Error messages trying to compute ensemble statistics

[knlowman]

Setup Information

• Xclim version: 0.52.1

Context

I am trying to compute the ensemble statistics for the dataset ctrl_ens, which looks like this:

Steps To Reproduce

My code for computing stats for the entire dataset is

ctrl_ens = ensembles.create_ensemble([ctrl_mem_1,ctrl_mem_2,ctrl_mem_3])
ensembles.ensemble_mean_std_max_min(ctrl_ens)

The corresponding error message is
UFuncTypeError: ufunc 'add' cannot use operands with types dtype('<M8[ns]') and dtype('<M8[ns]').

I also tried computing stats just for the variable temp with ensembles.ensemble_mean_std_max_min(ctrl_ens.temp)
and I got the error AttributeError: 'DataArray' object has no attribute 'data_vars'.

[aulemahal]

Hi @knlowman,

For both errors, I think one issue is that you didn't expect that ensemble_mean_std_max_min takes ONLY Datasets as input. We don't have a real reason for this, the function could easily be made to accept DataArray as well (a good first contribution maybe ;) )?

For the second error, try doing :

ensembles.ensemble_mean_std_max_min(ctrl_ens[['temp']])

Here I am using xarray's feature of selecting a list of (data) variables, which returns a Dataset with only that (data) variable. Keeping all relevant coordinates and attributes.

And the first error. dtype('<M8[ns]') means that one of the data variables in your dataset is made of datetimes. Are you expecting of taking the mean/std/max/min of dates ? I am guessing you are not, so the solution would be to exclude these variables from the computation. Using the same trick as above, you could filter out only the meaningful variables.

If you are indeed expecting stats on dates, then the problematic stat is std. You could make the calculation outside the xclim function for the particular datetime variables, skipping std as it is ill-defined for datetime.

[knlowman]

Thanks very much @aulemahal! That solved the second error, and you're correct that I don't want to average a datetime variable, so I excluded it.