Simulation and Analysis of Evapotranspiration in the Madu River Basin in the Three Gorges Reservoir Region, China

As a key component of regional eco-hydrological processes and the terrestrial water balance, evapotranspiration (ET) plays a crucial role in regional water resources management. In this study, the spatiotemporal dynamics of ET in the Madu River Basin (MRB), an important watershed of the Three Gorges Reservoir region in China, were simulated and analyzed for the period 1981–2018 using the Soil and Water Assessment Tool (SWAT) model, together with scenario analysis, Mann-Kendall (MK) trend test, and multiple regression analysis. Results showed that: 1) the SWAT model demonstrated satisfactory performance in multi-site calibration, with Nash-Sutcliffe efficiency (NSE) and R² values exceeding 0.80, confirming the suitability of the Moderate Resolution Imaging Spectroradiometer (MODIS) ET dataset (MOD16A2) for SWAT calibration at the sub-basin scale. Compared with ET observations from the eddy covariance (EC) system, the ET simulation yielded R² and NSE values of 0.82 and 0.61, respectively, verifying the reliability of the SWAT model for the MRB. 2) Annual mean ET showed a significant increasing trend (2.0 mm/yr) from 1981 to 2018, with a basin-wide mean of 642.6 mm/yr. The long-term mean values and change trends of ET before and after 2001 exhibited different patterns: the long-term mean value increased from 615.9 mm/yr before 2001 to 675.6 mm/yr after 2001; the decreasing trend (–0.0097 mm/yr) before 2001 reversed to an increasing trend (0.0025 mm/yr) after 2001. 3) Higher ET values were generally observed in sub-basins with higher elevations and fewer canyon landforms, whereas downstream areas showed comparatively smaller temporal fluctuations. ET decreased with increasing slope gradient and varied according to soil type. Multi-year mean annual ET among different land-use types followed the order: water > forest > grassland > cropland > built-up area. 4) ET was positively correlated with both air temperature and precipitation, while demonstrating an ‘evaporation paradox’ in October and a significant response to reduced precipitation. Seasonal variation and vegetation phenology joint modulate the effects of land use on ET patterns, with cropland ET exceeding forest ET from August to December and surpassing grassland ET in winter. This study presents a practical approach for ET estimation and provides new insights into the relationships between environmental factors and the spatial and temporal distribution of ET.