We use cookies to improve your experience on our site. By using our site you consent to cookies. Learn more
OK
When Winter Arrives At Once
MetSwift Claros · European windstorm
Our lowest call in fourteen winters.
Every August, before the season starts, we forecast whether Europe’s winter storms will
arrive spread through the season or all at once. For the winter now beginning, the call sits below all
thirteen we have scored.
September 2026
We are not expecting another February 2022. What follows is the record that call is scored
against.
2 of 2
the two winters we called highest were the two most concentrated of the thirteen
4 of 5
of the five we called highest, four finished in the five most concentrated
46%
of the variation in how bunched a winter is, explained by an August forecast
0.1%
explained by the long-run historical trend
What we forecast
Not how much bad weather is coming. Whether it arrives together. A second call, issued in
October, asks the other question: how severe the storms themselves will be.
Same amount of severe weather, and two completely different winters for an aggregate, a
reinstatement or an hours clause. February 2022 is the reference case: Dudley, Eunice and Franklin inside a
single week, and PERILS settled the series at €3,851m.
Nothing else tells you this
Every other way of guessing, measured on the same thirteen winters.
Knowing, with hindsight, how windy a winter was or how severe its storms were tells you
nothing about how concentrated it was.
How it was built
The relationships were found in the observed record first, then tested as forecasts.
Found in the recordAcross 53 winters of station observations, rain days track storm clustering at
+0.59 and warm days at +0.53. Across 41 winters, wind activity tracks our independent storm
catalogue at +0.62.
Then forecastOnly after those links held were the same instruments used to make a forward call.
Thirteen August forecasts later, that call tracks how concentrated the winter turned out at +0.68.
Pointed at the right wintersRun the concentration measure back across the 53-year record and
1989/90 sits with 2018/19 and 2021/22 at the top: the winter of Daria, Vivian and Wiebke.
The record
Thirteen winters, each called in August before the season and never revised. Ordered by what we
said, strongest at the top.
Four of the five winters we called highest finished among the five most concentrated.
Concentration is the share of a winter’s severe weather that landed in its single worst week, as a
multiple of an evenly spread season, and the worst week is found from station records rather than from a storm
list.
Two alarms, thirteen winters
Two alarms on top, what the winter actually did underneath. Reading down a column gives one winter.
An alarm fired in eight of the thirteen winters, and all eight were bad on at least one
measure. The five quiet winters were bad on neither. In six of the eight the alarm that fired was the one
that came in. In two, 2013/14 and 2020/21, the other measure was the one that went: a warning delivered rather
than a call proven. Concentration scores 4 right, 1 false and 1 missed; extremity 2, 2 and 2.
An alarm here fires when the call comes in above its own long-run average, which is a placeholder rather than a
calibrated level. Concentration is the August index; extremity is the three October measures averaged. A bad
winter is one of the five most concentrated of the thirteen, or one of the four most extreme in our storm
catalogue. The catalogue has no reading yet for 2022/23 or 2025/26, so extremity is scored on eleven
winters.
February 2022, found from the weather alone
The worst week of each winter is identified from station records with no reference to a storm
list. For 2021/22 it lands exactly where the market remembers.
16–22 Feb 2022
Dudley, Eunice, Franklin
Concentration
4.04× an even season
Rank of thirteen
1st
Stations hit hard that week
59%
PERILS industry loss
€3,851m
Our August 2021 call
2nd highest of thirteen
Seasonal storm severity
bottom quarter of 75 years
Seasonal wind total
barely above average
Flagged by an activity screen
no
A thin season on every conventional measure, and one of the costliest weeks on record inside it.
And where
What we expected on the left, what arrived on the right.
Built fromWeather station records and our own reanalysis catalogue. No loss data is used
anywhere in building, fitting or scoring the forecast.
Scored onThirteen winters, 2013/14 to 2025/26. Each call issued in August, before the season,
and never revised.
What it isA view on the shape of a season, for northwest Europe. Not a loss estimate, not a date,
and regional rather than local.
MetSwift Claros, European windstorm concentration signal. September 2026. Full method, station
selection, reference periods and validation are set out in the technical report, available on request.