EC-mean time averages based on unweighted monthly means

[jhardenberg]

On the EC-Earth dev Portal Twan noticed the following:

I noticed that the (multi-)yearly means produced by EC-mean are calculated as the unweighted means of the monthly mean time series. See this piece of code in scripts/global_mean.sh:

for (( year=$year1; year<=$year2; year++)); do
    indir=$DATADIR/Post_$year
    $cdonc cat $indir/${exp}_${year}_${vv}.nc $WRKDIR/full_field.nc
done

$cdo fldmean -timmean $WRKDIR/full_field.nc  $WRKDIR/mean_${vv}.nc

I suppose that for many purposes this will be sufficiently accurate, but it may also lead to confusion because the reported values are not the true (multi-)annual means. Wouldn't it be better to -- at some point -- include the proper weighting factors?

[plesager]

This is fixed by @klauswyser in a046037.

But @noije noted that the (lot less important) case of leap year when doing the average across years:

$cdo fldmean -timmean $WRKDIR/full_field.nc $WRKDIR/mean_${vv}.nc

is not accounted for. I suspect we can live with that one.