Estimation of Evapotranspiration from Remote Sensing based SEBAL model in Central Valley, California
Sagarika Roy Graduate Student Montclair State University New Jersey
Outline
Introduction
Objective
Method v Image Classification, Post Classification & Masking v SEBAL Theory and Method-Actual Evapotranspiration estimation
Results & Discussion
Validation of remote sensing based SEBAL model to ground based Penmann-Montheith Eq from CIMIS and Field for ET estimation.
Conclusion
References
Acknowledgments
Introduction •The Central Valley is a large, flat valley that dominates the central portion of the California •Extent: 400 miles from north to south •Sacramento drains the northern of the Central Valley. In the southern, the San Joaquin flows 330 miles (530 km) north from valleys. •Annual rainfall: 20 inches (arid to semi arid climate). •Agriculture: Tomatoes, almonds, grapes, cotton, apricots, and asparagus •Economy: 17 billion USD from agriculture.
Objective
Raw Satellite
Pre-Corrected image Geometric and radiometric
To estimate actual evapotranspiration from from Remote Sensing tool based on Surface Energy Balance Algorithm (SEBAL) Model of Almond class using Image classification and Mask
Corrected Pixel
Classification
Land Use/Land Cover
SEBAL
Evapotranspirat ion
Maskin g
Individual actual ET of Almond Class
Final Actual ET map
Energy Balance Equation
Metho d
Pistachi os
Land use/Land Cover classification
So Urba il n
Almon Wate d r
Road s
NP V
Image Classification
Identification of individual pixels or groups of pixels with similar spectral responses (spectral signatures) to incoming radiation.
Unsupervised No training data. Model inference and application both rely on test data exclusively
K-means
•Classifications use statistical techniques to group ndimensional data into their natural spectral classes. •Uses a cluster analysis approach
Supervised Use training data to infer model, compared with model to test data
Maximum Likelihood
•Assumes that the reflectance values for each class in each band are normally distributed and calculates the probability that a given pixel belongs to each
Spectral Angle Mapper
•An automated algorithm in ENVI that compares image spectra to reference spectra (endmembers) from ASCII files, ROIs, or spectral libraries. •It calculates the angular distance between each spectrum in the image and endmember in n-dimensions, where n is the number of
Comparison of Unsupervised with Supervised classification Unsupervised Supervis ed
KMeans
Spectral Angular Mapper Almond Non Photosynthesis plant Urba
n Wate rSo ilPistachi os Other Green Plants
Supervis ed
Maximum Likelihood Almond Non Photosynthesis plant Urba n Wate r So il Pistachi os Other Green Plants
Post Classification-Confusion Matrix of Maximum Likelihood Using Ground Truth ROI
Masking
Almond class Non Almond class Ground Reference data point
Evapotranspiration using SEBAL Model
Evapotranspiration (ET) is the loss of water to the atmosphere by the combined processes of evaporation (from soil and plant surfaces) and transpiration (from plant tissues)
SEBAL (Surface Energy Balance Algorithm for Land) is a one-layer energy balance model that estimates latent heat flux and other energy balance components without information on soil, crop, and management practices
A specific feature of SEBAL is that DT ((vertical air temperature difference between the z1 and zm) is determined from the hot (dry) and cold (wet) pixels with assumed values of sensible heat flux (H). H is estimated at extreme dry (H=Rn−G) and wet locations (H=0),
SEBAL THEORY
ET is related to surface energy balance
Go = 0.3811 exp− 2.3187NDVIRn H = Cp DT /Rah
Rn : net radiation flux at the surfa (W/m2) Go : soil heat flux (W/m2), H : sensible heat flux to the air (W/m2 λE : latent heat flux density (W/m2 Λ λ : Latent heat of vaporization J kg−1
ET24 = Rn24 / λ (mm/day)
Data Requirements for IDL Code Output data from IDL code for SEBAL Input data to SEBAL IDL Code Meteorological parameters (CIMIS): •Wind speed (miles/h) •Humidity (F) •Solar radiation (Ly/m) •Air Temperature (F) •Albedo MASTER data: •leaf area index ((from Emily) •vegetation index (NDVI) (Callie) •surface temperature (Cassie) Referred Literature parameters •Emmisivity (e) •albedo •Specific heat at constant pressure Cp (J/kg/K) • • •
model
•Soil heat flux (G) (W/m2) •Sensible heat flux (H ) (W/m2) •Crop Coefficient (Kc) (W/m2) •Latent heat flux density (λE ) (W/m2 ) •Evaporative fraction ( ) •Net Radiation (Rn) (W/m2) •Actual Evapotranspiration (mm/h)
Actual ET Map
(Using IDL code applied to ENVI to estimate pixel by pixel crop ET)
Eta 0.669 to 0.681 mm/h Eta 0.699 to 0.703 mm/h Eta 0.682 to 691 mm/h
ETa (mm/h)
Hourly actual evapotranspiration (ETa) of Almond on August 24, 2009
Result and Discussion
Fig 1 F ig3
Fig 2
Validation of SEBAL estimated Actual ET with Penmann-Montheith from CIMIS
Avarage ETa (SEBAL) : 0.6745 mm/h Avarage ETo(PM-CIMIS) : 0.7519
The correlation coefficient of Eta estimates from remote sensing with ETo are 0.8571 The regression coefficient 0.7347 The mean difference between actual ETa from SEBAL in almond and Penman-Monteith for over all observations associated with ETa is 0.77 mm
Comparison of actual ET from SEBAL and Penmenn-Monteith Equations
Mean Percent difference for ETa PM-SEBAL/SEBAL=0.109%
Conclusion
The types/land use classes were identified from the MASTER image using multi-layered maximum likelihood classification shows 97% accuracy to mask only almond class from the Image.
The results of the regression between land surface temperature (Ts), NDVI and, evapotranspiration (Eta) show negative (-) correlation. On the other hand Ts possessed a slightly stronger negative correlation with the ETa than with NDVI for Almond class.
The actual evapotranspiration (ET a) estimated from SEBAL is 0.639 to 0.703 mm/h. Avarage is 0.6745 mm/h
The average actual ET estimated from SIMIS using crop coefficient (Kc) 1.02 is 0.774 mm/h
The correlation coefficient of actual ET (ETa) estimates from remote sensing with Reference (Eto) from Penmann Monteith are 0.8571
The mean difference between actual ETa from SEBAL in almond and Penman-Monteith for over all observations associated is 0.77
Hence the avarage ETa from CIMIS is marginally higher than ETa estimated from SEBAL model. Mean percentage difference is 0.109%
Reference Allen, R.G., A. Morse, M. Tasumi, R. Trezza, W. G.M. Bastiaanssen, J.L. Wright, and W. Kramber, 2002. Evapotranspiration from a Satellite-BASED Surface energy balance for the Snake Plain Aquifer in Idaho, Proc. USCID conference San Luis Obispo, July 2002 Bastiaanssen, W.G.M., M. Menenti, R.A. Feddes and A.A.M. Holtslag, 1998. A remote sensing Surface Energy Balance Algorithm for Land (SEBAL), part 1: formulation, J. of Hydr., 212-213: 198-212 Bastiaanssen, W.G.M., M. Ud-din-Ahmed and Y. Chemin, 2002. Satellite surveillance of water use across the Indus Basin, Water Resources Research, vol. 38,
Acknowledgement
I I would like to thank Prof Susan Ustin & Shawn for supporting the presentation of this research output, and for possible incorporation into the project and the CIMIS for provision of climatic data for stations in the study area. I would also like to thank the
I am also thankful to Student Airborne Research Program co-ordinators / team/ mentors for smoothly conducting the research program.
Thanks to all SARP student team for the brilliant team work and data sharing.
Thank You
Questions?