PyPSA
clean out all nuts3 references
The goal is to eliminate shape reprojection from NUTS3 to model regions. Instead, use gridded data from
https://datadryad.org/stash/dataset/doi:10.5061/dryad.dk1j0
and
https://www.worldpop.org/
Checkout:
https://github.com/martacki/pypsa-ua_gdp_data
So using 5 arcmin dataset looks not quite close to NUTS3-level dataset
gridded dataset -> NUTS3 aggregated -> NUTS3 original
Using 30 arcsec dataset is quite slow.
Maybe mismatching units? NUTS3 dataset was pop in abs. values IIRC, the gridded pop data should be pop density.
dunno if you intend to use another formula now. For ua and md, that seems to be the currently best solution, however I have the feeling that for germany the current nuts3 data provides a much better distribution...
nuts3 for germany:
suggested data sources (according to the formulas used in my private repo that you're quoting):
Using 30 arcsec dataset is quite slow.
ah, but in the repo I'm not using the 30 arcsec data, I think. Sorry, overseen this comment... is unifying the data source with ua and md the only motivation for this pr? (ie. why clean out nuts3?)
I just wanted to sketch how it could work based on what you did for UA/MD and explore how it would compare to NUTS3 data. The motivation would be that it simplifies the code (the whole nuts3 function would be superfluous) and we could possible go to higher resolutions than nuts3. Also: opportunity to remove yet another vresutils dependency.
@FabianHofmann Can we reuse the atlite code used for the ExclusionCalculator to implement the calculation faster than what xagg has to offer? (Determine overlaps between regions / shapes and a raster / the population raster data).