Remote Sensing based Evapotranspiration Estimation and Sensitivity Analysis

Evapotranspiration (ET) is an important factor to calculate the water loss to the atmosphere and water demand for crops. Global and regional estimates of daily evapotranspiration are essential for our understanding of the hydrologic cycle. Remote sensing methods have many advantages in estimating daily ET for a large heterogeneous area.  In the present study, the sensitivity of ET with respect to different remote sensing-derived variables has been quantified while using the energy balance algorithm for land (SEBAL) method to estimate daily ET. The sensitivity of SEBAL-based ET has been determined for NDVI, LST, albedo, and SAVI using Extended Fourier Amplitude Sensitivity Test (eFAST) method. Relative changes in ET estimates for a range ± 20% of important parameters i.e., NDVI, albedo, SAVI, and LST have been determined and the sensitivity coefficient was estimated. Further, the sensitivity of SEBAL estimated ET has been investigated for different land cover and land use classes i.e., cropland, barren land, settlement, forest, and sparse vegetation. Results show that ET is significantly sensitive to the albedo and LST, however, other LULC classes have a different level of sensitivity. For cropland, ET is sensitive to NDVI. The sensitivity coefficient also indicates a significant effect of albedo and LST on the SEBAL estimated ET. For cropland, a 20% decrease in albedo and LST resulted in a 4.24% and 4.19% reduction in ET, and a 20% increase leads to an increase in ET by 13% and 5.53%, respectively. For sparse vegetation, a 20% reduction in albedo leads to an increase in ET by 7.46% while a 20% increase in albedo may reduce the ET by 15.70%. SAVI has an inverse relationship with ET for forest, barren land, settlement, and sparse vegetation as compared to other variables. The study concludes that SEBAL estimated ET is sensitive to albedo and LST significantly. The study helps in understanding the scope of uncertainty in remote sensing-based ET estimation.