Global Estimation of Canopy Water Content Susan Ustin (PI), UC Davis

1

Global Estimation of Canopy Water ContentSusan Ustin (PI), UC Davis

E. Raymond Hunt (Co-PI) USDA Water Lab

Vern Vanderbilt (Co-PI) NASA Ames Research Center

Goals: (1) Test and Validate Retrieval of Water Content (2) Evaluate Ecological Value of Water Content Index

►Theoretical Evaluations at Leaf and Canopy Scales • Evaluate effect of cover, vegetation type, and soil background

►Empirical Evaluations• Compare to Field Data• Compare to AVIRIS EWT• Compare to VIs under Different Land Cover Conditions

►Testing Ecological Information• Plant Water Stress/Drought Indicator• Estimate LAI at High LAI sites (>4)• Agricultural Irrigation Scheduling• Fuel Moisture Estimates for Wildfire Risk Prediction• Soil Moisture (SMOS) Corrections for Vegetation

2

Field Research Sites:

Wind River Ameriflux Site (mature conifer)SMEX 04 southern Arizona and Northern Mexico (semiarid)SMEX 05 agriculture, Ames, Iowa (corn, soybean)Agriculture, San Joaquin Valley, CA (cotton)

Analysis of MODIS Time Series Data at Ameriflux Sites:

Howland, MEHarvard Forest, MAWLEF-Tall Tower, WIWind River, WACentral California-Western Nevada (mixed semiarid vegetation)Bondville, IL

3

0.00

0.10

0.20

0.30

0.40

0.50

0.60

400 900 1400 1900 2400Wavelength (nm)

Ref

lect

ance

Cm=0.001g cm-2

Cm=0.015g cm-2

Cm=0.030g cm-2

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

0.40

0.45

400 900 1400 1900 2400Wavelength (nm)

Refle

ctanc

e

N=1.5

N=1.0

N=0.5

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

0.40

0.45

400 900 1400 1900 2400Wavelength (nm)

Ref

lect

ance

Cab=20

Cab=40

Cab=80

Chlorophyll Structure Parameter Dry Matter

Effect of Leaf Biochemistry on Leaf Reflectance

Y-B. Cheng, P.J. Zarco-Tejada, D. Riaño, C. Rueda, and S.L. Ustin

4

0.00

0.10

0.20

0.30

0.40

0.50

0.60

0.70

400 900 1400 1900 2400Wavelength (nm)

Ref

lect

ance

Dark Soil Medium Soil Bright Soil

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

0.40

0.45

400 900 1400 1900 2400Wavelength (nm)

Ref

lect

ance

Dark soil

Medium soil

Bright soil

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

0.40

400 900 1400 1900 2400

Wavelength (nm)

Ref

lect

ance

Medium soil

Dark soil

Bright soil

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

400 900 1400 1900 2400

Wavelength (nm)R

efle

ctan

ce

Dark soil

Bright soil

Medium soil

Soil background effect on canopy spectra simulated by (a) PROSPECT-SAILH, (b) PROSPECT-rowKUUSK, (c) PROSPECT-FLIM

Variation in Soil Reflectance

Y-B. Cheng, P.J. Zarco-Tejada, D. Riaño, C. Rueda, and S.L. Ustin

5

0

0.05

0.1

0.15

0.2

0.25

0.3

0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16

medium

dark

bright

0

0.02

0.04

0.06

0.08

0.1

0.12

0.14

0.16

0.18

0.2

0 0.05 0.1 0.15 0.2

medium

dark

bright

0

0.02

0.04

0.06

0.08

0.1

0.12

0.14

0.16

0.18

0.2

0 0.05 0.1 0.15 0.2

medium

dark

bright

Soil background reflectance on Simulated EWT and Canopy Water Content

(a) PROSPECT-SAILH (b) PROSPECT-rowKUUSK (c) PROSPECT-FLIM

Cw*LAI (cm)

EW

T

Y-B. Cheng, P.J. Zarco-Tejada, D. Riaño, C. Rueda, and S.L. Ustin

6

y = 7.9019x + 88.181

r2 = 0.90

0

1000

2000

3000

4000

0 100 200 300 400 500Canopy EWT (um)

AV

IRIS

EW

T (

um)

<20%

20-40%

>40%

n=19 n=15 n=10

n=20n=12

n=12

n=22

0

500

1000

1500

2000

2500

3000

3500

4000

EW

T (

um)

Comparison of Field Measured EWT and AVIRIS at Walnut Gulch, AZ

Variation in EWT-AVIRIS By Vegetation Type

y = 878.87x + 428.47

r2 = 0.82

0

1000

2000

3000

4000

0 1 2 3 4 5LAI

AV

IRIS

EW

T (

um)

Agriculture

Native species

(b)

Yen-Ben Cheng, Susan L. Ustin, and David Riaño

Hunt et al.

7

NDVIy = 1.0652x - 0.081

r2 = 0.6081

EVIy = 0.9483x + 0.0137

r2 = 0.7081

SIWSIy = 0.9222x + 0.0337

r2 = 0.7137

NDWIy = 0.9494x + 0.004

r2 = 0.433

0

0.2

0.4

0.6

0.8

1

0 0.2 0.4 0.6 0.8 1Cross-calibrated AVIRIS indexes

MO

DIS

inde

xes

NDWI

SIWSI

NDVI

EVI

Cross Calibration between AVIRIS and MODIS

8

y = 0.0002x + 0.3348

R2 = 0.5044

y = 0.0002x + 0.0624

R2 = 0.6358

y = 0.0001x + 0.0173

R2 = 0.5732

y = 7E-05x - 0.0701

R2 = 0.5365

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

0 500 1000 1500 2000 2500 3000 3500

NDWI

SIWSI

NDVI

EVI

Linear (NDVI)

Linear (EVI)

Linear (SIWSI)

Linear (NDWI)

y = 0.0002x + 0.1528

R2 = 0.8865

y = 8E-05x + 0.1146

R2 = 0.823

y = 0.0001x - 0.1807

R2 = 0.812

y = 4E-05x - 0.129

R2 = 0.6186

-0.3

-0.2

-0.1

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0 500 1000 1500 2000

NDWI

SIWSI

NDVI

EVI

y = 5E-05x + 0.7182

R2 = 0.4852

y = 5E-05x + 0.405

R2 = 0.5505

y = 8E-05x + 0.11

R2 = 0.6612

y = 4E-05x - 0.0821

R2 = 0.5853

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

0 500 1000 1500 2000 2500 3000 3500

NDWI

SIWSI

NDVI

EVI

Linear (NDVI)

Linear (EVI)

Linear (SIWSI)

Linear (NDWI)

Walnut Gulch, AZ on 25 August 2004

Relationship between EWT-AVIRIS and MODIS Indexes at 3 sites

AZCAL Properties, CA on 16 July 2002

Howland forest, ME on 23 August 2002 

Yen-Ben Cheng, Susan L. Ustin, and David Riaño

9

Walnut Gulch, AZ

(a) EWT (AVIRIS) (b) NDWI (MODIS) (c) NDII (MODIS)

Howland Forest, ME

AZCAL Properties, CA

Y-B Cheng, S.L. Ustin, and D. Riaño

10

MODIS-NDWI Time Series

Variation with Land CoverClasses

Time, 2000-2005

MO

DIS

ND

WI

Inde

x

Palacios-Orueta et al.

11

Neural Net Prediction (ANN) of EWT

Training Dataset

Validation Dataset

Real DataMODIS

Leaf Training Leaf Validation Application

Input

Input

Input

Output ANN

Input

Input

Input

Output ANN

Input

Input

Input

Output ANN

LOPEX dataPROSPECT

Both

LOPEX dataPROSPECT

PROSPECT-SAILHAVIRIS

D. Riaño, M.A. Patricio, P. Zarco-Tejada, C. Rueda, L. Usero, S.L. Ustin

12

ANN trained with Real Data at Leaf Levelfor EWT

• Trained with all LOPEX samples • Leave one out cross-validation• 420 input layers: 210 and 210

420 Input Layers

Hidden Layer with varying numbers of neurons

Output Layer

EWT

Riaño et al. (r2=0.95)

13

Analysis at canopy level• Trained with PROSPECT-SAILH: 600 random

samples • Validation with PROSPECT-SAILH: 7400

samples independent of training

210 Input Layers

Hidden Layer with variant number of neurons

Output Layer

PROSPECT-SAIH

EWT, LAI, DMN, Cab, LIDF, Soil

canopy

1. Radiative Transfer model

EWT*LAI

2. Training ANNcanopy

3. Validationcanopy ρ

EWT*LAI

D. Riaño, M.A. Patricio, P. Zarco-Tejada, C. Rueda L. Usero, S.L. Ustin

14

Analysis at Canopy Level with MODIS

• ANN trained with PROSPECT-SAILH

to generate EWT*LAI• ANN run on MODIS product

MOD09A1 • AVIRIS EWT Used for Validation

AVIRIS MODIS NDWI

Walnut Gulch in AZ

NDVI, NDWI, NDW6

MODIS EWT

AV

IRIS

EW

T

R2 = 0.82

D. Riaño, M.A. Patricio, P. Zarco-Tejada, C. Rueda, L. Usero, S.L. Ustin

15-1.0

1.0

3.0

5.0

7.0

9.0

11.0

-1.0 1.0 3.0 5.0 7.0 9.0 11.0

Measured %FMC

1:1 line

Training

Prediction

-0.01

0.00

0.01

0.02

0.03

0.04

0.05

0.06

-0.01 0.00 0.01 0.02 0.03 0.04 0.05 0.06

Measured EWT

Est

imat

ed E

WT

1:1 line

Training

Prediction

-0.001

0.004

0.009

0.014

0.019

-0.001 0.004 0.009 0.014 0.019

Measured DM

Es

timat

ed

DM

1:1 line

Training

Prediction

y = 0.8243x + 0.0006

R2 = 0.9478

0.00

0.01

0.01

0.02

0.02

0.03

0.03

0.04

0.04

0.05

0.05

0 0.01 0.02 0.03 0.04 0.05 0.06

Equivalent Water Thickness (g/cm2)

Measu

red

EW

T (

g/c

m2)

Mea

sure

d D

ry M

atte

r (g

/cm

2)

Dry matter (g/cm2)

y = 0.8308x + 0.0009

R2 = 0.342

0.00

0.01

0.01

0.02

0.02

0.03

0 0.002 0.004 0.006 0.008 0.01 0.012 0.014 0.016

Predicting Fuel Moisture Content for Wildfire Risk Assessment

Estimated by PROSPECT from LOPEX Fresh Leaf Data

Generalized additive algorithm-partial least square regression, GA-PLS

Lin Li, Susan Ustin, and David Riaño

P-value<0.0001