One of my favorite bits of irony is that the primary defenders of using surface temperature measurement over space-based satellite measurements is … the Goddard Institute of Space Studies at NASA, and James Hansen (its director and friend-of-Al) in particular. If find it amazing that people still want to use the GISS surface temperature numbers in preference to satellite figures, despite their proven biases and lack of consistent coverage. But the GISS numbers give a higher number for warming (since they are biased upwards both by measurement biases and GISS-added adjustment factors) and that is what is important to global warming alarmists. Its the "fake but accurate" meme brought to the realm of science.
But, since we do have to keep reminding people of the problems in surface temperature measurement, here is a study by Ren et al in 2008:
What was done
Noting that "a major divergence of views exists in the international climatological community on whether the urbanization effect still remains in the current global and regional average surface air temperature series," the authors employed a dataset obtained from 282 meteorological stations, including all of the ordinary and national basic and reference weather stations of north China, in order to determine the urbanization effect on surface air temperature trends of that part of the country over the period 1961-2000, dividing the stations into the following categories based on city size expressed in millions of people: rural (<0.05), small city (0.01-0.10), medium city (0.10-0.50), large city (0.50-1.00) and metropolis (>1.00).
What was learned
Ren et al. report that mean annual surface air temperature trends for the various station groups of north China over the 1961-2000 period — in degrees C per decade — were 0.18 (rural), 0.25 (small city), 0.28 (medium city), 0.34 (large city), 0.26 (metropolis), and 0.29 (national), which makes the urban-induced component of the warming trend equal to 0.07 (small city), 0.10 (medium city), 0.16 (large city), 0.08 (metropolis), and 0.11 (national), all of which results are significant at the 0.01 level.