The Myjava region in western Slovakia is a hilly landscape characterized by rapid runoff responses and a high susceptibility to soil degradation. Historically, this area has undergone significant anthropogenic transformation including large-scale agricultural collectivization which has exacerbated soil erosion and led to frequent "muddy floods". This study focuses on two specific catchments, Svacenicky Creek and Turá Lúka, to evaluate degradation processes under diverse climatic scenarios, ranging from short-term intense rainfall to long-term continuous precipitation. While physically-based models offer a process-oriented alternative to empirical methods, their application in degradation-prone areas is often hindered by extreme model complexity and the difficulty of obtaining high-quality on-site parameters. A critical unresolved issue is the "skin factor" a parameter representing soil crusting and biological activity which is nearly impossible to measure directly and requires extensive, site-specific calibration to ensure model relevance. Furthermore, current models often treat soil as a homogeneous, rigid element, failing to fully capture the natural variability and complexity of infiltration processes. The application of the EROSION-3D model revealed that land management is the primary driver of soil stability in the region. Winter wheat emerged as the most effective conservation practice, reducing surface runoff by 73% and net erosion by 76% compared to fallow land. Conversely, fallow land was the most vulnerable, yielding a mean net erosion of 7 tons per hectare. Sensitivity analyses identified initial soil moisture as a critical factor, showing a strong correlation (>0.70) with runoff and sediment volume across all soil types. To overcome current data limitations and reduce over prediction or underestimation, this study proposes a technical transition toward hybrid modeling. By integrating remote sensing for real-time parameter estimation and machine learning (e.g., ensemble models and deep learning) to analyze complex erosion factors, the predictive accuracy and robustness of the EROSION-3D model can be significantly enhanced.