Researchers from the Civil Engineering Department at IIT Delhi have introduced a groundbreaking tool, ML-CASCADE, aimed at Automated Mapping of Landslide extents using satellite data. This cloud-based machine learning tool is designed to provide rapid and accurate mapping of landslides, which is essential for post-disaster damage assessment.
ML-CASCADE simplifies the process of Automated Mapping of Landslide Events by requiring only two inputs: the approximate date and location of a landslide. Once these inputs are provided, the tool can map a complex cluster of landslides in five minutes, while simple landslides are mapped in just two minutes. This swift mapping process improves on traditional manual methods, which are time-consuming and often inaccurate, particularly in remote or vegetation-sparse areas.
Also Read – Contour Connection Method: An Automated Method for Landslide Identification with LiDAR
According to Prof. Manabendra Saharia from IIT Delhi, who co-developed the tool with PhD scholar Nirdesh Kumar Sharma, existing models based on vegetation indices fail to map landslides in regions with minimal vegetation. ML-CASCADE, however, enhances the Automated Mapping of Landslide events by integrating data from various sources, including satellite imagery, terrain data, and soil information, into a machine-learning model.
The tool uses a decision-tree-based classification approach to detect landslides through both pixel and object-based methods. By combining pre- and post-landslide Sentinel-2 satellite bands with slope data from NASA’s Digital Elevation Model, it offers a highly accurate solution for the Automated Mapping of Landslide areas. Unlike pre-trained models, ML-CASCADE adapts to each landslide event based on local terrain and environmental factors, creating a custom model in real time.
Designed to be user-friendly, ML-CASCADE’s interface was developed in consultation with both technical and non-technical users. The results are generated within minutes via Google Earth Engine, with no need for downloading data, and can be directly used in GIS systems for further analysis.
Validated through case studies in India’s Himalayas and Western Ghats, this tool is already aiding in the development of a national landslide inventory. The Automated Mapping of Landslide tool is expected to be a critical component in future landslide early warning systems.
Paper Link- https://link.springer.com/article/10.1007/s10346-024-02360-3
Source: IITK
