Is Catalonia getting hotter? What eight years of data can and cannot show

The question, and the eight years we hold

Every climate chart invites the same question, so it is worth asking it directly of this one. Onviure carries per comarca figures for tropical nights (nights that do not drop below 20°C) and days above 32°C, from the ERA5 reanalysis, for 8 years: 2018 to 2025. Aggregating to Catalonia gives one number per year, weighted by population, so the figure describes what residents experience rather than what the average comarca does. An unweighted mean would let a comarca of four thousand people count as much as the Barcelonès.

Plotted, the series looks like it is climbing. Days above 32°C go from 12.3 in 2018 to 25.6 in 2025, and the fitted slope is 1.7 more days each year. Most readers would stop there, and this article exists because stopping there is not safe.

The test, and what it says

A slope is not a trend until you know how much of it could be chance. The standard check asks how large the slope is against the scatter around it, which gives a t statistic, and compares that against the value expected at this sample size. With 8 observations that threshold is 2.45.

Days above 32°C come in at t = 2.04. It does not clear the bar. Tropical nights are nowhere near, at t = 0.28 against the same threshold, with a slope of 0.4 nights a year that is indistinguishable from flat.

The reason is visible in the chart: the year to year swings are far larger than the climb. The gap between the coolest year in the series (2020, at 10.0 hot days) and the hottest (2022, at 27.4) is 17.4 days. At the fitted rate it would take about 10 years of trend to cover the distance that two ordinary years already cover between them.

And the answer flips if you remove one year

2022 sits furthest from the fitted line, 9.9 days above it. The obvious next question is whether that one summer is producing the apparent trend. It is not, and the answer is more awkward than that.

Remove 2022 and the slope barely moves, from 1.7 to 1.6 days a year. But the scatter around the line shrinks, so the t statistic RISES, from 2.04 to 2.77, which does clear the 2.57 threshold that applies once you are down to 7 observations. On the full series there is no significant warming trend. Drop the single most extreme year and there is one.

Both answers come from the same data, and neither is wrong. That is what a sample of 8 looks like: the verdict rides on one observation, so it is not a verdict. Tropical nights behave the same way in the other direction, staying insignificant either way (t = 0.28, or 0.18 without 2022), which is at least consistent.

The honest statement is that Onviure cannot yet tell you whether Catalonia is warming. Not that it is not: the wider scientific record on Mediterranean warming is not in question here, and nothing in this article speaks to it. The claim is narrower and only about this dataset. Roughly three decades is the conventional minimum for separating a climate trend from weather, and 8 years of reanalysis is a weather sample.

What the same data does answer

None of this makes the climate figures useless, because comparing YEARS and comparing PLACES are different problems. Comparing the 43 comarques against each other within the same years uses 43 independent observations rather than 8, and a cross section that size resolves any correlation above about 0.30. The differences between comarques are far larger than that.

So the climate comfort ranking does what this article will not: it tells you, reliably, that some parts of Catalonia are meaningfully hotter at night than others, and by how much. That is a question this data can settle. "Is it getting hotter over time" is not, yet.

The methodology page lists the source and reference period behind every figure on this site. When the reanalysis series is extended far enough back, this article gets rewritten with an answer instead of a caveat.