A cooling-demand refrigerant-bank prior does not improve population for HFC-134a inversions (a negative result)

A reproducible negative result for the F-gas atmospheric-inversion community. It tests whether a refrigerant-bank prior weighted by population × cooling-degree-days (CDD) improves on the population proxy used by EDGAR as a spatial prior for European HFC-134a inversions. It does not. Result: against the 2020 ICOS PARIS HFC-134a posterior, the bank prior scores pooled spatial correlation 0.270 versus 0.634 for population and 0.671 for EDGAR, winning in 0 of 16 countries; CDD weighting degrades the population baseline by zeroing out the large refrigerant banks of cold-climate countries. HFC-134a is a diffuse, population-coupled source with no point-source structure to resolve, the opposite of CF4 from aluminium smelters (DOI 10.5281/zenodo.20617486). Scope: one proxy (population × CDD) over Europe in 2020. The test is conservative because the inversion's own prior is EDGAR/population-weighted, yet the bank prior still loses to plain population by 2.3×. Contents: the candidate and baseline gridded priors (NetCDF), the CDD field, the method note, the full validation report, and the reproduction code. Useful takeaway: keep the population/EDGAR proxy for HFC-134a spatial priors. Code: https://github.com/zachdissington/climate-solutions-research-kit