Validation of Spatially Continuous EDEN Water-Surface Model for the

Everglades, Florida withEcosystem Applications

Zhongwei Liu, Ph.D.1*

Aaron Higer1

Frank Mazzotti, Ph.D.1

Leonard Pearlstine, Ph.D.2

1. Ft. Lauderdale Research & Education Center, University of Florida2. Everglades and Dry Tortugas National Parks

* Liuz@ufl.eduAAG annual conference

Boston, MA, 4/2008

2

OutlineOutline Introduction

– EDEN network – EDEN water-surface model

Methodology – Study area– Data collection– Analysis methods

Results

Ecosystem Applications

Conclusions and Discussions

3

I. IntroductionI. Introduction Efforts to measure and

link surface-water depths to biotic communities in the Everglades (Loveless, 1959; Craighead, 1971; McPherson, 1973; Cohen, 1984; Newman et al., 1996; Busch et al., 1998; Ross et al., 2000; Hendrix and Loftus, 2000; Gawlik, 2002; Chick et al., 2004; Palmer and Mazzotti, 2004; Trexler et al., 2005).

Hydrologic models could provide spatially continuous hydrologic information.

Two research objectives: model validation and applications

alligator holes

Source: www.broward.edu.

sawgrass marsh

sloughalligator holes

tree islandswet

prairiewet prairie

4

Everglades Depth Estimation Network ((EDEN)

Integrated network of real-time water level monitoring, ground elevation modeling, and water-surface modeling

Daily water level/stage data (2000 - present) from 253 gage stations

Two types of gage stations: marsh stations and canal stations

a) satellite gageGulf of Mexico

Atlantic Ocean

Florida

Area of Enlargement

A marsh gage station

XWXW

XW

XW

XW

XW

XW

XWXW

XW

XW

XW

XW

XW

XWXW XW

XW

XWXWXW

XWXW

XW XWXW

XW

XW XW

XW

XW

XW XW

XWXW

XW XWXW XW XW XW

XWXW

XW XWXW XW

XWXW

XW

XW XW

XWXW

XW XW

XWXW

XW

XWXWXW

XW XWXWXW

XW

XWXW

XW

XW

XW

XW

XW

XWXW

XWXW

XWXW

XW XW

XW

XW XW

XWXWXWXW

XWXW

XW

XWXW XWXW

XW

!!! !

!

!

!!

!

!

!

!

!

!

!

!

!!

! !

!

!

!

!

!!!!

!

!

!

!

!!

!

!!

!

!!

!

!

! !!!! !

!!

!

!

!!

!

!

!!

!

!

!

!!!

!!

! !

!

!

!!!

!!

!!

!!

!!

!

!

!

!!

!!

!

!

!! !

# ##

## ###

##

###

#

##

#

#

#

#

#

##

#

#

###

#

#

#

# #

#

##

#

#

#

#

##

# #

##

#

#

#

!(

!( !(

!(

!(

!(

!(

!(

!(

!(

!(

!(!(!(

WCA 1

WCA 2A

WCA 3AN

WCA 3AS

WCA 3B

WCA 2B

ENP

BCNP

±0 10 20 305

Kilometers

EDEN Water Gagesby Operating Agency

!( BCNP

# ENP

! SFWMD

XW USGS

Canals and Rivers

Study Area

Eden Boundary

South Florida

5

EDEN Water-Surface ModelEDEN Water-Surface Model Developed by Pearlstine et

al. (2007)

Spatial interpolation of 240 gage stations in ArcGIS: radial basis function (RBF)

Model outputs– Water surface of 400m x 400 m grid

spacing, to match the EDEN ground digital elevation model (DEM)

– Water depth = water surface - DEM

150

120

30

300 270

150

120

210

LegendContour

Water surface on 7/10/2007 (cm)High : 457.7

Low : 1.4

±0 10 20 305

Kilometers

6

II. II. MethodologyMethodology Study area

– WCAs 3A South and 3B

– Florida Department of Environmental Protection (FDEP) benchmark network

"Îi

"Îi

"Îi

"Îi

"Îi

"Îi"Îi"Îi

"Îi

"Îi

"Îi

"Îi

"Îi

"Îi

"Îi

"Îi "Îi "Îi "Îi

"Îi"Îi

"Îi

"Îi"Îi "Îi"Îi

"Îi"Îi

"Îi"Îi

"Îi

"Îi"Îi

"Îi"Îi"Îi"Îi

"Îi"Îi

"Îi

"Îi

"Îi

"Îi

"Îi

"Îi

"Îi

"Îi"Îi

"Îi "Îi

"Îi"Îi

"Îi"Îi

!( !( !(

!(

!(!( !(

!(!(

!( !(

!(!( !(

!( !(!(

!( !( !(!(!(!(

!(!(!(!( !( !(!(

WCA 3A South

WCA 3B

28 W5

W2

W18

W15

W14W11

TI-9TI-8

SRS1

3AS+

3A9+

3A-5S9A-T

3ASW+

S344+TS344+H

S340+TS340+H

S336+TS336+H

S335-TS335-H

S334-TS334-H

S333+TS333+H

S31M-H

S151+TS151+H

L28S2+

EDEN_8

EDEN_7

EDEN_5

EDEN_4

3B-SE+

SITE_76

SITE_71

SITE_69

SITE_65

SITE_64

S343B+T

S343A+T

S337M-T

S12D_UPS12C_UPS12B_UPS12A_UP

EDEN_14

EDEN_12

EDEN_103BS1W1+

3AS3W1+

9

87

6

5

43

2

31

302927 26

25

24

2322

212019

18

1716

1514

13

1211

10

0 4 8 122Kilometers

±Legend!( Benchmark Sites

"Îi Water Level Gages

Study Area

I-75 (Alligator Alley)

Tamiami TrailTamiami Trail

7

Data CollectionData Collection

Field water-level data collected by Florida Atlantic University (FAU)

At 24 benchmarks

Apr. - Sept., 2007

Via airboat and helicopter

Totalobs.: 91

Southern WCA 3A75

WCA 3B8

Dry season16

Wet season75

8

Analysis MethodsAnalysis Methods GIS

– Spatial analysis

Error statistics– MAE (Mean Absolute Error)=

– MBE (Mean Biased Error)=

– RMSE (Root Mean Squared Error)=

Nonparametric statistical methods– Spearman’s rank correlation– Wilcoxon signed rank test– Kruskal-Wallis nonparametric analysis of variance (ANOVA)

N

iii OP

N 1

1

N

iii OP

N 1

1

2/12

1

1

N

iii OP

N

9

III. ResultsIII. Results DIFF_STAGE =

predicted water stage – observed water stage.

Underestimates and overestimates are represented by negative and positive values, respectively.

"Îi

"Îi

"Îi

"Îi

"Îi

"Îi"Îi"Îi

"Îi

"Îi

"Îi

"Îi

"Îi

"Îi

"Îi

"Îi "Îi "Îi "Îi

"Îi"Îi

"Îi

"Îi"Îi "Îi"Îi

"Îi"Îi

"Îi"Îi

"Îi

"Îi"Îi

"Îi"Îi"Îi"Îi

"Îi"Îi

"Îi

"Îi

"Îi

"Îi

"Îi

"Îi

"Îi

"Îi"Îi

"Îi "Îi

"Îi"Îi

"Îi"Îi!(

!(

!(

!(

!(!(!(

!(

!( !(!( !(

!(

!(

!(!(

!(!(

!(

!(

!(

!(!(

!(!(

!(!(

!(

!(

!(

!( !(!(!(!( !(!(!(!( !(!(

!(

!(!(!(

!(!(!(!(!(

!(!(

!(

!(

!(

!(

!(

!(

!(

!(

!(!(

!(!(!(

!(

!(!(!(

!(

!(

!(

!(

!(

!(

!(

!(

!(

!(!(!(

!(

!(

!(!(

!(

!(

!(!(

!(

!(

WCA 3A South

WCA 3B

I-75 (Alligator Alley)

Tamiami Canal

L-30

Can

al

L-28

Can

al-5.1

-11.1

-3.0

1.7

1.1

3.0-1.5-0.6

-1.7

-0.9

-6.3

-1.5

-0.8-3.2

-1.5-3.4

-0.1-2.8

-0.2

-0.5-0.6

-17.6

1.7

1.7

0.4

3.4

0.52.6

1.23.5

2.13.0

0.20.4

3.64.0

4.34.9

0 4 8 122Kilometers

± LegendDIFF_STAGE (cm)

!( -4.9 - 0.0

!( -9.9 - -5.0

!( -17.6 - -10.0

DIFF_STAGE (cm)!( 0.2 - 4.9

"Îi Water Level Gages

Canals

Study Area

10

Results (cont.)Results (cont.) Major statistics of interpolation

errors for water stageType N MIN

(cm) MAX (cm)

Standard deviation

Standard error b

MAE Mean

(MBE) RMSE

WCA 3A South, 3B 91 -17.6 4.9 3.32 0.35 2.38 -0.08 3.30 WCA 3A South 83 -5.1 4.9 2.47 0.27 2.11 0.32 2.48 WCA 3B 8 -17.6 1.7 6.97 2.46 5.15 -4.2 7.76

a Interpolation error (water-stage difference) = predicted water stage – observed water stage. b Standard error = standard deviation / N .

11

Spearman’s rank correlation analysis

Spearman’s rank correlation analysis for temporally detrended water stage data (Rangel et al., 2006)

Type Spearman’s rank

correlation coefficient

Corrected df a Corrected p-value

WCA 3A South, 3B 0.909 17.29 <0.001 WCA 3A South 0.882 5.542 <0.001 WCA 3B 0.833 5.277 <0.004 a Corrected degrees of freedom.

Type Variable Normality test

p-value (Shapiro-Wilk)

Spearman’s rank correlation

coefficient (rs)

P-value of rs

Observed_Stage <0.0001 WCA 3A South, 3B Predicted_Stage <0.0001 0.98 <0.0001

Observed_Stage <0.0001 WCA 3A South Predicted_Stage <0.0001 0.98 <0.0001

Observed_Stage 0.7133 WCA 3B a Predicted_Stage 0.2472 0.83 0.0102 a For WCA 3B, the parametric Pearson’s correlation coefficient is 0.8998 (p = 0.0023).

12

Kruskal-Wallis nonparametric ANOVA

Wilcoxon’s signed rank tests Type Variable

Normality test p-value

(Shapiro-Wilk)

Wilcoxon’s signed rank test statistic

P-value (Wilcoxon)

WCA 3A South, 3B Difference_Stage a <0.0001 166.5 0.5129 WCA 3A South b Difference_Stage 0.10 263.0 0.2346 WCA 3B c Difference_Stage 0.079 -8.0 0.2969

a Difference_Stage = predicted water stage - observed water stage. b For WCA 3A South, the test statistic of pairwised t-test is -1.19 (p = 0.2371). c For WCA 3B, the test statistic of pairwised t-test is 1.7 (p = 0.1321).

a Median difference = the median of water stage differences (predicted – observed).

Source Class N Median

Difference a (cm)

df Kruskal-Wallis test statistic (H) P-value of H

Dry (November – May) 16 -2.15 Season Wet (June – October) 75 0.6 1 7.0428 0.008

WCA 3A South 83 0.4 Region WCA 3B 8 -1.2 1 3.4761 0.0623

3: Sawgrass 72 -0.2 4: Upland 13 3.6 Vegetation 5: Exotics and Cattail 6 1.55

2 12.4469 0.002

13

IV. EDEN Water-Surface Model Applications

Estimation of Ground Elevation:

Ground elevation =predicted water

stage -observed water

depth.

14

Estimation of Water-Depth Hydrographs:

Water depth =predicted water stage -derived ground

elevation.

15

V. Conclusions and V. Conclusions and DiscussionsDiscussions

We found there are no statistically significant differences between model-predicted and field-observed water-stage data (p-value = 0.5129). Overall, the model is reliable by a RMSE of 3.3 cm. By region, the RMSE is 2.48 cm and 7.76 cm in WCAs 3A and 3B, respectively.

Two applications of the validated EDEN water-surface model to investigate the relationship between water depth and fresh water marsh habitats.

16

DiscussionsDiscussions Boundary problems

–When a benchmark reports high interpolation errors, it is likely to be near boundaries.

Data collection

Missing gage data–Localized impacts on the water surface

The method of estimating water depth–More accurate–More cost-effective

17

Further StudiesFurther Studies More field observations of dry and wet

seasons, and in other areas

Examine WCA 3B to improve the model

Rainfall data

A better regional ground DEM

Other interpolation techniques

18

AcknowledgementsAcknowledgements Dr. John Volin from University of Connecticut (formerly

FAU) and Dr. Dianne Owen and Jenny Allen from FAU

Pamela A. Telis (USGS/USACE Liaison), Dr. John Jones, Paul Conrads, Heather Henkel, and Michael Holmes from USGS, and Roy Sonenshein from ENP

Elmar Kurzbach, U.S. Army Corps of Engineers (USACE), and Dr. Ronnie Best, USGS Priority Ecosystem Science

Dr. Laura Brandt, Kevin Chartier, Adam Daugherty, and Wingrove Duverney, Joint Ecosystems Modeling (JEM)

Others…