How to deal with different time dimensions?

[dhohn]

I want to look at several experiments branching off from the 1pctCO2 runs. One annoyance that I came across is that the branching times are (by necessity) different across the various models. Thats inconvenient when I want to compare some metric at branching time vs some years later.

Is there a preprocessor to "align" the time coordinates of datasets to a common event such as the branching time in those 1pct runs? If not what would be a smart way to deal with it?

[zklaus]

With the latest release (2.5.0), we have drastically improved the handling of time ranges in ESMValTool and I believe this could help in your case as well. Please have a look at the timerange section in https://docs.esmvaltool.org/projects/esmvalcore/en/latest/recipe/overview.html#datasets, though I just noticed there seems to be a problem with the formatting. @sloosvel, I think you have applied this for a somewhat similar use-case in decadal predictions, right? Would you mind having a look? Perhaps this is something we could pick up at the monthly video call (see ESMValGroup/Community#23). In any case, @dhohn, you are very welcome to join that meeting if you like!

[dhohn]

Thanks @zklaus ,
That functionality is indeed useful. Although I also ran into the issue that auto-download doesn't work with that syntax.

It doesn't quite help with my problem though. I've now gone to not selecting the time with preprocessors but first reading the branching times from the metadata and then selecting time ranges in the diagnostic script itself.

The question in #2647 is similar to my question I think. Where a common event doesn't quite happen at the same time.

[zklaus]

#2647 is different in so far as in that case the runs that should be aligned have no prior relation. In your case, if I understand you correctly, in principle the runs that have been branched off, the children, (could/should) have the same time coordinate, but the parent time coordinate must be different because of the different branching points. This sounds more similar to me to what @znicholls suggested in ESMValGroup/ESMValCore#1572. Perhaps we could add handling of time coordinates to the discussion there?

In any case, until a new preprocessor for this is available, I think a workable solution would be to do this grafting in a diagnostic. Indeed, we might implement this in a special diagnostic to start with and then migrate it in stages to shared diagnostic code, then a preprocessor and then fully upstream into iris.

Of course, if you and @znicholls could team up to write a preprocessor that addresses this in ESMValGroup/ESMValCore#1572, that would be fantastic! I'll be happy to provide some help.

[dhohn]

Yes I would be happy to. Is ESMValGroup/ESMValCore#1572 the right number though?

[zklaus]

Sorry, the number is right, but the repository is not: This is being discussed in the core, so the correct link is this: ESMValGroup/ESMValCore#1572. I'll edit above comments to correct the link.