Objective

This project focuses on creating a data-processing pipeline and an algorithm for automated assessment of inland flooding from satellite observations. The pipeline collects satellite images captured before/during/after flood events and utilizes an image segmentation technique, specifically the K-means clustering algorithm, to automatically identify flooded areas.

Satellite data, particularly from Sentinel-2 with its 10-meter resolution, provides a powerful tool for observing and analyzing flood events in detail. While weather conditions such as heavy cloud cover during flood events can be challenging, the use of satellite imagery and machine learning techniques remains valuable. Satellite imagery offers broad and near real-time coverage that helps enhance the ability to monitor and assess flooded areas.

Moreover, this approach seeks to offer insights into enhancing flood detection using drone-based measurements which are not affected by cloud cover.

Area of Interest

The project initially focused on Maine. However, due to limited flood event observations in Maine, it has been expanded to include other states in the New England Region (primarily Vermont) which shares similar flood characteristics (brown water bodies caused by sediment).