What's happening here? Is it due to a lat-lon conversion, for example, from a Mercator-Projection to a Peters Map or something like that? You know, like, when map projections amplify the high latitudes too much so Greenland appears to be the size of South America but it really is only 1/4 the size of Brazil. Reproduce error: Download t2m from Era5 monthly https://cds.climate.copernicus.eu/datasets/reanalysis-era5-single-levels-monthly-means?tab=download January 1950-2026 and open in Panoply-app. For both formats, GRIB and netCDF, Panoply states a Global Mean of around 285 kelvin for each January (Global Mean shows up as footnote, below the colour scale). ClimateReanalyzer has the same-ish global mean °C for their Era5 Januaries. But: a python export from netCDF via xarray, doing a global annual January average in awk , produces global average January about 10K colder, between 275 and 277 K. The same dubious global average results come with awk on a "labelled-text" export from Panoply for both formats. Am I being stupid? Why is it 10 degrees colder in my global average? This is the awk snippet I use for averaging all the time and never get dubious results with it: divfield4[key]+= ($4=="" ? 0 : 1) field4[key]+=$4+0 … END … print (divfield4[a] ==0 ? "" : field4[a]/divfield4[a]) Grmpf.