A simple and actionable solution to flood forecasting

As climate change intensifies, hazards such as floods are becoming increasingly prevalent and catastrophic. With an adequate amount of notice, a community has time to prepare for a disaster and help prevent flood-related damage. In recent years, flood forecasting has improved with advancements in upper atmospheric satellite observation networks and increased computational power, but creating a useful early warning system remains challenging due to uncertainty in modeling, calibrating, and validating data.

Professor Shafiqul Islam of the Department of Civil and Environmental Engineering is developing flood prediction models that will help prepare communities to mitigate floods upfront. Islam and a team of researchers including Tufts alumni Wahid Palash, EG19, and Ali Akanda, EG11, created an accurate, easy-to-use data-driven prediction model called Requisitely Simple (ReqSIM). The team outlined the model in a recent paper in Scientific Reports, a Nature portfolio journal, titled “A data-driven global flood forecasting system for medium to large rivers.” 

The real-time flood forecasting model is easily customizable to any medium or large rain-fed river around the world. It requires easily accessible and minimal data points, making it a viable option for under-resourced areas. To test their model, the team implemented ReqSIM in 51 watersheds in 13 major river basins across five continents. They found useful forecasts up to ten days in advance of weather events for some of the most flood-prone areas in the world such as the Ganges and Brahmaputra rivers in South Asia. 

True to the model’s name, researchers prioritized simplicity in creating ReqSIM and relied on a simple modeling structure and readily available data to make predictions. Their goal was to make a model that’s simple enough to realistically be implemented in a range of situations without compromising accuracy. While there are flood models in existence, very few offer real-time long-range forecasting. Many models that do offer such capabilities are so complicated that it is not feasible to implement them in resource-constrained places where they could be most helpful.

ReqSIM builds on Islam’s previous work in flood forecasting, which has been presented at the 2023 IEEE International Conference on Big Data and published in the Journal of Hydrology. He continues to develop solutions in AI-motivated flood forecasting and leverage technology to develop pragmatic solutions for issues related to climate change. Palash is now a water resources engineer at Integrated Sustainability Consultants and Akanda is an associate professor of civil and environmental engineering at the University of Rhode Island; both earned a PhD in civil and environmental engineering while studying with Islam at Tufts.