Incorporating an iterative energy restraint for the Surface Energy Balance System (SEBS)

The Surface Energy Balance System (SEBS) has proven itself as an effective remotely sensed estimator of actual evapotranspiration (ETa). However, it has several vulnerabilities associated with the partitioning of the available energy (AE) at the land surface. We introduce a two stage energy restraint process into the SEBS algorithm (SEBS-ER) to overcome these vulnerabilities. The first offsets the remotely sensed surface temperature to ensure the surface to air temperature difference reflects AE, while the second stage uses a domain based image search process to identify and adjust the proportions of sensible (H) and latent (λE) heat flux with respect to AE. We effectively implemented SEBS-ER over 61 acquisitions over two Landsat tiles (path 90 row 84 and path 91 row 85) in south-eastern Australia that feature heterogeneous land covers. Across the two areas we showed that the SEBS-ER algorithm has: greater resilience to perturbed errors in surface energy balance algorithm inputs; significantly improved accuracy (p < 0.05) at two eddy covariance flux towers in heavily forested (RMSE 62.3 W m− 2, R2 0.879) and sub-alpine grassland (RMSE 33.2 W m− 2, R2 0.939) land covers; and greater temporal stability across 52 daily actual evapotranspiration (ETa) estimates compared to a temporally stable and independent ETa dataset. The energy restraint within SEBS-ER has reduced exposure to the complex errors and uncertainties within remotely sensed, meteorological, and land type SEBS inputs, providing more reliable and accurate spatially distributed ETa products.