Improving Landslide Susceptibility Assessments to Protect California’s Transportation Infrastructure

Research Team: Shabnam J. Semnani (lead) and Yi Han

UC Campus(es): 

Additional Research Partners: UC San Diego

Problem Statement: Climate change increases the risk of droughts, wildfires, washouts and landslides, which pose a major threat to the state’s transportation infrastructure. Despite developments in landslide susceptibility mapping and hazard assessment, accurate landslide forecasting remains a major challenge, mainly due to the complexity of landslide triggering mechanisms and insufficient real-time measurements of surface conditions, groundwater conditions, and soil and rock properties. Machine learning techniques have been widely applied in recent years to assess landslide susceptibility over regions of interest. However, many challenges limit the reliability and performance of machine learning-based landslide models. In particular, class imbalance in the dataset, selection of landslide conditioning factors, and potential extrapolation problems for landslide prediction under future conditions need to be carefully addressed.

Project Description: This project introduces methodologies to address challenges using machine learning-based landslide models. XGBoost was used to train a landslide prediction model. Data resampling techniques were adopted to improve the model performance with the imbalanced dataset. Various models were trained, and their performances evaluated using a combination of different metrics. The results show that synthetic minority oversampling technique combined with the proposed gridded hyperspace sampling technique performs better than the other imbalance learning techniques with XGBoost. Subsequently, the extrapolation performance of the XGBoost model was evaluated, showing that the predictions remain valid for the projected climate conditions. As a case study, landslide susceptibility maps in California were generated using the model and compared with the historical California landslide catalog. These results suggest that the model can be of great significance in global landslide susceptibility mapping under climate change scenarios.

Status: In Progress

Budget: $75,908