How airports like BWI help set outlier high temperature records

Just to add to what you were saying, I am about 30 – 45 mins southeast of RDU in the country. It was 30F when I got up yesterday (5AM) and there was ice on the water tanks compared to your “… 5 am, the temperature outside was 35 F…”
Interesting. To the southeast, too — the exact direction it is often a bit warmer (closer to the coast, over in the sandhills). We just missed a frost the last three nights running (low of 37F last night, fortunately for my peach and plum trees:-) which makes this, in the end, a rather normal year, if anything running relatively cool at the moment however warm March was. I generally consider 4/15 to be “last frost”, although I’ve had killing frosts in the first week of May in the last fifteen years (it killed all of my azaleas, that are winter hardy but very vulnerable to frost after they bloom) and spent an absolutely frigid Easter in mid-April three or four years ago all the way out on Ocracoke (right on the Atlantic, mind you, generally moderated by the ocean) fishing in a mix of snow and sleet in a 40 mph freezing wind (that yes, killed all of my peaches and plums back in Durham). But many years back in the 80s I used to plant my tomatoes 4/1 and get away with it.
Personally, I think the most interesting way to compute average temperature in the US would be to:
a) Contact the Weather Underground, or any other similar group that receives regular input from user-contributed weather stations.
b) Take the data from these stations. It typically is an hourly or better timeseries, and available backward for many years per station.
c) Cover the US with an icosahedral grid at a granularity such that most grid cells have at most one station.
d) Sum local temperature times area for a suitable (adaptive) area per station, per time, applying a simple accept/reject criterion to “mark out” stations that are obviously down.
e) Do an actual time average of the temperature per site! One has the data, why not? Indeed one can do this and compare the result to the average obtained from (max + min)/2 and get a clear idea of the bias inherent in the latter (if any).
f) Compute the spatiotemporal average temperature by dividing out by area and time, using the actual data over the actual year.
Come on, folks, it isn’t that hard. No need for any corrections, just compute the damn averages by summing and dividing. The only adjustable parameter, the average area assigned per site, can be computed using an algorithm that can be precisely normalized a variety of unbiased ways. Indeed, one measure of a good algorithm for doing the spatial integration required in the average is that it should produce the same result as all of the other good algorithms for doing so (there are several one can use) so that the result does not depend on algorithm or any personal choices that could lead to bias. Either Willis or I could come up with numerically sound algorithms for estimating an integral over a 2D surface from irregularly spaced data points, if only because we would actually look them up and test them with e.g. Monte Carlo data to be sure that they give the right answers (and at the same time, estimate their probable error bars).
The key to doing this is to not correct for anything but a station being offline entirely or emitting obviously spurious results (a constant but incorrect/inconsistent temperature all day long). That way one ends up with a dynamic spatial map of temperatures. In fact, one could make a damn movie out of it — a false color scheme for temperatures, an hour a second, and you could watch the US temperatures evolve at the granularity of the stations from overhead over an entire year in a little less than three hours. UHI and probable spurious spikes would be perfectly evident. Child’s play.
rgb

“For a brief 10 minutes, the steady NW wind that persisted all day at BWI shifted to a westerly direction. That allowed the HEAT from the nearby runway to provide a quick 3 degree warm-up between hourly obs. Once the winds shifted back to a NW direction, the temperature fell back to 59 degrees. ”
There is one big problem with this. Looking at the source linked in the article I noticed rather a lot of “things” on the supposed runway. Source: http://binged.it/HC0WPG
Now most runways I’ve seen have been devoid of any object larger than a piece of DC-10 engine cowling. (No, I’m not joking but that’s another story).
So I went to the link and zoomed out one step.
http://www.bing.com/maps/?v=2&cp=qhw0dm8m8q77&lvl=19.64&dir=0.14&sty=b&where1=39.1666,%20-76.6833&for
m=LMLTCC
Ah! there’s “another” runway to the north of the ASOS location. So the ” steady NW wind that persisted all day at BWI ” was, in fact, coming directly over the REAL runway (not the smaller access road to the south of the sensor).
Now I don’t know much about weather but our host knows a thing or two so perhaps there’s an explanation in different air masses or the part of the city over which the air just past.
I would have thought that a notable change in wind direction would usually be accompanied by a change in temp. I’ll let Anthony go into that in more detail.

Prof. Brown and E M Smith – just a hurrah for your posts. What is so baffling is that fixing this problem is neither difficult nor expensive.
It may not matter much in the world of Mr Mosher’s arcane data collection/processing (we had a brush about Paris elsewhere where he said in answer to a question about why the numbers didn’t fit Paris – we didn’t include Paris) – but people living in cities are being forced into all kinds of expensive policy measures based on these shoddy statistics.