Thinking · Method

81% of weather services offer health data. 23% of health ministries use it. The gap is a missing decision.

Abhijith Magal · 24 September 2026 · 8 min read

The case for AI weather forecasting in health arrived in September with one pair of numbers attached. Per the World Meteorological Organization's 2023 stocktake, quoted in a new report from the University of Chicago's Institute for Climate and Sustainable Growth, 81 per cent of national meteorological services say they provide climate services for health. Only 23 per cent of health ministries operate a surveillance system that uses meteorological information. Four in five weather agencies are producing for a customer that, three times out of four, is not consuming.

The report, AI Weather Forecasts for Health: The Case for a Decision-First Approach, published on 22 September with support from the Rockefeller Foundation and input from experts across 20 organisations, is led by Amir Jina at the Harris School. Its technical premise is that the old constraint has fallen. The first revolution in forecasting, seventy years ago, ran on physics models that needed supercomputers costing on the order of a hundred million dollars, so tailored forecasting belonged to the wealthiest weather services. AI models now produce a ten-day forecast in minutes on a single chip. The capability that was rationed by hardware is, in the report's phrase, available to anyone.

The gap was never mainly a delivery gap

The easy reading of 81-against-23 is a dissemination failure: forecasts exist, health systems just need the pipe. The report rejects that reading, and its alternative is the piece worth carrying. Forecasts went unused because they were never built for the decisions health officials face. Too coarse to tell one district from another. Too late to act on. Centred on a weather variable that does not track what makes people ill. Judged, throughout, on weather accuracy rather than on health outcomes. Health decisions were never specified, the report notes, and health ministries are the ones who can specify them.

That is the same diagnosis the Andean plan's launch compressed into a sentence a week later, that an alert without a team, a protocol and resources remains just information, which I unpacked in the anatomy of the Americas' plans. The Chicago report supplies the method that sentence implies: start from the decision, then procure the forecast that serves it.

Worked example: the decision before the heatwave

Take the chronic-care arm of a heat response, the measure I argued is this El Nino season's highest-yield unfunded one. Written as a decision spec, it fits on a page. Trigger: a forecast hot spell crossing the local heat-action threshold with five days' lead, from IMD products that already exist. Action: refills brought forward and outreach calls made for every listed patient over 65 or on cardiac, renal or respiratory treatment in the affected blocks. Owner: the programme's district coordinator, by name. Count: share of listed patients reached before day one of the spell, and treatment days maintained through it.

Notice what the spec did. It converted a forecast from information into a budget line with a performance measure, and it told the meteorological side exactly which product matters, block-level heat thresholds at five days, rather than everything at once. The same template writes the dengue version, pre-monsoon larval-control surges on rainfall outlooks, which Chennai's September makes topical in the anticipatory piece, and the drought version, worksite screening circuits timed to employment-guarantee demand. Which districts deserve the first specs is a ranking question; the district index is my answer to it.

The Chicago group's own India connection makes this more than theory. The institute already works with Indian partners on locally designed forecasting, and the Rockefeller Foundation, which backed the report, funds climate-health action globally. For an Indian programme that can show up with a written decision spec, the data side of the collaboration has rarely been this open.

One limit. The 81 and 23 per cent figures are the WMO's 2023 self-reported stocktake, a survey of agencies rather than an audit of services, and the report is a design argument, not a trial; evidence that decision-first services improve health outcomes is exactly what it calls for building. The method stands on its own logic, and the outcomes column of the spec is where each programme supplies its own proof.

Common questions

What does the 81 per cent against 23 per cent gap describe?

Per WMO 2023 figures quoted in the University of Chicago report, 81 per cent of national meteorological services provide climate services for health, while only 23 per cent of health ministries run surveillance systems that use meteorological information. The report argues this is mostly a design failure, forecasts never specified for health decisions, rather than a delivery failure.

How has AI changed weather forecasting for health?

AI models produce ten-day forecasts in minutes on a single chip, work that previously required supercomputers costing on the order of a hundred million dollars. Locally tailored forecasting is now affordable to the low- and middle-income countries facing the greatest climate-health risks.

What is a decision-first approach to climate-health services?

Specifying the health decision before procuring the data: the forecast trigger with threshold and lead time, the action taken when it fires, the named owner, and the health outcome counted. The service is then judged on the decision it moves, not on meteorological accuracy.

What would a decision spec look like for an Indian heat programme?

Trigger: a forecast hot spell crossing the local threshold with five days' lead. Action: advance refills and outreach calls to listed patients over 65 or on cardiac, renal or respiratory treatment. Owner: the district programme coordinator. Count: patients reached before the spell and treatment days maintained through it.

Abhijith Magal

Founder, Syntropy Earth. Nine years across two global pharmaceutical multinationals in patient access and commercial roles, with health-equity work alongside the WHO-Foundation and UNICEF. He works on climate-access: where climate disruption breaks the link between patients and care.

Last updated: 3 October 2026