What every frame is made of.

• ERA5 reanalysis. Hourly, 0.25°, read from the public ARCO ERA5 Zarr store, gs://gcp-public-data-arco-era5/ar/full_37-1h-0p25deg-chunk-1.zarr-v3. The newest months are ERA5T, the preliminary release; final ERA5 ends earlier and is shown on its own row.
• ECMWF IFS analysis. From ECMWF open data: the step-0 field of each 6-hourly cycle plus its +3 h field, so the bridge between ERA5 and now is 3-hourly.
• Forecast. The newest IFS run, out to 360 h, or the newest AIFS run, both from ECMWF open data.
• Between fields. Frames that fall between two fields are a linear blend in time of the neighbours on either side.

Frames are palettised 8-bit PNGs on a 2880-pixel Web Mercator raster, two pixels to each 0.25° grid column, with azimuthal caps over each pole so a globe view has no hole in it. A frame once published is never rewritten: when ERA5T, the preliminary stream, is replaced by the final one, the day is drawn again under a new revision and the index points at it.

An anomaly is a departure from the 1991–2020 normal for the same day of the year. In the hourly step that comparison is only honest for fields without a strong daily swing — 500 hPa height and precipitable water — because an instant minus a 24-hour-mean normal is mostly the difference between noon and midnight. Two-metre temperature therefore has an anomaly in the daily and monthly steps and not yet in the hourly one.

Past forecast runs are read from ECMWF's public open-data archive at 0.25°. That archive begins on 1 March 2024.

Ghost Forecast and Run-to-Run cannot go earlier. Runs from before then, such as the ones issued ahead of Sandy in 2012, are not openly archived, so there is nothing to pin over those events.

• One 12 UTC ERA5 sea-level-pressure field per day since 1940, box-averaged to 2.5°.
• Anomalies are taken against a 1991–2020 day-of-year climatology, smoothed over 31 days.
• The match is a pattern correlation over the chosen region, weighted by the cosine of latitude.
• The outcome shown for each analog uses 2 m temperature anomalies from the same index.

It is a climatology of outcomes, not a forecast. And because anomalies are measured against 1991–2020, older analogs read colder: the warming trend is in the baseline.

The Old Weather theme channel is styled after Philip Brohan's 20CRv3 diagnostics: grey sea, grey land, heavy black labelled isobars.

Here it draws the single ERA5 or analysis field as those heavy labelled isobars. The ensemble-member spaghetti and the observation dots of the original are not drawn, because no ensemble or observation data is wired into Forecast Replay.

Neither the European Commission nor ECMWF is responsible for any use of this information.

Open Replay Studio