process_cesm_data not suited for all CESM outputs

[anoukvlug]

I tried with @gyjls to use the process_cesm_data function to process the 400BP-1990CE files of the CESM trace simulation. Unfortunately the function gives an error, because the files have unlike the CESM-LME no time_bnds variable.

time:bounds = "time_bnds" ;
        double time_bnds(time, nbnd) ;
 time_bnds:long_name = "time interval endpoints"

[fmaussion]

It should be quite easy to fix right? Do you plan to submit a fix?

[anoukvlug]

Yes, it should be not to complicated. I could look for a solution and submit it. Though I would prefer to first deal, with PR #814, to limit the versions I have of the code. (It is not right timing for me to start experimenting with having multiple branches at the same time ;) )

[anoukvlug]

I started looking into how to solve this issue. It seems like it might be less straight forward than I thought.

Here is a list of issues that I encountered so far:

• Not all netcdf files have the same time units.
• The most recent files have the "days since 0001-01-01 00:00:00" time unit. Thought this time unit might be a mistake in the assignment, as it doesn't look like that when I try to open it.
• Other files have a "ka BP" time unit. This unit is not supported by netCDF.date2num. Therefore I think it might be impossible to process these climate files and save them with the standard gcm_data.nc format that is being used. If saving in this standard format turns out to be impossible, than changes in other parts of the model need to be made in order to be able to solve this issue.

@fmaussion, what do you think?

@gyjls

[fmaussion]

@anoukvlug thanks for looking into this.

The solution depends on the scale of the data you are looking at: the best of course would be to creat compatible files before feeding them to OGGM, i.e. by adding/correcting the files attributes once for all in simple separate notebooks.

A second way is to build all the dirty if .. then logic in a process_dirty_cesm function and use this one. In this case I'm relunctant to add this function to OGGM, but I don't see a big issue here since it is possible to use the process_gcm_data anyway.

[anoukvlug]

Yes, you are right. I thought 'days since ....' couldn't be negative. As they can be negative there is indeed not as much of a problem as I thought and process_gcm_data should be able to deal with these files. once they are in the proper format.

[anoukvlug]

I started making a notebook, to preprocess the CESM trace data and use this as climate data to run OGGM. I generated a climate file to test this and tried to use it in the run_from_climate_data function. This works fine up to going back to ~6700BP.

When going further back in time I get the error:

ValueError: julian day must be a positive integer

In the following part of the code:

with ncDataset(fpath, mode='r') as nc:
    time = nc.variables['time']
    time = netCDF4.num2date(time[:], time.units)

I am not sure how to best deal with this issue.

[anoukvlug]

A (ugly) work around for this issue is to:

1. Process the data through process_gcm_data
2. Open the gcm_data*.nc file that you generated.
3. Increase the time variable, so you don't have dates before 4712 BC.
4. Save the data and the shifted time with write_monthly_climate_file
5. Use the climate file, with for instance run_from_climate_data
6. Don't forget to shift the time back once analysing the output.