Assessment of two temporal-information-based methods for estimating evaporative fraction over the Southern Great Plains

Abstract

Evaporative fraction (EF) is an important index for partitioning surface available energy. Temporal information derived from multi-times observations of remote sensing can reduce the uncertainty of EF estimates caused by the retrieval error of land surface parameters. Two temporal-information-based EF methods were assessed using in situ measurements from the Energy Balance Bowen Ratio system and the outputs from the second phase of the North American Land Data Assimilation System (NLDAS-2) over the Southern Great Plains (SGP). One is a newly developed method for estimating daily EF using a semi-empirical parameterization based on the day–night differences of land surface temperature (LST), air temperature, and incoming solar radiation; the other is the triangular feature space method constructed by the day–night difference of LST and fractional vegetation cover. The results showed that the two methods could reasonably estimate EF over the SGP region in spatial distribution. The EF estimated from the newly developed method was closer to the in situ measurements with a bias of −0.028 and a root mean square error (RMSE) of 0.194, and it was also closer to the outputs of the NLDAS-2 with an RMSE of ~0.13 compared with the results from the feature space method. A strong relationship between the results from the two methods was found with a coefficient of determination up to 0.824 and an RMSE of 0.143. The RMSE for the EF estimates caused by different vegetation indexes in the feature space could reach 0.15, and larger bias mainly occurred on surfaces with low evapotranspiration surface.