The potential of remote sensing-based models on global water-use efficiency estimation: An evaluation and intercomparison of an ecosystem model (BESS) and algorithm (MODIS) using site level and upscaled eddy covariance data

Ecosystem water-use efficiency (WUE) is a critical indicator to investigate the interaction between the terrestrial ecosystem carbon and water cycles. WUE, estimated from gross primary productivity (GPP) and evapotranspiration (ET) based on remote sensing (RS)-based ecosystem models and algorithms (e.g., MODIS (MODerate resolution Imaging Spectroradiometer), BESS (Breathing Earth System Simulator)), have been used to quantify the spatiotemporal dynamics of WUE and its responses to environmental changes. However, few studies have assessed the ability of RS-based ecosystem models and algorithms on global WUE estimation. In this study, we evaluated 8-day and annual WUE from MODIS and BESS among different sites, land cover types and climate zones using the FLUXNET2015 dataset as reference, and conducted spatial intercomparisons of annual WUE between MODIS, BESS and an upscaled FLUXNET dataset (MTE). The site level evaluation results showed that BESS WUE had better performance than MODIS WUE at both 8-day and annual scales. Among different land cover types and climate zones, MODIS and BESS WUE performed unsatisfactorily, especially for MODIS WUE in open shrublands and savannas and for BESS WUE in closed shrublands. Additionally, both MODIS and BESS WUE performed poorly in the hot semi-arid climate zone. The spatial intercomparisons over 2001-2011 revealed that BESS WUE had similar spatial patterns of annual WUE and linear trends with MTE WUE over the globe, except at the high latitudes. However, the spatiotemporal patterns of MODIS WUE were different from those of MTE and BESS WUE, particularly in the (sub) tropical arid and semi-arid regions. Our evaluations results suggested that coupling carbon and water cycles into RS-based models could improve their performance on global WUE estimation. Moreover, the performance of MODIS and BESS on global WUE estimation should be further improved, especially for their performance on temporal variation and their performance at the (semi) arid areas and the high latitudes.