Climate change intensifies extreme events, causing damage to agricultural ecosystems and exacerbating issues such as erosion and desertification, particularly in geologically fragile regions like Italy. This study develops a methodology for site-specific management of agricultural areas prone to hydrogeological instability, utilizing unsupervised machine learning along with multispectral and Light Detection And Ranging (LiDAR) data acquired by Unmanned Aerial Vehicles (UAVs). The study area, located in the province of Avellino, features diverse cropping systems and a high risk of landslides. The methodology produced maps highlighting the most vulnerable areas, revealing that chestnut groves are especially stressed and at high hydrogeological risk. Inappropriate agricultural practices, such as stubble burning, further exacerbate these risks. The results recommend targeted management interventions to enhance vegetation health and mitigate risks. Additional agronomic studies and an extended temporal dataset are necessary to refine the analyses and improve result accuracy.