Mar 18 2008
Hay Springs Surface Station
In his latest post on how not to measure temperature, Anthony Watts profiles the Hay Springs surface stations. He actually provides photos of two COOP stations in Hay Springs, but it appears only one is included in the USHCN. The ‘better’ sited of the two is included in the USHCN. There is some interesting things about this station when compared to its surrounding stations.
Using the same technique that I’ve previously used, I compare the temperature at Hay Springs (COOP ID: 253715) with the 10 closest surrounding stations. I find the 10 closest stations via a brute force method, using the latitude and longitude of each station to calculate the great circle distance between them. The temperature of the 10 stations are then averaged and subtracted from the Hay Springs temperature.
|
In this figure, the red vertical dashed lines represent station moves from the NCDC station history. What’s interesting is that Watts says that “the USHCN station was converted from a Stevenson Screen to a MMTS in 1989″, but this is not included in this station history. I’ve included 1989 as a vertical dashed cyan line.
Immediately after the 1989 switch to the MMTS this station has a positive bias compared to the surrounding stations. After a few years, this bias decreases, but is always greater than zero. All the other station moves associated with this station appear to have been adequately corrected. If it’s true that there was a switch to a MMTS in 1989, it would appear that this is not correctly corrected.
This plot also includes a green line, which is the linear fit of the bias between this station and surrounding stations. At the beginning of the time series, the Hay Springs station is negatively biased compared to surrounding stations. But at the end of the time series, it is positively biased. I’ve included the linear fitted trend to the subtitle. This trend is not significantly different than zero, but if this station is considered rural, it would appear that it would bias the temperature record towards more positive trends.
(No Ratings Yet)
Loading ...Related Posts:
11 Responses to “Hay Springs Surface Station”
To reduce spam, comments are automatically closed 30 days after the last comment. If you would like to comment on any closed thread, please use the contact form at the top of this page.
Hi Atmoz,
I’m unconvinced by the MMTS adjustment done by NOAA based on Quayle. My sense would be that some of the adjustments are going to be high and others low (that’s probably a testable hypothesis) After all Quayle merely estimated a mean difference and then that mean (which has error) is merey used to adjust records. So, One could assume (but test of course) that some adjustments would be a wee bit high and others a wee bit low.
if you get my drift. Also, Need to look at the siting in quayles study. perhaps as time as gone on the MMTS have actually been place closer to buildings (altering the cooling bias found in the original quayle study) .
I agree with Mosh. I believe the Quayle study mostly touched on instrumental differences, not siting differences.
The MMTS shield is more efficient at blocking direct and reflected sunlight and LWIR. So it is not surprising to see a cooling bias for it.
I don’t think Quayle took into account that the COOP manager has a shovel and an afternoon as the only tools available to install these MMTS units. Time, tools, availability of location to place it, and impassable barriers like sidewalks and driveways, plus varying degrees of effort/motivation all conspire here during installation.
One of the best photos that illustrates the problem is from Bainbridge, Georgia, where the old and new stations and the difference plus barriers to installation can be clearly seen.
http://gallery.surfacestations.org/main.php?g2_itemId=5645
[...] I was surprised to see that many of the listed stations have been identified as being extremely poorly sited. In fact, the surface station at Hay Springs fits the above criteria for “goodness” [see also Watts Up, Atmoz]. [...]
First of all you have to dig deeper to find the equipment changes and a whole lot more of the nitty gritty.
Second, and more interesting, did you look and see whether the nearby stations went to mmts? The other station in Hay did so in 1990 for example.
Using your list of NE stations, Albion did so in 1985, Beaver City in 87, wtf Auburn 5ese is I do not know, etc. It’s not as simple as the equipment change.
Those direct links don’t work. But station moves, and equipment changes should already be accounted for in the temperature record. The switch to MMTS should also be in the accounted for in the temperature record. Whether the deviation is because of an undocumented change at this station, an undocumented change at a surrounding station, or a combination is immaterial.
The other Hay Springs station is not included in the USHCN. Thus, it wasn’t included in this analysis. I’m not sure why you brought it up.
“Auburn 5ese” is USHCN COOP ID 250435.
Also, the “regional average” temperature is found by taking the average of the 10 closest stations, regardless of “goodness“, and regardless of which state they are in. The stations which were compared are listed in the figure of this post. It’s a little hard to read, but they are: Alliance, NE; Lodgepole, NE; Bridgeport, NE; Rapid City, SD; Wray, CO; Hettinger, ND; Hot Springs, SD; Cheyenne Wells, CO; Cottonwood, SD; and New England, ND.
Okay, go to the link for the station say the one you posted for Auburn 5ese. Hit the link for station history. Hit the next link below the station name for Additional Station History. Hit the tab for equipment. Other tabs and links in on the Additional Station History page are interesting.
FWIW, Alliance got an MMTS in 86, Lodgepole in 89, Bridgeport in 89, Rapid City in 83, Wray in 84, Hettinger in 88, and you get the idea. So whatever you have found is not a systematic effect from the MMTS since all the other stations were MMTS
It may not be the MMTS. But surface station temperature time series have been shown to be highly correlated for stations within 1000km of each other (Hansen and Lebedeff, 1987, and me). Given the high correlation we expect at distances less than 1000km, and the fact that all 10 of these stations are within at least 500km, what is causing the 2F temperature bias immediately after the MMTS was installed at this location?
Reference:
Hansen, J.E., and S. Lebedeff, 1987: Global trends of measured surface air temperature. J. Geophys. Res., 92, 13345-13372.
Highly correlated does not mean perfectly correlated.
[...] post on the Hay Springs surface station a few days ago has attracted the attention of Eli Rabbett. I was drawn to this particular surface [...]
Anthony, There is a 20 year side by side study of MMTS versus LIG in a CRS.
http://ams.confex.com/ams/15AppClimate/techprogram/paper_91613.htm
this study establishes the instrument Bias at a single site.
I think it also shows the seasonaility of the effect, which might
help sort things out.
QUAYLE, if I read him correctly, analyzed feilded systems. So whatever he found would be a function of instrument bias
AND siting difference. ( if there was any) I need to re read his
paper to be sure of this, but i think I’m right on this.
The relevant question, I think, is what was the siting like in quayle’s study and is the siting today like it was when quayle studied, or have a majority of the newly installed systems
moved closer and closer to structures. That’s not a simple question to answer, especially of qualyle didnt document the sites he used in his analysis.
Also, there is the nasty little fact that every adjustment is in fact
an ESTIMATE with an error.
Atmoz,
After reviewing data from a number of USHCN stations looking for temporal and spatial differences in temperature trend patterns, I make the following observations: 1) the monthly mean temperatures for a station were not normally-distributed and are not stationary over the 109yr period that I evaluated, 2) trends in the deviation of normalized monthly mean temperature from the period average temperature were readily discernable but differed in amplitude and duration from month-to-month, 3) after calculating the annual mean temperature from the monthly mean temperatures, the trends were readily discernable but diferent from all of the monthly trends and 4) the station-to-station differences were quite significant. These observations add up to one thing — be careful about interpreting trends. I realize that averages are averages, whether they be arithmetic means, geometric means, RSS means, etc., but combining non-stationary data is problematic.
To get a better handle on the effects of non-normal and non-staionary data, I created 12-tuple vectors for the annual data and compared the difference in length and the angle between the annual vector for each year and the 109yr vector. These differences were significant: 1 ) the difference in length for a specific site ranged from 12 to 13 deg F with a STDEV of 3 to 4 and 2) the anglulardifference with respect to the 109yr vector for a specific sight ranged from about 0.07 radians (3.8 deg) with a STDEV of 0.018 (1.03 deg) to 0.12 radians with as STDEV of 0.01 radians (0.6deg). Small differences? Perhaps, but significant to a control systems engineer.
All of this leaves me a bit uncomfortable with the way that you determine regional averages.
BTW, I used data for the following sites: 1) Ft. Morgan, CO, 2) Ft. Collins, CO 3) Boulder, CO, 4) WRAY, CO, 5) Cheyenne, WY and 6) Imperial, NB, all being within about 100 miles from Ft. Morgan