Strengthening Coastal Communities’ Resilience Through Better Forecasting and Projecting Compound Flood Risk

Between March 4 and 15, 2019, tropical cyclone Idai hit Mozambique twice. This storm, of rare intensity, was one of the worst disasters to hit the southern hemisphere. Torrential rainfall resulted in devastating riverine and flash floods, that combined with coastal flooding due to storm surge, led to an unprecedented crisis requiring international and humanitarian assistance. Within Mozambique, 602 people lost their lives, and hundreds of thousands were left with no recourse but to be displaced, internally or to neighboring countries. 

This event was catastrophic, and unfortunately it is not an isolated occurrence. Coastal areas are often at risk of compound flooding — flood events due to two or more drivers, such as storm surge, high tides, rainfall and river flows. Mozambique is at high risk of compound floods due to tropical cyclones, making it one of the world’s most disaster-prone countries.  

Better flood forecasting and improved warnings could limit the loss of life and damage.  Supported by an AXA-UNESCO fellowship, Dr Andrea Ficchì, a hydrologist, and environmental engineer based at Politecnico di Milano, has been working on PRINTFLOODS, a project aimed at improving models predicting compound flood risk and identifying high-risk areas. Drawing from interviews with local stakeholders (including humanitarians, disaster managers and forecasters), local ground observations and satellite data, he has been working to refine existing models and datasets to better support humanitarian action and increase the resilience of communities in coastal areas of Mozambique​.

Together with his team, Dr Ficchì has been assessing the accuracy of state-of-the-art tropical cyclone rainfall forecasts and flood prediction models. His work has led to adjusting the models’ scores to better tailor them to decision-making for humanitarian action and emergency response. To improve predictions, the team, working in collaboration with the EU-H2020 CLINT Project consortium, also developed a deep learning framework that can enhance flood hazard datasets like tropical cyclone rainfall and flood maps to increase their resolution, correct spatial errors, and adjust for any potential bias in the data. Their analysis of AI-enhanced forecasts was tailored to the needs of the Mozambique Red Cross and has proven its potential, as the team were able to demonstrate the increased effectiveness for early actions, in terms of cost reductions and loss mitigation. They also worked on other case studies, building an AI model for rapid storm surge and sea-level projections to inform coastal adaptation planning, focusing on the New York City coastline, a metropolitan area highly exposed to coastal flood risk due to sea level rise and storm surge. ​Moreover, the team explored the potential of index-based drought insurance in Italy, based on different indexes and contracts, and are now evaluating the effectiveness of different insurance pooling schemes for floods. 

In Octo​ber 2024, to ensure knowledge sharing and foster collaboration, Dr Ficchi also organized and led a 4-day workshop and training at the Universidade Eduardo Mondlane (UEM) in Maputo, in collaboration with Dr Ascenso from the CLINT team. The workshop brought together 25 participants from UEM and local organisations (national hydro-meteorological agencies, the Mozambique Red Cross and World Food Programme) to help strengthen existing technical capacities for climate services using AI tools. This event laid the foundation for further collaborations between Politecnico di Milano (POLIMI), UEM, national institutes, and humanitarian organisations in Mozambique.  

The approaches developed in the PRINTFLOODS project are designed to be transferable across geographies. The tropical cyclone rainfall and flood forecast data used are globally available, while the AI techniques tested in Mozambique can be deployed in other flood-prone regions. For the flood forecast enhancement techniques, the transfer to other regions would require some AI model tuning efforts using any available local flood data or satellite images. Humanitarian organisations and hydro-meteorological agencies in Mozambique have expressed great interest in the AI-based forecast enhancement tools developed in PRINTFLOODS, to support their anticipatory action planning, improve disaster preparedness, and allocate resources more effectively. The insights and tools first shared during the workshop in October 2024 are still being discussed within local organisations in Mozambique and online meetings with POLIMI are being held to support the transfer from research to operation, so that local communities will benefit directly from faster and more effective actions when flooding is forecast. 

Related Links:  

• Discover the Maputo workshop: https://www.printfloods.eu/2024/10/06/printfloods-in-maputo-workshop-and-training-on-ai-for-climate-services-and-environmental-applications-at-the-universidade-eduardo-mondlane/ 

• Learn more about the PRINTFLOODS project: https://www.printfloods.eu/  

• Read more about using deep learning to improve flood forecasts for humanitarian decision making: Ficchì, A., Fayaz Mir, M., and Castelletti, A.: “Enhancing global flood forecasts in Southern Africa using Deep Learning: A user-oriented evaluation for anticipatory actions”, EGU General Assembly 2025, https://doi.org/10.5194/egusphere-egu25-17698, 2025. 

• Read more about bias correction and enhancements of Tropical Cyclone rainfall datasets: Ascenso, G., Ficchì, A. et al. (2024). “Downscaling, bias correction, and spatial adjustment of extreme tropical cyclone rainfall in ERA5 using deep learning”. Weather and Climate Extremes, https://doi.org/10.1016/j.wace.2024.100724  

• Read more about using deep learning for coastal flood risk and storm surge extremes projections: Longo, E., Ficchì, A., Muis, S., Verlaan, M., and Castelletti, A.: “Projecting storm surge extremes with a deep learning surrogate model”, EGU General Assembly 2025, https://doi.org/10.5194/egusphere-egu25-9088, 2025. 

• Read more about user-oriented evaluation of global flood forecasting systems for anticipatory actions: Hossain, S., Cloke, H., Ficchì, A., et al. (2023). “A decision-led evaluation approach for flood forecasting system developments: An application to the Global Flood Awareness System in Bangladesh”. Journal of Flood Risk Management. https://doi.org/10.1111/jfr3.12959  

• Follow Dr Ficchi on LinkedIn: