Estimating water used by vegetated areas is very important for water resources management and water rights. Traditionally the amount of water delivered to an area is calculated by installing some measuring device (flumes, weirs, flow meters, etc.). The alternative approach presented here estimates the actual water use in a vegetated areas based on ground surface energy balance concept using the ReSET model (Remote Sensing of ET – ReSET developed by IDS group in Colorado state university) that uses satellite and Arial imagery with visible and thermal bands along with weather data to estimate daily actual crop Evapotranspiration (ET) for vegetated areas. Surface energy balance models have been proven to be a robust approach for estimating vegetation evapotranspiration. One of the main limitations of wider application of these models in water resources and irrigation management is the requirement of extensive back ground in surface energy modeling. This presentation shows the development and the application of an ArcGIS toolbox that runs an automated version of the ReSET model. The tool is compatible with NASA/USGS Landsat Legacy Project. The presented ArcGIS tool automates the model in all stages and requires minimum interference from user. The tool presented accommodates both basic and advanced users. The results using the tool were tested and validated using results from manual ReSET model runs.

1. Automation of surface energy balance model
{ReSET}
Aymn Elhaddad and Luis Garcia,
Colorado State University, Fort Collins, CO.
1

2. Crop Water Use (ET) Monitoring System
•Ground, air- and space- borne
RS of ET
"Courtesy: National Science Foundation"
2

3. Traditional ETc Calculation
3

4. Advantages of Surface Energy Balance
Models Over Other Methods
Classic calculation of crop ET relies on using
reference ET and a crop coefficient.
Unfortunately, this methodology does not take
into account the local conductions such as:
• Water shortage/Waterlogging
• Crop Type
• Planting dates
• Salinity
4

5. Remote Sensing of ET Models
• Most common ones use the energy
balance equation.
• Have large footprint (regional coverage).
• Measure actual, not potential ET.
• Several models have been developed:
SEBAL, METRIC, ReSET, SAT,
ALARM, etc.
5

6. Remote Sensing of Evapotranspiration
Using Surface Energy Balance ET is calculated as a
“residual” of the energy balance
LE = Rn – G – H
H (Heat to air)
ET
Rn
(Radiation from sun)
G (Heat to ground)
6

7. Description of Energy Balance Models
The use of the energy balance equation:
Rn = LE + G + H
Net Radiation (Rn), Soil Heat Flux (G), Sensible Heat Flux
(H), and Latent Energy consumed by ET (LE).
Model Rn, G and H, then determining LE as a
residual.
LE = Rn – G – H
7

8. SEB Models Limitations
Imagery availability
If no images are available for a region SEB models
can not be used.
Cloud cover
SEB models are sensitive to clouds since clouds
impact the thermal band.
Calibration
In the model without ground calibration if advection
occurs it will introduce errors in the results.
In the calibrated mode high quality weather data is
required.
8

9. Surface Energy Balance Using A Raster
Concept (ReSET-Raster)
• The ReSET-Raster model is a surface energy
balance model that uses Surface Energy Balance
to calculate actual ET for every pixel.
• Unlike METRIC or SEBAL, ReSET generates
surfaces of every variable (wind, hot, cold, ETr)
• Model main inputs
1. Aerial imagery with visible and thermal bands.
2. Ground weather data from available weather
stations.
9

10. ET Modeling
Aerial images with visible
and thermal bands
ET 24 hour grid
ReSET
Model
10

11. Irrigation Event Captured by the ReSET
Model 5/27/2006
11

12. ET Variability within Fields Detected by
ReSET Model 7/30/2006
Landsat image
ReSET ET
12

13. Running a Surface Energy Balance Based
Model Manually
To manually run SEB models users need to have a
background in:
•
•
•
•
•
•
•
•
Hydrologic science,
Behavior of soil, vegetation and water systems,
Environmental physics,
Radiation,
Aerodynamics,
Heat transfer,
Vegetation systems,
Image Processing.
The need for these expertise limits the number of model users and
eventually hinders the expansion of surface energy balance models
usage and applications.

14. EB Model
Manual Mode
1dem.mdl
2ls7_1.mdl
2ls5_1.mdl
ArcGis for Albedo
Coef
2ls_2.mdl
2ls7_3_savi.mdl
2ls5_3_savi.mdl
3_11_14r_savi.mdl
3_11_14r_savi.mdl
ArcGIS to obtain cold and
hot points shp
3.mdl
pre_it_rain_r1.mdl
pre_it_rain.mdl
pre_it_r1.mdl
H_r.mdl
pre_it.mdl
H.mdl
Rah.mdl
4.mdl
SI_r1.mdl
50_51_reset1_
point.mdl
SI.mdl
H_r.mdl
H.mdl
50_evapo_fr
_1.mdl
Rah.mdl
4.mdl
SI.mdl
50_evapo_fr_2.mdl
H.mdl
H_r.mdl
50_evapo_fr_2r.mdl
50_evapo_fr_2r1.mdl

15. Automated ReSET Model Under ArcGIS 10.1
15

16. Surface Energy Balance Model
Components
Location
characteristics
Surface Albedo
Anchor points
selection
Energy
balance
calculations
16

17. Model First Two Components
Image header
file
Digital
Elevation
map
Location Geographic
Characteristics
Multi layer satellite
imagery
•Reflectance of light energy
•Vegetation indices
(NDVI,TOA-ALBEDO)

18. Surface Albedo calculation
α =
α toa − α path _ radiance
τ sw
2
(Chen and Ohring, 1984; Koepke et al., 1985)
αpath_radiance ~ 0.03
τsw = 0.75 + 2
Bastiaanssen (2000)
10-5
z
FAO 56
18

19. Iterative Process for Calculating
Surface Albedo
TOA Albedo
Calculate Surface Albedo
using initial albedo
coefficient
Surface Albedo
Mean value of surface
albedo within NDVI
mask
Surface Albedo
NDVI
mask
Check
Recalculate
Surface Albedo
Update Albedo
coefficient
Fail
Pass
Final Surface
Albedo

20. NDVI image (left) and NDVI mask (right) for Albedo
20

21. NDVI mask for Surface Albedo
21

22. Modeling of Fluxes
(Rn = LE + G + H)
• In the energy balance equation various fluxes are modeled as
follows:
• Net Radiation (Rn) = Gain – Losses
Broad band emissivity
Rn = (1 – α) Rs + RL – RL – (1-εo) RL
• Soil Heat Flux (G) – Empirical equation
G = (Ts/α (0.0038α + 0.0074α2) (1 – 0.98NDVI4)) Rn
• Sensible Heat Flux (H)
H = ρaCp (T1 – T2)/rah
Cp = air specific heat
ρa = air density
Z2
rah
dT
H
Z1
rah = aerodynamic resistance to heat transport
22
(T1 – T2) = dT = near surface temperature diff.

23. Modeling of Fluxes (cont.)
H = ρaCp (T1 – T2)/rah
• Due to the difficulty of computing T1 – T2 the model
uses a dT function.
Z
2
rah
dT
H
Z1
rah = aerodynamic resistance to heat transport
(T1 – T2) = dT = near surface temperature diff.
H = ρaCp (dT)/rah
Cp = air specific heat
ρa = air density
23

24. dT Function
• To determine dT, two extreme pixels are selected (cold and hot
anchor pixels).
• The cold pixel is selected from a well water vegetated area where
dTcold= 0
• The hot pixel is selected as a very dry land area where it is
expected that LE = 0.
dThot = (Rn - G) rah / (ρaCp)
H = Rn – G – (LE = 0)
• A linear relationship is assumed to exist between the surface
temperature and dT, which then enables the calculation of H for
each pixel.
24

25. Selection of Hot and Cold Anchor Points
Surface
temperature
NDVI Filter
Surface Albedo Filter
Clumping
Sieving
Weather
Stations
Proximity
Sorting
Anchor
points
25

26. Checking the Value of the Cold Anchor
Points
26

27. Proximity Check for Cold Anchor Points
27

28. Solving for H
• H = ρaCpa (dT)/rah
T and dT
25
We find H for every pixel
LE = Rn – G – H
20
dT
15
10
5
0
280
-5
290
300
310
320
330
340
T (Kelvin)
is latent heat of vaporization (j/kg) representing the
energy required to evaporate a unit mass of water
To extrapolate from instantaneous ET to 24-hour
ET, ETrF is multiplied by 24-hour ETr.
28

29. LE and EF Calculation
• Using LE the evaporative fraction (EF) is calculated:
EF = LE/(Rn – G) Evap./Available energy
• It assumes that this fraction remains constant
throughout the day, therefore can be used in
determining daily ET as shown below:
ET24 = 86,400 * EF * (Rn24 - G24)/ λv
• Under calm weather conditions or moderate
advection for non-irrigated areas, the assumption of
EF being constant can be acceptable.
29

30. Model Inputs
(Acquired data sets)
• Imagery data:
Satellite imagery and header file.
• Weather data:
Daily and hourly reference ET as single point values or grids
and daily wind run.
• Location data:
Shapefile of weather station locations and shapefile for the
area of interest.
30

31. Model Inputs
(User entered parameters)
• User enters several values for the creation of
masks and the filtering process.
• Those processing parameters are critical but do
not require advanced background.
• The model initial run uses default parameter
values (which are suitable under normal crop
growing conditions).
• Model parameters are fine-tuned in a learning
process through a feed back from the area of
interest the model is being applied on.
31

32. Advantages of Automating the Use of
SEB Models
• Help the expansion of surface energy balance
models applications.
• Facilitate the mass production of actual ET
maps on several temporal and spatial
resolutions (current and historical).
• Provide valuable inputs to near real time
irrigation scheduling models and ground
water models.
32

33. Limitations of Automation of the Use of
SEB Models
• The automated model was used in different
areas in the US (Colorado, California, Texas)
and overseas (Spain, Egypt and Saudi Arabia)
with performances that matched or surpassed
the manual mode. However, the automation
process is unable to accommodate these two
limitations:
• Pixels affected by clouds.
• Bad pixel data ( Landsat 7), image boundaries.
33

34. Overcoming Limitations of Automation
1. Imagery preprocessing can be done to overcome the
previously mentioned limitations.
– Clouds:
Image pixels affected by clouds should be masked and replaced
with no data values.
– Bad pixel data:
• Satellite images should be clipped at the borders to eliminate
pixels with bad data at the edges of the image.
• Bad stripes in satellite imagery (Landsat 7) should be filled with
same image date adjacent data or masked.
2. Running the model in a semi automated mode
where the user can manually select hot and cold
anchor points.
34

35. ReSET Raster Applications
•
South Platte Basin – Estimation of irrigation efficiency
and validation of pumping records, development of Kc.
•
Arkansas River Basin – Estimation of actual crop water
use, Aquifer recharge/depletion, impact of salinity on ET,
calculation of canal service area water use.
•
Palo Verde Irrigation District – Estimation of actual crop
water use.
•
Aragona Irrigation District in Spain – Estimation of
actual crop water use.
•
Nile Delta in Egypt- Evaluate irrigation canal system
efficiency.
35

36. ReSET Model Modes
• ReSET is available in full automatic and semi
automatic mode for all Landsat satellites
imagery (LS 5, LS 7, LS 8) and MODIS.
• Colorado State University will be offering a
three day training on the ReSET model this
fall, for more information please email
Aymn.elhaddad@colostate.edu
36

37. Questions
37