Just a check-in of the exact code and process used to generate the data. The climate data file was downloaded from: https://crudata.uea.ac.uk/cru/data/hrg/cru_ts_4.06/cruts.2205201912.v4.06/tmp/cru_ts4.06.2011.2020.tmp.dat.nc.gz It should remain downloadable as-is in the future, though there have been cases of the files being reissued without a new point release if they're found to have defects. The source names (city and country) were copy-pasted from table S4 of https://www.thelancet.com/cms/10.1016/S0140-6736(14)62114-0/attachment/3daac933-d843-4742-95cd-07cae706f14f/mmc1.pdf The script finds the appropriate wikidata entry with user assistance, then uses the wikidata property https://www.wikidata.org/wiki/Property:P625 (coordinate location) to look up the nearest cell in the climate dataset. The average temperature is computed over all datapoints in the provided temperature NetCDF file (so, for the one I picked, over all 120 months from 2011 to 2020 inclusively). The output.csv file is the result of this process. Can be useful in identifying if something was done incorrectly.