanced remote sensing methods
for mangrove mapping and monitoring is presented in this thesis. More precisely, an
object based image analysis method to discriminate mangrove and non-mangrove
areas, appropriately parameterised support vector machine algorithm for classifying
mangroves at species level, individual tree crown delineation, sub-pixel mapping
algorithms for analysing spatial variations of mangrove biophysical characteristics,
and a relative risk model for assessing ecological risk of Darwin Harbour mangroves
are the main outcomes of this thesis. Overall the study represents a significant
contribution to the current knowledge in the context of mangrove mapping and
monitoring.
7.4 Limitations and recommendations for future research
Chapters 2 – 6 addressed the limitations to methods used and results obtained
separately. This final section briefly addresses key areas that remain critically
281
challenging for mapping any wetland environment with broad band high spatial
resolution data.
7.4.1 Mapping with high spatial resolution image data
The difficulties with integrating field and image data within heterogeneous
environments should not be underestimated. In this study, errors due to mismatch
between image pixels and field samples were addressed by including additional
information collected in the field for mangrove tree identification. However, even
with 2 m pixel resolution, it is still difficult to confidently locate sampled trees in the
imagery. Work should commence to address the challenges of linking field sampled
mangrove tree positions with image data. For instance, for future studies, we would
recommend a transect method for field sampling for more successful results. Two
ends of transects should be established outside the mangrove forest with a highly
accurate positioning system. Then, all other measurements that based on the
established transect lines will be with greater accuracy. This is not an issue that is
specific to mangrove remote sensing, but is a critical area of research that should be
considered when mapping any forest environment.
When separating mangrove and non-mangroves and identifying individual
tree crowns, the spectral resolution of remotely sensed images played a more crucial
role than spatial resolution. Although this conclusion is specific to this study only,
the main reason would be the lack of integration of spatial features such as shape and
texture or colour of remotely sensed images for object based image analysis. Within
the shape criterion, the degree of smoothness of the object border and the
compactness of the objects can be altered. Having said that, one example is to
282
introduce the shape criterion from very high resolution aerial images together with
the digital surface model to improve the accuracy of the individual tree crown
delineation.
Another limitation of this study was the ability to transfer developed methods
to other locations around the world with different mangrove environment settings.
The results from mangrove species mapping using the support vector machine
algorithm were site specific – that is the algorithm was parameterised according to
the test site. As such, the framework for using this classification can be applied in
other areas, however specific parameters will need modifying. Every attempt was
made to create transferable variables within the object based image analysis
environment for separating mangroves from other features, and ultimately
delineating tree crowns. However modifications within the ruleset for other locations
and conditions are inevitable. Therefore, in order to apply these methods to vast and
different species-rich mangrove forests, further studies are required.
The recent advances in remote sensing imaging technology provide a wide
range of image datasets available both commercially and for research purposes.
Hence, the other limitation associated with this study was the investigation of limited
number of remotely sensed images; that is the main images used for this study over
the Rapid Creek mangrove forest were WorldView-2 satellite images and aerial
photographs obtained from the UltraCamD camera. When analysing individual tree
crowns, it was found that the higher spatial resolution such as 1 m or more would
provide higher accuracy in identifying individual shrub crowns, gaps between trees
and foliage clumping. Future work needs to investigate the utility of both high spatial
and spectral resolution datasets towards these aspects. For instance, when
283
considering towards the recent advances in imaging; Unmanned Aerial Vehicles
(UAV) would be an option. They can be used for optical, infrared and hyperspectral
imaging. They are incorporated with on-board high precise gyroscopes and GPS
receivers, and thus the stability of these systems is high; positioning is accurate.
Although UAV can use for rapidly and frequently surveying inaccessible areas and
obtaining high spatial resolution images, the spatial scale of the phenomena under
investigation is important.
The spatial scale of analysis (suitable spatial resolution of remote sensing
data) should match the scale of phenomena under investigation. Remote sensing data
with too many details may result in lower overall accuracy due to an increase in
within-class spectral variability. Although imaging using UAV’s open the path for
acquiring data with very high resolution even up to a spatial resolution of 1 cm,
future studies are necessary to understand the effect of scale variation in mangrove
environment. For instance, these very high spatial resolution images (1 cm to 20 cm)
allow analysing mangrove leaf structures and leaf orientation that enable managers
for sustainable mangrove conservation. However, as discovered by this study, they
may not provide better results in terms of mangrove canopy chlorophyll or above
ground biomass variations.
7.4.2 Relative Risk Modelling
According to this study, a relative risk model is a robust application that is
suitable for a large geographic area where multiple stressors are of concern. In this
application, data uncertainty is the major limitation associated with the model. Some
topographic data that we used for the analysis was more than 20 years old. They do
284
not correctly represent current situation of the Darwin Harbour, for example, there is
no correct information about recently cleared mangroves and recently populated
suburbs. An investigation of up-to-date data for relative risk modelling remains a
challenge. Further, rather than informal discussion, it is necessary to organise a series
of stakeholder meetings at the first phase of the modelling process for problem
formulation. This will reduce the effect of subjective opinions of experts regarding
interaction between stressors and their effect on ecological endpoints.
7.5 Contribution to Knowledge
This study signifies the operational use of advanced remote sensing
techniques such as object based image analysis methods and pixel based image
processing algorithms (support vector machine, random forest and partial least
squares regression) for mangrove mapping. The study also demonstrates the utility of
a regional ecological risk assessment method to assess the ecological risk associated
with mangroves. This model evaluated the interaction of sources that release
stressors, the habitat where the receptors reside and impacts to the assessment
endpoints for mangrove monitoring purposes.
The contributions of this thesis to the body of scientific knowledge are;
• An innovative method developed and tested to integrate satellite data, a
digital surface model derived from aerial photographs and an object based
image analysis techniques for separating mangroves and non-mangroves,
285
• The first ever effort to identify the most suitable remote sensing data source
with the support vector machine algorithm for mapping individual mangrove
species,
• Development of a method to isolate individual mangrove tree crowns or
groups of trees using remote sensing data and object based image analysis
techniques,
• The first assessment of spatial variability of mangrove canopy chlorophyll
over a broad area,
• The first ever application of digital photography method for calculating
mangrove leaf area index for calibration and validation purposes,
• Demonstration of the utility of remote sensing data and image processing
techniques for analysing spatial variability of mangrove above ground
biomass and leaf area index, and
• The first ever effort to develop a relative risk model to assess the ecological
risk associated with mangroves in the context of rural and urban
development.
In summary, this research presents unique applications of advanced remote
sensing methods for analysing up-to-date spatial information of mangrove forests,
and signifies the use of remotely sensed data against field observations. These
286
methods and the risk model can be used as a baseline for mangrove monitoring and
management.
287
Appendices
Appendix 1
Re: Requesting a formal written permission to re-use an aticle in a thesis Remote Sensing Editorial Office [[email protected]] Sent:Thursday, 5 November 2015 6:33 PM To: Muditha Kumari Heenkenda
Dear Mr./Ms. Heenkenda,
Thank you for your email. According to our copyright statement, MDPI
maintains Open Access publishing, which means that the authors hold the
copyright, readers can freely read, download articles published in MDPI
journals. So you can reuse the materials you published with us without
obtaining our permissions. You may find more details at
http://www.mdpi.com/about/openaccess
Wish you good luck with the thesis preparation and getting approval
smoothly.
Best Regards,
Xuanxuan Guan
Senior Assistant Editor
Remote Sensing Editorial Office
(http://www.mdpi.com/journal/remotesesing)
/Remote Sensing/ Impact Factor (2014), 3.180
Top Open Access Journal in Remote Sensing
Remote Sensing at Twitter: https://twitter.com/RemoteSens_MDPI
---
Ms. Xuanxuan Guan, M.Sc.
MDPI Branch Office, Beijing
Room 2207, Jincheng Center, No. 21 Cuijingbeili
Tongzhou District, Beijing 101101, China
Tel./Fax +86 10 8152 1170
E-Mail: xuanxu [email protected]
**************************************
MDPI AG
Klybeckstrasse 64, CH-4057 Basel, Switzerland
Tel. +41 61 683 77 34; Fax +41 61 302 89 18
E-mail: [email protected]
http://www.mdpi.com/journal/remotesensing/
**************************************
On 2015/11/6 0:34, Muditha Kumari Heenkenda wrote:
> /Remote Sensing/Editorial Office,
> MDPI AG, Klybeckstrasse 64, 4057 Basel,
> Switzerland.
>
>
> Dear Sir or Madam,
>
>> As a main author of the following paper, I would like to request a
> formal written permission from the publisher to incorporate this article
> in my thesis as PhD by publication. I would highly appreciate your
> response.
>
> /Heenkenda, M., Joyce, K., Maier, S. & Bartolo, R. 2014, Mangrove
> Species Identification: Comparing WorldView 2 with Aerial Photographs,
> Remote Sensing, vol. 6, 6064-6088, doi:10.3390/rs6076064.
> http://www.mdpi.com/2072-4292/6/7/6064/
>
>
>
> Thanks,
>
>
> Muditha K. Heenkenda
>
Appendix 2
Muditha K. Heenkenda,
Research Institute for the Environment and Livelihoods,
Charles Darwin University, Darwin, Australia
ASPRS grants you permission to reproduce the following ASPRS material in your
PhD dissertation Charles Darwin University, Darwin, Australia:
Heenkenda, M. K., Joyce, K. E. & Maier, S. W. 2015, Mangrove Tree Crown Delineation
from High Resolution Imagery, Photogrammetric Engineering and Remote Sensing, vol. 81,
471-479; doi: 10.14358/PERS.81.6.471. Please acknowledge the reprinted material with the following citation:
“Reproduced with permission from the American Society for Photogrammetry and
Remote Sensing, Bethesda, Maryland, www.asprs.org.”
If you require additional permissions or would like to use the material in a different format or
as part of another work, please let me know.
Sincerely,
Matthew
Austin Publications Production
Assistant (301) 493-0290 ext.
108 (301) 493-0208
(Fax)
Appendix 3
ELSEVIER LICENSE TERMS AND CONDITIONS
Nov 03, 2015
This is a License Agreement between Muditha Heenkenda ("You") and Elsevier ("Elsevier")
provided by Copyright Clearance Center ("CCC"). The license consists of your order details, the
terms and conditions provided by Elsevier, and the payment terms and conditions. All payments must be made in full to CCC. For payment instructions, please see
information listed at the bottom of this form. Supplier Elsevier Limited
The Boulevard,Langford Lane Kidlington,Oxford,OX5 1GB,UK
Registered Company 1982084 Number Customer name Muditha Heenkenda Customer address RIEL
Darwin, NT NT909 License number 3741190851314 License date Nov 03, 2015 Licensed content publisher Elsevier Licensed content publication ISPRS Journal of Photogrammetry and Remote Sensing Licensed content title Quantifying mangrove chlorophyll from high spatial resolution
imagery Licensed content author Muditha K. Heenkenda,Karen E. Joyce,Stefan W. Maier,Sytze de
Bruin Licensed content date October 2015 Licensed content volume 108 number Licensed content issue n/a number Number of pages 11 Start Page 234 End Page 244 Type of Use reuse in a thesis/dissertation Portion full article Format both print and electronic Are you the author of this Yes
Elsevier article? Will you be translating? No
Title of your A transition from traditional mangrove remote sensing to recent thesis/dissertation advances: Mapping and Monitoring Expected completion date Jan 2016 Estimated size (number of 300 pages) Elsevier VAT number GB 494 6272 12 Permissions price 0.00 CAD
VAT/Local Sales Tax 0.00 CAD / 0.00 GBP
Total 0.00 CAD
Terms and Conditions
INTRODUCTION 1. The publisher for this copyrighted material is Elsevier. By clicking "accept" in
connection with completing this licensing transaction, you agree that the
following terms and conditions apply to this transaction (along with the Billing and
Payment terms and conditions established by Copyright Clearance Center, Inc.
("CCC"), at the time that you opened your Rightslink account and that are
available at any time at http://myaccount.copyright.com).
GENERAL TERMS
2. Elsevier hereby grants you permission to reproduce the aforementioned material
subject to the terms and conditions indicated. 3. Acknowledgement: If any part of the material to be used (for example, figures)
has appeared in our publication with credit or acknowledgement to another source,
permission must also be sought from that source. If such permission is not
obtained then that material may not be included in your publication/copies.
Suitable acknowledgement to the source must be made, either as a footnote or in
a reference list at the end of your publication, as follows: "Reprinted from Publication title, Vol /edition number, Author(s), Title of article /
title of chapter, Pages No., Copyright (Year), with permission from Elsevier [OR
APPLICABLE SOCIETY COPYRIGHT OWNER]." Also Lancet special credit -
"Reprinted from The Lancet, Vol. number, Author(s), Title of article, Pages No.,
Copyright (Year), with permission from Elsevier." 4. Reproduction of this material is confined to the purpose and/or media for
which permission is hereby given. 5. Altering/Modifying Material: Not Permitted. However figures and illustrations may be
altered/adapted minimally to serve your work. Any other abbreviations, additions,
deletions and/or any other alterations shall be made only with prior written authorization
of Elsevier Ltd. (Please contact Elsevier at [email protected]) 6. If the permission fee for the requested use of our material is waived in this instance,
please be advised that your future requests for Elsevier materials may attract a fee.
7. Reservation of Rights: Publisher reserves all rights not specifically granted in
the combination of (i) the license details provided by you and accepted in the
course of this licensing transaction, (ii) these terms and conditions and (iii) CCC's
Billing and Payment terms and conditions. License Contingent Upon Payment: While you may exercise the rights licensed immediately upon issuance of the license at the end of the licensing process for the transaction, provided that you have disclosed complete and accurate details of your proposed use, no license is finally effective unless and until full payment is received from you (either by publisher or by CCC) as provided in CCC's Billing and Payment terms and conditions. If full payment is not received on a timely basis, then any license preliminarily granted shall be deemed automatically revoked and shall be void as if never granted. Further, in the event that you breach any of these terms and conditions or any of CCC's Billing and Payment terms and conditions, the license is automatically revoked and shall be void as if never granted. Use of materials as described in a revoked license, as well as any use of the materials beyond the scope of an unrevoked license, may constitute copyright infringement and publisher reserves the right to take any and all action to
protect its copyright in the materials. 9. Warranties: Publisher makes no representations or warranties with respect to the
licensed material. 10. Indemnity: You hereby indemnify and agree to hold harmless publisher and
CCC, and their respective officers, directors, employees and agents, from and
against any and all claims arising out of your use of the licensed material other
than as specifically authorized pursuant to this license. 11. No Transfer of License: This license is personal to you and may not be sublicensed,
assigned, or transferred by you to any other person without publisher's written permission.
12. No Amendment Except in Writing: This license may not be amended except in a writing
signed by both parties (or, in the case of publisher, by CCC on publisher's behalf). 13. Objection to Contrary Terms: Publisher hereby objects to any terms contained in any
purchase order, acknowledgment, check endorsement or other writing prepared by you,
which terms are inconsistent with these terms and conditions or CCC's Billing and
Payment terms and conditions. These terms and conditions, together with CCC's Billing
and Payment terms and conditions (which are incorporated herein), comprise the entire
agreement between you and publisher (and CCC) concerning this licensing transaction.
In the event of any conflict between your obligations established by these terms and
conditions and those established by CCC's Billing and Payment terms and conditions,
these terms and conditions shall control. 14. Revocation: Elsevier or Copyright Clearance Center may deny the permissions described in
this License at their sole discretion, for any reason or no reason, with a full refund payable to
you. Notice of such denial will be made using the contact information provided by you. Failure to receive such notice will not alter or invalidate the denial. In no event will Elsevier or
Copyright Clearance Center be responsible or liable for any costs, expenses or damage
incurred by you as a result of a denial of your permission request, other than a refund of the
amount(s) paid by you to Elsevier and/or Copyright Clearance Center for denied permissions.
LIMITED LICENSE The following terms and conditions apply only to specific license types: 15. Translation: This permission is granted for non-exclusive world English rights
only unless your license was granted for translation rights. If you licensed
translation rights you may only translate this content into the languages you
requested. A professional translator must perform all translations and reproduce
the content word for word preserving the integrity of the article. 16. Posting licensed content on any Website: The following terms and conditions
apply as follows: Licensing material from an Elsevier journal: All content posted to
the web site must maintain the copyright information line on the bottom of each
image; A hyper-text must be included to the Homepage of the journal from which you
are licensing at http://www.sciencedirect.com/science/journal/xxxxx or the Elsevier
homepage for books at http://www.elsevier.com; Central Storage: This license does
not include permission for a scanned version of the material to be stored in a central
repository such as that provided by Heron/XanEdu. Licensing material from an Elsevier book: A hyper-text link must be included to the
Elsevier homepage at http://www.elsevier.com . All content posted to the web site
must maintain the copyright information line on the bottom of each image.
Posting licensed content on Electronic reserve: In addition to the above the following clauses
are applicable: The web site must be password-protected and made available only to bona fide
students registered on a relevant course. This permission is granted for 1 year only.
You may obtain a new license for future website posting. 17. For journal authors: the following clauses are applicable in addition to the above: Preprints: A preprint is an author's own write-up of research results and analysis, it has not
been peer-reviewed, nor has it had any other value added to it by a publisher (such
as formatting, copyright, technical enhancement etc.). Authors can share their preprints anywhere at any time. Preprints should not be
added to or enhanced in any way in order to appear more like, or to substitute for,
the final versions of articles however authors can update their preprints on arXiv or
RePEc with their Accepted Author Manuscript (see below). If accepted for publication, we encourage authors to link from the preprint to their
formal publication via its DOI. Millions of researchers have access to the formal
publications on ScienceDirect, and so links will help users to find, access, cite and
use the best available version. Please note that Cell Press, The Lancet and some
society-owned have different preprint policies. Information on these policies is
available on the journal homepage. Accepted Author Manuscripts: An accepted
author manuscript is the manuscript of an article that has been accepted for
publication and which typically includes author-incorporated changes suggested
during submission, peer review and editor-author communications. Authors can share their accepted author manuscript:
- immediately via their non-commercial person homepage or blog by updating a preprint in arXiv or RePEc with the accepted manuscript via their research institute or institutional repository for internal institutional uses
or as part of an invitation-only research collaboration work-group directly by
providing copies to their students or to research collaborators for
their personal use for private scholarly sharing as part of an invitation-only work group
on commercial sites with which Elsevier has an agreement - after the embargo period
via non-commercial hosting platforms such as their institutional
repository via commercial sites with which Elsevier has an agreement In all cases accepted manuscripts should:
- link to the formal publication via its DOI - bear a CC-BY-NC-ND license - this is easy to do - if aggregated with other manuscripts, for example in a repository or other site,
be shared in alignment with our hosting policy not be added to or enhanced in
any way to appear more like, or to substitute for, the published journal article. Published journal article (JPA): A published journal article (PJA) is the definitive
final record of published research that appears or will appear in the journal and
embodies all value-adding publishing activities including peer review co-ordination,
copy-editing, formatting, (if relevant) pagination and online enrichment. Policies for sharing publishing journal articles differ for subscription and gold open
access articles: Subscription Articles: If you are an author, please share a link to your article rather than the
full-text. Millions of researchers have access to the formal publications on ScienceDirect, and so
links will help your users to find, access, cite, and use the best available version.
Theses and dissertations which contain embedded PJAs as part of the formal
submission can be posted publicly by the awarding institution with DOI links back to
the formal publications on ScienceDirect. If you are affiliated with a library that subscribes to ScienceDirect you have additional
private sharing rights for others' research accessed under that agreement. This includes
use for classroom teaching and internal training at the institution (including use in course
packs and courseware programs), and inclusion of the article for grant funding purposes. Gold Open Access Articles: May be shared according to the author-selected
end-user license and should contain a CrossMark logo, the end user license, and
a DOI link to the formal publication on ScienceDirect. Please refer to Elsevier's posting policy for further information. 18. For book authors the following clauses are applicable in addition to the above: Authors are
permitted to place a brief summary of their work online only. You are not allowed to download
and post the published electronic version of your chapter, nor may you scan the printed edition to
create an electronic version. Posting to a repository: Authors are permitted to post a summary
of their chapter only in their institution's repository. 19. Thesis/Dissertation: If your license is for use in a thesis/dissertation your thesis
may be submitted to your institution in either print or electronic form. Should your thesis
be published commercially, please reapply for permission. These requirements include
permission for the Library and Archives of Canada to supply single copies, on demand,
of the complete thesis and include permission for Proquest/UMI to supply single copies,
on demand, of the complete thesis. Should your thesis be published commercially,
please reapply for permission. Theses and dissertations which contain embedded PJAs
as part of the formal submission can be posted publicly by the awarding institution with
DOI links back to the formal publications on ScienceDirect.
Elsevier Open Access Terms and Conditions You can publish open access with Elsevier in hundreds of open access journals or in nearly
2000 established subscription journals that support open access publishing. Permitted third
party re-use of these open access articles is defined by the author's choice of Creative
Commons user license. See our open access license policy for more information. Terms & Conditions applicable to all Open Access articles published with Elsevier: Any reuse of the article must not represent the author as endorsing the adaptation of the
article nor should the article be modified in such a way as to damage the author's honour or
reputation. If any changes have been made, such changes must be clearly indicated. The author(s) must be appropriately credited and we ask that you include the
end user license and a DOI link to the formal publication on ScienceDirect. If any part of the material to be used (for example, figures) has appeared in our publication with
credit or acknowledgement to another source it is the responsibility of the user to ensure their
reuse complies with the terms and conditions determined by the rights holder. Additional Terms & Conditions applicable to each Creative Commons user license:
CC BY: The CC-BY license allows users to copy, to create extracts, abstracts and new
works from the Article, to alter and revise the Article and to make commercial use of the
Article (including reuse and/or resale of the Article by commercial entities), provided the
user gives appropriate credit (with a link to the formal publication through the relevant
DOI), provides a link to the license, indicates if changes were made and the licensor is not
represented as endorsing the use made of the work. The full details of the license are
available at http://creativecommons.org/licenses/by/4.0. CC BY NC SA: The CC BY-NC-SA license allows users to copy, to create extracts,
abstracts and new works from the Article, to alter and revise the Article, provided this is not
done for commercial purposes, and that the user gives appropriate credit (with a link to the
formal publication through the relevant DOI), provides a link to the license, indicates if changes
were made and the licensor is not represented as endorsing the use made of the work.
Further, any new works must be made available on the same conditions. The full details of the
license are available at http://creativecommons.org/licenses/by-nc-sa/4.0. CC BY NC ND: The CC BY-NC-ND license allows users to copy and distribute the Article,
provided this is not done for commercial purposes and further does not permit distribution of
the Article if it is changed or edited in any way, and provided the user gives appropriate
credit (with a link to the formal publication through the relevant DOI), provides a link to the
license, and that the licensor is not represented as endorsing the use made of the work. The
full details of the license are available at http://creativecommons.org/licenses/by-nc-nd/4.0.
Any commercial reuse of Open Access articles published with a CC BY NC SA or CC BY NC
ND license requires permission from Elsevier and will be subject to a fee. Commercial reuse includes:
- Associating advertising with the full text of the Article - Charging fees for document delivery or access - Article aggregation - Systematic distribution via e-mail lists or share buttons
Posting or linking by commercial companies for use by customers of those companies.
20. Other Conditions:
v1.8 Questions? [email protected] or +1-855-239-3415 (toll free in the US) or
+1-978-646-2777.
Appendix 4
Permissions
T & F Reference Number: P111615-02 11/16/2015 Muditha Heenkenda Ellongewen Drive Casuarina Campus NT909 Australia [email protected] Dear Sir or Madame, We are in receipt of your request to reproduce your article Muditha K. Heenkenda & Renee Bartolo Regional Ecological Risk Assessment Using a Relative Risk Model: A Case Study of the Darwin Harbor, Darwin, Australia Human and Ecological Risk Assessment: An International Journal (Online) DOI: 10.1080/10807039.2015.1078225 For use in your dissertation and to post to your institutional repository You retain the right as author to post your Accepted Manuscript on your departmental or personal website with the following acknowledgment: “This is an Accepted Manuscript of an article published in Human and Ecological Risk Assessment: An International Journal online [August 20, 2015], available online: http://www.tandfonline.com/doi/full/10.1080/10807039.2015.1078225 This permission is all for print and electronic editions. For the posting of the full article it must be in a secure, password-protected intranet site only. An embargo period of twelve months applies for the Accepted Manuscript to be posted to an institutional or subject repository. We will be pleased to grant you permission free of charge on the condition that: This permission is for non-exclusive English world rights. This permission does not cover any third party copyrighted work which may appear in the material requested. Full acknowledgment must be included showing article title, author, and full Journal title, reprinted by permission of Taylor & Francis LLC (http://www.tandfonline.com). Thank you very much for your interest in Taylor & Francis publications. Should you have any questions or require further assistance, please feel free to contact me directly.
Sincerely, Mary Ann Muller Permissions Coordinator Telephone: 215.606.4334 E-mail: [email protected]
530 Walnut Street, Suite 850, Philadelphia, PA 19106
Phone: 215-625-8900
Fax: 215-207-0050 Web:
www.tandfonline.com