An integrated FR–MCDA framework on Google Earth Engine for automated landslide susceptibility mapping and near real-time early warning in the Chittagong Hill Tracts. It combines terrain, vegetation, and rainfall data to generate dynamic, district-level landslide risk assessments and alerts.
Run Landslide Risk AnalysisA landslide is the downward movement of rock, soil, or debris under the influence of gravity, which can occur slowly over time or rapidly as a sudden and potentially catastrophic event. They are among the world's deadliest geohazards.
Bangladesh and surrounding regions face heightened risk due to steep hill terrain, intense monsoon rainfall, deforestation, and rapid urbanization in vulnerable zones.
Landslides are a recurrent and highly localized hazard in the Chittagong Hill Tracts (CHT) of Bangladesh, driven by steep terrain, intense rainfall, and ongoing land-use changes. Traditional landslide susceptibility models are largely static and do not adequately capture the dynamic nature of triggering factors, limiting their effectiveness for operational early warning and real-time decision-making. This project introduces an integrated Frequency Ratio (FR)–Multi-Criteria Decision Analysis (MCDA) framework, implemented within Google Earth Engine (GEE), to enable automated landslide susceptibility mapping and near–real-time risk monitoring.
The system leverages multi-source geospatial datasets, including terrain attributes derived from SRTM (e.g., slope), vegetation condition from Sentinel-2 NDVI, and precipitation data from CHIRPS.
A binary training dataset is developed using documented landslide inventory points alongside randomly generated non-landslide samples. Key conditioning factors are systematically reclassified into four susceptibility classes, and their relative importance is quantified using the Frequency Ratio (FR) method. These statistically derived weights are then integrated through an MCDA approach to produce a normalized Landslide Susceptibility Index (LSI).
To enhance temporal relevance, the framework incorporates a dynamic rainfall-triggering mechanism by integrating short-term cumulative rainfall (3-day CHIRPS) with the susceptibility layer. This results in a spatially explicit, multi-level risk surface categorized into four operational warning levels: No Warning, Caution, Alert, and Alarm.
An interactive GEE-based dashboard enables users to select specific locations and dates to automatically generate risk maps and warnings, supporting timely and localized decision-making. The overall system provides a scalable and reproducible geospatial intelligence solution that bridges static susceptibility assessment with dynamic hydro-meteorological triggers.
Our models are trained on historical landslide data, satellite imagery, elevation, slope, soil type, and land-use patterns to generate accurate risk predictions.
Powered by Google's planetary-scale geospatial analysis platform, our tool processes terabytes of satellite data in real time to map vulnerable zones.
We deliver timely alerts via loudspeaker announcements, SMS notifications, and emergency service coordination to protect at-risk communities.
Interact with the live geospatial map — explore risk zones, run analysis, and view alert layers across the region.
Masters Student
Department of Meteorology
University of Dhaka
Undergraduate Student (4th Year)
Department of Meteorology
University of Dhaka