>The techniques used to smear individual observations over vast swathes of territory tend to leave much to be desired.

I am by no means an expert, but after farting around with climate data, even today I'm curious how they smear data across comparatively smaller swathes of territory.

I've noticed, in my area (south west), that many of the stations are located in seemingly unique microclimates--like the tops of mountains, on the shores of rivers, etc.

It roadblocked my layman research because distance to a given station was not very indicative of how that area's climate is.

For one example, note https://data.giss.nasa.gov/gistemp/graphs_v4/ which leads to http://www.columbia.edu/~mhs119/Temperature/ which leads to https://data.giss.nasa.gov/gistemp/ which leads to https://data.giss.nasa.gov/gistemp/sources_v4/ which should give you some non-source controlled Python code translated from Fortran code which tried to use a weighted average from stations in something like a 1250 mile radius based on the unproved assertion that deviations from a 30-year mean were correlated well even though levels were not.

Representative comment:

    In Step 5: 8000 subboxes are combined into 80 boxes, and ocean data is
    combined with land data; boxes are combined into latitudinal zones
    (including hemispheric and global zones); annual and seasonal anomalies
    are computed from monthly anomalies.

I can't believe I missed this: https://data.giss.nasa.gov/gistemp/sources_v4/gistemp.html:

> In 2007, David Jones' team, Climate Code Foundation, reprogrammed the whole procedure in python, as a product of the Clear Climate Code project and the Climate Code Foundation, http://clearclimatecode.org/

That's all the source control you'll ever get. Don't bother clicking on that link.