Last week I showed some graphs like the ones below, which demonstrate that the TOBS (time of observation bias) adjustment is bogus. Stations which took their readings during the morning on July 15, 1936 are actually cooling slightly relative to the July 15, 1936 afternoon stations. This behavior is the opposite of what TOBS was created to correct.

So this morning I tried the same experiment on the raw monthly USHCN data using the identical set of stations as was used in the daily analysis. In this case, something very unexpected appeared. The morning stations are warming much faster than the afternoon stations, which is what TOBS theory predicts.

This discrepancy makes no sense, because I am using the identical set of stations for both the monthly and daily data. The monthly data is supposed to be the average of the daily data.

So the next obvious experiment was to compare the monthly data for…