HadCRUT5 CMORizer odd tas

[malininae]

Hi all,

I might have done something wrong, but I was trying to run HadCRUT5 cmorizer on the data I've downloaded yesterday and I ended up with the weirdest tas. Surprisingly, the resulting temperature timeseries goes down with the time, at the same time tasa seems to be perfectly fine. I will double check what's wrong in a couple of weeks (maybe), but maybe someone knows what the deal is?

[malininae]

This is a plot of the tas after applying mean area_statistics and mean annual_statistics

[malininae]

And this is tasa

[remi-kazeroni]

Thanks for opening this issue, @malininae! Could you maybe share your test recipe so I could see if I can reproduce these plots with the data we have at DKRZ (i.e. the data downloaded at the time this CMORizer was merged)? @LisaBock - do you remember seeing something like this when working on the CMORizer?

[LisaBock]

Hi @malininae, I haven't used the climatological data of HadCRUT5 but only the anomaly data. The climatological data is tricky. As the global coverage of the data changed a lot from former days to now, it is not quite possible to use the time evolution of the global mean. In the beginning there is no data over the polar regions included, that's why the global temperature is decreasing with time as more and more polar regions (with lower temperatures) became covered. For which application do you want to use the data?

[malininae]

@LisaBock @remi-kazeroni thanks for the replies! I hoped to use the 1940-1959 anomaly to make it comparable with my ERA5 data, so I decided to use tas originally in order not to make an anomaly of anomalies. It's not a super big deal to use the 1961-1990 anomalies in my case, since HadCRUT in my code is just an auxiliary dataset, which I use as a covariate to fit my non-stationary GEVs on the ERA5 data. I just wanted to point the issue out, because someone else might come across that too.

I don't have an example recipe, to reproduce the figures, I just used gmst = eprep.annual_statistics(eprep.area_statistics(raw_gmst_cb, 'mean'), 'mean') and plotted that cube with iplt. raw_gmst_cb is the cmorized tas from OBS.

[LisaBock]

Hi @malininae, in principle it is possible to use the clim (tas) data to calculate the 1940-1959 anomaly. But you have to calculate the anomalies first gridpoint wise and afterwards average over the globe (see Jones et al., 1999 or a ncl example). But practically the clim (tas) data is already calculated from the anomaly (regarding 1961-1990) and one climatology field. You could check in the cmorizer.