Anthony Watts has three new sites: Lovelock, Electra and Fallon, and there’s something unexpected in them (Aside from the jet fighter discharging on the weather station – which is disappointing but almost expected by now.)
These three sites were included in a priority list because they are among the 10 nearest neighbors of Reno airport and Reno airport was chosen as a type case to show that USHCN v2 homogenization (scheduled for rollout in July 2007) deals with urbanization issues. USHCN here purported to show that their algorithm netted out little difference between Reno and its 10 nearest neighbors, as shown below.
Figure 1. Difference between annual minimum temperatures at Reno, Nevada and the mean from 10 nearby stations. The red line indicates TOB adjusted data; the green line is based on the fully adjusted data. Units are °F.
The 10 nearest sites to Reno are: Tahoe City; Lake Spaulding; Fallon Experimental Station, Susanville, Colfax, Quincy, Electra PH, Lovelock, Yosemite Park Headquarters, Marysville. In modern climate science, apparently it is not necessary to either visit the sites or even to have pictures of them in order to draw these conclusions. Although “professional” climate scientists do not need to visit sites to ensure quality control, professionals in other fields have different standards of due diligence and Anthony has interrupted his latte drinking to place 3 of the above 10 sites on the photographic record, each with a different surprise.
There are two sites at Lovelock (note Anthony’s comment below): one at Lovelock AP and another in town. Lovelock AP is used in GISS and GHCN but not USHCN, while the town site is part of both USHCN and th other networks.
Anthony has posted up separately on the Lovelock AP observing:
“In addition to the air conditioners, there’s nothing like an occasional jet blast or propwash to complement your high temperature measurements. Thats not a museum piece, its a working jet. The aviation ramp is in fact just 10 feet from the Stevenson Screen. This station has since been replaced with a more modern ASOS, about 200 yards south of this location, but this Stevenson Screen measured data for years in the exact same spot. And yes, the data from it is in the GISTEMP database…. Oh, and they like light bulbs in the screen here too. Notice this screen is the old kind with doors that open side by side. At least somebody had the good sense to put in a timer.
The values for GISS raw, GISS adjusted and GHCN raw for Lovelock airport are identical. (There is no GHCN adjusted series). All three series are plotted in the figure below, but only the last one shows due to over-writing. BTW I’ve been wondering why the GHCN adjusted data set was about 30% smaller than the GHCN raw dataset – I guess that this is one example of a difference. I compared this to the GISS pot and the geometry is the same (although I’ve plotted it here in anomaly version). There is an observation gap between 1981 and 1989, but no adjustment. How would one go about ensuring that the later readings were not inadvertently 0.3 deg C biased relative to the earlier readings? It doesn’t seem possible to me. Given that there is no need to go out of one’s way to include compromised data, I don’t see why GISS or GHCN would use this station. It doesn’t even have a complete normal period from 1961-1990.
With his usual CSI diligence, Anthony located both the present Lovelock location in a CWO observer’s backyard shown below, together with the prior location at a post office shown in Anthony’s comment below. For more details see surfacestations.org. Anthony observes: “Its in a CWO’s backyard. I also have the Lovelock previous station location at the Lovelock Post Office. The CWO asked for it to be moved to backyard from the PO dues to vandalism.”
Here’s a collation of Lovelock airport and town information (keep in mind that identical versions overwrite). There is a strong divergence between Lovelock AP and Lovelock town in the 2000s, amounting to over 1 deg C. The Lovelock town station moved from the Post Office to the observer’s backyard. It doesn’t look like this has been picked up in the adjustment algorithms. What a mess this stuff is.
Here is a different sort of oddity. The station is at a power station, in a “rural” but hardly a rustic setting, as shown below. (For ful details, as always, surfacestations.org)
Again, this site does not show warming and measurements are not reported after 1994. The GISS homogenized version is shown below. (I think that Anthony mentioned that the caretaker said that measurements were discontinued in 1999?)
Now here is a plot collating USHCN, GHCN and GISS versions, showing values later than the 1994 GISS termination. It also shows high values in the early part of the 20th century truncated by GISS – doubtless for a “good” reason, but a reason that, to my unknowledge, is unreported.) I checked the monthly data and there are no values in 6 of the 7 series for Electra PH after Dec 1994. However the USHCN adjusted values (Filnet) continue to the most recent USHCN period in this data (Dec 2002). If Electra PH is in the roster of 10 nearest stations for the Reno comparison, it is there despite having no values for the last 13 years.
Now for the most interesting one and it didn’t look interesting right away. Anthony reports that Fallon was hard to find: “At Fallon the actual site is ten miles from the USHCN recorded Lat/Long,and it took three hours of CSI work to find the station.” (I think that Anthony would have been good on The Amazing Race, where they had to find weird things all over the world.) Again see surfacestations.org for details. This picture “looks” fine at first glance, [but see comments in the thread about crop use].
Here is a plot of the various versions (the early part is truncated at GISS). Here we finally have a series that shows some recent warming (although it closes on an unexceptional value). USHCN data ends in Dec 2002 in the archived collation, but the GISS and GHCN data go to March 2006 (why not more updated, I don’t know). The surprising thing here is the situation is the opposite of what one expected: I, for one, was expecting to see exacerbated increases at the “bad” sites, but here’s what appears to be the opposite: a “good” site with strong warming commencing around 1993 – but stronger warming than the nearby “bad” sites.
Just to be thorough, I checked the Station History online here . Sure enough, there was a station move in November 1992, supposedly with a change in altitude of 1 foot and no change in lat-longs. At this point, I don’t know how much weight can be put on this sort of detail. Another move is reported in 1999 from 39°27’00″N 118°46’59″W to 39°27’25″N 118°46’51″W, southwest 0.3 miles. We’ve seen at De Bilt (and elsewhere) that seemingly innocuous moves of a few hundred meters can introduce measurement differences equal to total global warming in a century. In this case, no adjustment for the move has been made in any of the data sets. Should there be an adjustment? Because the bias is small relative to annual fluctuations, it’s hard to see how to find how to decide.
The comparison to other sites indicates to me that the Fallon move has resulted in an upward bias to this series. My suspicion is that the USHCN adjustment methods (which remain opaque to me) are biased against picking up this sort of upward shift, but will pick up downward shifts like dogs on a bone. I think that the reason for this is the inclusion of series with known UHI in the comparanda, which then bias the results. If none of the series are going up (e.g. east Colorado) the net result may not be very problematic, but if there are some bad apples, I suspect that the properties of the method change. Increasingly, it seems to me that USHCN is doing something very Mannian. They throw all kinds of data into a hopper and hope that “advanced” methods can recover something from it, without using a method known to the outside statistical world whose properties were understood and without many people even understanding how the method worked. Subtle biases can creep into this sort of method.
Whether the shift at Fallon is enough to trigger some sort of statistical test, I don’t know. However, it was enough that it caused me to look for a change around 1993 and sure enough, there was a change in November 1992. So there’s probably some test that would pick up this visual perception. It’s not easy saying how you can adjust for these things when everything in the network has problems and you’re trying to separate out UHI bias as well. My hunch is that you need to focus on quality. The CRN network is fine as far as it goes, but it gives information from 2003 on (and CRN data is not available to third parties as far as I can tell.) It doesn’t solve the historical problem. My guess is that one needs to extract the “best” 300 or 200 or 100 stations and that won’d be easy, since it seems that small station moves at “good” sites can affect a record as thoroughly as incinerators at bad sites. By falsely claiming that USHCN stations were all “high quality” stations, Karl and Hansen have deflected attention away from the enterprise of determining what can actually be salvaged from this network. Clearly something can be salvaged from it, but the work should have been done long ago.