As I've demonstrated (to my own satisfaction, at least) the strange, arcane adjustments for HIE made by CRU and GISS are not even necessary. One gets the same trends whether they are made(GISS Data Set 2) or not (GISS Data Set 1) if an unbiased and well understood method for determining temperature anomalies is used. What exactly they are doing with those adjustments is up in the air, but the question remains to what extent the HIE is affecting the results that we are seeing for global temperatures.
Analysis of the GISS station data had begun to make me think that the heat island effect did not exist. But yesterday I was looking over station data from the local Houston area, and it did seem that there was an obvious HIE. Individual stations in the heart of the city, inside the loop, were reading about 3 or 4 degrees above those out in the suburbs and beyond. It was striking to me, though, how quickly the readings came back down to match those of the surrounding countryside getting even a little away from the city center.
Looking at other cities I found the same thing. wunderderground.com enables one to look at dozens of individual stations in an area simultaneously. Note how in Chicago the temperature readings decline as one gets away from the city center:
And here is the New York City area:
Compare to the surrounding regions:
Note the two stations in the middle of Manhattan. They are located 500 feet up.
Unless stations were actually within the city they would not detect the effect.
So, looking back over the maps of GISS station locations, it hit me that the vast majority of those stations marked as urban are not urban at all, at least not to the extent that was seemingly needed to find a HIE. All this time the suspicion has been that the stations marked rural are too urbanized when it could be exactly the opposite. Those marked urban are often not that urban.
Thus, I selected stations from the 85 largest urban centers in the world. In order the qualify the station had to be within the urban area, urban sprawl for 0.1 degrees latitude or longitude in every direction from the designated station location (which is provided to the nearest tenth of a degree). The temperature anomalies for these, as determined by the statistical algorithm used in previous posts, is compared to the published GISS data below:
To make a long story short, as more cities are added in to the list of city stations included in the analysis above, from smaller and smaller cities, the trend line comes down to be much closer to that of the GISS data. And this is still well short, only about 10%, of the stations listed as urban in the giss.dset1 data. As I've shown previously, if all the "urban" stations are included the HIE disappears.