Using Machine Learning Applied to Landslides Hazard Map
The Company
The Problem
The Solution
The Results
The institution´s mission is to monitor natural hazards in risk areas in Brazilian municipalities, and conduct research and technological innovation that can improve its early warning system.
The southeastern region is significantly related to landslides representing the second most significant type of disaster that causes deaths in this region.
The development and use of methods that assist in predicting these processes can be great allies in urban planning in municipalities. Integrating geospatial analysis and machine learning (random forest) approaches, and using satellite radar images and drone data could produce new insights that improve the prediction of landslide-prone areas. We have applied this methodology for a small area situated in Campos do Jordao, Matiqueira Ridge, Southeastern Brazil.
The Random Forest method presented a good performance for the project area showing that machine learning techniques can be tools considered in disaster risk management, improving the detection of high-risk landslide areas compared with traditional methods.