A new study of climate extremes since 1988 finds that many regions have seen increases in deaths due to floods, storms and extreme temperatures. In human terms, the harm comes not just from deaths, but also from lost labor and property damage. (And this doesn’t consider damage to species and ecosystems.) A new look at trends and outliers has been published in Geophysical Research Letters.
How the study was conducted
B. B. Cael, of the Department of Geophysical Sciences at the University of Chicago, used data from the publicly available Emergency Events Database (EM?DAT). He selected from a subset of 1,974 disasters by trimming the data in four ways: Events before 1988 were not considered, nor events after 2024.
Only events with 30 or more deaths were included, which Cael found accounted for 95% or more of all deaths; he found that the thresholds were not significantly sensitive to these limits. Only climate-related hazards were considered—climatological, hydrological and meteorological—with earthquakes, volcanoes, pandemics and infestations not included.
Wildfires, droughts and glacial lake outbursts were also excluded, as they make up only 2% of remaining events. Events were then grouped by type—extreme temperatures, floods and storms—and by five continents that included Latin America and the Caribbean together, while South America and Australia were excluded.
After this filtering, Cael finds the count is “300 extreme temperature events, 1,088 floods, and 586 storms, totaling 940,895 deaths. … By region, there are 270 events in Africa, 1,215 in Asia, 181 in Europe, 220 in Latin America, and 88 in North America.”
Statistical methods and findings
To determine the magnitude of deaths, Cael used a Generalized Pareto Distribution (GPD), because these events represent the upper tail of deaths from all climate hazardous events, where death events are distributed along a timeline.
A GPD is often used to analyze the tail ends (large or small values) of other distributions, such as a normal (Gaussian) distribution (sometimes called a Bell Curve), an exponential distribution of some value or a continuous uniform distribution. Mathematically, it can be proven that the tails of many distributions converge to a GPD, and a GPD is often used for climate hazards, for example for hurricane damage.
Since the events considered in Cael’s research are only the deadliest events, this mathematical result ensures that the GPD will be an accurate approximation of the true distribution of deaths.
Cael found that the filtered observations are well?described as GP distributed, with a GPD for each region of each type of event. These distributions were then used to estimate whether events of a given type in a particular region are becoming more or less deadly over time.
Regional trends and notable events
Cael’s analysis examined three areas in particular. Extreme floods and storms in Asia have not become less frequent, but they have become less deadly due to a reduced vulnerability due to an increase in adaptive capacity, saving about 350,000 lives over the years 1988 to 2024. (In particular, his statistical analysis found there was a 95% probability that the number was between 220,000 and 560,000 lives saved.)
In other words, Cael writes, “Development?driven vulnerability reduction in Asia over 1988–2024 has resulted in hundreds of thousands few[er] people dying from floods and storms compared to a scenario where vulnerability stayed constant and the Asian population continued to grow at the pace at which it did.”
This is a large number of lives saved, both absolutely and relatively. From the EM-DAT, Cael further finds that Asian flood and storm events with 30 or more deaths were responsible for 511,875 recorded deaths, “meaning the above result corresponds to a 40% (with a 95% confidence interval of 30% to 52%) reduction in Asian flood and storm mortality over 1988–2024, even without accounting for intensifying precipitation extremes due to climate change.”
Storm Daniel struck the Mediterranean area in September 2023 and was the deadliest Mediterranean tropical-like cyclone in recorded history. Responsible for about 13,200 deaths according to EM-DAT, most of them in Libya where the Abu Mansour Dam and the Derna Dam collapsed, releasing 30 million cubic meters of water. It caused at least €20 billion (US$21 billion) in damages.
Cael found deadly floods in Africa have become more frequent due to population events, but excluding Daniel from African storm deaths over 1988-2024 shows no trend in such deaths, making it a one-in-two-century outlier event, and not part of the trend.
Cael’s analysis also found, as has been previously reported, that heat waves in Europe have become deadlier over time, because heat waves are occurring more often than cold snaps. Over the time period studied, the timing of fatalities has trended from autumn and winter (October to February) to spring and summer (May to August).
Heat waves are becoming more common relative to less deadly cold snaps; but they’re apparently not due to exposure, Cael concludes, because Europe’s population has grown less than 4% since 1988.
Furthermore, Cael found that there are no statistically significant trends in the Americas.
Cael told Phys.org that for the decline in Asian deaths due to floods and storms, “I was a bit surprised at how large that estimate turned out to be.” He added, “It’s of course important to remember that these results are very high-level and continent-scale, though they don’t at all mean that Asian flood and storm risk due to climate change isn’t a massive concern.”
Written for you by our author David Appell, edited by Sadie Harley, —this article is the result of careful human work. We rely on readers like you to keep independent science journalism alive.
If this reporting matters to you,
please consider a donation (especially monthly).
You’ll get an ad-free account as a thank-you.
Publication details
B. B. Cael, Climate Hazard Mortality: Diagnosing Trends and Outliers, Geophysical Research Letters (2025). DOI: 10.1029/2025gl119218
The content is provided for information purposes only.