[Illustrations, footnotes and references available in PDF version]
Dan intriguing article regarding the influence of the number of stations on the global temperature measurement has been written by Ross McKitrick and published on his website here: http://www.uoguelph.ca/~rmckitri/research/nvst.html. Looking at his graph, the eye is instantly drawn to the apparent correlation between the stations and temperature.
There is a huge step in the raw mean temperatures around 1990 with a synchronous change in the number of stations. The adjustments and gridding methods employed to create the accepted global temperature records, are reportedly satisfactory to deal with such data aberrations and on visual inspection there is no immediately obvious problem. However, that does not rule out the possibility that the gridded end product contains artefacts of the processing and/or character from the raw data that are not related to climate and that could potentially distort the overall picture. This brief communication describes a method, utilising the dataset from Ross McKitrick and Joe D’Aleo, to calculate a historic temperature record through modelling the relationship between raw mean temperatures and the number of stations.