Yesterday, I posted up a first look at differences between station histories classified as CRN=1 (good) versus CRN=5 (bad) – a simple comparison of averages, noting that other factors may well enter into the comparison.
A couple of other points that I’ve made consistently as we look at these results which I’d like people to keep in mind:
(1) the elephant in the room in these station studies is the difference in trends between the US history with 1930s at levels more or less similar to the 2000s and the ROW with a pronounced trend (where’s Waldo?);
(2) the US network has a large representation of rural stations which have records stretching back to the 1930s, a different situation than for the ROW;
(3) NOAA and NASA have quite different procedures, with NOAA showing a much more pronounced trend than NASA in the US 48
(4) whatever the warts on the NASA methodology, they at least make a more concerted effort to adjust for urbanization in their U.S. network (relative to NOAA) and we need to keep both networks in mind. In particular, NASA already has a high 1934 relative to 1998, especially as compared to NOAA.
While NASA has been taking the brunt of recent criticism, it is actually NOAA rather than NASA that has made highly publicized announcements about 2006 being the “warmest year” and we need to keep this in mind as our understanding of these methods and data improves.
First, reviewing the bidding: here is a simple comparison of the averages of the CRN1(good) and CRN5(worst) USHCN stations – a first cut making no attempt to disaggregate regionally or to check on ASOS instrumentation or things like that. It shows a noticeable difference between CRN1 and CRN5 results.
John V has carried out some useful analyses of the data, noting that the regional distribution of CRN1 and CRN5 stations was not homogeneous: CRN1 stations turn out to be skewed to the east, especially the southeast, while CRN5 stations are skewed to the west. I haven’t verified this point, but it seems plausible.
The median longitude in the USHCN network is 95W. As a coarse cross-check, I split the stations into groups east and west of 95W and compared CRN1 to CRN5 stations and secondly CRN1,2 to CRN5 stations. Doubtless many other variations and crosscuts can and will be identified, but this seemed like a pretty simple first check on regional issues.
East of 95W
The two graphics below compare (first) CRN1 to CRN5 ; (second) CRN 1,2 to CRN 5 for stations E of 95W. As you see, there is a strong increase of CRN5 relative to either CRN1 or CRN1,2 stations (over 0.4 deg C). (There are a number of ASOS stations in the CRN1 network.) Another thing to keep in mind is that the surfacestations.org quality classification does not coincide with the GISS lit-unlit classification. Of the 27 CRN1,2 stations in this group, only two were lights=0 and only 9 were dim/dark; 18 were classified as bright.
Nonetheless, there appears to be a difference between CRN1,2 and CRN5 stations in this eastern group. In fact, as seen below (and somewhat surprisingly), the difference is greater in the eastern stations than the western stations, an issue that I’ll return to below.
West of 95W
Here’s a similar calculation for west of 95W. Here there is surprisingly relatively little difference between CRN1,2 sites and CRN5 sites. Again the QC standards somewhat crosscut the urbanization standards, with some urban sites in the CRN1,2 classification (San Antonio WSFO, Berkeley). There’s not much trend over the full record, but there is a pronounced difference in the CRN1 records between levels in the 1930s and 2000s. There are not very many CRN1 stations in this grouping, which may affect things – but we’re also told by Gaivn Schmidt and others that a relatively small network of good stations should suffice for a global network and the number of CRN1 stations in the west would be sufficient within these standards, without the CRN2 stations. (And one would need to ascertain whether the CRN1-CRN2 differential was regional climatic or quality as well.)
The results, at a first pass, are opposite to a number of expectations. Eli Rabett, in one of his many sniggers against the mere idea of checking station quality, hypothesized that for every station in the west failing QC due to warming asphalt, there was an offsetting station in the east failing QC due to cooling tree growth (the Halpern Hypothesis of Offsetting QC Failures). Yet here we seem to have a greater difference between CRN1 and CRN5 sites in the east, where vegetation growth is an issue, relative to the west, where asphalt is more of an issue.
Secondly, as noted above, the QC classification crosscuts the traditional UHI issue (the nocturnal inversion caused by an urban setting, distinct from microsite issues). In this first pass analysis, there has been no attempt to cross-stratify these issues and that’s definitely something that should be done.
NOAA and NASA
Regardless of the above, the differences between the NASA temperature history for the U.S. (with its relatively warm 1930s) and the CRN1,2 averages (here averages of east and west stations are done first and then averaged) does not show a marked trend, a point noted by John V. Actually, it’s somewhat downward. (Below: dotted – post Y2K NASA version;solid – pre Y2K):
However, the situation is quite different with NOAA, as shown in the next graphic, which shows the difference between NOAA and NASA temperature histories for the U.S. (NOAA taken from http://www1.ncdc.noaa.gov/pub/data/cirs/drd964x.tmpst.txt). As you see, the NOAA trend relative to NASA is about 0.33 deg C per century since 1941 [ Note: this is a revised version of earlier graphic.
NOAA minus NASA annual versions for US Lower 48. Calculation and plot script is here.
Obviously, the NOAA trend relative to CRN1,2 stations is going to be over 0.6 deg C per century. As you’ll recall, it’s actually been NOAA that’s made a point of issuing press releases about 2006 being the “warmest year”, rather than NASA, although NASA’s been taking the brunt of recent criticism.
The profound differences between NOAA and NASA results obviously point to substantial differences in their adjustment methods. While we’re gradually pinning down what NASA did, the process of disentangling NOAA results hasn’t really begun.
While I’ve been critical of NASA (and plan to make further criticisms of the procedures involved in their September adjustments), I’ve noted at all times that the U.S. is unique in having a large population of rural sites reaching back to the 1930s and that NASA has at least attempted to adjust for urbanization. Based on regional disaggregation – an approach that I endorse, John V suggested that the relatively similarity of NASA and CRN1,2 histories was a vindication of NASA methodology – a point that reader in a comment below asked me to note here. However, John V failed to observe that NASA used different methodologies outside the U.S. than in the U.S. and that the rural content of ROW networks was completely different than the U.S. and thus, using his approach, he could not argue that NASA methodology for the ROW was vindicated, as he suggested.
I’ve noted the worry that the QC ratings from the first cut of Anthony Watts ratings include a lot of stations classified by NASA as being in “unlit” areas. This strongly suggests the need to do a further cross-cut of the analysis, which will take a bit of time. The TOBS adjustment also needs to be looked at.
On the other hand, my guess is that such a cross-cut won’t change the similarity between CRN1,2 and NASA’s U.S. temperature history very much. I agree that this may well end up supporting (and perhaps even “vindicating”) the NASA analysis method for the U.S., which after all, resulted in the conclusion that 1934 was the warmest year. If it turns out that:
(1) ROW countries have a similar framework of rural “unlit” sites with records stretching back to the 1930s and continuing up to the present;
(2) NASA coerces the trends at urban stations in the ROW to these rural “unlit” stations
then one might also extrapolate that the methods applied to the U.S. might work reasonably on those countries. Turning John V’s point against him somewhat, I think that the analysis might even be held to demonstrate the necessity for an analysis of the type carried out in the U.S. by NASA.
The first casualty of such a process is obviously NOAA – whose results are inconsistent with NASA’s results. The greater similarity of the NASA temperature history with the CRN1,2 stations shows that the choice between NASA and NOAA histories is not completely arbitrary, but that, in this case, the NASA history for the U.S. looks more reasonable.
The evidence from our quick reconnaissance to date of the ROW suggests that NASA does not meet the above standards in how it handles the ROW on a number of counts. First, instead of using one integrated record at each station (as with the USHCN stations), NASA has a perverse splicing of station records, introducing a potential bias at every ROW station in which MCDW data is spliced to historical data, the effects of which have not been evaluated. Second, we’ve seen little evidence (where’s Waldo?) of a framework of long rural records: indeed, the evidence from Antarctica, South America, Africa and India is that there either are no such records or that they don’t show any material trend. The “Bias Method” used outside the U.S. has very different statistical properties. To the extent that the NASA approach in the U.S. has been vindicated – an approach that, once again, showed that 1934 was the warmest year on record, it merely highlights that this approach is not applied in the ROW so that one is left to speculate as to whether the difference in the ROW results from the failure to apply the “vindicated” method in the ROW – and perhaps it’s impossible to do so – or whether the U.S. has simply had a different climate history than the ROW, one in which, for some peculiar reason, present U.S. temperatures are not much different than the 1930s, while ROW temperatures have increased noticeably. (If U.S. temperatures diverge from world history during this observed period, one may then plausibly wonder as to why U.S. bristlecone growth should be held to have magic qualities for detecting world temperature.)