Large areas of cultivated pure forests on the Chinese Loess Plateau (CLP) face the risk of drought-induced mortality. Despite potential advantages of mixed forests in optimizing limited water resources, their evapotranspiration dynamics and climate change responses remain poorly understood relative to pure forests. We examined pure Pinus tabuliformis (PT) and Robinia pseudoacacia (RP) forests and a mixed PT & times; RP forest of comparable age and stand density in central CLP's Ansai County. During the 2022-2023 growing seasons, we measured soil water content, leaf water potential, and daily evapotranspiration rates, and calculated actual evapotranspiration of growing season for all three forests. We validated and applied a modified BBGC-SPERRY model to simulate these hydrological processes. We then projected soil water content, soil and leaf water potential, and actual evapotranspiration of growing season under three Representative Concentration Pathways (RCP2.6, 4.5, and 8.5). The model accurately simulated soil water content (mean difference: 0.01 cm(3) cm(-3)), leaf water potential (0.02 MPa), and daily actual evapotranspiration (0.05 mm) compared with measured values. Both simulated and observed evapotranspiration of growing season followed the pattern: RP > PT & times; RP > PT. Under future scenarios of warming and increased precipitation, soil water content and water potential are projected to decline for all forests relative to the baseline period (1997-2021), while growing season actual evapotranspiration increases from RCP2.6 to RCP8.5. This indicates intensifying soil desiccation across all forest types due to excessive water loss. Compared with RP, PT & times; RP improves projected soil water conditions and reduces growing season actual evapotranspiration across all RCPs. Our findings demonstrate that mixed forests possess a greater capacity to conserve soil water under climate change, providing practical guidance for forest managers on the Loess Plateau to mitigate drought-induced forest mortality.