Quantification of the effect of hydrological drivers on actual evapotranspiration using the Bayesian model averaging approach for various landscapes over Northeast Asia

Abstract

Actual evapotranspiration (AET) is a critical component of hydrological processes. Accurate estimation of AET variation and its driving hydrological factors is vital for natural hazard adaptation and water resource management. This study first used AET observations from FLUXNET at 12 flux tower sites representing four different land cover types and three climate zones over Northeast Asia. Then the relationships and seasonal patterns among AET and seven hydrological variables of gross primary production (GPP), wind speed (WS), air temperature (Ta), net radiation (RN), soil moisture (SM), precipitation (P), and air pressure (Pa) were assessed using Bayesian model averaging (BMA). Finally, the performance of BMA was examined at all flux tower sites, and the impact of factor importance to predicting performance was analyzed. Generally, the results demonstrated higher AET values at sites associated with temperate and cropland areas. Analysis of the contributions of key elements to AET based on BMA suggested that RN and GPP were the two most sensitive factors for AET variation over all the selected sites. With the exception of RN and GPP, WS strongly influenced AET estimations in the wetland and cold regions. AET variation in the cropland and grassland regions was significantly affected by SM. Regardless of landscape type and climate zone, RN was the most important variable in each season, followed by GPP, WS, Ta, and SM. Ta was found to be important in spring, while WS showed a strong influence on AET in winter. In terms of prediction performance, the mean correlation coefficient and root mean square error (RMSE) of the BMA model was 0.92 and 0.49 mm/day, respectively. Using more factors with a posterior inclusion probability larger than 0.9 improved the performance of BMA. Overall, the strong correlation and low error demonstrate that BMA is an effective approach to capturing AET and exploring its interaction with climatological anomalies across different regions, especially under varied agricultural conditions, which could improve crop water management. © 2022 Elsevier B.V.