Multiscale Landslide Susceptibility Mapping in Myanmar, Southeast Asia

Abstract:

Landslides frequently occur in Myanmar due to a combination of geomorphological, hydrological, cultural, and rainfall-related factors. While less common in some regions and more prevalent in others, the lack of data regarding landslide events makes it difficult to gather information in order to mitigate and anticipate future events. Communities and disaster managers face difficult decisions when presented with information of different scales and certainty. We explore the implications of these uncertainties by using two landslide inventories of different extents alongside with two hazard mapping methods. The resulting susceptibility maps can help guide future inventory work to provide the most effective means of mapping landslide susceptibility.

The two landslide inventories were digitized using high resolution satellite imagery in Google Earth Pro. These inventories were created both within a small subsection of Chin State and on a larger, more national scale but are not fully comprehensive landslide inventories. With these inventories, Weight of Evidence (WoE) and Logistic Regression (LR) approaches were used in order to create landslide susceptibility maps.

In order to compare the accuracy of the susceptibility maps, methods such as area under the ROC curve (AUC) and contingency tables will be used to evaluate performance. Based on preliminary results for the national-scale analysis, LR has produced a higher AUC value than WoE. Currently, work is being done in order to assess how both hazard-mapping methods perform using the sub-state landslide inventory. This presentation will contain our final results regarding the performance of LR and WoE on varying scales of measurement.