This thesis explores the application of machine learning to predict the environmental impact of human trampling on natural terrains. The project builds upon the TRAIL simulator, a tool developed to model the effects of human locomotion on natural environments. Due to the computational limitations of the simulator, a U-Net model was trained to predict terrain deformation. The machine learning model successfully predicted vegetation changes but struggled with soil compression and accumulation. In addition to the machine learning approach, the thesis introduces a vegetation recovery model based on logistic growth, which simulates plant recovery after trampling. The model was validated against empirical data from field studies, and the recovery model results were visualized in renderings and animations. Overall, this thesis contributes to more scalable and realistic models for simulating human impact on natural landscapes.