Extreme events and what they cost
Losses are the least ambiguous climate dataset available, because someone has to write the cheque. They are also incomplete in a specific and important way: they measure insured value, which is concentrated in rich countries, so they systematically understate what happens everywhere else.
Insured losses by year
Six consecutive years above $100bn. The trend is driven by a combination of more intense events, more property in harm's way, and higher rebuild costs — the attribution between those three is genuinely contested, and this chart does not settle it.
2025 was a secondary-peril year
A record 92% of insured losses came from wildfire, severe convective storm and flood rather than from tropical cyclones. The industry calls these "secondary" perils. In 2025 they were the entire story, and the reinsurance market is repricing on that basis.
US billion-dollar disasters — a dataset that no longer exists
NOAA's billion-dollar disaster database was the most-cited series in this field: 45 years, inflation-adjusted, methodologically consistent. It was retired in May 2025 with no successor. The series below therefore ends at 2024, and it is going to keep ending at 2024.
Why this matters for anyone paying for data. When a public statistical series is discontinued, the commercial alternatives are neither free nor consistent with what came before, and the break is easy to hide. Any chart you see with a 2025 US billion-dollar disaster count is using a different method than the years before it. This site would rather show a series that stops than a series that quietly changes definition.
Notable events, 2023–2025
Selected for scale, for being a record of some kind, or for showing a peril doing something new — Helene flooding Appalachia hundreds of miles inland, Melissa hitting Jamaica at category 5.
Latest coverage
Headlines from third-party publishers, linked in full. Nothing here feeds the figures above — it is context, not source data. All headlines.