As climate change accelerates, societies face increasing exposure to natural disasters. This highlights the need for early warning systems (EWS) that go beyond mere monitoring and assessment of impacts on the environment and people’s lives, provide solutions to improve risk communication and address the decision-making process in a documented manner.
If a few weeks ago a research team led by the Image and Signal Processing (ISP-IPL) published a review on the use of artificial intelligence to improve understanding of extreme weather phenomena, with a view to developing more reliable prediction systems, the same team from this institute of the Science Park of the University of Valencia is now co-leading a new work that explores the transformative potential of integrated AI models. Both articles are published in Nature Communications.
This second study highlights the role of AI in developing early warning systems (Early Warming Systems-EWS), which not only predict extreme weather events but also assess their impacts on vulnerable communities and specific ecosystems. These are multimodal AI-based systems, which allow the integration of real-time geospatial, meteorological and socio-economic data, and process large volumes of information, including satellite imagery, in situ data and climate simulations, to assess the relationship between climate events and direct impacts on populations, infrastructure and ecosystems.
"This new dimension of climate warning may be crucial for planning resilient infrastructure and long-term climate change adaptation policy," said Gustau Camps-Valls, IPL researcher and project leader
"The AI systems we are working on not only anticipate the event but aim to simulate possible scenarios, helping communities and response agencies to prepare better and make more informed decisions," explains Max-Planck-director Marcus Reichte in Institute for Biogeochemistry (Jena, Germany) and leader of the work.
The article also introduces the concept of 'decadal warning systems', a key innovation that would enable detailed spatial predictions to be generated with unprecedented anticipation. "This new dimension of climate warning can be crucial for planning resilient infrastructure and long-term climate change adaptation policy," says IPL researcher Gustau Camps-Valls, who Professor of Electronic Engineering and responsible for the project by the Universitat de València.
While the power of AI to model, anticipate and communicate climate risks is growing every day, the study emphasizes the challenges that this technology must continue to face, such as the limited diversity of extreme event samples. This makes the robustness of predictive models difficult. "However, this reality is changing as the worsening climate change and the increasing frequency of extreme events provide new opportunities to train AI models with more diverse and representative data," says Camps-Valls.
As in previous work, the article emphasizes the importance of integrating climate scientists, humanitarian actors and policy makers into research. "Only then can we have effective models," the scientist insists.
Achieving the objectives of the 'Early warnings for all' (UN) initiative, which seeks to ensure access by 2027 to effective warning systems for all vulnerable populations, requires, according to the team, the implementation of these technological advances. However, for this goal to become a reality, greater interdisciplinary collaboration and an approach focused on the needs of affected communities will be crucial. " Integrating AI into warning systems is a kind of catalyst to transform the way we manage climate risks," says Giulia Martini, World Food Programme (WFP) Data Specialist. "This approach represents a paradigm shift in the way we address climate risk, providing data-driven and science-based solutions that can save lives and strengthen global resilience," concludes the expert.
Source: UV News
Don’t miss our podcast The Livermore Bulb with IPL researcher Gustau Camps-Valls and Elena Jiménez, CEO of Helenis Biocosmetics