From artificial intelligence to climate intelligence
Artificial intelligence (AI) is rapidly transforming many sectors – and climate services is no exception. “From improving short-term weather forecasts to enabling long-range climate projections, AI is opening new pathways for understanding and responding to climate extremes,” says Andrea Castelletti, a professor at the Polytechnic University of Milan(opens in new window). Helping to blaze those pathways is the EU-funded CLINT(opens in new window) project. The project developed an innovative AI framework that improves the detection, causation analysis, attribution and future projections of such extreme events as cyclones, heatwaves and droughts. “By combining machine learning innovation, physical interpretability and operational deployment, CLINT successfully bridges the gap between cutting-edge climate science and actionable climate services,” adds Castelletti, who serves as the project coordinator.
A suite of AI-based solutions
Bringing together more than 60 researchers from across Europe, the majority of whom are early-career scientists, the project set out to address one of climate science’s key limitations: the ability to detect and attribute extreme weather events. To do so, it turned to a range of AI techniques, including convolutional neural networks, autoencoders and causal discovery approaches, amongst others. The net result of this work is a suite of solutions, such as an AI tool for detecting tropical cyclones, a hybrid AI-physics framework for attributing heatwaves, and deep learning methods for reconstructing incomplete precipitation datasets. “By leveraging the versatility of AI, we significantly enhanced the spatial and temporal detection of extremes and strengthened the attribution of individual events to anthropogenic climate change,” explains Castelletti. “We also reconstructed incomplete observational datasets and deepened our understanding of compound and concurrent extreme events.”
Turning climate data into climate action
But these tools aren’t just theory, they’re ready to deliver decision-ready climate services capable of translating complex climate information into actionable insights. For example, the project’s AI-enhanced hydrological post-processing solution is significantly improving streamflow simulations across more than 35 000 European sub-basins and enabling climate-informed energy planning. In the agricultural sector, an AI surrogate model is being used to reduce the computational cost of crop simulations by four orders of magnitude, all while maintaining a high level of predictive accuracy. Furthermore, the project’s impact-based drought indices revealed a northward shift in Europe’s drought hotspots – information that authorities can use to implement effective adaptation measures. “Together, these achievements demonstrate not only scientific excellence but also strong dissemination performance, interdisciplinary collaboration and a high level of operational readiness for real-world application,” notes Castelletti.
Embedding artificial intelligence into climate services
By embedding AI into climate services, CLINT has contributed to more robust, forward-looking and climate-informed decision-making across Europe. “Most importantly, CLINT demonstrated that AI can meaningfully strengthen Europe’s adaptive capacity under climate change, not only through technological innovation, but also through knowledge sharing, open science practices and the empowerment of the next generation of climate researchers,” concludes Castelletti. Researchers are now working on the full operationalisation of the project’s AI-based climate services, ensuring their long-term sustainability and uptake by public authorities, climate service providers and private stakeholders.
