Long-term evapotranspiration (ET) information is important for water-resource assessment in basins associated with large-scale water-transfer projects. This study characterized ET variability and ET–environment associations across grassland, cropland, and forestland in the Longchuan River Basin using multisource remote-sensing data from May to September during 2001–2020. Temporal ET variability was assessed using all MOD16 8-day composites overlapping this period, while 13 environmental variables were evaluated using pixel-wise Pearson correlations and land-use-specific ET estimation models were developed using selected predictors. Mean daily ET generally increased from May to August and declined in September. The long-term May–September mean daily ET was highest in forestland (3.12 mm d⁻¹), followed by the basin average (2.24 mm d⁻¹), cropland (1.87 mm d⁻¹), and grassland (1.33 mm d⁻¹). Gross primary productivity (GPP) was the predominant correlate across all three land-use types and remained predominant after removal of the monthly climatological cycle and further detrending. Validation R², RMSE, and NRMSE were 0.84, 0.23 mm d⁻¹, and 17.49% for grassland; 0.84, 0.32 mm d⁻¹, and 16.94% for cropland; and 0.78, 0.41 mm d⁻¹ and 13.22% for forestland, respectively. These findings provide a long-term baseline for land-use-specific ET assessment and water-resource management.