DEVELOPING REGIONAL-SCALE LANDSLIDE HAZARD AND RISK MODELS IN A DATA-SCARCE ENVIRONMENT – A CASE STUDY IN CENTRAL VIETNAM

The repeated occurrence of destructive landslides poses a significant geological hazard in Vietnam, resulting in human fatalities and substantial financial losses. Predicting the location, timing, and size of future landslides is a fundamental challenge with real societal impacts. This study aims to develop (i) probabilistic landslide hazard models incorporating the spatial, temporal, and size probability of landslides and (ii) proposes a methodology for developing a landslide risk model by introducing the topographic connectivity of the region. A multi-temporal landslide inventory of central Vietnam was used to develop a landslide susceptibility model using a Random Forest algorithm to delineate the spatial probability of landslides. Temporal probability was calculated based on the Poisson distribution model, and subsequently, landslide size probability was calculated by implementing a Probability Density Function. The final hazard map was obtained as the joint probability of landslides, spatial, temporal, and size probabilities. Nine hazard maps show landslide probabilities spanning three orders of magnitude (100 m², 1000 m², and 10,000 m²) for the next 2, 5, and 10 years. The landslide risk model was developed for the study area's road network, integrating the topographic connectivity index with the spatio-temporal probability of landslides of all sizes. The topographic connectivity of the roads was derived from the 30m SRTM DEM and processed using the SedInConnect open-source software. The model's outcome quantifies landslide risk on a scale from 0 to 1, where 1 represents the highest-risk areas prone to landslides.