Abstract: Against the backdrop of climate change and multiple forms of ecological degradation, geological disasters such as debris flows—characterized by their sudden onset and destructive power—have become increasingly frequent, and pose severe threats to ecological security. Therefore, it is of great significance to accurately assess debris-flow susceptibility and ecological damage, so as to optimize resource allocation for ecological restoration strategies and promote ecological management toward more proactive prevention and targeted intervention. First, a Random Forest model was employed to achieve automated identification of debris-flow impact areas, providing data support for subsequent analyses. Second, a long short-term memory (LSTM) model was used to integrate nine key factors, including elevation, slope, precipitation, and topographic position index, to conduct pixel-level susceptibility analysis across the study area. Subsequently, a standardized anomaly detection method based on NDVI was adopted to quantitatively evaluate the extent of ecosystem damage following the debris-flow event. The results indicated that the affected areas of the debris flow were highly consistent with valley gullies, with sediment and rocks mainly transported along the gullies. Although areas classified as very low risk accounted for 53.7% of the study region, areas of very high risk still required particular attention. Overall, regions with high elevations, steep slopes, and abundant precipitation exhibited significantly greater susceptibility to the debris flow. Among all factors, the topographic position index contributed the most, with an average contribution value of 0.13, indicating that gully catchment areas and high-potential-energy regions were the primary hotspots for disaster occurrence. Based on the NDVI time-series analysis results, significant NDVI anomalies appeared in the disaster-affected areas at an average of 15.2 d after the disaster, i.e., the average remote sensing response time was 15.2 d; the average standardized loss was-3.0, and the local maximum loss reached −24.1, indicating that the danger of ecosystem damage was relatively severe.