Mapping daily and seasonal evapotranspiration from irrigated crops using global climate grids and satellite imagery: Automation and methods comparison

Abstract

AbstractThe surface energy balance algorithm for land (SEBAL) estimates land surface evapotranspiration (ET) from radiometric surface temperature (TR), but requires manual selection of calibration pixels, which can be impractical for mapping seasonal ET. Here pixel selection is automated and SEBAL implemented using global climate grids and satellite imagery. SEBAL is compared with the MOD16 algorithm, which uses remotely sensed data on vegetation condition to constrain reference ET from the Penman‐Monteith equation. The difference between the evaporative fraction (Λ, range 0–1) from SEBAL and six eddy flux correlation towers was less than 0.10 for three of six towers, and less than 0.24 for all towers. SEBAL ET was moderately sensitive to surface roughness length and implementation over regions smaller than ∼10,000 km2 provided lower error than larger regions. Pixel selection based on TR had similar errors as those based on a vegetation index. For maize, MOD16 had lower error in mean seasonal evaporative fraction (−0.02) compared to SEBAL Λ (0.23), but MOD16 significantly underestimated the evaporative fraction from rice (−0.52) and cotton fields (−0.67) compared with SEBAL (−0.09 rice, −0.09 cotton). MOD16 had the largest error over short crops in the early growing season when vegetation cover was low but land surface was wet. Temperature‐based methods like SEBAL can be automated and likely outperform vegetation‐based methods in irrigated areas, especially under conditions of low vegetation cover and high soil evaporation.