Application of a multiple model integration framework for mapping evapotranspiration with high spatial–temporal resolution in the Haihe River Basin, China

Abstract

Evapotranspiration (ET) is a key component of the water and carbon cycles. Because it cannot be observed directly on large regional scales at present, many satellite-based ET datasets have been widely used for different purposes. However, their use has been limited at regional and field scales because of their coarse spatial and temporal resolution. In this study, the Bayesian model averaging (BMA) method was used to simulate daily ET values with 500 m spatial resolution in the Haihe River Basin (HRB) from 2000 to 2019. Validation performed with the in-situ observations showed that the BMA ET values had higher accuracy (R2 = 0.69, RMSE = 1.15 mm/day) than those found with individual models. The BMA ET values also had higher accuracy and more credibility based on a water balance equation validation in the HRB. We used interpolated meteorological datasets, reanalysis net radiation products, and remote-sensing datasets to drive the BMA ET model. The mean annual BMA ET in the HRB from 2000 to 2019 was about 601.8 mm/year. The monthly change characteristics of this BMA ET product reflected water consumption characteristics and irrigation regularity of the winter wheat and summer maize rotation system. The BMA ET in the HRB showed a significant increasing trend from 2000 to 2019 of 3.39 mm/y2. The BMA ET values showed a strong positive trend at 80 % of HRB areas. The increasing trend of the BMA ET values was mainly attributed to an increase in the leaf area index. The high spatiotemporal resolution product showed variations in ET trends more accurately than the coarse resolution products. This high spatiotemporal product has great potential for applications in studies of regional microclimate change, interactions between human activities and climate change, drought disaster monitoring, agricultural policy making, and water security.