Evaluating Vegetation Evapotranspiration - South Dakota ...

Evaluating Vegetation
Evapotranspiration (VegET)
Modeling Results in South
Dakota
Gabriel Senay 1 and Geoffrey Henebry 2
1 SAIC, contractor to the U.S. Geological Survey (USGS) Center for Earth Resources
Observation and Science (EROS); South Dakota State University (SDSU)
Work performed under USGS contract 03CRCN0001
senay@usgs.gov
2 Geographic Information Science Center of Excellence (GIScCE),
South Dakota State University
Geoffrey.Henebry@sdstate.edu
U.S. Department of the Interior
U.S. Geological Survey

Outline
• Summary of results
• VegET modeling background and objective
• VegET model outputs
• VegET evaluation
• Soil Moisture (NRCS: SCAN)
• Crop Yield (NASS County)
• Flux Data (AmeriFlux)
• Vegetation Water content (AMSR-E VWC )
• NDVI
• Conclusion

Summary
• VegET is a water balance model that produces landscape ET
(water use) on a pixel-by-pixel basis on rainfed systems.
• Correlated well with AmeriFlux latent heat flux (ET) and SCAN
soil moisture data.
• Promising results with crop yield data.
• Show interesting relationship with AMSR-E Vegetation Water
Content.
• Operational setup of VegET is possible for crop and grassland
monitoring.

VegET Modeling:
Background and Objective
• VegET is a new modeling approach that integrates Land Surface Phenology
(LSP) and commonly used water balance modeling algorithms to estimate
actual vegetation ET (water use) in primarily non-irrigated crop and grassland
environments for agro-hydrological applications.
• Key inputs to the model:
• 1) Daily rainfall data at 5-km resolution from NOAA/NWS (a new product
since 2005, a blend of NEXRAD and station data)
• 2) Daily reference ET calculated at EROS for global applications from the
Global Data Assimilation (GDAS) climate parameters
• 100 km, downscaled to 10 km
• 3) “Climatological” remotely sensed NDVI time series to define “optimum”
water- use patterns of landscapes throughout the year; 1-km resolution
• 4) Soil water holding capacity from STATSGO (in case of US)
• from 1:250,000 scale soils map
• Objective:
• Produce daily ETa and soil moisture to monitor crop and grassland
performance for early assessment of yield and biomass production.

PRECIPITATION
Reference ETo
VegET
ETa = Ks * Kp * ETo
SOILS
PPT i
Soil Stress Coefficient
Water Balance
ETa i
LSP Water-Use Coefficient
0.8
"Climatological" NDVI Pattern:Nebraska
-97.8, 41.4
0.7
Water
Balance
Model
WHC
Surplus
Runoff
SW i
ETa = ETp ETa = Kp * ETo
ETa < ETp
ETa = Ks * Kp * ETo
Drainage
SW i
= SW i-1
+ PPT i
– ETa i
– RF i
-DD i
NDVI
0.6
0.5
0.4
0.3
0.2
0.1
0
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35
dekad ( 1-36)
Series1
Series2
Series3
Series4
Series5
Series6
Land Surface Phenology
(LSP)

Evaluation Datasets
• Soil Moisture (vol/vol)
• Soil Climate Analysis Network (SCAN): Soil moisture (hourly, July
2005, point), EROS, South Dakota
• Crop Yield (Bu/ac)
• National Agricultural Statistics Service (NASS): (yearly, 2005, 2006,
county-average), South Dakota.
• Latent Heat Flux (ET)
• AmeriFlux (daily, 2005, Point) Brookings, South Dakota; also
Arizona
• Vegetation Water Content (kg/m 2 )
• AMSR-E (daily, 2005: 25 Km), Faulkton, South Dakota
• NDVI
• AVHRR: (weekly, 2005,2006; 1 km), Faulkton, SD

VegET Model Outputs
Conterminous USA
South Dakota

Illinois Flash Drought in 2005?

2006 drought in much of South Dakota, Nebraska, Southeastern US?

Yearly total shows more water use in the western US.

Seasonal and annual ETa show similar spatial patterns.
Seasonal patterns show more differences in magnitude between
the two years.

VegET: May – Sep

VegET: May – Sep
Major
ETa reduction

VegET Evaluation
Soil Moisture: SCAN Data

Hourly Soil Moisture: EROS SCAN
Hourly Average SM data
SCAN EROS
34
SM (v/V)
32
30
28
26
24
storm event
Day 11
Day 12
22
700
1000
1300
1600
1900
2200
100
400
700
1000
1300
1600
1900
2200
100
400
Hours
hours
on
on
July
11 and
1112, and
July,
12,
2005
2005

90
VegET Soil Water vs SCAN soil moisture
South Dakota Site, EROS SCAN Site
34
VegET SW (mm)
80
70
60
50
40
30
20
10
0
VegET SW (mm)
SCAN SM (V/V)
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31
Days in July, 2005
32
30
28
26
24
22
SCAN Soil Moisture (v/V)
SCAN: Avg of 2, 4, 8, 20 and 40 inch depth data
VegET soil water level change captures the temporal variations in soil moisture
as measured by the SCAN Site.
Note: VegET (5 km), SCAN (Point)

SCAN Soil Moisture (6 am) vs VegET Soil Water Estimate
South Dakota/EROS Site (SCAN ID 2072) (31 days in July 2005)
VegET Soil Water (mm)
90
y = 7.5618x - 159.68
80
R 2 = 0.9063
70
60
50
40
30
20
10
0
20 22 24 26 28 30 32 34
SCAN Soil Moisture (vol/Vol)
Senay, 2007

Evaluation: Crop Yield
USDA: NASS Data

2005 County Wheat Yield vs Seasonal Total ETa
48 counties in South Dakota
County Wheat Yield vs ETa
South Dakota: 2005; n = 48
Wheat Yield (Bu/Ac)
70
60
50
40
30
20
10
0
y = 0.0945x + 6.3324
R2 = 0.41
0 100 200 300 400 500 600
VegET ETa (mm) (May – Sep)
Some counties had ETa more than 400 mm

2006 County Wheat Yield vs ETa
46 counties in South Dakota
Wheat Yield (Bu/ac)
70
60
50
40
30
20
10
y = 0.2035x - 31.266
R2 = 0.50
County Wheat Yield vs ETa
South Dakota 2006; n = 46
0
0 100 200 300 400 500 600
VegET ETa (mm) (May – Sep)
Almost all counties showed ETa less than 400 mm
ETa is better correlated with yield when water is a limiting factor.

2005/2006 County Wheat Yield vs ETa
Counties in South Dakota
Yield (Bu/ac)
70
60
50
40
30
20
10
0
y = 0.1433x - 12.481
R2 = 0.50
County Wheat Yield vs ETa
South Dakota: 2005 & 2006; n = 94
0 100 200 300 400 500 600
VegET ETa (mm) (May – Sep)
ETa is reliable spatially and temporally

Comparing Yield and ETa Differences
(between 2005 and 2006, county level)
Yield Difference
(Bu/Ac) [2006 - 2005]
Wheat Yield Difference vs ETa difference: 2006 minus 2005
(South Dakota County Average, n = 36)
40
30
20
10
0
-200 -150 -100 -50 -10 0 50 100 150 200
-20
-30
-40
ETa Difference (mm) [2006 - 2005]
Most counties showed a large drop both in Yield and ETa from 2005 to 2006

Wheat Planted Area Distribution

2005

2006

Possible Sources of Error
• Spatial mismatch
• ETa = county-wide average
• Yield = sub-county aggregate, not all of the counties
produces wheat
• Proposed Solution: use crop layers to redefine ETa average
• Temporal mismatch
• ETa = May – Sep
• Yield = (all seasons total: winter wheat, spring wheat)
• Solution: analyze spring and winter wheat data separately
• Management differences in yield that is not reflected
in ETa (water-base)
• Solution: develop temporal relationship using more years of
historical data and work with yield anomalies.

Evaluation: Latent Heat Flux (ET)
AmeriFlux Data

Latent heat flux (ET) from AmeriFlux tower and VegET ETa
South Dakota, Brookings
VegETa vs Flux LE: South Dakota 2005
Cover: crop/grassland
12
0
10
10
20
ETo/ETa (mm)
8
6
4
Flux LE (ETa)
GDAS ETo
VegETa-5k
Flux Rain
30
40
50
60
70
Rainfalll (mm)
2
80
90
0
100
1 20 39 58 77 96 115 134 153 172 191 210 229 248 267 286 305 324 343 362
Days (2005)
VegETa captured both the magnitude and temporal variations of measured flux
at the tower site, including gross primary production (data not shown)

Latent heat flux (ET) from AmeriFlux tower and VegET ETa
Arizona, 2005
VegETa vs Flux LE: Arizona 2005
cover: grassland
12
0
10
ETo/ETa (mm)
10
8
6
4
Flux LE (ETa)
GDAS ETo
VegETa-5k
Flux Rain
20
30
40
50
60
70
Rainfalll (mm)
2
80
90
0
100
1 20 39 58 77 96 115 134 153 172 191 210 229 248 267 286 305 324 343 362
Days (2005)
A stronger correspondence between VegETa and tower latent heat flux.

Evaluation Vegetation Water Content (AMSR-E)
The Advanced Microwave Scanning Radiometer - EOS (AMSR-E)
=>passive microwave data (brightness temperature) converted
into Vegetation Water Content (VWC)
=>acquired twice daily: pre-dawn and early afternoon
=> 25-km resolution

VegET ETa vs AMSR-E VWC
Faulkton, South Dakota, 2005
AMSR-E VWC vs VegET ETa
Faulkton, SD 2005
8
0
ETa (mm), VWC
(kg/m2)
7
6
5
4
3
2
1
Rainfall
ETa
ddVWC
dVWC
aVWC
5
10
15
20
rainfall (mm)
0
-1
-2
10-Jan
10-Feb
10-Mar
10-Apr
10-May
10-Jun
10-Jul
10-Aug
Days (7-day Interval)
10-Sep
10-Oct
10-Nov
10-Dec
25
30
ddVWC = dVWC - aVWC
Large negative difference between dVWC and aVWC
seems to occur during optimum water supply and high ET

VegET ETa vs AMSR-E VWC, NDVI
Faulkton, South Dakota, 2005
VegET ETa vs AMSR-E VWC, NDVI
Faulkton, SD 2005
8
0
ETa (mm), NDVI (x10),
ddVW (kg/m2)
7
6
5
4
3
2
1
0
-1
-2
10-Jan
10-Feb
Rainfall
ETa
NDVI
ddVWC
10-Mar
10-Apr
10-May
10-Jun
10-Jul
Low ETa
10-Aug
10-Sep
10-Oct
Days (7-day interval based on weekly NDVI)
10-Nov
10-Dec
5
10
15
20
25
30
rainfall (mm)
ddVWC = dVWC - aVWC
-Both ETa and ddVWC show immediate stress to low rainfall
-NDVI show similar stress at a lesser degree with a lag time

VegET ETa vs AMSR-E ddVWC
Faulkton, South Dakota, July 2005
Faulkton: VegET daily ETa vs ddVWC (kg/m2)
2005, July (31 days)
7
ETa (mm), VegET
6
5
4
3
2
y = -2.1925x + 2.6246
R 2 = 0.8465
Physical significance
of this relationship
needs to be
investigated.
1
0
-1.50 -1.00 -0.50 0.00 0.50 1.00 1.50
ddVWC (kg/m2)

Evaluation: NDVI
ETa vs NDVI
Faulkotn, SD (2005 & 2006)
Faulkton, SD (2005 & 2006)
ETa (mm)
8
7
6
5
4
3
2
1
0
ETa2005
ETa2006
NDVI2005
NDVI2006
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
NDVI (-)
3-Jan
3-Feb
3-Mar
3-Apr
3-May
3-Jun
3-Jul
3-Aug
Days (7-day interval)
3-Sep
3-Oct
3-Nov
3-Dec
-Both ETa and NDVI show a reduction in 2006
-The relative reduction of ETa is much higher and immediate
than the reduction in NDVI.

Relative Reduction of ETa and NDVI
w.r.t to Crop Yield
NDVI (x 1000) / ETa (mm)
500
450
400
350
300
250
200
150
100
50
0
NDVI (avg) vs ETa (sum)
Faulkton, SD (2005 & 2006)
78% of
2005
2005 2006
Year
95% of
2005
ETa
NDVI
Faulk County: Wheat yield reduction from 2005 to 2006 = 61%
ETa more sensitive to wheat (crop) yield reduction than NDVI

Conclusion
• VegET ETa showed encouraging performance with respect to:
• SCAN soil moisture
• AmerifFlux Latent Heat Flux from Brookings and Arizona
• Non-irrigated county wheat yield
• Interesting relationships between VegET and AMSR-E VWC was observed. The diel
difference VWC (ddVWC) showed a strong negative relationship with ETa. Further
research is required to understand the physical relationship between ddVWC and
ETa
• VegET ETa was shown to be more sensitive than NDVI in detecting yield reduction
from 2005 to 2006.
• Use of historical crop yield data will improve the yield prediction capability of
VegET.
• VegET can be setup to provide daily estimates of sub-county soil moisture, ETa
and crop yield performance and general plant biomass production.
• Caution should be taken in applying VegET on irrigated areas and forested
landscapes that draw water from ground sources or transported surface water.