Estimating global evapotranspiration and its components based on a modified Priestley–Taylor jet propulsion laboratory model

Estimates of evapotranspiration (ET) and its components from satellite-based ET models have large uncertainties that hinder the understanding of the global terrestrial water cycle and its feedback to climate change. In this study, a modified Priestley–Taylor Jet Propulsion Laboratory (MPT-JPL) model was proposed by adjusting the relative surface wetness and plant moisture constraints. The MPT-JPL model was validated using the global FLUXNET dataset and compared with the Remote Sensing Penman–Monteith (RS-PM), Global LAnd Surface Satellite (GLASSv4.2), Global Land Evaporation Amsterdam Model (GLEAMv4.2a), and a Simple Terrestrial Hydrosphere model (SiTHv2). The MPT-JPL is more accurate (daily: r = 0.78, RMSE = 0.87 mm d−1; yearly: r = 0.66, RMSE = 0.55 mm d−1) than the RS-PM and comparable to GLASSv4.2 at the site level. The ET and its component estimates from the MPT-JPL were highly related to field observations and had a higher accuracy than those from the original PT-JPL. The MPT-JPL improved the underestimation of transpiration (Tr) and overestimation of interception evaporation (Ei) in tropical forests. The MPT-JPL ET and Tr were spatially consistent with GLEAMv4.2a and SiTHv2 ET and Tr; however, large dissimilarities in soil evaporation and Ei were noted among the three products. The MPT-JPL Ei was closer to the GLEAMv4.2a Ei and larger than the SiTHv2 Ei. The MPT-JPL estimated Tr better in the Amazon forest and tropical non-forest and underestimated ET and Tr for tropical forests in Southeast Asia. This study provides a feasible method for increasing the estimation accuracy of ET and its components.