Manupulations on data from GeoData object about esmvaltool

[bulli92]

@koldunovn , thanks for this.
The point is, that the diagnostic currently in ESMValTool was done with an older version of geoval.
@BenMGeo developed this diagnostic.

Since then we have c�hanged the default behavior of many geoval functions. Instead of numpy arrays, they now return by default a GeoDataobject. I think this is the problem for the fldmeanfunction. You can see current implementation here.

A fix would be easy I think. Instead of

  File "./diag_scripts/aux/LMU_ESACCI-diagnostics/diagnostic.py", line 859, in _p_stat
    _std_data=np.sqrt((D2.fldmean()-_mean_data**2))

It should be

  File "./diag_scripts/aux/LMU_ESACCI-diagnostics/diagnostic.py", line 859, in _p_stat
    _std_data=np.sqrt((D2.fldmean(return_data=False)-_mean_data**2))

Can you check? It might be that similar problems also occur at other places. Perhaps you can check if it works for this particular problem and make a PR. In general @BenMGeo would need to revise the diagnostic to make sure that everything works with the currently officially published version of geoval.

In the future we will easily capture these issues through automatic testing of all diagnostics. As we are in a transition phase at the moment it is good to have such Master users like you :-)

[koldunovn]

Thanks for nice words, Alex. I will try to do what I can :)

I have applied the fix in several places and made a PR #49 . It worked to the level of passing thought the script and ending successfully. However the plots that are generated have only coastlines on them. I had to hack geoval itself in couple of places to make the plots work.

[BenMGeo]

@koldunovn can we close this? as we drop V1 and I need to recode everything for V2 either way.