Early on in my www.surfacestations.org project I got complaints like “why don’t you ever show a cooling bias?”. Problem was, there weren’t a lot of cold bias stations to show. Other than shading, and in one case in Washington state where the station was getting mist from a dam spillway, there just weren’t many examples. This also explains why “Eli Rabett” was never able to keep up his “cool station of the day” series that he started as a counter to my published photos (he used mine), some months ago.
Given human energy use, and artificial surfaces such as concrete, asphalt, and roofing tend to be near where many stations have been located, the bias has been predominantly warmer. Another reason the predominate bias is warm is that stations are often placed for the convenience of the observer, which means closer to human habitation.
So it was with some intrigue that I found this station yesterday, with the help of my very first volunteer, Russ Steele. The NOAA COOP station simultaneously demonstrates a cold bias and a warm bias. This is the infrared photo I took of it:

The warm bias seem obvious, but can you guess what the cold bias is?
Ok if you didn’t guess it, don’t feel bad. Infrared photos don’t show detail well.
Here is the answer:

Snow on the roof of the shelter.
At the time, the sun had just passed below the tree line to the west, and the snow was melting. The cold melt water was dribbling down the slats on the rear of the Stevenson Screen shelter. You can see some of the slats are just a little bit darker in the IR photo. Of course any air passing through those slats would lose some heat to that water, cooling the air reaching the thermometer. The result would be a cool bias for daytime Tmax when the snow is melting, but only for as long as the snow is there. The effect would likely be greater on a warmer day with more sunshine as more snow melted.
You might say; “but all similar stations in the area would have snow on the roof, so the bias would be nearly equal and would be caught in adjustments.” Well….maybe. You see the problem is, the NWS doesn’t seem to have a published procedure for COOP observers related to snow removal from Stevenson Screens (At least none I’ve found, if you know of one please advise). So some observers might clean snow from the roofs, while others may not. So some stations in the area might have a cool bias, others may not. But how would you know? Note the square hole in the snow. That is for a snow depth board for which there are procedures.
This station has some other biases, one of which Russ and I are examining in detail. It turns out this particular station figures significantly into a peer reviewed paper. Stay tuned as we learn more from comparative measurements done in parallel with this station.