Europe's coldest January since 2010 happened inside the fifth-warmest January on record

The Copernicus Climate Change Service published its January bulletin on 6 February, and it contains two facts that are usually reported separately. January 2026 was the fifth-warmest January globally, with an average surface air temperature of 12.95°C, 0.51°C above the 1991–2020 average and 1.47°C above the estimated 1850–1900 pre-industrial level [s1]. It was also Europe's coldest January since 2010, with an average temperature over European land of −2.34°C, 1.63°C below the 1991–2020 average [s1].

Both are true of the same month. In the final weeks of January, severe cold affected large parts of the Northern Hemisphere — North America, Europe and Siberia — driven mainly by a wavier-than-usual polar jet stream that allowed frigid Arctic air to spill into the mid-latitudes [s1]. Widespread cold conditions occurred across Fennoscandia, the Baltic States, eastern Europe, Siberia, and the central and eastern United States [s1]. Meanwhile the largest warmer-than-average anomalies were in the Arctic, most notably across much of the Canadian Arctic Archipelago, Baffin Bay, Greenland and the Russian Far East [s1].

"January 2026 delivered a stark reminder that the climate system can sometimes simultaneously deliver very cold weather in one region, and extreme heat in another," said Samantha Burgess, Strategic Lead for Climate at ECMWF, adding that resilience and adaptation to increasing extremes are key to preparing society for heightened climate risk [s1].

The World Meteorological Organization, summarising the month on 10 February, attributed the mid-latitude cold intrusions to a weakened polar vortex [s2]. Its account includes the snow that came with it: Japan's Aomori prefecture recorded a snow depth of 1.7 metres on 3 February, the most in 40 years, while Russia's Kamchatka Peninsula received over 2 metres of snow in early January on top of 3.7 metres in December — one of the snowiest periods the peninsula has seen since the 1970s [s2]. North America saw life-threatening cold and ice, with widespread cancellations and power outages affecting hundreds of thousands [s2].

Elsewhere the same month ran the other way. Ceduna, South Australia reached 49.5°C on 26 January, a new local record, and multiple locations across South Australia, northwest Victoria, inland New South Wales and southwest Queensland exceeded 45°C [s2].

What cold does, measured

The health literature on temperature and mortality has skewed heavily toward heat in recent years. Two studies published this month add specificity on the other tail.

The first, in the International Journal of Biometeorology on 6 February, examined cold spells in a setting rarely studied — a low-latitude subtropical plateau [s3]. Mortality and meteorological data were collected from 129 counties in Yunnan Province, China, between 2014 and 2020 and grouped into 16 administrative regions, with a quasi-Poisson distributed lag non-linear model applied across cold seasons from October to March [s3].

Model fit was best when a cold spell was defined as a daily average temperature below the 10th percentile for three consecutive days [s3]. Pooled relative risks for non-accidental mortality were 1.01 (95% CI 1.01–1.02) at lag 0 days, 1.08 (1.04–1.12) across lag 0–7, 1.11 (1.06–1.17) across lag 0–14 and 1.13 (1.06–1.20) across lag 0–21 [s3].

The lag structure is the finding. Cold's mortality effect is small on the day and accumulates for three weeks — which is why event-day counts systematically understate it. Cause-specific risks were higher still: 1.14 (1.06–1.23) for cardiovascular mortality and 1.22 (1.13–1.32) for respiratory mortality [s3]. Populations in the more easterly regions of Yunnan, and individuals aged 75 and over, showed higher risk of non-accidental mortality [s3].

The second study, published in Vascular Health and Risk Management on 17 February, looked at a single acute condition across a whole country [s4]. Using CDC WONDER mortality data for aortic dissection from 1999 to 2023 alongside monthly air temperature from the North America Land Data Assimilation System, the authors characterised trends and temperature associations [s4].

In 2023 there were 4,418 aortic dissection deaths in the United States, an age-adjusted mortality rate of 1.73 per 100,000 (95% CI 1.68–1.78) [s4]. Rates were highest in the Midwest (1.96), among men (2.19 versus 1.24 in women), among non-Hispanic Black individuals (2.75) and in adults aged 85 and over (11.67) [s4]. Across the full period, 86,943 deaths occurred; the overall age-adjusted rate declined (average annual percent change −0.37) but reversed to sustained increases after 2013 [s4].

Across all four census regions, monthly age-adjusted mortality was lowest in summer and highest in winter throughout 1999–2023 [s4]. The authors report that lower temperatures and greater thermal variability are linked to higher mortality, and argue for seasonally and regionally tailored prevention and health-system preparedness [s4].

What these studies cannot show

Both are ecological time-series or registry analyses. They associate ambient temperature with population mortality; they do not follow individuals, do not capture indoor temperature, fuel poverty, housing quality or heating access, and cannot separate cold exposure from the behavioural and infectious changes that accompany winter. The Yunnan analysis covers one province across seven years [s3]. The aortic dissection analysis uses death certificate data, which depends on cause-of-death coding, and its 20-year forecasts are model projections rather than observations [s4].

The Copernicus figures, meanwhile, describe a month, not a trend. Arctic sea ice extent in January was 6% below average — the third lowest on record for the month — and Antarctic extent 8% below, while global sea surface temperature over 60°S–60°N averaged 20.68°C, the fourth-highest January value on record [s1]. A cold European month sits inside those numbers without contradicting them.

What to watch

Whether excess winter mortality estimates for the European cold spell appear from national statistical agencies in the coming months, and how they compare with the lag-extended effects the Yunnan model implies [s3]. And whether the post-2013 reversal in US aortic dissection mortality holds as more years are added — a trend the authors flag without explaining [s4].

Sources

1. [s1] Copernicus: Fifth-warmest January sees 2026 start with weather extremes across both hemispheres. Copernicus Climate Change Service / ECMWF, 6 February 2026. https://climate.copernicus.eu/copernicus-fifth-warmest-january-sees-2026-start-weather-extremes-across-both-hemispheres
2. [s2] Extreme heat, cold, precipitation and fires mark the start of 2026. World Meteorological Organization, 10 February 2026. https://wmo.int/media/news/extreme-heat-cold-precipitation-and-fires-mark-start-of-2026
3. [s3] Impacts of cold spells on mortality risk in low-latitude subtropical plateau regions. International Journal of Biometeorology, published online 6 February 2026. https://doi.org/10.1007/s00484-025-03115-y
4. [s4] Spatiotemporal Mortality Patterns and Temperature-Associated Risk of Aortic Dissection in the United States: A National CDC WONDER Database Analysis. Vascular Health and Risk Management, 17 February 2026. https://doi.org/10.2147/vhrm.s569639