In any case, here's what my limited data collection has in it:
I mentioned to Ira having seen something on the difference between NOAA's and NASA's (GISS) use which I found at Climate Audit: NOAA versus NASA-1/20/09. McIntyre also refers to a post a few days earlier with a bit more data: Climate Audit: NOAA versus NASA-1/16/09
The figure above suggests that the NOAA temperature dataset from about 1930-1940 to be 0.1°C colder than the GISS dataset - and by 2000, 0.2°C warmer than the GISS. If I'm interpreting this correctly, the NOAA chart would suggest more warming than GISS of about 0.3°C (0.5°F) over the last 70 years. This 0.3°C difference MIGHT be the effect of GISS attempting to correct for UHI more extensively than NOAA.
The NOAA UHI correction apparently can be described as: "I'm not actually going to tell you what we do." They refer to Smith and Reynolds 2005 and I could not find any reference to UHI or any station corrections based on location.
There's a helpful UHI related summary at CO2Science.org.
I'll give you the 2 cent tour of this, as there's some interesting information. But first, I'd like to summarize a bit about UHI. Urban heat islands are the readily observable increases in temperature (both day and night) seen in cities. At night, heat absorbed in pavement and buildings can warm the city (and any NOAA temperature sensors that happen to be located there). But in the daytime, the lower albedo of cities can apparently cause warming. Wind flow is interrupted as well, probably reducing the dissipation of city heat. Finally, the heat released as energy is used (can we blame Nicolai Carnot?) is not insignificant.
So anyway, CO2 Science summarizes several papers at the location above. They point out that the UHI effect can be measured either via satellite measurements or surface thermometers. Harbor seawater temperatures show this effect as well (Boston harbor warmed 3.6°C over 100 years). He refers to a Changnon study that might show that the majority of the reported warming in Illinois from 1890 to 1950 could well be UHI errors.
In a large (~4600 residents) native community in the Arctic near Barrow, Alaska, a grid of thermometers within and well outside the community suggested that the mean UHI warming was 2.2°C and during the daytime, differences of 6°C to 8°C and above were seen.
UHI differences of 8°C have been seen in San Juan, Puerto Rico as well. The author of a study of cities in Mexico concluded that the huge magnitude of the UHI effect probably swamps any CO2 AGW signal.
The CO2 science author concludes: "it is presumptuous in the extreme to believe that the global surface air temperature record of the last few decades has been adequately adjusted for small-town and large-city heat island effects."
You might also enjoy following a blogger that has recently been plotting UHI related dataset information at: The Sextant, January 10, 2010.