Mapping the multi-hazards risk index for coastal block of Sundarban, India using AHP and machine learning algorithms

Global climate change, climate extremes, and overuse of natural resources are all major contributors to the risk brought on by cyclones. In I West Bengal state of India, the Pathar Pratima Block frequently experiences a variety of risks that result in significant loss of life and livelihood. In order to govern coastal society, it is crucial to measure and map the multi-hazards risk status. To depict the multi-hazards vulnerability and risk status, no cutting-edge models are currently being applied. Predicting distinct physical vulnerabilities is possible using a variety of cutting-edge machine learning techniques. This study set out to precisely describe multi-hazard risk using powerful machine learning methods. This study involved the use of Analytic Hierarchical Analysis and two cutting-edge machine-learning algorithms - Random Forest and Artificial Neural Network, which are yet underutilized in this area. The multi-hazards risk was determined by taking into account six criteria. The southern and eastern regions of the research area are clearly identified by the multi-hazards risk maps as having high to extremely high hazards risk levels. Cyclonic hazards and embankment breaching are the main dominant factors among the multi-hazards. The machine learning approach is the most accurate model for mapping the multi-hazards risk where the ROC result of Random forest and artificial neural network is more than the conventional method AHP. Here RF is the most validated model than the other two. The effectiveness, root mean square error, true skill statistics, Friedman and Wilcoxon rank test, and area under the curve of receiver operating characteristic tests were used to evaluate the prediction capacity of newly constructed models. The RMSE values of 0.24 and 0.26, TSS values of 0.82 and 0.73, and AUC values of 88.20% and 89.10% as produced by RF and ANN models, respectively, were all excellent.