Uncertainty Analysis on an Operational Simplified Surface Energy Balance algorithm for Estimation of Evapotranspiration at Multiple Flux Tower Sites

Abstract:

Current regional to global and daily to annual Evapotranspiration ( ET) estimation mainly relies on surface energy balance (SEB) ET models or statistical empirical methods driven by remote sensing data and various meteorology databases. However, these ET models face challenging issues—large uncertainties from inevitable input errors, poorly defined parameters, and inadequate model structures. The eddy covariance measurements on water, energy, and carbon fluxes at globally available FLUXNET tower sites provide a feasible opportunity to assess the ET modelling uncertainties. In this study, we focused on uncertainty analysis on an operational simplified surface energy balance (SSEBop) algorithm for ET estimation at multiple Ameriflux tower sites with diverse land cover characteristics and climatic conditions. The input land surface temperature (LST) data of the algorithm were adopted from the 8-day composite1-km Moderate Resolution Imaging Spectroradiometer (MODIS) land surface temperature. The other input data were taken from the Ameriflux database. Results of statistical analysis indicated that uncertainties or random errors from input variables and parameters of SSEBop led to daily and seasonal ET estimates with relative errors around 20% across multiple flux tower sites distributed across different biomes. This uncertainty of SSEBop lies in the error range of 20-30% of similar SEB-based ET algorithms, such as, Surface Energy Balance System and Surface Energy Balance Algorithm for Land. The R² between daily and seasonal ET estimates by SSEBop and ET eddy covariance measurements at multiple Ameriflux tower sites exceed 0.7, and even up to 0.9 for croplands, grasslands, and forests, suggesting systematic error or bias of the SSEBop is acceptable. In summary, the uncertainty assessment verifies that the SSEBop is a reliable method for wide-area ET calculation and especially useful for detecting drought years and relative drought severity for agricultural production. Reduction of errors from input variables (i.e., LST from MODIS) and key parameters (i.e., hot temperature boundary) in topographically complex areas and high albedo and emissivity surfaces can significantly improve the ET calculation of the SSEBop model.