Glacial lake outburst flood characteristics and susceptibility assessment prediction in High Mountain Asia based on Black Kite Algorithm optimisation

Glacial lake outburst disasters occur frequently in the high-mountain regions of Asia. Since 1900, more than 180 glacial lake outburst events have been recorded in these regions, posing significant threats to the lives and property of downstream residents. Existing glacial lake evaluation methods exhibit deficiencies in assessing and predicting outburst susceptibility, which limits their applicability in engineering planning and risk management. Therefore, this study focuses on the High Mountain Asia (HMA) region and proposes a glacial lake outburst susceptibility assessment and prediction framework based on Support Vector Machine (SVM), Random Forest (RF), and Multilayer Perceptron (MLP), and optimized using the Black Kite Algorithm (BKA). Utilizing historical glacial lake outburst event data, parameters including dam type, glacial lake area, distance between the glacial lake and the glacier, rear-slope gradient, upstream catchment area, mean annual temperature, and annual precipitation were selected as inputs, with the glacial lake outburst susceptibility classification as the output. The model achieved a prediction accuracy of 87.10% upon validation. Under the SSP245 scenario, the susceptibility to glacial lake outbursts in the High mountain Asian regions may increase to 1.3–1.6 times the current level.