AI Model Revolutionizes Flood Evacuation Forecasting

**Concordia researchers** Mohamed Almetwally Ahmed and Samuel Li have introduced a **machine-learning model** that significantly improves the accuracy of **short-term river discharge predictions**, crucial for flood evacuation procedures. The model utilizes historical data along with nine predictors, including **weather parameters** like rainfall, temperature, and humidity collected over decades. These inputs enable the model to deliver precise **sub-diurnal forecasts**—essentially forecasts over periods of less than 24 hours, which provide more reliable predictions than those made over longer time frames. **The novel method** developed sorts and groups data combinations, constantly iterating until the most accurate predictive model is achieved. This method adjusts based on the specific time frame and river characteristics. For example, a prediction for 12 hours ahead will differ from those for shorter durations. Testing on rivers such as the Ottawa, Boise, and Missouri demonstrated the model's versatility and reliability. Ahmed intends for the model to be integrated into flood evacuation preparedness strategies, providing authorities with vital data to make real-time decisions on evacuation routes and potentially saving lives and property. The research signifies a potential shift in how flood risk is managed, with the possibility that such tools will become commonplace in personal forecasting, similar to weather updates.