When climate models are wrong, it doesn't help at all to have bad data. Of obvious importance to climate data is of course, temperature measurements. Countless stations around the country monitor all kinds of statistics, which ideally should be completely free of interference so that objective conclusions can be made.
These stations were originally placed at rural locations throughout the country, but due to our gorging on cheap sources of fossil fuels, suburbia has expanded far into the countryside, often engulfing weather stations in their wake. There is a common weather phenomenon associated with cities, known as the urban heat island. Because of the vast amounts of concrete, asphalt, pavement in cities, and the relative lack of greenery, cities tend to much hotter in sunshine than the country. The heat island effect is noticeable to the point where even a child can scientifically measure its existence. So one would think that having stations that go from being in a green zone to being in an urban zone would be problematic. The IPCC assures us that this does not matter: if a sensor has been in a heat zone for very long amount of time, it should display a consistently higher temperature, but not constant increase. I'm not certain if I buy this explanation. To see how massively urban areas have expanded, compare this map of Chicago around when temperatures began to be recorded, to this today:
This is not to say that global warming isn't happening, it's that doubt shouldn't be cast on the extremity of predictions. Normally when a scientist conducts an experiment, and their data proves to be corrupted, they meet massive ridicule. Instead the IPCC has become indignant, saying the heat islands don't matter, as long as their measurements for ground temperatures match their data on oceans temperatures. However, independent sources say that conclusions based on oceanic data are ambiguous at best.
Lesson learned? If a set of data is unreliable, it's better to use data that's unambiguous and hard to fool--statistics can only repair data up to a point. This is why satellite data should be used as often as possible. Data can be recorded from anywhere on the globe, and it samples temperature form the atmosphere and is therefore prone to less variability than surface temperatures. Unfortunately, satellite data only goes as far back as mankind has had satellites. Weather balloon measurements serve as a good proxy, but even then those go back about fifty years. However, proponents of using surface temperatures don't like satellite measurements (which record temperatures in the atmosphere), because--surprise--they record less severe warming trends than for surface temperatures. Could this be the urban heat island effect in play? You be the judge.
Addendum 4/21/2008:In this post I neglected the man who is behind the verification of the climate data station--making sure that they are a minimum distance away from potential sources of the urban heat island effect--so that they give a more reliable measure of the temperature record. His name is Anthony Watts, a meteorologist who runs a blog that describes his research as well as SurfaceStations.org which keeps track of information about each station.