Artificial intelligence
Artificial intelligence (AI) is changing how weather, climate, water and environmental information is produced and used. WMO is working with its Members and partners to ensure that AI strengthens forecasts, services and early warnings while remaining scientifically sound, transparent, equitable and grounded in trusted international standards.
Overview
AI refers to computer systems that can analyse large amounts of data, identify patterns and generate predictions or other useful outputs. In meteorology, hydrology and related fields, AI can draw on satellite and radar data, surface and river observations, historical records and model outputs to support weather, climate, water and environmental services.
AI is already being applied across the information and warning value cycle. In operational forecasting, AI-based and hybrid systems are increasingly being used to support daily and medium-range forecasts, including the European Centre for Medium-Range Weather Forecasts (ECMWF)’s Artificial Intelligence Forecasting System and systems developed by national meteorological services such as those in Germany, Canada and the United States.
AI can also help improve tropical cyclone track prediction, support nowcasting for rapidly developing hazards such as heavy rainfall, lightning and strong winds, and assist with flood forecasting, impact assessment, decision support and communication. Applications for nowcasting, sub-seasonal and seasonal prediction are developing rapidly, but many remain under testing, evaluation or pilot implementation. Within Early Warnings for All (EW4All), AI can contribute to disaster risk knowledge; detection, observation, monitoring, analysis and forecasting; warning dissemination and communication; and preparedness to respond.
AI-based weather prediction systems can generate forecasts rapidly and, once trained, often require much less computing power to run than traditional numerical weather prediction models. This changes the economics of forecasting: although training and maintaining advanced systems still require extensive data, computing infrastructure and specialised expertise, the cost of producing forecasts can be much lower. AI could therefore help make advanced prediction capabilities more accessible to National Meteorological and Hydrological Services (NMHSs) that cannot operate large traditional modelling systems, while supporting more frequent updates, finer-resolution information and services better tailored to local needs.
AI does not remove the need for observations, physical science or traditional forecasting systems. Many AI weather models are trained using decades of observations, reanalyses and numerical model outputs produced through international cooperation. Physics-based models remain important for representing processes, providing training data and assessing whether AI results are scientifically plausible. AI therefore works as part of a wider forecasting system, alongside established technologies, operational infrastructure and human expertise, rather than replacing them.
Impact
AI can strengthen forecasts and early warnings by processing information quickly, combining diverse data sources and recognising complex patterns. In nowcasting, deep-learning techniques can help follow fast-developing storms and predict conditions over the next few minutes or hours. In hydrology, AI and machine learning can support river flood forecasting, water availability assessments and decisions about how water resources are distributed among different users and sectors.
AI also has potential beyond prediction. It can support the integration of hazard, exposure and vulnerability data, help translate forecasts into expected impacts and assist with multilingual or more accessible warning information. Used across an end-to-end warning system, these tools may help connect risk knowledge and hazard monitoring with communication, preparedness and early action.
AI may lower some barriers to advanced forecasting. NMHSs could gain access to skilful prediction tools without operating the largest traditional modelling systems. Regionally led projects can also develop methods suited to local conditions. For example, the Advancing Nowcasting with Deep Learning Techniques (ANDeL) project, endorsed by the WMO World Weather Research Programme, uses satellite data to develop short-term rainfall forecasts for West Africa, where weather radar coverage is limited.
The benefits are not automatic or equally distributed. AI models depend on the quality and coverage of the data used to train them. They may perform less well in observation-sparse regions, during rare events or under conditions that differ from those data. Some outputs may be difficult to explain or physically inconsistent. Access to data, computing services and specialised skills also remains uneven.
For warnings that protect lives and property, speed must be matched by reliability, accountability and public trust. Human experts remain responsible for evaluating AI outputs, interpreting uncertainty, considering local vulnerabilities and turning technical information into action. NMHSs must remain the authoritative source of public weather, climate and hydrological warnings.
WMO's response
WMO provides an international framework for integrating AI into operational weather, climate and water services. Its role is to help Members assess new systems, establish standards, share knowledge and move promising tools from research and pilots into sustainable national operations.
In October 2025, an Extraordinary session of the World Meteorological Congress endorsed the integration of AI into WMO’s global observation, data processing and forecasting infrastructure. The Congress also called for stronger collaboration among the public, private and academic sectors and requested a new WMO Integrated Processing and Prediction System (WIPPS) strategy incorporating AI. It also emphasised open data, open-source tools, ethical safeguards, support for Members with fewer resources and the continued authoritative role of NMHSs. The Joint Advisory Group on Artificial Intelligence provides oversight and guidance for this work across WMO.
Rigorous verification is essential before AI products can be used operationally. WMO expert bodies are developing approaches for comparing AI-based, physics-based and hybrid prediction systems, including their forecast skill, physical consistency, robustness and performance across regions and events. Shared benchmarks, intercomparisons and transparent documentation can help users understand their strengths and limitations.
WIPPS pilot projects are testing how AI can support services in practice. In Malawi, Data-Driven Weather Forecasting for All combines Bris, a high-resolution AI forecasting model developed by the Norwegian Meteorological Institute, with ECMWF’s Forecast-in-a-Box system, which packages the tools needed to run forecasts locally. The project enables Malawian forecasters to produce AI-based forecasts, compare them with existing forecasts and assess their accuracy and usefulness for early warnings.
Other initiatives include the AI for Nowcasting Pilot Project, which assesses AI-based nowcasting products, their operational value, real-time dissemination and opportunities for technology transfer. The ECMWF-WMO AI Weather Quest meanwhile is developing an open and standardized framework for evaluating AI-based subseasonal-to-seasonal forecasts.
WMO is also expanding work on AI in operational hydrology. Together with Google and the NMHSs of the Czech Republic, Nigeria, Uruguay and Viet Nam, WMO has carried out a pilot study exploring AI and machine learning approaches to river flood forecasting under different geographical and hydroclimatic conditions. The study is intended to inform how new approaches could complement national forecasting capacities.
In May 2026, WMO began implementing a five-year initiative supported by the Ministry of Climate, Energy and Environment of the Republic of Korea to strengthen flood forecasting in selected countries. The project will explore approaches aligned with the WMO Flood Forecasting Framework, including AI and machine learning, and draw on the Republic of Korea’s operational experience.
WMO leads the inter-agency Working Group on Digital Transformation for Hydrology and Water Resources. With United Nations entities, scientific and technology partners, the group supports collaboration, capacity development, standards and enabling frameworks for emerging technologies in operational hydrology and water resources management, including AI.
WMO is also a founding member and co-leader of the Global Initiative on Resilience to Natural Hazards through AI Solutions, which brings together United Nations agencies and expert communities to advance research, innovation, standards and governance frameworks for the responsible use of AI in disaster risk management. Through this initiative, WMO supports trustworthy, safe and interoperable AI applications that can strengthen early warning systems and resilience to natural hazards.
WMO is helping shape the governance of AI for weather, climate, water and environmental services. The WMO-UAE Conference Statement on AI for Weather Prediction set out a shared vision for trusted, transparent and interoperable AI systems based on open data and tools, coordinated benchmarks, human-centred service design, capacity development and the authoritative role of NMHSs. In 2026, WMO also convened a governance consultation with developing-country NMHSs ahead of the United Nations Global Dialogue on AI Governance, focusing on trust, authority, data stewardship, equitable access and public-private roles in AI-enabled weather and climate services.
Open and high-quality Earth system data remain fundamental to these efforts. WMO promotes international data exchange, transparent evaluation and open tools so that AI can strengthen public weather, climate and water services across its Members.
