EGU24-3187, updated on 08 Mar 2024
EGU General Assembly 2024
© Author(s) 2024. This work is distributed under
the Creative Commons Attribution 4.0 License.

Real-time flood and water level forecasting using AI-based models for early warning and disaster risk reduction

Stefano Bagli, Paolo Mazzoli, Koen van der Brink, valerio luzzi, and mario papa
Stefano Bagli et al.
  • GECOsistema srl, Rimini, Italy (

The increasing frequency and intensity of floods pose a significant threat to lives, property, and infrastructure. Real-time flood forecasting is crucial for early warning systems and disaster risk reduction. However, traditional forecasting methods often have limitations in terms of accuracy and timeliness.

This paper, developed under the framework of AI4Copernicus 5th Calls project, presents a data-driven approach for real-time flood and water level forecasting using AI and machine learning algorithms. The proposed system is based on a hybrid model that combines multiple machine learning algorithms, including DLinear/NLinear, LSTM Hindsight Modelling, and FLEX. The system is trained on historical data on hydrological and meteorological features, and is able to predict water levels at river gauging stations up to the next 9 hours.

The system has been tested on data from the Lamone River in Italy, and has been shown to achieve a mean-absolute-error of only a few (<5) centimetres. This is a very low error margin for this kind of river, and is comparable to or better than the performance of other alternative forecast approaches.

The system has been integrated into GECOSistema's Flood Risk Intelligence platform, named SaferPlaces ( This platform provides a user-friendly interface for accessing flood risk information, and includes features such as real-time flood maps, early warning alerts, and detailed flood risk assessments.

The proposed system has the potential to be a valuable tool for flood forecasting and disaster risk reduction. It can be used to support decision-making at both the local and regional levels, and can help to save lives and property.

How to cite: Bagli, S., Mazzoli, P., van der Brink, K., luzzi, V., and papa, M.: Real-time flood and water level forecasting using AI-based models for early warning and disaster risk reduction, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-3187,, 2024.