Flood hazard using random forest: a pilot study in Sao Paulo City, Brazil
The Institutions
The Problem
The Solution
The Results
This project was developed within the postgraduate program in Natural Disasters (UNESP and Cemaden partnership), São José dos Campos, Brazil.
São Paulo is one of the largest metropolises in the world and has had a series of events related to flooding associated with the main rivers that cross the city. With the growth of the city, many people occupy risk areas. The flood risk maps do not have the details to identify the areas with the most significant risk potential.
Our solution helps to better detail the areas with the highest risk of flooding by using images with higher spatial resolution, a data-driven solution, and spatial statistics/geostatistics. Hydroclimatological surveys associated with floods require expensive equipment that needs to be installed in the field and is subject to theft and vandalism. Our solution seeks to optimize high-detailed images (LiDAR) and analytical methods based on artificial intelligence and machine learning.
The project is in progress. The analysis of flood points from spatial statistics/geostatistics in the last twenty years has helped establish the areas where we will conduct more detailed investigations.