As an exercise, I’ve plotted the locations of the GISS Code 1 stations in the US, color coded them to show the ones that end early and then examined the Code 1 stations in California where there is a combination of both a strong GISS trend anomaly and station survey completion.
First, here is an interesting diagram from Hansen et al 2001 showing 1900-1999 trends based on “Unlit” stations – showing results for different USHCN versions. I presume that the interpolation here uses something like the R package akima, which I used in a plot last year. For now, don’t worry too much about the differences in trends for the different USHCN versions – just notice the hot spot in southern California, as that’s what we’re going to look at in more detail.
Next here is a plot showing the location of all the Code 1 stations in the GISS file, color coded so that blue ends by 1990 and red goes after 2000. A couple of things to notice here – first, none of the GISS Code 1 stations in Canada or Mexico has any information after 1990. C’mon, Jim and Reto – get off your LaZee-Boys and get the data. So to the extent that post-1990 information affects the Hansen two-legged adjustment, it is all from US data.
Next here is a zoom-in showing the stations that would presumably contribute to the southern California hotspot. There are only 10 of them. Since Hansen has identified them as Code 1, I guess that we can interpret that as some kind of voucher that these stations are the crème de la crème of U.S. stations, that here we will find the distilled essence of station history at its finest, NASA’s top guns.
Cuayamaca (surface survey here) is station nearest the SW California corner – let’s work up from here. Cuyamaca has the temperature history shown below. I’ve shown the 1900-2000 trend per Hansen et al 2001 together with a trend from 1925-2000 in green. There is actually a slight decline in the 75-year period, though there is a noticeable trend in the 100-year history. (Post-2000 values are not included in this analysis because they were not available for Hansen et al 2001.) The station history notes a move in 1926 – see dashed red line. The trend for this station thus depends on what we know about Cuyamaca station in the early part of the century – do the low early 20th century values represent well-measured temperature or is some portion an artifact of station moves?
San Nicolas Island
This is an offshore location, which is in GHCN but not in USHCN. Observations, as indicated in the history below, are spotty. This seems like a peculiar selection as one of NASA’s finest as the record is extremely discontinuous. Surely there are better alternatives.
This is another non-USHCN (ending in the 1980s) which is a peculiar inclusion as a Top Gun. It appears that there may have been a station discontinuity around 1940. As it is not a USHCN station, there is no station history so it’s hard to tell. The available information obviously is not strong support for Hansen’s hot spot, support for which must obviously be derived from other Top Guns.
Sandberg is another non-USHCN site included in the Top Guns. As shown below, it only has a short squib of a record and it’s hard to see why it should be accorded Top Gun status.
Fairmont and Tejon Rancho
These two sites are near each other and shown below. One has a slight increase since the 1920s, the other a slight decrease. One has no early 20th century values, the other has lowish values along the lines of Cuyamaca, and opposed to Beaumont.
Needles, another USHCN site, is the last Code 1 site in southern California. It has a long history and considerable visual similarity to Lemon Cove. In the late 19th century, it has quite high values, but it also looks like there is a discontinuity.
Childs is one of only two Code 1 sites in Arizona (the other is Lees Ferry in the very north of Arizona.) Hansen uses distances up to 1000 km (625 miles) for weighting though the weight declines with distance. Because it is relatively isolated from other Code 1 stations, it has a very strong influence in Arizona – an area that we’ve discussed at some length. It has rather low values in the 1930s, a pattern opposite to Needles and Lemon Cove. One wonders at how solid this information relative to the leverage of the site.
Lee’s Ferry has a still different appearance. Values before 1940 are similar to values after 1940. The post-2000 pattern is different than Childs and one wonders whether there is a discontinuity at one site or the other.
While I endorse the concept of trying to find quality within the USHCN network and I think that they can be identified with objective criteria, I’m not convinced that the GISS unlit methodology has developed a very good network of Top Guns in the southwest U.S. Obviously some of the Top Guns have very spotty histories and should be disqualified on those grounds alone. Once they are excluded, we are left with a rather small population and it’s relatively easy to do some QC on each site. On a quick pass, it seems unlikely that all of the sites could have been properly QCed and that the remaining differences are climatic. It seems far more likely that there are still material QC problems with the Top Guns in southwest U.S. and that 1900-1999 tend quantification (a la Hansen et al 2001) is to a considerable extent far too dependent on QC issues.
The 1900-1999 trend is extremely dependent on the pre-1920 portion of the Code 1 record and, in this case, the early part of the century is highly dependent on only a few stations, with one of the stations (Beaumont) having very inconsistent results to the others.