AI helps turn meteorological data into early action
As the impacts of extreme weather intensify, data needs to be translated into decisions – and fast. This is a challenge made for artificial intelligence (AI).
Celeste Saulo, Secretary-General of the World Meteorological Organization (WMO), and Monique Kuglitsch, Chair of the Global Initiative on Resilience to Natural Hazards through AI Solutions led by the International Telecommunication Union (ITU), reflect on how AI is reshaping disaster management, where progress is being made, and the challenges that remain.
How do you see the role of AI evolving in disaster preparedness over the next decade?
Celeste Saulo: What is becoming increasingly clear is that AI is not a single tool applied at one stage of the process. It is beginning to shape the entire chain, from Earth and weather observation to decision-making. Starting with observation, AI has a huge role to play, particularly in quality control, managing different data formats, and aggregating different data sources.
Forecasting is also changing. Machine-learning models are now, in some cases, matching or even outperforming traditional dynamical models. Physical models remain essential, but AI can significantly improve their quality.
Another important aspect is accessibility. The outputs of numerical weather prediction models are not easy to access for everyone, and even the best forecast is only useful if it can be understood and acted upon. Translating data into something meaningful for decision-making is a vital area where AI can help.
Monique Kuglitsch: As Celeste mentioned, there is no shortage of ways that AI can contribute to disaster preparedness. To benefit from AI, however, policy must adapt. That means the institutions and experts developing international standards and building policies must stay agile, following the latest research and anticipating future trajectories without losing sight of values. Responsible, trustworthy and inclusive AI has to remain the anchor.
What recent developments in AI-based weather forecasting stand out to you?
Celeste Saulo: With AI, forecasting tools are becoming less expensive and less difficult to maintain, which can help reduce gaps between countries that are currently lagging behind. Initiatives such as the “forecast-in-a-box” from the European Centre for Medium-Range Weather Forecasts (ECMWF) – essentially a set of algorithms, often cloud based – show how forecasting capabilities can be adapted to different national contexts, including in countries where resources are more limited.
Monique Kuglitsch: One clear trend we’re seeing is a move towards greater transparency: open data, shared methods, and interpretable or explainable AI. This is closely tied to an emphasis on end-user needs. We looked closely at the topic of explainable AI in a study with ECMWF, published last year in Nature Geoscience.
In high-stakes disaster contexts, AI systems will only be adopted if people understand how they work and why they produce certain results. Echoing what Celeste said previously, AI’s value lies in its ability to convert data into meaningful information. But what counts as “meaningful” depends on the user. That’s why we advocate designing systems with users in mind – ideally through co-creation.
What are the main challenges or limitations we should be aware of?
Celeste Saulo: The main limitation is data. AI is only as good as the data we use to feed it. In many of the most vulnerable regions, data is incomplete or unreliable, and this is a serious constraint.
Another issue, from the WMO perspective, is that the global infrastructure that supports these developments is not always well understood or sufficiently supported. Data collected in one country enters global forecasting systems because WMO facilitates this exchange. These are not bilateral relationships; it is a multilateral system. If we lose these building blocks – especially data sharing – everything can be at risk.
Monique Kuglitsch: When we began our standardization work at ITU, the technology landscape looked very different. This was before the surge of generative AI, before major regulatory frameworks like the European Union’s AI Act, and before many of the AI systems we now take for granted. At the time, there were no coordinated international efforts to develop standards for AI in disaster risk reduction, and outside policy and regulatory circles, very few people were even aware that this gap existed.
This was a huge challenge – encouraging researchers, scientists and stakeholders to participate in standards development. We had to show why this work matters.
Since then, we’ve been thrilled to see increasing interest from a wide range of communities. It’s encouraging to finally see programmes like Horizon Europe funding opportunities for projects directly linked to disaster risk reduction and standards, such as our recently launched “ARTEMis” project to lay the groundwork for such protocols in Europe.
Bringing together expertise from geoscience, AI, telecommunications, and policy has been essential. The quality and relevance of standards depend on the diversity of perspectives behind them, particularly given how much AI readiness and regulatory environments vary across countries. We need to support these opportunities for exchange.
Celeste Saulo: The priority is not only to improve individual components, but to strengthen the system as a whole. Disaster resilience depends on many elements working together: observation, forecasting, communication, preparedness, and early action.
What steps should be prioritized over the next decade to strengthen disaster resilience?
We need to build and share best-practice models and raise awareness of why institutions and global collaborations matter. The frequency of extreme events is increasing, and we are witnessing avoidable losses of lives and livelihoods. There is no time to lose.
We have a huge opportunity and also a huge responsibility.
Monique Kuglitsch: Many hands make light work. We have a mammoth task and we need everyone’s support.
This article is based on a discussion at the 3rd meeting and workshop of the Global Initiative on Resilience to Natural Hazards through AI Solutions in December 2025.
The latest annual theme for World Meteorological Day (23 March) promotes Observing Today, Protecting Tomorrow.
The conversation around AI and disaster resilience will continue at the AI for Good Global Summit between 7 and 10 July 2026 in Geneva, Switzerland. Find out more.
Header image credit: WMO
