Better forecasts could offset the deaths climate change itself will cause
As the climate warms and heat grows more lethal, a team of economists has found an unexpected lever of protection: the humble weather forecast. Research published in PNAS suggests that improving short-term temperature predictions to levels meteorologists consider achievable could reduce American heat deaths by 18 to 25 percent by 2100 — enough to offset the additional mortality that climate change itself would bring. The study, drawing on two decades of forecasting data and mortality records, reminds us that how well we see danger coming may matter as much as the danger itself.
- Heat-related deaths in the US are projected to rise sharply as climate change makes extreme temperatures more frequent and severe.
- Researchers found a direct link between forecast errors — especially underestimating heat — and spikes in mortality, revealing that bad information costs lives.
- By surveying professional meteorologists about the future of their field, the team modeled three forecasting scenarios shaped by AI, funding, and climate disruption.
- In most climate futures tested, better forecasts substantially cut the death toll, with the benefit growing larger as warming intensified.
- A quiet alarm runs through the findings: declining investment in forecasting infrastructure could reverse these gains, turning a tool of protection into a source of preventable loss.
As the planet warms, heat will kill more Americans. But a team of economists, led by Derek Lemoine of the University of Arizona alongside colleagues from Columbia, Oregon, and Princeton, has identified something that could soften the blow: better weather forecasts.
Published in PNAS, the research suggests that improving short-term temperature predictions to levels meteorologists consider technologically feasible could reduce US heat-related deaths by 18 to 25 percent by 2100 — enough to cancel out the additional mortality climate change would cause. "We would still rather not experience the climate change," Lemoine says. "But at least we can find ways to potentially cancel out the increased mortality."
The team built their case on two decades of real data — day-ahead National Weather Service forecasts matched against actual temperatures from thousands of stations, layered with county-level death records from the CDC. A clear pattern emerged: when forecasts underestimated how hot it would be, more people died. Accurate warnings gave people time to seek shelter and adjust their plans.
To look forward, the researchers surveyed professional meteorologists in early 2025, asking them to envision their field in 2100. From those responses — shaped by expectations about AI, funding, and climate impacts — the team constructed optimistic, pessimistic, and perfect-accuracy scenarios, then ran them against multiple warming futures. In most combinations, better forecasts meaningfully reduced the death toll, with the benefit growing as heat waves became more frequent.
The study also carries a warning: if investment in forecasting infrastructure declines, forecast quality could deteriorate — and deaths would rise accordingly. Lemoine points out that when governments assign a dollar value to each life saved, the economic case for sustained forecasting investment becomes overwhelming. The question, he suggests, is whether that case will be heard.
As the planet warms, heat will kill more Americans. That much is certain. But a team of economists has found something that might soften the blow: better weather forecasts.
The research, published in PNAS, suggests that improving short-term temperature predictions to match what meteorologists think is technologically feasible could reduce heat-related deaths in the United States by 18 to 25 percent by the year 2100. That reduction would be large enough to cancel out the additional deaths climate change itself would cause. Derek Lemoine, an economist at the University of Arizona, led the work alongside researchers from Columbia, Oregon, and Princeton. "We would still rather not experience the climate change," Lemoine says. "But at least we can find ways to potentially cancel out the increased mortality."
The team built their analysis on two decades of real data. They pulled day-ahead forecasts from the National Weather Service going back to summer 2004, then matched those predictions against actual temperatures recorded by thousands of weather stations across the country. They layered in county-level death records from the Centers for Disease Control and Prevention, filtering out deaths unrelated to weather. What emerged was a clear pattern: when forecasts got the temperature wrong—especially when they underestimated how hot it would be—more people died. Accurate warnings, by contrast, gave people time to adjust their plans, seek shelter, or take other protective steps.
To project forward, the researchers did something unusual. In early 2025, they surveyed professional meteorologists and asked them to imagine the future of their own field. What would forecasting look like in 2100? The meteorologists offered a range of possibilities, shaped by their expectations about artificial intelligence, climate impacts, funding levels, and staffing. From those responses, the team constructed three scenarios: one optimistic, one pessimistic, and one in which forecasts became perfectly accurate.
They then ran the numbers under multiple climate futures. What if temperatures stayed roughly where they were in 2015 to 2020? What if the contiguous United States warmed by 1.6 degrees Celsius, 2.7 degrees, or 3.8 degrees? In most combinations, better forecasts substantially reduced the death toll. The benefit grew larger as warming increased, because more frequent heat waves meant more opportunities for accurate warnings to save lives.
But the research also contained a warning. If investment in weather forecasting declined and forecast quality deteriorated, the opposite would happen. Worse predictions would contribute to more deaths. Lemoine emphasizes that this is not abstract. When the government evaluates whether to fund a program, it assigns a dollar value to each life saved. That value is enormous—large enough to dominate most policy analyses. Apply that value to the thousands of lives that better forecasting could save, and the economic case for sustained investment becomes very strong. "Once you apply that value to the number of lives saved through better forecasting, you end up with a substantial benefit," Lemoine says. The question now is whether that case will be heard.
Notable Quotes
We would still rather not experience the climate change—but at least we can find ways to potentially cancel out the increased mortality.— Derek Lemoine, University of Arizona economist
If investment in forecasting declines and forecast quality deteriorates, lower-quality predictions could contribute to more heat-related fatalities.— Research findings
The Hearth Conversation Another angle on the story
So the study is saying that better forecasts could offset climate deaths. But doesn't that sound like we're just adapting our way out of a problem instead of solving it?
That's exactly right to notice. The researchers are clear: they'd rather not have climate change at all. But they're working in the world as it is. Heat waves are coming. People will die. The question becomes: given that reality, what tools do we have right now to reduce that harm?
And forecasts are a tool that already exists.
Exactly. We have weather stations, we have meteorologists, we have the infrastructure. We just need to invest in making it better. The study found that when forecasts underestimate heat, people don't prepare. They stay outside, they don't check on elderly relatives, they don't open cooling centers. Accuracy changes behavior.
But why does this matter economically? Why does Lemoine keep talking about the dollar value of lives saved?
Because government agencies actually do this calculation when they decide whether to fund something. They assign a number to a statistical life. It's how they compare whether a new safety regulation is worth the cost. In this case, the number of lives that better forecasting could save is so large that it justifies the investment in the forecasting system itself.
So if we don't invest, we're essentially choosing to let people die when we could have prevented it.
That's the hard implication, yes. The research shows that forecast quality isn't inevitable. It depends on funding and staffing. If those decline, accuracy declines, and more people die in heat waves. It's a choice.