New Report Warns AI Could Close a 70-Year Gap in Weather Forecasting for Health or Widen It Without Deliberate Action

A 10-day forecast now takes minutes to produce, on a single computer chip, putting life-saving warnings within reach of the world's most vulnerable communities.

NEW YORK, NY | September 22, 2026 - A new report from the University of Chicago's Institute for Climate and Sustainable Growth, supported by The Rockefeller Foundation, finds that Artificial Intelligence (AI) could put locally tailored weather forecasts within reach of the low- and middle-income countries facing the greatest health risks from climate change, at a cost they can afford. However, the report warns that without deliberate action, AI forecasting will leave behind the same populations that conventional forecasting did. The status quo, it argues, is now a choice rather than a necessity.

This is possible because artificial intelligence is transforming weather forecasting. AI weather models now match or beat the best physics-based models behind most forecasts today, and they run at a fraction of the cost and time. For decades, producing a reliable weather forecast required a supercomputer that cost around $100 million. A trained AI model now produces a 10-day forecast in minutes on a single computer chip.

"AI weather models offer the opportunity to democratize forecasting, so that 70 years of progress can finally reach everyone," says report author Amir Jina, an assistant professor at the University of Chicago Harris School of Public Policy. "By putting this tool in the hands of health experts and meteorologists around the world, forecasts can now be tailored to the decisions they need to make to protect health and ultimately save lives."

Eighty-one percent (81%) of national meteorological services report providing climate services for health, while just 23 percent of health ministries operate a surveillance system that uses meteorological information (World Meteorological Organization, 2023). The report finds this is not only a delivery failure. Forecasts have rarely been built around the decisions they are meant to serve, leaving health officials with information they cannot act on. Until public health authorities state which decisions they need weather information for, and at what lead time, forecast producers have nothing to build toward.

Many of the health decisions that matter most fall at lead times where forecasts barely exist. Among the potential applications that would benefit from this kind of advance warning are heat response and vector-borne disease control: Two to four weeks of advance warning would let a health system stage its response to a heat episode instead of reacting to one, and would let a malaria program time spraying and larval source management to when they would be most effective rather than to a fixed calendar. The report identifies that window as the least invested-in range in forecasting.

Evidence from the Field: Nigeria

A case study in the report documents the mismatch. In a cross-sector exercise in Nigeria, nearly every preparedness action participants identified required weeks of notice, while the forecast products they requested were available only in days or in seasons. No participant asked for anything in the intervening range, because no such product had ever been available to them.

"AI-powered weather forecasting can bring critical information to communities facing the greatest health risks from climate change and the least capacity to prepare," said Dr. Naveen Rao, Senior Vice President of Health at The Rockefeller Foundation. "But the technology alone isn't enough - we need to invest in the data, training, and local ownership that makes these early warnings worth acting on."

Four Priorities for Action

Better forecasts on their own will not be enough. Fixing forecasts that were too coarse, too late or too poorly tailored is a real achievement, the report says, but it is not the same as improving health decisions or outcomes. Accurate, locally tailored forecasts must be paired with the health data to evaluate them, the institutions to act on them, and clear responsibility for who acts when a warning arrives.

The report sets out 22 recommendations across four types of investment. It identifies a top priority in each category:

• Discovery: Build benchmarking that evaluates forecasts on whether they support a real health decision, beginning with extreme heat. How models are evaluated determines which models are developed next.
• Evidence: Require every service to start by stating the health decision it serves, then fund evaluation against that decision from the first cycle. Scaling funders need evidence that services improve health before they will commit, and so far there is very little.
• Scaling: Create a dedicated mechanism to carry proven climate-health services to national scale, brokering between governments and development banks. Small grants at this stage determine whether far larger loans include a well-designed health component.
• Sustainability: Write health into the national AI and data policy frameworks now being drafted, which are on short cycles and difficult to reopen. Climate policy is the cautionary example: all 59 national adaptation plans reviewed in 2025 name health as a vulnerable sector, yet fewer than half name a lead agency for it and 0.2 percent of international adaptation finance goes primarily to health.

The report also argues these changes have to be built where they are most needed. That means situating decision-making, budgets, ownership and long-term control of AI weather tools in the countries carrying the greatest burden, starting with the training that lets weather forecasters and health practitioners there test, adapt, use and eventually build AI forecasts themselves.

Human-Centered Weather Forecasts Initiative

These approaches are already being tested. Through the Human-Centered Weather Forecasts Initiative, University of Chicago researchers have worked with the Indian government and are now working with the Ethiopian government to build forecasts around locally identified needs, disseminate them and measure whether they change outcomes.

The same team, with partners, is training meteorologists from 30 low- and middle-income countries on how to build AI weather forecasts and tailor them for their local needs.

The AI Weather Forecasts for Health report received input from experts in meteorology, public health, climate, technology, and the social sciences representing 20 organizations. The organizations included the World Health Organization/World Meteorological Organization Joint Office for Climate and Health, NVIDIA, the African Centre of Meteorological Application for Development and the Indian Institute of Tropical Meteorology.

The full report is available here.

About the Institute for Climate and Sustainable Growth

The Institute for Climate and Sustainable Growth leverages the University of Chicago's unique legacy and resources to balance the risks of a changing climate with the essential need for human progress. It does so by combining frontier research in economics and policy, climate systems engineering, and energy technologies with a pioneering approach to energy and climate education. The Institute also seeds interdisciplinary research that explores new topics in this ever-evolving field and deploys practical, effective solutions in countries central to this challenge. For more information, visit climate.uchicago.edu and follow us on LinkedIn @uchicago-climate-energy-institute, X @UChiClimate and Instagram @uchiclimate.

About The Rockefeller Foundation

Investing $30 billion over the last 113 years to promote the well-being of humanity, The Rockefeller Foundation is a pioneering philanthropy built on unlikely partnerships and innovative solutions that deliver measurable results for people in the United States and around the world. We leverage scientific breakthroughs, artificial intelligence, and new technologies to make big bets across energy, food, health, and finance, including with our public charity, RF Catalytic Capital (RFCC). For more information, sign up for our newsletter at www.rockefellerfoundation.org/subscribe and follow us on X @RockefellerFdn, Instagram @rockefellerfdn, and LinkedIn @the-rockefeller-foundation.

A 10-day forecast now takes minutes to produce, on a single computer chip, putting life-saving warnings within reach of the world's most vulnerable communities.

NEW YORK, NY | September 22, 2026 - A new report from the University of Chicago's Institute for Climate and Sustainable Growth, supported by The Rockefeller Foundation, finds that Artificial Intelligence (AI) could put locally tailored weather forecasts within reach of the low- and middle-income countries facing the greatest health risks from climate change, at a cost they can afford. However, the report warns that without deliberate action, AI forecasting will leave behind the same populations that conventional forecasting did. The status quo, it argues, is now a choice rather than a necessity.

This is possible because artificial intelligence is transforming weather forecasting. AI weather models now match or beat the best physics-based models behind most forecasts today, and they run at a fraction of the cost and time. For decades, producing a reliable weather forecast required a supercomputer that cost around $100 million. A trained AI model now produces a 10-day forecast in minutes on a single computer chip.

"AI weather models offer the opportunity to democratize forecasting, so that 70 years of progress can finally reach everyone," says report author Amir Jina, an assistant professor at the University of Chicago Harris School of Public Policy. "By putting this tool in the hands of health experts and meteorologists around the world, forecasts can now be tailored to the decisions they need to make to protect health and ultimately save lives."

Eighty-one percent (81%) of national meteorological services report providing climate services for health, while just 23 percent of health ministries operate a surveillance system that uses meteorological information (World Meteorological Organization, 2023). The report finds this is not only a delivery failure. Forecasts have rarely been built around the decisions they are meant to serve, leaving health officials with information they cannot act on. Until public health authorities state which decisions they need weather information for, and at what lead time, forecast producers have nothing to build toward.

Many of the health decisions that matter most fall at lead times where forecasts barely exist. Among the potential applications that would benefit from this kind of advance warning are heat response and vector-borne disease control: Two to four weeks of advance warning would let a health system stage its response to a heat episode instead of reacting to one, and would let a malaria program time spraying and larval source management to when they would be most effective rather than to a fixed calendar. The report identifies that window as the least invested-in range in forecasting.

Evidence from the Field: Nigeria

A case study in the report documents the mismatch. In a cross-sector exercise in Nigeria, nearly every preparedness action participants identified required weeks of notice, while the forecast products they requested were available only in days or in seasons. No participant asked for anything in the intervening range, because no such product had ever been available to them.

"AI-powered weather forecasting can bring critical information to communities facing the greatest health risks from climate change and the least capacity to prepare," said Dr. Naveen Rao, Senior Vice President of Health at The Rockefeller Foundation. "But the technology alone isn't enough - we need to invest in the data, training, and local ownership that makes these early warnings worth acting on."

Four Priorities for Action

Better forecasts on their own will not be enough. Fixing forecasts that were too coarse, too late or too poorly tailored is a real achievement, the report says, but it is not the same as improving health decisions or outcomes. Accurate, locally tailored forecasts must be paired with the health data to evaluate them, the institutions to act on them, and clear responsibility for who acts when a warning arrives.

The report sets out 22 recommendations across four types of investment. It identifies a top priority in each category:

• Discovery: Build benchmarking that evaluates forecasts on whether they support a real health decision, beginning with extreme heat. How models are evaluated determines which models are developed next.
• Evidence: Require every service to start by stating the health decision it serves, then fund evaluation against that decision from the first cycle. Scaling funders need evidence that services improve health before they will commit, and so far there is very little.
• Scaling: Create a dedicated mechanism to carry proven climate-health services to national scale, brokering between governments and development banks. Small grants at this stage determine whether far larger loans include a well-designed health component.
• Sustainability: Write health into the national AI and data policy frameworks now being drafted, which are on short cycles and difficult to reopen. Climate policy is the cautionary example: all 59 national adaptation plans reviewed in 2025 name health as a vulnerable sector, yet fewer than half name a lead agency for it and 0.2 percent of international adaptation finance goes primarily to health.

The report also argues these changes have to be built where they are most needed. That means situating decision-making, budgets, ownership and long-term control of AI weather tools in the countries carrying the greatest burden, starting with the training that lets weather forecasters and health practitioners there test, adapt, use and eventually build AI forecasts themselves.

Human-Centered Weather Forecasts Initiative

These approaches are already being tested. Through the Human-Centered Weather Forecasts Initiative, University of Chicago researchers have worked with the Indian government and are now working with the Ethiopian government to build forecasts around locally identified needs, disseminate them and measure whether they change outcomes.

The same team, with partners, is training meteorologists from 30 low- and middle-income countries on how to build AI weather forecasts and tailor them for their local needs.

The AI Weather Forecasts for Health report received input from experts in meteorology, public health, climate, technology, and the social sciences representing 20 organizations. The organizations included the World Health Organization/World Meteorological Organization Joint Office for Climate and Health, NVIDIA, the African Centre of Meteorological Application for Development and the Indian Institute of Tropical Meteorology.

The full report is available here.

About the Institute for Climate and Sustainable Growth

The Institute for Climate and Sustainable Growth leverages the University of Chicago's unique legacy and resources to balance the risks of a changing climate with the essential need for human progress. It does so by combining frontier research in economics and policy, climate systems engineering, and energy technologies with a pioneering approach to energy and climate education. The Institute also seeds interdisciplinary research that explores new topics in this ever-evolving field and deploys practical, effective solutions in countries central to this challenge. For more information, visit climate.uchicago.edu and follow us on LinkedIn @uchicago-climate-energy-institute, X @UChiClimate and Instagram @uchiclimate.

About The Rockefeller Foundation

Investing $30 billion over the last 113 years to promote the well-being of humanity, The Rockefeller Foundation is a pioneering philanthropy built on unlikely partnerships and innovative solutions that deliver measurable results for people in the United States and around the world. We leverage scientific breakthroughs, artificial intelligence, and new technologies to make big bets across energy, food, health, and finance, including with our public charity, RF Catalytic Capital (RFCC). For more information, sign up for our newsletter at www.rockefellerfoundation.org/subscribe and follow us on X @RockefellerFdn, Instagram @rockefellerfdn, and LinkedIn @the-rockefeller-foundation.