Flood Prediction for Costa Rica

flooding
remote sensing
google earth engine
machine learning
global south
Published

December 11, 2023

Overview

As part of a broader UN-Habitat initiative, I have been developing a methodology to predict urban flooding probabilities using open source remote sensing data and machine learning models in Google Earth Engine. This work will help urban planners in Costa Rica mitigate biodiversity loss due to urban and agricultural expansion.

A raster map of flood probabilities across Costa Rica

Data and Model

This work builds on a UN-Spider workflow that extracts flooding data for a given date, a paper on build a municipal flood index in Costa Rica, and two papers on building deep learning-based flood prediction models in Seoul and Brisbane. Data come via Google Earth Engine and are used to train a random forest model in Google Earth Engine via the Python API.