AI-Assisted Flood Early Warning System (2024 Fall)
The Challenge
Communities in the Terai region are highly vulnerable to seasonal floods, but previous warning systems were limited by flawed data and inaccurate rainfall reports. A robust and accurate model was needed to provide reliable evacuation warnings.
Our Solution
A capstone team developed a robust AI-assisted flood prediction framework using the XGBoost model. The system achieved a remarkable 92.6% accuracy in predicting river height and correctly identified 86% of actual flood events without any false alarms.
How It Works
The system analyzes historical data from Nepal’s Department of Hydrology and Meteorology. It identifies 7-day cumulative rainfall as the strongest predictor of rising water, allowing the model to be used for multiple different rivers across the country.
The Outcome
This AI model empowers local communities with advanced warning of dangerous flood conditions. By providing accurate information, the project enables faster disaster response and directly helps save lives in flood-prone regions.