Researchers design more effective AI-based early warning systems to enhance decision-making in response to climate change impacts

As climate change accelerates, societies face increasing exposure to natural disasters. This highlights the urgent need for early warning systems (EWS) that go beyond merely monitoring and assessing environmental and human impacts. These systems must also enhance risk communication and provide well-documented support for decision-making processes.

Just a few weeks ago, a research team led by the Image and Signal Processing (ISP-IPL) research group of the UV published a review on the use of artificial intelligence to enhance the understanding of extreme weather events, with a view to developing more reliable prediction systems. Now, the same UV team is co-leading a new study that explores the transformative potential of integrated AI models. Both articles have been published in Nature Communications.

This second study underscores the role of AI in developing early warning systems (EWS) that not only predict extreme weather events but also assess their impacts on vulnerable communities and specific ecosystems. These AI-enabled systems are based on multimodal AI, which allow the real-time integration of geospatial, meteorological and socio-economic data. They process vast amounts of information — including satellite images, on-site data and climate simulations — to evaluate the relationship between weather events and their direct consequences on populations, infrastructure and ecosystems.

“The AI systems we are working on not only anticipate events but also aim to simulate potential scenarios, helping communities and response agencies to prepare more effectively and make better-informed decisions”, explains Marcus Reichtein, director of the Max Planck Institute for Biogeochemistry (Jena, Germany) and leader of the study.

The article also introduces the concept of decadal warning systems, a key innovation that would allow for highly detailed spatial predictions with unprecedented advance notice. “This new dimension of climate warning could be crucial for planning resilient infrastructure and formulating long-term climate adaptation policies”, says Gustau Camps-Valls, professor of Electronic Engineering, researcher at the Image Processing Laboratory (IPL) and the project lead at the University of Valencia.

The New ‘Thinking AI’: Opportunities and Challenges in Impact Prediction

Although AI’s ability to model, anticipate and communicate climate risks is advancing rapidly, the study highlights the ongoing challenges this technology must address. One such challenge is the limited diversity of extreme weather event samples, which hinders the robustness of predictive models. “However, this reality is changing, as worsening climate change and the increasing frequency of extreme weather events provide new opportunities to train AI models with more diverse and representative data”, assures Camps-Valls.

As in previous studies, the article emphasises the importance of integrating climate science experts, humanitarian actors and policymakers into research efforts. “Only by doing so can we develop truly effective models”, insists the scientist.

Towards a Truly Global and Inclusive Early Warning System

Achieving the objectives of the UN’s Early Warnings for All initiative — aimed at ensuring that all vulnerable populations have access to effective warning systems by 2027 — requires the implementation of these technological advances. However, for this goal to become a reality, greater interdisciplinary collaboration and a community-centred approach will be essential. “Integrating AI into warning systems acts as a catalyst for transforming how we manage climate risks”, says Giulia Martini, data specialist at the World Food Programme (WFP). “This approach represents a paradigm shift in the way we address climate risk, offering data- and science-driven solutions that can save lives and strengthen global resilience”, concludes the expert.

Reference:

Early warning of complex climate risk with integrated artificial intelligence. Markus Reichstein,Vitus Benson,Jan Blunk,Gustau Camps-Valls,Felix Creutzig,Carina J. Fearnley,Boran Han,Kai Kornhuber,Nasim Rahaman,Bernhard Schölkopf,José María Tárraga,Ricardo Vinuesa,Karen Dall,Joachim Denzler,Dorothea Frank,Giulia Martini,Naomi Nganga,Danielle C. Maddix&Kommy Weldemariam. Nature Communicationsvolume16, Articlenumber:2564(2025)

Some related news:

Artificial intelligence will be key to mitigating the impacts of climate change

World experts in climate change and artificial intelligence gather this week in Valencia

New AI methodology to improve understanding of the planet in a climate change situation