ECMWF has today discovered an error that affects 2m temperature values depicted on its standard (web-based) 10-day meteograms ("10-day ENSgrams"). The "15-day ENSgrams" and "15-day ENSgrams with climate" are unaffected.
The error is as follows:
To derive 2m temperature values to plot on its meteograms ECMWF applies a height-based correction to model 2m temperature output, that is based on a fixed lapse rate (-6.5K/km) assumption. In short we compute the altitude difference between the site in question, and the nearby model orography, and then use this as a multiplier for the lapse rate adjustment. Due to different horizontal resolutions, the HRES orography dataset differs from the ENS orography dataset. The error is such that, for the 10-day meteograms, these orography datasets were accidentally switched around. This happened when we moved our operations to Bologna from Reading, on 18 October 2022. In relatively flat areas the impact on displayed 2m temperature is negligible, but in very mountainous regions where HRES and ENS orography heights generally differ by a large amount it can be substantial. There are further potential complications at coastal sites.
This error has been addressed quickly, and with effect from the 12UTC run of 4 January 2022 all meteogram 2m temperatures will be correctly computed.
We sincerely apologise to all ECMWF users for the inconvenience that the error will have caused.