<p>High-altitude regions are prone to various debris flow hazards, particularly in areas where glacier retreat has occurred. Glacial debris flows demonstrate intricate characteristics and harbor considerable destructive capabilities. The occurrence of these events is likely to pose a significant hazard to the lives and property of the local residents. Due to insufficient observational data, the monitoring and issuance of probabilistic forecasts in high mountain areas pose challenges. To address the above issues, this study investigates the monitoring model based on multi-source remote sensing data and machine learning techniques. By undertaking this approach, a scientific foundation can be established to support initiatives aimed at disaster prevention and mitigation. As an illustration, this research investigated the Guxiang Gully glacial debris flow in Tibet, which stands out as the largest, most severe and challenging to manage within the Par lung Zangpo River Basin. Support Vector Machines and GeoDetector (SVMGD) methods were employed to develop the hazard assessment model. Additionally, a probabilistic model was created using a multi-parameter Weibull distribution. The results indicate that the SVMGD method achieves the best hazard assessment performance, with an AUC accuracy of 0.8513, while the multi-parameter probabilistic model predicts debris flow occurrences with an accuracy of 83.33%. The approach addresses the absence of observational data concerning alpine glaciers. It offers a transferable framework for monitoring and conducting probabilistic research on glacial debris flow.</p>