The replacement stations are not necessarily any better than the originals. The replacement for LOA, for example is Salina. Take a look at the picture at
http://members.dslextreme.com/users/juanslayton/salina.JPG
and tell me a sensor on top of a multi-story building almost on top of what appear to be large air conditioning units is better than the LOA station. (You can see my pictures for Loa on the surfacestations gallery.)

For the first part of the analysis I limited my trend measurements to those stations with complete data for the 1920-2006 period of interest. I then looked at those stations with trend differences between V1 and V2 of greater than 0.1 and less than -0.1 degrees C per decade. I compared those stations by CRN rating as determined from the Watts team’s latest reported findings and by the USHCN population designation (urban, suburban and rural) as reported in the Watts team spreadsheets. The table below lists the results of this analysis.

The numbers are small for concluding much from the CRN ratings comparisons, but I can conjecture that the CRN 1 and CRN2 category portions show greater positive trends than does their respective population portions and that CRN5 portion shows a greater negative trend than does its population portion. The three population designation comparisons shows a very apparent bias for the largest V1 and V2 trend differences, both positive and negative, to occur with the rural stations. A somewhat surprising finding, since the rural stations are sometimes and in some ways considered a standard for adjustments to other stations.

The results from the second part of the analysis is shown below where I have presented time series for V1 and V2 at a given USHCN station for two general cases where (1) the trend differences are large and denoted by the trend lines in the graphs and (2) the trend differences between V1 and V2 are very small but the trend for both versions are relatively large. I have dispersed the graphs in such a way that the examples of the two cases can be seen on top of one and another for ease of comparison.

Nothing jumps out for me in the way of differentiating why the stations with large trend differences between V1 and V2 have a V1 appearance different than that of the V1 time series that have small differences between V1 and V2. Perhaps the readers of this post can offer some differentiating features.

My second point of confusion on this matter involves visually understanding from these examples in the graphs how those corrections might have been applied. I believe that the algorithm whereby the breakpoints are determined are available but as I recall not in a packaged form for direct application to a time series. I do not recall whether sufficient information is available to apply an algorithm for adjusting the time series given the breakpoints.

It is important to keep in mind that the breakpoints are not determined based on the stations time series but rather based on the difference time series of the station in question compared two stations at a time with correlated and nearby stations. The explanation is listed below:

First, a series of monthly temperature differences is formed between numerous pairs of station series in a region. Specifically, difference series are calculated between each target station series and a number (up to 40) of highly correlated series from nearby stations. In effect, a matrix of difference series is formed for a large fraction of all possible combinations of station series pairs in each localized region. The station pool for this pairwise comparison of series includes U.S. HCN stations as well as other U.S. Cooperative Observer Network stations.

Tests for undocumented changepoints are then applied to each paired difference series. A hierarchy of changepoint models is used to distinguish whether the changepoint appears to be a change in mean with no trend (Alexandersson and Moberg, 1997), a change in mean within a general trend (Wang, 2003), or a change in mean coincident with a change in trend (Lund and Reeves, 2002)

Presented below are links to relevant sources of data used in this analysis.

Link to the Watts team data on CRN ratings and population designation for the USHCN stations:
http://www.surfacestations.org/USHCN_stationlist.htm

V1 (Urban USHCN) data from this link:
http://cdiac.ornl.gov/epubs/ndp/ushcn/ndp019.html#tempdata

V2 data from these links:
ftp://ftp.ncdc.noaa.gov/pub/data/ghcn/v2
ftp://ftp.ncdc.noaa.gov/pub/data/ushcn/v2/monthly

V2 adjustments explained from this link:
http://www.ncdc.noaa.gov/oa/climate/research/ushcn

In its place, USHCNv2 “adds” Defiance OH. However, the USHCNv2 station list shows that its “Defiance” is in fact a “composite” of Defiance (Coop 332098) and two other stations, Coop 335664 and 335669. However, these are simply the codes for Napoleon, before and after a small move on 5/1/62.

Since the Defiance record runs from 3/1/1893 to the present, there was no need to splice its early or later period onto another station, so it is not clear why or how Napoleon was merged with the Defiance record. The USHCNv2 station list merely names the additional stations that are part of “composite” records, but does not indicate when which station is used.

So to the extent USHCNv2 has “dropped” bad stations, it may in fact have retained them, simply by averaging them into “new” stations. This should be studied in greater detail.

Steve, but it’s Science. And we’ve finally joined the age of Science-Based Policy.

Do you know if the methods are known inside climate science? Not a sarcastic question.

I agree 1000% with the need to work from the best quality data – “best” being defined in some objective way. Anthony’s CRN1-2 stations very much represent that.

The problem is the lack of equivalent information elsewhere – particularly when the US history is rather discordant from ROW temperature history.

The problem with all these data sets seems to be that they are a rancid stew of good and bad data, seemingly with negligible effort at quality control, instead using weird multivariate methods to supposedly adjust the bad data – methods that are invariably poorly explained and documented and unknown outside climate science.

I think there’s potential that the majority of the warming is due to bad sensor networks. I read a post here from some time ago which showed the sparseness of the data back in the 1900′s – an eye opener to say the least. Combine that with the terrible siting issues, inconsistencies of measurement times and UHI, I’m not sure we can conclude the amount warming to 1C accuracy from this information. If we take the best sites in the world from the last 50 years uncorrected, that might be our best bet to know what happened.