Publishing this post now in advance of completion to have it available in another post. More details of the companion KML file to be added.
I started this blog in 2010, discussing shortcomings in the station location metadata used in Gistemp at that time, derived from the GHCN-M v2 station inventory. GISS has moved on to use the GHCN-M v3 inventory, and now the GHCN-M v4 inventory. But the station location metadata in these versions is still problematic. It is not possible to correctly classify a station as urban or rural if you try to make the classification on the basis of night time luminance from a satellite image if you do not actually know the correct location of that station.
Limassol was one of the first mistakes I spotted when looking for stations located at sea which should be found in reality on land. In this case the error seems to be a simple data entry one, 32° East entered instead of 33° East, leading to identification by night time luminance at a location guaranteed to be dark at night rather than in the port of Limassol.
Le Puy de Dome and El Tor are two further examples where I have personally visited the true location. These three examples illustrate two ways erroneous location metadata can be detected in the inventory itself with mapping (location of a land station at sea and location of a high altitude station at low altitude), as well as by personal knowledge of the location.
The majority of stations used by GISTEMP v3 which currently report data are WMO stations (1397 of the 2117 stations reporting in 2018). Many of these have erroneous station location metadata in the GHCN v3 inventory. The higher precision coordinates provided by the WMO programme to update Vol A can be used to find such errors. Not all such location errors will lead to erroneous classification. Looking at a rural location many kilometres away from a true rural location will still give a “rural” classification despite the error. But 434 WMO stations in the GHCN/GISS v3 inventory will change urban/rural classification if the erroneous location metadata is replaced by the WMO coordinates. A KML file displaying the WMO and erroneous locations of these 434 stations can be found in a companion post.