Day after day Seattle-Tacoma Airport's temperature gauge was warmer than observations in its neighborhood. Was it the impact of the third runway? Or the nearby concrete and gravel?
The mystery grew. The game was afoot.
And then on 15 March 2014, the Seattle-Tacoma Airport temperature sensor went crazy, reporting a high temperature of 76F while four nearby observing sites had highs of 58, 59, 60, 61. The next day the NWS meteorological technicians replaced the temperature sensors and some of the electronics, and guess what happened?
Much of the warm bias disappeared.
Here is a plot of the difference of the temperature between Seattle-Tacoma Airport and a nearby weather site (Weatherunderground SEAT4, see map below, that station is SW of the Sea-Tac observation--the one with the cloud). The vertical red line shows the day the equipment was replaced and the horizontal green lines show the average temperatures before and after the switch. UW's Mark Albright produced this plot. There is clearly a significant drop in temperature after the sensor replacement.
Further evidence is provided by plotting temperatures at Seattle-Tacoma Airport and at nearby Tacoma Narrows Airport in Tacoma. Before March 16th, Sea-Tac has a higher maximum temperature than Tacoma Narrows (TIW) nearly every day, but after the repair they were consistenly very similar.
Is this an isolated incident? Unfortunately not. I can give you a long list of airport weather sensors that were reading warm, and then cooled considerably when they were replaced (e.g., Douglas, Arizona). Or sensor that began reading warm when their fans failed or slowed. The trouble is that most failures seems to lead to warmer, not cooler temperatures. Thus, problematic sensors and equipment are contributing a warm bias in official temperature records. How big a problem is this? I am not sure we have any kind of handle on this.
And there is another problem leading to unrepresentative warming: urbanization and development around temperature sensors. The Wenatchee temperature sensor, part of the Historical Climatological network, is shown below as an example of poor exposure--with a hot parking lot nearby as well as nearby buildiing.
We are fortunate to have a number of meteorological sensors that are less prone to such problems (e.g., satellites), but their records are quite short compared to the surface instrumental record.