Increasing Climate Resilience Through Improved Predictions Using Machine Learning
News

Increasing Climate Resilience Through Improved Predictions Using Machine Learning

May 3rd, 2023

Featured image 1

Andrea Ficchì, post-doctoral researcher at the Environmental Intelligence Lab of the Department of Electronics, Information and Bioengineering of the Politecnico di Milano, is one of the seven selected fellows of the AXA – IOC/UNESCO joint call launched in 2021 as a part of the Ocean Decade.

His research project, named PRINTFLOODS (Prediction Intelligence for Floods), has been endorsed by the Intergovernmental Oceanographic Commission of UNESCO as part of the UN Ocean Decade. Launched in November 2022 and developed under the supervision of Prof. Andrea Castelletti, PRINTFLOODS seeks to improve compound flood forecasting and understanding of uncertainties in future projections. Andrea Ficchì’s work deals with forecast evaluation, machine learning, forecast-based action and humanitarian applications, among other aspects. The goal is to advance the understanding of flood forecast predictability and flood drivers, and consequently improve resilience to natural hazards, especially in communities in sub-Saharan Africa.

The project will have a particular focus on Mozambique. Situated in Southern Africa, it is one of the world’s most natural disaster-prone countries, with a high risk of compound floods caused by tropical cyclones.

Faced with the inevitability of rising sea levels and episodic flooding events, local and national coastal authorities around the world have historically pursued two possible courses of action. “Soft-path” measures, such as early warning and early action systems, real-time emergency management, insurance and disaster financial risk hedging mechanisms, are examples of short-term solutions to increase coastal communities’ resilience to climate change. On the other hand, long-term solutions typically rely on “hard-path” measures. These consist of coastal protection structures – barriers, seawalls and revetments – the reinforcement of houses and infrastructures, as well as the implementation of nature-based solutions, such as land-use planning to reduce impervious surfaces and restore coastal ecosystems. However, as sea levels continue to rise, so will the cost to maintain and improve those defenses, and so will the cost of failure.

Such challenges can be overcome by modulating investments over time, and integrating hard-path measures with soft-path solutions as hedging mechanisms, using decision analytics methods and climate data, to identify robust and optimal pathways.

During his two-year AXA Research Fund fellowship, Andrea Ficchì will employ machine learning to better predict compound flood risk and identify high-risk areas. He will base his work on climate services, i.e. on climate information and products generated to inform and assist in decision-making processes related to climate risk management. Thanks to a multi-source flood extent and impact database, Andrea Ficchì will assess the skill level of current state-of-the-art predictions and guide the machine learning algorithms to enhance them. He will demonstrate the potential value of existing climate services and of the improved predictions by focusing on their capacity to support humanitarian emergency management and weather-based insurance.

His research will help tackle the UN Ocean Decade Challenge 6, which aims to enhance multi-hazard early warning services for all geophysical, ecological, biological, weather, climate and anthropogenic related ocean and coastal hazards, and mainstream community preparedness and resilience.