Why would the "time" dimension read differently by "ncdump -h" and "Python" for the same nc file?

[Hemraj]

Hi,
I find it surprising that the same CESM2.2 CLM model output NetCDF file displays different "time" values when opened on different systems.

Opening it using ncdump shows JANUARY:
time:units = "days since 2015-01-01 00:00:00"



while opening the same file by Python shows FEBRUARY
cftime.DatetimeNoLeap(2015, 2, 1, 0, 0, 0, 0, has_year_zero=True)



Shouldn't it be the same? I came across this while using "groupby('time.month')" function in Python. Wondering, if would this data be read as JAN or FEB by this function (more likely FEB), although it is JAN. Why would it show different date for the same file?

Many thanks.
Hemraj

 

[strandwg]

The "time" coordinate from monthly average data in CLM denotes the 1st timestep of the month *following* the average, since that's the time value the model has at that moment.

This quirk is being corrected.

 

[Hemraj]

Dear @strandwg
Thanks for your response. Sorry, I am still not clear. I understand the time shown represents the FIRST timestep for the monthly average output. In that case for JAN shouldn't it be 2015-01-01 instead of 2015-02-01?
Now, I am doubting my analysis and plotting where I show monthly variation. Maybe JAN is shown as FEB, FEB is shown as MARCH and so on shifting by one month ahead.
Is there any suggestion to me where I can read "time" correctly like JAN as JAN (not FEB)?

 

[strandwg]

Dear @strandwg
Thanks for your response. Sorry, I am still not clear. I understand the time shown represents the FIRST timestep for the monthly average output. In that case for JAN shouldn't it be 2015-01-01 instead of 2015-02-01?
Now, I am doubting my analysis and plotting where I show monthly variation. Maybe JAN is shown as FEB, FEB is shown as MARCH and so on shifting by one month ahead.
Is there any suggestion to me where I can read "time" correctly like JAN as JAN (not FEB)?

The correct value for "time" can be found by averaging the values for "time_bounds" - that will properly show the midmonth value for "time".

 

[oleson]

This might also be helpful:

Frequently Asked Questions — ESDS 0.1 documentation

See:

Dealing with CESM monthly output - is there something wrong with time​

 

[strandwg]

This might also be helpful:

Frequently Asked Questions — ESDS 0.1 documentation

See:

Dealing with CESM monthly output - is there something wrong with time​

Thanks, @oleson !

 

[Hemraj]

Thank you @oleson and @strandwg
Thats really pointing my current problem related to time. I tried exactly the same but doesn't work as shown. I am not sure if it needs some extra environment.
End error message is:
KeyError: "Dataset.cf does not understand the key 'time'. Use 'repr(Dataset.cf)' (or 'Dataset.cf' in a Jupyter environment) to see a list of key names that can be interpreted."




And my ncdile read is:

Not sure how that worked in your system but not on mine! Looking forward.

Many thanks.

 

[oleson]

We can't really help with analysis tools such python, xarray, etc. on this Forum. I suggest you try some of the resources that are listed on the ESDS page or others.

 

[slevis]

ESDS info here:

About Us — NCAR-ESDS 0.1 documentation

 

[Hemraj]

This might also be helpful:

Frequently Asked Questions — ESDS 0.1 documentation

See:

Dealing with CESM monthly output - is there something wrong with time​

This is helpful. Thanks lot @oleson