Combining generalized complementary relationship models with the Bayesian Model Averaging method to estimate actual evapotranspiration over China

Abstract

Evapotranspiration, commonly classified as actual evapotranspiration (AET), wet evapotranspiration (WET), and potential evapotranspiration (PET), is an important component of hydrological processes. Several models have been developed for estimating AET, but none are applicable under all conditions due to the complexity of the algorithms and limitations in parameterization. However, PET and WET can be easily calculated using several simple equations based on available climate data. In this study, a regional reanalysis dataset, the China Meteorological Assimilation Driving Dataset for the SWAT model (CMADS), was used to estimate PET and WET, and a generalized complementary relationship (GCR) was used to convert from PET and WET to AET. Then a Bayesian Model Averaging (BMA) method was applied to merge the GCR-based AETs to improve the accuracy of AET estimations in China. After examining the performance of GCR and BMA models at the flux tower sites in different climate zones, we found that other PET models were better than the PM model for predicting AET using GCR. In addition, the BMA estimations were closer to flux tower observations than were those of the GCR-based AET in each climate zone. The average RMSE of BMA-merged AET was reduced to 0.66 mm/day, compared to 1.82 mm/day in the original GCR model. Moreover, the performance of AET estimated by CMADS and other AET products, such as the Global Land Data Assimilation System and the Moderate Resolution Imaging Spectroradiometer Global Evapotranspiration Project was examined at the flux tower sites. The results showed that the CMADS AET product performed better than other AET products in each climate zone with better statistical values. In general, the results showed that the GCR estimates are promising when combined with BMA for future studies to characterize global AET.