cis col still seems to be an issue for me. This issue might have something in common to the previous one but I encountered it in a different way. I did this;
cis col Mean:Mean_Vari_Qtiles_Dust_TOTAL_conc_sfc_26aer_2008jul.nc /group_workspaces/jasmin2/crescendo/Data/AERONET/AOT/LEV20/Fake_data/monave/AOT_440_Izana_2008-07_monave_nco.nc:variable=AOT_440 -o Means_Dust_TOTAL_conc_sfc_26aer_2008jul_Izana.nc
This gives me warning message like these;
2017-12-31 11:13:40,256 - WARNING - /usr/lib/python2.7/site-packages/cis/collocation/gridded_interpolation.py:375: RuntimeWarning: divide by zero encountered in divide
norm_distances.append((x - coord[i]) / (coord[i + 1] - coord[i]))
2017-12-31 11:13:40,257 - WARNING - /usr/lib/python2.7/site-packages/cis/collocation/gridded_interpolation.py:351: RuntimeWarning: invalid value encountered in add
value += np.asarray(values[edge_indices]) * weight[vslice]
and Mean does not have a value;
Mean = _ ;
Because there are no same value in longitude or latitude, the problem might be time which has only one value in each of the input and the reference data? Previously the problem seemed to have been caused by the name of time dimension 't' instead of 'time'. But I had also switched from monthly to 3 hourly model data when it worked. Also the aeronet data had multiple time steps then.
Now I want to collocate the monthly data (created by me using python) without collocating in time because I just aim to get values at the stations. So I created *fake* aeronet data which has value only at the middle of a month. I made fake aeronet data which has the same structure to the real monthly data for which cis col worked except for the size of time dimension.
I tried the option ":collocator=lin,kernel=nn_t" in attempt to stop interpolating in time but this gave me the same result.
Please could I have your advice on this issue? I copied the data used here to my home directory on JASMIN just in case;
/home/users/myoshioka/temp/
Thank you.
Masaru
Hi Masaru, sorry for the delay in replying.
The issue is indeed the extra length-1 time dimension in Mean_Vari_Qtiles_Dust_TOTAL_conc_sfc_26aer_2008jul.nc which CIS tries to interpolate over - leading to the NaN. You're right that the nearest neighbour approach would have fixed this, but for some reason this option isn't getting passed through correctly - I've raised an issue for this here: https://jira.ceh.ac.uk/browse/JASCIS-410.
I have a plugin which removes these length-1 dimensions when I don't want them here which might be useful: https://github.com/duncanwp/cis_plugins/blob/master/squeeze-netcdf.py
Then the following worked for me: cis col Mean:/home/users/myoshioka/temp/Mean_Vari_Qtiles_Dust_TOTAL_conc_sfc_26aer_2008jul.nc:product=squeeze_netcdf /home/users/myoshioka/temp/AOT_440_Izana_2008-07_monave_nco.nc:product=cis -o Means_Dust_TOTAL_conc_sfc_26aer_2008jul_Izana.nc
Hi Duncan,
Thank you. How do I use squeeze_netcdf? I copied it in my plugin directory. Should I somehow specify this in .bashrc ? Currently it gives me an error when I try to use it.
Masaru