We would like to inform Free and Open data users that some files from the 00 UTC IFS medium-range forecast of 22 October 2026 were corrupted.
During that forecast run the model job failed at step 324 because of a hardware fault, and it was rerun from step 288. Steps 294 to 324 had already been disseminated from the first attempt before the rerun replaced them in our archive. From step 330 onwards the data were from the rerun.
As a result, all parameters at steps 294 to 324 come from a different model run than the steps that follow, for both the atmospheric (oper) and the wave (wave). This is most visible in accumulated fields: total precipitation, for example, can decrease between steps 324 and 330 by up to about 130 mm.
If you downloaded the data before 13:00 UTC on 23.10.2026, for consistency we suggest re-downloading the full range from step 288 to step 360. Other forecast runs are not affected.
We apologise for the inconvenience.