Urban green spaces (UGS) are vital for mitigating the urban heat island (UHI) effect, yet the specific contribution of tree species diversity to cooling efficiency across different UGS types and seasons remains insufficiently quantified, especially in semi-arid regions. This study integrates multi-temporal Landsat-8 and Sentinel-2 imagery during 2021-2022 with field-based ecological surveys conducted in 2021 to evaluate how tree diversity regulates the cooling performance of urban green spaces in Xi'an, China. Using downscaled 10-m land surface temperature (LST), cooling effects were quantified using temperature-drop amplitude (TDA) and cooling range (CR). Results show that park green spaces exhibit the strongest cooling capacity (annual mean TDA: 1.06 degrees C; CR: 229 m), significantly outperforming community and street green spaces. Tree-species richness, measured by the Patrick index, shows a strong positive association with cooling intensity (TDA; P < 0.001). Multivariate analysis further demonstrates that the cooling benefit of diverse tree communities is mainly via enhance canopy density. Seasonal variation emerges as the dominant control on cooling performance, with peak effects observed in summer and minimal cooling in winter. At the annual scale, water coverage and season are the most influential factors regulating cooling intensity (R-2 = 0.69), followed by vegetation coverage, while impervious and bare surfaces suppress cooling effect. Notably, building density within a 105-m buffer shows a positive contributing to cooling intensity. Overall, the findings demonstrate that beyond expanding green-space area, optimizing tree species composition and land-cover configuration is essential for strengthening UGS cooling performance.