[cryptica]

That's why I said to look at any place. It you want to make the argument that the temperature differences are so small that we shouldn't expect to notice any difference from looking at many different before and after charts then perhaps those differences should be considered statistically insignificant... More likely to be something wrong with the measuring instruments (maybe changes in the manufacturing process of these instruments), or people are becoming more deceptive when entering their measurements or being more selective in how they combine different measurements.

Why should I trust a heavily processed, combined result like the Wikipedia page you posted more than random sampling of raw data directly from local meteorologists? It's pretty obvious that the more 'processing' of the data is carried out, the more inaccurate it's going to be. By the time we get to the global mashup of the data, it's been so heavily tempered by people along the hierarchy that it's basically useless. The global statistics are not to be trusted, they're the result of many meteorologists and statisticians playing 'Chinese whispers' at a global scale.