revise tropomi modules for monetio updates

[mlirenzhenmayi]

update drivers, tropomi no2 - wrfchem paring

[mbruckner-work]

From what I can tell, the changes are fine for ensuring the TROPOMI NO2 capability works before the tutorial.

I would suggest adding a scaling to the paired observations and model data like was done for MOPITT CO (eg. linked codeblock). An argument to specify ylabel_plot (ylabel_plot: '$10^{18} molec/cm^{2}$' ) then should go in the yaml file with the variable name under the obs label (see control_mopitt.yaml). The axes on the Taylor plot will actually be readable this way.

MELODIES-MONET/melodies_monet/util/satellite_utilities.py

Lines 107 to 112 in c9a1ec1

	# Apply scaling to drop scientific notation (x10^{18} molec/cm2 instead of molec/cm2)
	## Taylor plot doesn't work if don't do this.
	ds[co_ppbv_varname+'_column_model'] /= 1e18
	ds[co_ppbv_varname+"_column_model"] = ds[co_ppbv_varname+'_column_model'].assign_attrs(units='$10^{18} molec./cm^{2}$')
	ds['column'] /= 1e18
	ds["column"] = ds['column'].assign_attrs(units='$10^{18} molec./cm^{2}$')

[blychs]

@mlirenzhenmayi , I would still go for keys(). This is already the solution in #284 , and .variables fails if it is DataFrame.
Otherwise, we would have to add checks to see whether it is a dataframe or a dataset.

[zmoon]

@blychs for membership it's the same without .keys() (if var in df_or_ds:). Will it not always be an xarray dataset for satellite?

[blychs]

If var in ds_or_df would work and is probably the best solution, I agree. Satellite is always ds, but that check is run for every case, including surface and aircraft, if I'm not wrong

[mlirenzhenmayi]

From what I can tell, the changes are fine for ensuring the TROPOMI NO2 capability works before the tutorial.

I would suggest adding a scaling to the paired observations and model data like was done for MOPITT CO (eg. linked codeblock). An argument to specify ylabel_plot (ylabel_plot: '$10^{18} molec/cm^{2}$' ) then should go in the yaml file with the variable name under the obs label (see control_mopitt.yaml). The axes on the Taylor plot will actually be readable this way.

MELODIES-MONET/melodies_monet/util/satellite_utilities.py

Lines 107 to 112 in c9a1ec1

	# Apply scaling to drop scientific notation (x10^{18} molec/cm2 instead of molec/cm2)
	## Taylor plot doesn't work if don't do this.
	ds[co_ppbv_varname+'_column_model'] /= 1e18
	ds[co_ppbv_varname+"_column_model"] = ds[co_ppbv_varname+'_column_model'].assign_attrs(units='$10^{18} molec./cm^{2}$')
	ds['column'] /= 1e18
	ds["column"] = ds['column'].assign_attrs(units='$10^{18} molec./cm^{2}$')

Thanks Maggie! Current tropomi modules can plot Taylor, although with messed units. But with the scaling applied, other plots like the map plot, box plot are messed up and need to be re-configured with units in the yaml file. I prefer to revise the Taylor plot module to make it more flexible to handle large units.

[blychs]

From what I can tell, the changes are fine for ensuring the TROPOMI NO2 capability works before the tutorial.
I would suggest adding a scaling to the paired observations and model data like was done for MOPITT CO (eg. linked codeblock). An argument to specify ylabel_plot (ylabel_plot: '$10^{18} molec/cm^{2}$' ) then should go in the yaml file with the variable name under the obs label (see control_mopitt.yaml). The axes on the Taylor plot will actually be readable this way.

MELODIES-MONET/melodies_monet/util/satellite_utilities.py

Lines 107 to 112 in c9a1ec1

	# Apply scaling to drop scientific notation (x10^{18} molec/cm2 instead of molec/cm2)
	## Taylor plot doesn't work if don't do this.
	ds[co_ppbv_varname+'_column_model'] /= 1e18
	ds[co_ppbv_varname+"_column_model"] = ds[co_ppbv_varname+'_column_model'].assign_attrs(units='$10^{18} molec./cm^{2}$')
	ds['column'] /= 1e18
	ds["column"] = ds['column'].assign_attrs(units='$10^{18} molec./cm^{2}$')

Thanks Maggie! Current tropomi modules can plot Taylor, although with messed units. But with the scaling applied, other plots like the map plot, box plot are messed up and need to be re-configured with units in the yaml file. I prefer to revise the Taylor plot module to make it more flexible to handle large units.

Hi Meng and Maggie, I tried to do that in the plotting class and was not able to (for the TEMPO tool). I think that the problem is with the AxisArtist class from Matplotlib that the MONET uses, but I was not able to find an easy solution for that.
In satplots_xr, I just added the option "normalize", which plots a normalized Taylor diagram... therefore having smaller numbers.
I agree that revising the Taylor plot module is a better solution, and maybe someone else can find a way to do it (maybe there is a way with the GridHelperCurveLine?).

Having the option for an offset as suggested by Maggie is a better solution than mine, we could add that to the plt_grp in the yaml (and not the observations) for Taylor (even better, we could have both an offset and the option for normalizing it). However, I'd say that we can merge TROPOMI for now and deal with that after the deadline.
Would that work?
Cheers
Pablo

[mbruckner-work]

Thanks Maggie! Current tropomi modules can plot Taylor, although with messed units. But with the scaling applied, other plots like the map plot, box plot are messed up and need to be re-configured with units in the yaml file. I prefer to revise the Taylor plot module to make it more flexible to handle large units.

I hadn't noticed an issue with maps, but generally am not using the box plot tool and haven't been checking it in my tests. I agree with @blychs that we can deal with solutions for this later.

[mlirenzhenmayi]

From what I can tell, the changes are fine for ensuring the TROPOMI NO2 capability works before the tutorial.
I would suggest adding a scaling to the paired observations and model data like was done for MOPITT CO (eg. linked codeblock). An argument to specify ylabel_plot (ylabel_plot: '$10^{18} molec/cm^{2}$' ) then should go in the yaml file with the variable name under the obs label (see control_mopitt.yaml). The axes on the Taylor plot will actually be readable this way.

MELODIES-MONET/melodies_monet/util/satellite_utilities.py

Lines 107 to 112 in c9a1ec1

	# Apply scaling to drop scientific notation (x10^{18} molec/cm2 instead of molec/cm2)
	## Taylor plot doesn't work if don't do this.
	ds[co_ppbv_varname+'_column_model'] /= 1e18
	ds[co_ppbv_varname+"_column_model"] = ds[co_ppbv_varname+'_column_model'].assign_attrs(units='$10^{18} molec./cm^{2}$')
	ds['column'] /= 1e18
	ds["column"] = ds['column'].assign_attrs(units='$10^{18} molec./cm^{2}$')

Thanks Maggie! Current tropomi modules can plot Taylor, although with messed units. But with the scaling applied, other plots like the map plot, box plot are messed up and need to be re-configured with units in the yaml file. I prefer to revise the Taylor plot module to make it more flexible to handle large units.

Hi Meng and Maggie, I tried to do that in the plotting class and was not able to (for the TEMPO tool). I think that the problem is with the AxisArtist class from Matplotlib that the MONET uses, but I was not able to find an easy solution for that. In satplots_xr, I just added the option "normalize", which plots a normalized Taylor diagram... therefore having smaller numbers. I agree that revising the Taylor plot module is a better solution, and maybe someone else can find a way to do it (maybe there is a way with the GridHelperCurveLine?).

Having the option for an offset as suggested by Maggie is a better solution than mine, we could add that to the plt_grp in the yaml (and not the observations) for Taylor (even better, we could have both an offset and the option for normalizing it). However, I'd say that we can merge TROPOMI for now and deal with that after the deadline. Would that work? Cheers Pablo

Sounds good to me. - Meng

[rschwant]

Is this ready now? Should we merge it in?

[blychs]

@mbruckner-work , I believe we still need you to approve the review?