Extreme Weather Events

Extreme weather event frequency and severity have risen with climate warming. Heatwave frequency has roughly tripled since 1970 in many regions. The fraction of major hurricanes (Cat 3-5) has risen as a share of total. Compound events — fire-heatwave-drought combinations — have become more common. Climate attribution science can now quantify the warming contribution to specific events within weeks of occurrence.

~3×
Heatwave frequency increase in mid-latitudes since 1970
+2.07°C
Northern Hemisphere extratropical summer 2023 vs the 1–1890 CE mean (Esper et al. 2024)
$137B
Global insured natural-catastrophe losses, 2024 ($107B in 2025, Swiss Re)
150×
Lower bound on how much warming raised the odds of the 2021 Pacific Northwest heat dome

Key insights

🥵

Heatwaves are the cleanest climate signal

Heatwaves show clearer climate signal than hurricanes or droughts — they're a direct consequence of warming. Frequency has risen 2-3× in mid-latitudes since 1970. 2021 Pacific Northwest heat dome (49.6°C in BC, Canada) was attributed by climate science to be ~150× more likely under current warming than pre-industrial. The July 2023 heatwaves in Europe and North America would have been virtually impossible without warming, and China's about 50× less likely (World Weather Attribution). Heat mortality has risen — though improved AC and warning systems have offset some of the rise.

🌀

Hurricanes are intensifying, not necessarily more frequent

Hurricane/typhoon total count globally is roughly flat. But the proportion that are major (Cat 3-5) is rising. Rapid intensification (gain of 35+ mph in 24 hours) — like Hurricane Helene (2024) — is becoming more common. Warmer ocean waters provide more energy; warmer air holds more moisture for rainfall. Sea-level rise amplifies storm surge. The composition shift — fewer minor, more major storms — drives most of the increase in damage.

🔬

Attribution science has matured

World Weather Attribution and similar groups now produce attribution studies within days/weeks of major events. Methodology: compare event probability in current climate model simulations vs counterfactual pre-industrial climate. For heatwaves and many flooding events, attribution confidence is high. For convective storms and tornadoes, attribution is harder. The science has informed legal cases (climate litigation), policy debates, and public communication.

Climate-related disasters by decade

Number of disasters (EM-DAT criteria), per decade

Show data table
Climate-related disasters by decade
Climate disasters reported
1970-79680
1980-891620
1990-992620
2000-093650
2010-194080
2020-242050

Key Finding: Reported climate disasters rose about sixfold between the 1970s and 2010s — partly a real increase, largely better reporting.

Global insured losses from natural disasters by year

USD billions

Show data table
Global insured losses from natural disasters by year
USD billions insured
201055
201440
2017150
202090
2022140
2023110
2024137

Key Finding: Insured losses have grown — driven by exposure (more assets in harm's way) and climate-amplified events.

Methodology & caveats

EM-DAT inclusion criteria

CRED EM-DAT counts disasters meeting one of: 10+ deaths, 100+ affected, declaration of emergency, call for international assistance. Reporting has improved over time — earlier decades undercounted. Comparing across decades requires acknowledging improving detection alongside any real increase.

Attribution methodology

World Weather Attribution and Climate Central protocol: run large ensembles of climate models with current GHG levels vs counterfactual pre-industrial. Calculate event probability in each ensemble. The ratio is the 'probability ratio'. Robust for heat and many precipitation extremes; weaker for convective storms and droughts. Studies produced within days/weeks of major events.

Insured vs total losses

Insured losses ($137B in 2024, $107B in 2025, Swiss Re) are a fraction of total economic losses ($220B in 2025). Insurance penetration varies — high in OECD, low in developing countries. Total losses are typically about 2-3× insured, depending on where disasters hit. Cost trends are influenced by exposure growth (more assets at risk), not just hazard intensification. Disentangling exposure from hazard is the methodological challenge.

Reading the disaster counts

EM-DAT counts reported events, so the sharp rise from the 1970s owes much to better reporting, satellites and communications. The same thresholds treat a small local flood and a continental heatwave as one event each. Heat deaths are poorly recorded in much of the world, so heatwaves are probably under-counted relative to storms. The latest bar in the chart covers only 2020-24, five years rather than ten, and should not be read as a decline.

Why dollar losses overstate the climate signal

Insured losses rise with the value of what is built in harm's way, so raw totals mix hazard and exposure. They also favour rich countries: a flood in a low-insurance economy can be deadlier and still barely register. In 2025 the Los Angeles wildfires alone cost insurers about $40 billion, showing how one event near expensive property moves a global total. Normalised series, which adjust for wealth and exposure, are the fairer test of trend, and they remain contested.

Cite this page

GlobeChart. “Extreme Weather Events.” Last updated 2026-10-31. https://globechart.com/co2-emissions/extreme-weather-events/

Please also credit the original publisher named under each chart: EM-DAT International Disasters Database, CRED; Swiss Re sigma, Munich Re NatCat Service. Spotted an error? Tell us.