This continues the series comparing the GHCN station temperature data used by most major indices and blog reconstructions with the larger, different SYNOP-based NCDC GSOD dataset which Ron Broberg has made available. Since GSOD is patchy pre 1973, that is the start year for this check.
It turns out (below) that for Global Land/Ocean, both datasets give very similar results. In a way, this is not surprising, because both use the same ocean dataset HADSST2, which covers most of the globe. But it does suggest that any lack of coverage by GHCN is not a big source of error.
Here is the metadata compared:
And here are the plots separately, and superimposed. The anomalies are mapped to the common base period 1979-2000.
Since we have already shown that the GHCN results compare well with the major indices, the GSOD results will also be very similar.