Abstract: The evolution of groundwater depth under climate change can dominate the future water resources utilization and long-term planning of the region. In this study, the measured data from the Wudaogou hydrological experimental station in the Huaibei Plain in Anhui Province was divided into the calibration period (1976-2010) and the validation period (2011-2018). An improved Ye Shuiting formula was proposed to improve the simulation accuracy of diving evaporation, and then to establish an iterative groundwater depth algorithm. After that, the dynamic groundwater change was simulated with the influence factors of precipitation and groundwater evaporation. The applicability of the algorithm was also assessed to consider the groundwater evaporation in the study area. The simulation performance of the climate models was evaluated using three assessment indicators, including the mean relative error, correlation coefficient, and coefficient of determination. Among them, the future precipitation of the three climate models in CMIP5 was taken as the main influence factor. At the same time, the simulations of the climate model were corrected to eliminate some systematic biases. A multi-model integrated approach was selected to predict the variations in the groundwater depth under future climate. The results showed that the depth of groundwater fluctuated greatly from year to year, reaching the historical maximum of 3.13 m in 1995 and the lowest depth of 1.71 m in 1980. The multi-year average groundwater depth was 2.55 m. The monthly variation within the year showed a trend of increasing, then decreasing and then increasing, with the lowest groundwater depth in July. The seasonal variation of the multi-year average groundwater depth was ranked in the order of spring (2.52 m) > winter (2.49 m) > autumn (2.21 m) > summer (2.17 m), indicating an overall spatial trend of gradual shallowing from north to south, and the lower groundwater depths in the southeast than that in the southwest. The improved Ye Shuiting formula had better performance in simulating the groundwater evaporation from lime concretion black soil and yellow moist soil, indicating a general consistence with the actual variation. The simulated monthly average changes of precipitation and evaporation capacity in the climate model were more consistent with the measured values, but the overall simulated values were small. Correspondingly, the climate model was significantly improved in simulating the precipitation and evaporation after bias correction. Under RCP4.5 scenario, the integrated multi-model approach predicted that the future precipitation in 2030-2100 would continue to increase by up to 9.1%, and the monthly variation of groundwater depth within the year would firstly decrease and then increase, where the shallowest depth in July was 0.73 m and the deepest depth in January was 4.22 m. In terms of seasonal variations, the shallowest depth in summer was 0.94 m and the deepest depth in winter was 3.98 m, but the overall annual variation showed a decreasing trend with relatively small ups and downs. The finding can provide important implications for the rational regulation of regional water resources.