What Is ERA5 Reanalysis, and Why We Use It for Weather History

Every temperature, rainfall total, and trend line on WeatherJourney traces back to one dataset: ERA5, the reanalysis produced by the European Centre for Medium-Range Weather Forecasts (ECMWF) for the European Union's Copernicus Climate Change Service. It is one of the most widely used climate datasets in the world, and it is worth understanding what it is — and what it is not — before you read too much into any single number.

What a reanalysis actually is

A reanalysis is not a pile of station records. It is the output of a modern weather model run over the past, continuously corrected by the observations that were available at the time — surface stations, weather balloons, ships, buoys, aircraft, and satellites. The model fills every gap in space and time with a physically consistent estimate, so the result is a seamless grid covering the whole planet, hour by hour, with no blank cells and no stations that move or close.

That completeness is the point. Raw station histories are patchy: instruments change, sites relocate, records start and stop. A reanalysis trades the "purity" of a single instrument for a consistent, gap-free picture that can be compared cleanly from one decade — and one place — to the next.

From 1940 to today, on one consistent grid

ERA5 currently reaches back to 1940, which is why the history on this site starts there. It is delivered on a global grid of roughly 0.25 degrees — about 28 km between points — with a value for every cell, everywhere, over that whole span.

We take the daily and monthly aggregates from that record, compute each city's long-term averages, and chart how they have shifted over more than eight decades. Because the grid and the method are the same for every year, a comparison between the 1940s and the last decade is a like-for-like one, not an artefact of a station being upgraded or relocated.

Modelled, not measured — and why that matters

Here is the honest caveat we keep front and centre: ERA5 is modelled, not a direct station measurement. A reading on this site is the model's best reconstruction for a grid cell, anchored to the observations of the day — not a thermometer you could have stood next to.

In practice that means two things. First, we avoid language like "the hottest day ever recorded." The right framing is "the hottest day in the record, since 1940" — true to both the modelled nature of the data and its length. Second, our baseline for "average" is the full record, not the 30-year "climate normal" you may see elsewhere. Both are legitimate; they are just different reference periods, and we label ours plainly so nothing is misread.

Where the grid gets the terrain wrong

A single 0.25-degree cell is about 28 km across, so it flattens everything inside it to one elevation and one surface. Over open plains that is barely noticeable. Over complex terrain it is — a valley town and the ridge above it fall in the same cell but have very different climates, and coastlines blur the land–sea boundary. Where a city sits far from its cell's average elevation, a raw reanalysis value can run too warm or too cold, which is why steep and coastal locations deserve a little extra scepticism than a number alone suggests.

How this shows up on the site

When you open a city, the long historical charts — monthly averages, the year-by-year trend, the warming stripes — are ERA5. The live "right now" figure is filled from Open-Meteo, because the reanalysis publishes with a short lag and we would rather show you a fresh current temperature than a five-day-old one. The two are labelled so you always know which is the modelled climate record and which is the live snapshot.

None of this is meant to talk you out of the numbers. ERA5 is a remarkable, rigorously produced record, and for the question this site exists to answer — how has this place's climate changed, and is today unusual for it? — a consistent 80-year grid is exactly the right tool. Knowing how it is built just lets you read it for what it is.