What is reanalysis data? ERA5, ERA5-Land and regional reanalyses explained

What is climate reanalysis?

Weather centres make forecasts by combining the previous forecast with newly arrived observations to get the best estimate of the current state of the atmosphere, a process called data assimilation. A reanalysis runs the same process over past decades, at lower resolution and with one version of the model, so the whole record is produced consistently. Because it is not racing to issue a forecast, it can use more observations, including ones that arrived late or were later reprocessed.

The result is a globally complete, physically consistent record of temperature, humidity, wind, pressure, radiation, cloud and precipitation for every grid cell and every hour, with no gaps.

ERA5 and ERA5-Land at a glance

• ERA5 is ECMWF's fifth-generation global reanalysis, produced for the Copernicus Climate Change Service with cycle CY41R2 of ECMWF's Integrated Forecasting System. It provides hourly data from 1940 to the present at a native resolution of about 31 km, distributed on a regular 0.25° grid.
• ERA5 is updated daily about five days behind real time. The early data, called ERA5T, can differ from the final release published two to three months later if problems are found.
• A lower-resolution 10-member ensemble estimates ERA5's uncertainty, which depends on how many observations were available; the observing system has changed a lot since 1940.
• ERA5-Land replays ERA5's land component at 9 km (on a 0.1° grid), hourly from January 1950. It is driven by ERA5 air temperature, humidity and pressure, corrected for the height difference between the two grids with a lapse-rate correction.
• Both datasets are distributed through the Copernicus Climate Data Store under a CC-BY licence.

Regional reanalyses: CERRA and NORA3

Finer grids help most where the climate changes quickly over short distances: mountains, valleys and coasts.

• CERRA, the Copernicus European Regional ReAnalysis, covers Europe from September 1984 to the present at 5.5 km. It uses the HARMONIE-ALADIN system with ERA5 as its boundary conditions; analyses are 3-hourly and hourly values come from short forecasts. Its added value comes mainly from a better description of terrain and the assimilation of more surface observations. C4B reads CERRA through Open-Meteo, whose copy ends in June 2021, so it uses complete years up to 2020.
• NORA3 is a 3 km hindcast from the Norwegian Meteorological Institute. It runs the HARMONIE-AROME model forced by ERA5, with each forecast starting from a state adjusted to surface observations. Strictly it is a hindcast rather than a full reanalysis, but its fine grid represents Nordic terrain and coastlines in far more detail than a global grid.

Urban heat-island effects

Cities are often warmer than the countryside around them, especially at night, because buildings and paving store heat and release it slowly. The effect varies from street to street. C4B does not account for it at any level greater than the use of reanalysis data: a grid cell several kilometres wide, or an airport weather station, captures it only partly, if at all. On dense urban sites, treat night-time temperatures as likely to be underestimated.

Reanalysis vs weather station data

The two answer different questions. A station measures the real air at one point, usually an airport, with its own surroundings and occasional gaps. A reanalysis gives a modelled but complete record for the grid cell containing your site.

• Location: the nearest station can be tens of kilometres away and at a different elevation; reanalysis is available at the site's coordinates, averaged over the grid cell.
• Completeness: station records can have missing variables or periods, and solar radiation in particular is often missing or modelled; reanalysis provides every variable every hour.
• Reliability: the World Bank's Climate Change Knowledge Portal describes ERA5 temperature as highly reliable, but recommends comparing several datasets for precipitation, which is harder to measure and model.
• Local effects: a grid cell smooths out terrain and land cover smaller than the cell, so slopes, shorelines and urban heat islands are only partly represented. A station reflects its own surroundings, which may not match the site either.

What reanalysis means for building design

For architects and engineers, reanalysis fills two gaps. It gives a climate record for sites without a representative station nearby, and it provides long, consistent hourly series from which a typical year can be built for the exact location.

Resolution sets the limit of what it can tell you. A 9 km or 5.5 km cell describes the climate of the area around the site, not the conditions against one façade or in one courtyard; that still needs on-site measurement or microclimate modelling.

A practical check is to compare reanalysis with the nearest station. Where the two disagree, the difference shows how far the site may depart from the station, and it is worth understanding why before choosing one.

How C4B uses reanalysis

In projects on paid plans, C4B compiles a site-specific typical year from gridded reanalysis at the project's exact coordinates: worldwide, Open-Meteo's default blend of ERA5-Land (0.1°, about 9 km), ERA5 (0.25°, about 25 to 31 km) and, from 2017, ECMWF IFS analysis (about 9 km); CERRA in Europe (5.5 km); and NORA3 in northern Europe (3 km). Temperature and humidity are corrected to the site's elevation using the Copernicus GLO-90 terrain model, and months are selected with ISO 15927-4 from the latest 15 complete years in the dataset (2006 to 2020 for CERRA).

The result represents the grid cell around the site, not a microclimate. In projects, you can switch between the site data and the nearest weather station file to compare the two.