The main thrust of our Climate Analysis page is that the global temperature we obtain from the surface weather stations contains systematic errors of order ±0.1°C/decade. In the period 1960 to 2000, the number of weather stations dropped from around 8000 to below 2000. The weather stations are not uniformly distributed about the globe. Both CRU and NCDC claim that their analysis of the weather stations accounts for urban heating, disappearance of stations, and poor geographical distribution. But we are able to obtain almost exactly the same curve as them by simply integrating the average station temperature derivative. We did so with only a few lines of code (for description of calculation see here). We took no account of urban heating, no account of station disappearance, and no account of geographical distribution. (See graph below, our trend in pink, CRU trend in blue.)
From this we conclude that whatever compensation CRU and NCDC make for urban heating, station disappearance, and geography, their compensation has no significant effect upon their result. But station disappearance does have a significant effect upon the underlying data, as we show here. Thus we conclude that CRU and NCDC performed no compensation for the disappearance of stations. Because station disappearance and urban heating are intimately linked by the expansion of cities, we suspect that CRU and NCDC performed no correction for urban heating either.