Soil erosion (SE) is a critical threat to the sustainable development of ecosystem stability, agricultural productivity, and human society in the context of global environmental and climate change. Particularly in tropical island regions, due to the expansion of human activities and land use/cover changes (LUCCs), the risk of SE has been exacerbated. Combining the RUSLE with machine learning methods, SE spatial patterns, their driving forces and the mechanisms of how LUCCs affect SE, were illustrated. Additionally, the potential impacts of future LUCCs on SE were simulated by using the PLUS model. The main results are as follows: (1) Due to LUCCs, the average soil erosion modulus (SEM) decreased significantly from 108.09 t/(km2·a) in 2000 to 106.75 t/(km2·a) in 2020, a reduction of 1.34 t/(km2·a), mainly due to the transformation of cropland to forest and urban land. (2) The dominant factor affecting the spatial pattern of SE is the LS factor (with relative contributions of 43.9% and 45.17%), followed by land use/cover (LUC) (the relative contribution is 28.46% and 34.89%) in 2000 and 2020, respectively. (3) Three kinds of future scenarios simulation results indicate that the average SEM will decrease by 2.40 t/(km2·a) under the natural development scenario and by 1.86 t/(km2·a) under the ecological protection scenario by 2060. However, under the cropland protection scenario, there is a slight increase in SEM, with an increase of 0.08 t/(km2·a). Sloping cropland erosion control remains a primary issue for Hainan Island in the future.