The New Zealand Climate Conversation Group have released their report and reanalysis of the NIWA 7-Station Review. CCG claim NIWA misrepresented the statistical techniques it used, and exaggerated warming over the last hundred years.
The CCG results (Figure 20 above) prove there are real problems in the adjustments to temperature measurements for moves and equipment changes in NZ (also seen in Australia).
As any trained scientist or engineer knows, failure to follow a well-documented and justified method is a sign of pseudoscience. The New Zealand Climate Conversation Group is correct in examining if Rhoades & Salinger (1993) has been followed, as advertised.
In 2010, NIWA published their review of their 7-station temperature series for New Zealand. The review was based upon the statistically-based adjustment method of Rhoades & Salinger (1993) for neighbouring stations. In this report, we examine the adjustments in detail, and show that NIWA did not follow the Rhoades & Salinger method correctly. We also show that had NIWA followed Rhoades & Salinger correctly, the resultant trend for the 7-station temperature series for New Zealand would have been significantly lower than the trend they obtained.
Despite searching, I cannot see a methodology section in NIWA’s report, which is a disjoint analysis of each of the seven sites, although it is clear in a number of places that they infer that Rhoades and Salinger (1993) forms the basis. For example, page 145 on Dunedin.
In February 2010, NIWA documented the adjustments in use at that time (see web link above). These adjustments to the multiple sites comprising the ‘seven-station’ series were calculated by Salinger et al. (1992), using the methodology of Rhoades and Salinger (1993), which extended the early work on New Zealand temperatures by Salinger (1981). Subsequent to 1992, the time series have been updated regularly, taking account of further site changes as circumstances required.
The Climate Conversation Group summarize the differences between Rhoades and Salinger (1993) and the method actually used by NIWA. The R&S method for comparing a station with neighbouring stations involves the use of:
– Monthly data
– Symmetric interval centred on the shift
– A 1-2 year period before and after the shift
– Weighted averages based on correlations with neighbouring stations
– Adjustments only performed if results are significant at the 95% confidence level
In contrast, the NIWA method uses:
– Annual data
– Asymmetric intervals
– Varying periods of up to 11 years before and after the shift
– No weighted averages
– No evidence of significance tests – adjustments are always applied.
Any of these methodological deviation could create substantial differences between the results, but the Climate Conversation Group (nor I) could not find a rationale or discussion in the NIWA review reports for not implementing the R&S method as stated.
What are the details of the methods? The CCG report compares a single station at Dunedin, using NIWA and R&S methods in their Table 1. There were five site moves — 1913, 1942, 1947, 1960, and 1997 — with five potential adjustments. The NIWA method adjusts at each of the moves, resulting in an increasing trend of 0.62C/century for Dunedin. The R&S method only implements two adjustments resulting in a 0.24C/century increasing trend.
The other six stations are similar (Masterton, Wellington, Nelson, Hokitika, and Lincoln), with the NIWA method doing generally more frequent, and more negative adjustments, and resulting in exaggerated trends, as shown in Figure 20 at the top of this post.
It would seem that significance tests and weighting of neighboring sites is very important. It ensures the nearby sites used to calibrate the site moves actually provide information on the site in question. A larger neighborhood of 11 years would probably confound short-term changes with the long-term warming trend, and may bias the adjustments to exaggerate the trend.
To ignore significance tests, weightings, and modify the method arbitrarily, whether sloppy or intentional, is bad practice, and would not be favorable to NIWA in their upcoming court case, brought by CCG.