Landslide Conditioning Factors in the Hilly Regions of Bangladesh

Landslides are a common occurrence in the mountainous southeastern region of Bangladesh, which results in significant economic and human casualties yearly. A commonly employed strategy to mitigate these effects involves creating a Landslide Susceptibility Map (LSM) through the use of appropriate methodology and choosing the appropriate conditioning factors. This study aims to identify and select the optimal landslide conditioning factors by evaluating the predictive capability of the bivariate statistical model— Frequency Ratio. A total of 730 landslides were acquired from earlier reports. 70% of the landslides were used as a training dataset for the model, while 30% of the landslides were used to validate the model. This model used sixteen conditioning factors that include proximity to streams, surface geology, proximity to fault lines, elevation, aspect, slope, profile curvature, general curvature, plane curvature, proximity to roads, average rainfall, topographic wetness index (TWI), normalized difference vegetation index (NDVI), stream power index (SPI), sediment transport index (STI), and topographic position index (TPI). The Receiver Operating Characteristics (ROC) were utilized to determine the level of accuracy of the prediction performance of the susceptibility map. The outcome demonstrates that the area under the ROC curve (AUC) for the FR model (AUC = 0.797). To determine the prevalent landslide-causative factors, the analysis is repeated by removing one factor at a time after considering all sixteen factors for landslide hazard mapping by the Frequency Ratio model. A total of eleven dominant factors are chosen from the sixteen factors. The hazard map is generated with an accuracy of 80.6% by employing only 11 dominant factors. From this analysis, 2.5%, 9.4%, 30.63%, 37.93%, and 19.57% of the landslides fall in the very low, low, medium, high, and very high landslide susceptibility classes, respectively. This study highlights the importance of selecting optimal conditioning factors and demonstrates the value of integrating a statistical model with GIS techniques and provides essential tools for disaster risk management and contributing to the safety and resilience of vulnerable communities in the region.