Cordex-EUR11 data has band patterns in data values

Hello,

I am currently working with both temperature and SPEI data from Cordex-EUR11.

All is working well, but I’ve noticed some kind of unnatural splits in the data values which form longitudinal lines in the Eastern part of the study area.

I was wondering if someone could shed light on the potential origin, and whether the data are still reliable given these seemingly artificial differences in data values.

For example, here’s the average SPEI-6 between 2067 and 2069 under RCP85:

I was wondering whether someone could take a look at this, thank you!

Dear Merijn,

Thanks for looking at the CDS datasets.

Before exploring the details, would you let me know, please which data did you use? Is that the C3S Atlas dataset’s SPEI-6 variable? Is that the mean for all the available models or a specific model?

Thank you, best regards

Andras Horanyi

Hello Andras,

This is the SPEI-6 data available from the dataset underlying the C3S Atlas of the Cordex-EUR-11 origin : Gridded dataset underpinning the Copernicus Interactive Climate Atlas

Indeed, this is the mean across all members of the ensemble model.

Just to clarify, I believe this is related to the Cordex-EUR-11 dataset itself, and not variable-specific.

Here’s a comparison of some of its SPEI-6 data and some data for the Monthly mean of daily maximum temperature.

Hello,

I was just checking whether anyone had an update on this! Thank you!

All the best

Hello,

I was just wondering whether anyone had time to look at this.

All the best

Sorry Merijn, I missed your answer, I am going to check this now and hope to come back soon. Thank you and sorry again!

Thanks so much Andras!

Dear Merijn,

We got a realistic explanation of the issue what you have detected. The different modelling centres use slightly different EURO-CORDEX domains (larger than the official EURO-CORDEX domain).

In the C3S Atlas viewer those areas are masked therefore we don’t see any discrepancy, however in the C3S Atlas dataset (what you use) we kep the information, which is outside the official EURO-CORDEX domain.

In practise this would mean that when you make the averaging then you would use different number of models outside the EURO-CORDEX domain and that causes these strange features.

Please have a look to the https://na-cordex.org/files/cordex_archive_specifications.pdf specificiation document, where you can find the details of the exact EURO-CORDEX domain. Then you might be able to consider only the EURO-CORDEX domain and therefore filter out those areas where we have less number of models available. This would make the mean more homogeneous.

I hope all this helps, sorry again for the delays!

Best regards

Andras Horanyi

Thanks so much Andras, it’s clear to me now!
I was able to clip my processed rasters to the original domain.

Thank you!