The two faces of a subsidence surge
There have been eight subsidence "surge years" in the UK since 1990 — years when insurers saw more than 50,000 claims land at once, mostly from clay-rich ground in the south east. We went back through 35 years of sunshine, rainfall and temperature data to see what these years actually had in common.
The answer: less than you'd think. And the split tells you something useful about 2026.
The list
1990 (disputed — see note below), 1991, 1996, 2003, 2006, 2018, 2022 and 2025 all show up in the industry record as elevated or confirmed surge years. We pulled Jan–Jun sunshine, rainfall and temperature for every one of them — plus 2023 and 2024 as non-surge reference points, and 2026 to date.
Plot them together and two distinct clusters appear.
Cluster one: the classic drought
2003, 2022 and 2025 sit exactly where you'd expect a "drought surge" to sit — hot, sunny, and dry all at once. 2022 is the driest point on the whole chart. These are the years the phrase "hosepipe ban summer" was built for: single-season heat and sun outpacing rainfall, drying clay from the top down within one growing season.
Cluster two: the ones that don't fit
1991 and 1996 are the surprise. Both are confirmed surge years — but both sit at the coolest point on the chart, nowhere near a hot, sunny start to the year. If you only looked at that year's own weather, you'd never flag either of them as a drought risk.
So what actually drove them?
1991 was the tail end of a three-year drought. England and Wales' rainfall ran below average for 1988, 1989 and 1990 in a row — the driest 28-month stretch since the 1850s. No single one of those years looks dramatic on its own. It's the cumulative deficit that mattered, and it bottomed out in 1991, which is exactly when the claims peaked.
A note on 1990: some industry lists cite 1990 (and 1992) as the surge years from this period rather than 1991. Our own read of the rainfall record suggests why. 1990 was the driest individual year of the three (604mm) — the obvious one to point to. But it's the cumulative shortfall across all three years that matters for clay, and that didn't bottom out until the end of 1991. 1992 then saw rainfall recover to above the long-run average, arguing against it being a genuine surge year in its own right. Citing 1990 as "the" year likely misses the lag between when the ground actually reaches its worst state and when the driest single year happened to fall.
1996 is a lag effect. Its own weather was unremarkable — but 1995's summer wasn't. August 1995 saw just a few millimetres of rain across the south east, an extreme desiccation event. Subsidence claims take time to surface: cracks appear slowly, and claims get filed and processed with a delay. Our read is that 1996's "surge" is really 1995's drought showing up in the claims data a year later.
Why this matters for trees, not just weather
There's a physical mechanism behind the lag, and it's one that's well documented in plant science, if not usually connected to insurance data. Trees under water stress don't just wilt and recover once the rain returns — they shift growth toward roots, particularly fine roots, and that shift persists. A tree that dried out one summer has, by the following spring, already built deeper, more extensive root architecture to exploit soil moisture it couldn't reach before.
That's not the tree predicting next year's drought, and it's not the tree drying out faster either — if anything, the deeper roots help it cope better. What changes is what the tree can now do to the ground around it: a tree that dried out last year has already built the root architecture to draw water from deeper down this year — which means the ground beneath it dries out faster, even as the tree itself survives more comfortably.
That's a plausible part of why 1996 surged off the back of 1995's summer, even though 1996 itself wasn't especially dry. It's also worth flagging as a genuine model limitation — most soil moisture models, including the one behind this analysis, assume a fixed root-zone water capacity every year. They can't see a legacy effect like this. If root systems really do deepen after a stress year, the model may be underestimating risk in any year that follows another dry one.
Where 2026 sits
2026 lands in the hot-and-sunny cluster — not the lagged one. Jan–Jun 2026 is the warmest start to the year of any in this record, including every confirmed surge year back to 1990, and the second-sunniest behind 2025. Rainfall, meanwhile, is unremarkable — close to 2018's total, nowhere near the driest years on record. April and May were among the driest on record in their own right, and even June's unusually heavy rainfall wasn't enough to arrest the rise in deciduous SMD — a sign that atmospheric demand has been outpacing supply for months, not weeks.
That combination matters. It says 2026 isn't a "no rain" story. It's a story of atmospheric demand — heat and sunshine pulling moisture out of clay-rich ground — outpacing even a fairly ordinary amount of rainfall. And it's arriving straight after 2025, itself a confirmed surge year. Two hot, sunny years back to back, with whatever root legacy 2025 left behind still in place, is a different risk profile to any single year in isolation.
We're calling 2026 a confirmed surge year — a view shared across multiple insurers we work with, and one backed by our own numbers too: enquiries have more than doubled since late July, sitting on top of the clearest meteorological signal in this whole record.
Methodology note: sunshine and rainfall are the England & Wales and South East England HadUK-Grid areal series respectively (Met Office), except for 2026 where the equivalent MORECS Square 161 site data was used to cover the most recent months. Temperature is the Central England Temperature series. The 1996-as-lag-effect and root-legacy hypotheses are our own reading of the physical evidence, not a claim we've seen made elsewhere — worth treating as informed interpretation rather than settled fact.
