Hi All I am monthly-averaging daily ACCESS diagnostics and saving to files (1 sample per output file). Then I add to file time and time_bounds variables and relevant attributes. I can proceed to build a cosima dataset without any error messages, but using that dataset to load the actual variables f…

Thanks for checking that out. Strangely I still can’t open your files using the analysis3-unstable environment with xarray unless I reset the time dimension: ncap2 -s 'time=double(time)' in.nc out.nc Even using doubles for time gives the same overflow error. Maybe something is encoding the files i…

I don’t think your issue has anything to do with the cookbook, it’s just how you’ve added the time dimension to your NetCDF files. If I try to open them directly with xr.open_dataset, it fails with an error like: ValueError: Failed to decode variable 'time': unable to decode time units 'days since …

Thanks Angus. I’m not sure why open_dataset returned an error for you. For me it has been working fine with these files and I verified again now that is the case. Anyway, I remade them with the time and time_bounds as type double, and they can be found here: /g/data/g40/access-om2/archive/01deg_jra5…

I have now added the variables average_T1,average_T2,average_DT and relevant attributes to the passive_adelie variable based on the way they appeared in the original dataset before monthly-averaging and adjusted to the monthly average (some of these variables were defined above in the thread): aver…

I’ll recreate the files in this manner Angus. Thanks!

Update: this has worked in that the created files are readable from the explorer or getvar, with the caveat that their time_frequency appears as “static”. Selecting end/start_time still appears to work well though.

Correction. Using cc.querying.getvar with frequency=‘static’ on the dataset built in above solution only loads one file i.e. month. To specify start_time and end_time I find I need to speicfy that the saved files contain the string monthly via the getvar argument ncfile = ‘%monthly%’, as in here: …

Sorry, I think I misled you on this. I’d forgotten that the time bounds are used in one instance: to determine the temporal frequency of a file with only a single timestep. Obviously this would be useful in this case! I’m glad you found the workaround to match only the relevant files instead.

Thanks Angus! That only worked after the modification in the file creation that you recommended btw.

Building a COSIMA dataset on time-averaged files

Technical

Aidan (Aidan Heerdegen, ACCESS-NRI Release Team Lead) 19 February 2023 22:12 8

Can you please mark the relevant reply as the solution, that way it makes it easier for others who might have the same problem to find the answer. Thanks!

Topic		Replies	Views
Issues loading ACCESS-OM2-01 data from cycle 4 Technical	5	587	15 February 2023
Datastore making - unidentified realm COSIMA	22	220	10 May 2026
Issues in loading 'ht' in latest conds envs Ocean help	7	97	20 April 2026
ACCESS-NRI Intake catalogue for CM2-025 simulation Earth System help , cosima , model-evaluation	18	210	24 February 2025
Xarray warnings while loading data using cosima cookbook Technical python , help , cosima	15	605	5 November 2024

Building a COSIMA dataset on time-averaged files

Related topics