Landslide Susceptibility in Data-Sparse Regions: A Hierarchical Bayesian Approach
Аннотация
Landslide susceptibility refers to the probability of landslides occurring in a region, which is vital information for effective risk management. To date, landslide susceptibility mapping in data-sparse regions has been challenging due to insufficient historical landslide records and unique region-specific characteristics. In this paper, a hierarchical Bayesian model (HBM) is applied to predict landslide susceptibility in data-sparse regions by effectively borrowing information from other regions. To evaluate the prediction capacity of this method, a database was compiled consisting of 4,696 landslides from Zhejiang and Fujian provinces in China. The HBM was applied to develop a landslide susceptibility model for a data-sparse region in the calibration database. Results indicate that the accuracy of the susceptibility model improves as more region-specific landslide data becomes available. The HBM was also compared against two alternative models across three situations: data-absent, data-sparse, and data-rich. Findings reveal that the HBM performs comparably to or better than the alternative models in data-absent and data-rich regions and outperforms them in data-sparse regions.