Upload
fao
View
326
Download
0
Embed Size (px)
Citation preview
Soil class = abstract = the name from
which we retrieve various
Characteristics related with the dynamics
(water and nutrients) of the body of soil
Point
Soil Spectroscopy to monitor the state of soil
resources in the present and in the future
José Alexandre M. Demattê University of São Paulo, Brazil
facilitate
Necessity: Food and Water
Precision agriculture
*Land use and Environment planning
*Soil and water Conservation
*Choose adequate areas for residue
disposal
*Allocation of the adequate plant in
the adequate soil with higher
Productivity
Can spectroscopy reach all this? Why?
Soil Attributes maps
Soil Class maps
Oxisol, Ultisol…
From one information =
several information
0,00
0,02
0,04
0,06
0,08
0,10
0,12
0,14
0,16
400 900 1400 1900 2400 2900
Refle
ctân
cia
Comprimento de onda (nm)
LR
Monitor soil quality
Soil hydrocarbons monitoring in petroleum installation by spectroscopy and hyperspectral image
Brazilian, Remote Sensing Simposium 2013, Foz do Iguacu, Brazil.
Pabon et al. (2013)
Geomorphological analyses coupled with Vis–NIR spectroscopy and geostatistics allowed to compare the
spatial distribution of SOM and water erosion processes in the Turbolo catchment.
2100 2150 2200 2250 2300 2350 2400 2450
Refle
câtn
cia -
contin
uo r
em
ovi
do
0.75
0.80
0.85
0.90
0.95
1.00
Comprimento de onda (nm)
2100 2150 2200 2250 2300 2350 2400 2450
0.88
0.90
0.92
0.94
0.96
0.98
1.00
2100 2150 2200 2250 2300 2350 2400 2450
0.86
0.88
0.90
0.92
0.94
0.96
0.98
1.00
K
Mg
Soil Chemistry: potassium
detection
Mineralogy
range
Ramirez-Lopes & Demattê (EJSS, in submission 2014)
HO
HO
Al+++
Al+++
Al+++
K+
K+
K+
K+
K+
K+ K+
K+
K+
R2 = 0,72
Soil chemical management: liming application
y = 1,0101x + 0,1685R² = 0,952
RMSE = 0,46
0
2
4
6
8
10
0 2 4 6 8 10
NC
est
imad
a (
t h
a-1
)
NC determinada (t ha-1)Determined Lime Necessity
Est
imate
d L
ime N
eces
sity
SLAUGHTERHOUSE
Blood goes to a drain
Wrong way
Soil chemical alteration due to slaughterhouse waste application as identified by spectral reflectance: an environmental
useful tool, J.A.M. Demattê, L.Ramirez Lopez, S.A. Araujo, F.S. Terra, J.C. Oliveirad, S.M.F. Maia, F.F.C. Mello
Submmission Geoderma, 2014
We have to monitor this!!
Soil samples were incubated with different doses of blood, and taken their spectra.
Sodium: altered from 0.8 to 9.8 mmolc dm-3
Different
concentration
Wavelength, nm
500 750 1000 1250 1500 1750 2000 2250
Refle
ctan
ce
0.0
0.1
0.2
0.3
0.4
0.5
33 mg kg-1
6729 mg kg-1
Wavelength, cm-1
40080012001600200024002800320036004000
Ref
lect
ance
0
2
4
6
8
10
12
68 mg kg-1
207 mg kg-1
1990 mg kg-1
6068 mg kg-1
Investigating Cr in soils applied with tannery sludge by Vis-Nir
and Mid-IR spectroscopy techniquesAraujo, S.; Vicente, S.; Demattê, J.A.M. (2014, submission Remote Sensing Environment)
0
600
1200
1800
2400
3000
3600
0 600 1200 1800 2400 3000 3600
Rv2 = 0.85
RMSEv= 324Rm
2 = 0.74
RMSEm= 408
Laboratory soil analysis Vis-NIR MID-IR
Determing soil properties in Amazonian Dark Earths by vis-nir-mid spectroscopy
S.R. Araújo, J.A.M. Demattê, M. Söderström, J.Eriksson, C.Isendahl – In submission EJSC
Exploratory (1:1.000.000)
Recognaissence (1:500.000)
Semidetailed (1:100.000)
Detailed (1:20.000) Most important for Agriculture
0,25%
mapped
500 samples
Areia % - Mapa digital
0 20 40 60 80 100
Are
ia %
- M
apa e
spectr
o-d
igital
0
20
40
60
80
100
Argila % - Mapa digital
0 20 40 60 80
Arg
ila %
- M
apa e
spectr
o-d
igital
0
20
40
60
80R
2
Ajust.= 0.96
RMSE = 3.26ME = -0.22m = 0.87Intercesção = 3.03
R2
Ajust.= 0.97
RMSE = 4.38ME = -0.04m = 0.86Intercesção = 10.03
A. B.
f = 8.06 + 281719.27 (1+e-(x+60.03)/-6.36)-1
R2Adj.= 0.85
Tamanho do grupo de amostras de calibração
0 100 200 300
SD
E -
SB
6
8
10
12
14
16
18
20
22
95% Intervalo de confiança
f = 1.88 + 0.02(1-e-0.017x)-2.37-0.017x)-2.37
R2Adj.= 0.98
Tamanho do grupo de amostras de calibração
0 100 200 300
SD
E -
Mg
1
2
3
4
5
6
7
95% Intervalo de confiança
f = 5.99 +1207029.3(1+e-(x+284.7)/-23.2)-1
R2Adj.= 0.81
Tamanho do grupo de amostras de calibração
0 100 200 300
SD
E -
Ca
5
6
7
8
9
10
11
12
95% Intervalo de confiança
f = 6.80 + 2.08 (1+e(-(x+77.53)/-24.34))-1
R2Adj.= 0.85
Tamanho do grupo de amostras de calibração
0 100 200 300
SD
E -
Are
ia
6.0
6.5
7.0
7.5
8.0
8.5
9.0
9.5
95% Intervalo de confiança
f = 5.86 + 1.07 (1+e(-(x+78.32)/-7.12))-1
R2Adj.= 0.85
Tamanho do grupo de amostras de calibração
0 100 200 300
SD
E -
Arg
ila
5.0
5.5
6.0
6.5
7.0
7.5
95% Intervalo de confiança
A B
DC
E
Detailed map
Lab: 500
Sensor: 166
Spatial attributes vs Optimal number of soil samples
6
16
26
36
46
56
66
10
30
50
70
90
N
0 1000 2000 3000 m
Argila, %
Areia, %
A. B.
C. D.
Sand
Clay
Observed Predicted
Ramirez-Lopes , Demattê, J.A.M., Terra, F.S. (2013- in submission Geoderma)
Wavelength, nm
Ref
lect
ance
fac
tor
Quantitave texture determination by spectral reflectance in Brazilian soils
for soil management M.H.D. Franceschini, M.V. Sato, J.A.M. Demattê, L.E.Vicente, C.R. Grego
In submission Brazilian Journal of Soil Science, 2013
Very clayey (>60% clay)
Clayey (35-60% clay)
Medium Clay (25-35% clay)
Very Sandy (<10% clay)
Sandy (10-15% clay)
Medium Sandy (15-25% clay)
- From top flat toward first shoulder
- Similar shape “in depth”)
low reflectance (rectilinear and flat ↑ weathering
-
Hematite
Hematite
Goethite
Crystalline
Crystalline
Crystalline
J.A. M. Demattê; F.S. Terra. Spectral Pedology: a new perspective on evaluation of soils along pedogenetic alterations. Geoder. 2014
- Shallowest unit (hillside): poor drainage
- Different intensities
- - Higher content of weathering minerals
Hematite
Soil profile spectral transformation along the toposequence (≠ pedogenesis)
- Foothill and second less steep hillside: restricted drainage (lower landscape position)
- Typic Argiudoll: argillic horizon (lessivage) and influence of Diabase
- Typic Hapludoll: feature at 1,900 nm larger than others (2:1 mineral predominance)
Hematite
Goethite
More crystalline
Less crystalline
Soil profile spectral transformation along the toposequence (≠ pedogenesis)
- 1st Terrain (flat depression): restricted drainage and accumulation of materials
- Slow weathering (large contribution of montmorillonite and vermiculite)
Intermediate crystallinity
Goethite
0,0
0,1
0,2
0,3
0,4
1 2 3 4 5 6Ban ds
Re
fle
cta
nc
e f
ac
tor
b.
different intensit ies
variat ions on the sp ect ral curve
tendency due to soil p rop erties
(mainly granulometric
and iron content differences)
AQ
PV
LE
PE
LR
T R
7
0 ,0
0 ,1
0 ,2
0 ,3
0 ,4
3 5 0 6 5 0 95 0 12 5 0 1 5 5 0 1 8 5 0 2 1 5 0 2 4 5 0
W av elen gth , n m
Re
fle
cta
nc
e f
ac
tor
p os it ive t endency
horiz ontal to negat ive t endency
a.
variat ion in the content and forms of iron
AQ
PVLE
PE
TR
LR
0 ,0
0 ,1
0 ,2
0 ,3
0 ,4
1 2 3 4 5 6
B a n d s
Re
fle
cta
nc
e
c .
c la y e y s o ils :
lo w e r in t e n s ity
s a n d y s o ils :
h ig h e r in t e n s ity A Q
P V
L i
L R
T R
P L
P E
L E
L V
7
Laboratory
vs Landsat
Demattê et al. (2009)
0.67Clay
0.72Fe2O3
0.9
0.95
clay
FeO2
Nanni and Demattê (2006) SSSAJ
Calibration of models
SPATIAL PREDICT OF CLASSES
THROUGH THE FUSION OF
INFORMATION
Elevation
Mean curvature
Inclination
etc...
TM Images
(Inf. superficial)
Band 1
Band 2
Band 3
...
Band 7
Fe2O3
TiO2
Ki
etc...
Parameter extraction
of relief by MDE
Coupling spectrally predicted soil weathering attributes, terrain parameters and satellite data for predictive mapping of tropical soils
J.A.M. Demattê, L.Ramírez-López, P.R. Fiorio. (Submitted Geoderma 2013)
Traditional map Predicted map
69% global accurracy
Demattê, J.A.M., Ramires-Lopez, L.; Fiorio, P.R. (2013-on going Geoderma)
Without EPO With EPO
RMSEP [g kg-1] 130 69
Bias [g kg-1] 19 95
RPD 2.0 1.0
Yufeg Ge and Cristine Morgan
showing the p-matrx can be used to correct field moist intact scans of soils
(Geoderma-in press 2014).
Jason Ackerson, Cristine Morgan Yufeng Ge, Budiman Minasny, and
Alex McBratney
Extraction of Water interference on soil spectra
(External Parameter Orthogonalization)
moist
Dried
EPO makes
correction
Soil properties mapping using Hyperespectral Spectir-image. Franceschini, M., Dematte, J.A.M. In submission 2014
- Bare Soil- Photosynthetic vegetation- Non-Photosynthetic vegetation
(a) Soil sampling in bare
soil sites (visual analysis)
(b) Estimation of fractional cover
using hyperspectral images
(c) Spectra near
sampling points
selected in pixels with
> 70% of bare soil
(d) Partial least squares
regression using spectra
extract from image as
predictor of soil properties
CTC 0.73 8.19 1.94 0.80 7.95 2.29
Clay 0.58 46.48 1.56 0.73 41.09 1.96
Sand 0.66 43.67 1.73 0.73 45.77 1.96
Calc. 0.44 7.31 1.36 0.60 6.40 1.61
Magn. 0.51 2.07 1.45 0.52 1.79 1.47
OM 0.53 2.85 1.47 0.47 3.86 1.40
Cross-valid.
(n = 60)Independent data
(n = 29)
R2 RMSE RPD R2 RMSE RPD
n = 89 samples
Vis-Nir: 357 bands
The first test with an on-the-go spectral sensor in Brazilian soils for liming. Franceschini, M.,
Dematte, J.A.M. –submission 2014
(a) Field plots where lime
was applied in variable rates
Lime Requirement (LR)
calculation = soil saturation
by cations
(b) On-the-go spectral measurements
and soil sampling
(c) Partial least squares
regression using spectra
measured on-the-go as
predictors
pH OM Calc. Magn. CTC V LR
R2 0.43 0.43 0.29 0.48 0.33 0.38 0.35
RMSE 0.51 1.97 8.78 4.51 6.15 14.47 1.36
RPD 1.34 1.34 1.20 1.40 1.23 1.28 1.25
Validation with independent data (n = 50)
n = 150
samples
USING A MOBILE REAL-TIME SOIL VISIBLE-NEAR INFRARED SENSOR FOR HIGH RESOLUTION SOIL PROPERTY MAPPING
Masakazu Kodaira, Sakae Shibusawa Geoderma, v. 199, p. 64-79, 2013.
Number of sample in database per state
Number of Institutions Involved 32
PA: 253
MA:400
MT : 2000
TO:3
GO: 10000
MG: 1550
DF: 50
MS: 3000
SP: 15.000
RS: 2000
AM: 200
AP: 1000
PB
PE:300
RJ
PR:4000
AC
SC: 1410
In
process
50 Researchers and Collaborators
Involved
Dinamics to use the
Data and a document
With agreement between researchers
11000 samples
Real time detection of soils
contamination at fieldReal time soil
mapping at field
Laboratories of soil analysis by spectral sensing with traditional methods
Real time on the go
spectral sensing
Merge spectroscopy
With other
parameteres
Real time detection by
satellite hyperspectral
sensing from orbital
or aerial data
Final considerations
Merge with Field. Most answers are there!
Continue efforts on statistics
Believe in Spectroscopy.
We are on the correct track
IMPORTANCE OF THE BIG PICTURE
SOIL MAPS DERIVED FROM THIS TECHNOLOGY FOR MORE FOOD
WITH SOIL ENVIRONMENT QUALITY