Creating grids using national downscaling when not all countries/territories have (GDP) data

[jkikstra]

I have a question: in previous projects, how did one deal with countries for which you had no GDP data available?

The file [ssp_comb_indexraster.nc](http://ssp_comb_indexraster.nc/) has 234 "countries" (incl. sdn_ssd, isr_pse, srb_ksv).

1. What should be done with countries not in the SSP GDP data (but that are in regionmapping, historical data, and in the gridding used)?
2. Could you point me to the data I need to reproduce this file and tweak it as necessary? I see the script "notebooks\gridding_data\generate_ceds_proxy_netcdfs.py" which points to gridding (on the RESCUE drive). Could you confirm this is the right place?

Another questions I wondered about (but less urgent): did you also consider using population maps (or other proxies beyond GDP), and what did you conclude?

@gidden @coroa
(FYI @ssmithClimate)

[gidden]

Hey @jkikstra - thanks for raising this.

In short, as input to the process, we need a raster that places all CEDS countries/regions on a relevant grid. In the past, we do that by looking exactly at how they did this with their Rdata files (you can find the function here.

Those country combinations specific to CEDS are defined in the config file

Concordia then simply calls into aneris'sDownscaler class. Population is supported, but you can actually add any method you like via the add_method() function here

And in fact, we do use population and other GDP weights to downscale some sectors, as enumerated in the defaul_methods() function. I realize that's not very transparent, and it would be great to have some improvements PR'd!

[gidden]

And yes, there should be data to reproduce on the RESCUE drive (I would need to look from my non-work computer to access that)

[jkikstra]

Thanks @gidden.
Now back from leave and looking at this again.
I found the country masks on the RESCUE drive and will check in with @ssmithClimate whether this is still up-to-date

In addition, I found the "country_location_index_05.csv" file is also used. Where did that originate (I couldn't find the source)?

concordia/notebooks/gridding_data/generate_ceds_proxy_netcdfs.py

Line 145 in 22bf33b

idxs = pd.read_csv(ceds_input_gridding_path / "country_location_index_05.csv")

[ssmithClimate]

country_location_index_05 originated with CEDS - we still use that actually (its in the public CEDS repo).

This is a "Matrix indices indicating locations of country masks in extent of a global map at 0.5 degree resolution." that I believe we generated from high resolution GPW country data.

[ssmithClimate]

And here is the documentation on the origin of those country boundary data summarized in that file:
The shapefile is manually created mainly based on GPW v3 ntblnds.shp with a small portion mixed using gdam v2.8 country boundary shapefile to provide boundary information of some newly assembled/dissolved countries/regions.

[gidden]

Yep, we used the publicly available CEDS processing data here. You can download that folder tree from CEDS!

[jkikstra]

Alright perfect, thanks both, found it https://github.com/JGCRI/CEDS/blob/master/input/gridding/gridding_mappings/country_location_index_05.csv

And the masks: https://zenodo.org/records/3249323