OK, I’m starting to get the feel of the new proxy network and have some ideas of what the new Mannomatic is doing. The manouevres have a lot of similarities to the moves in MBH98 and I think that a few main components can be singled out:
1. Once again, as in MBH98-99, the “proxies”, as a network, do not have a strong common “signal” and the HS-ness is contributed by a very small minority of the proxies.
2. In proper data analysis, if one knows that the load is actually carried by a small number of proxies, then the analyst has to put these under the miscroscope to ensure that these are really good proxies – well-studied and endorsed by their original authors as temperature proxies. Once again, we get the opposite. As in MBH98-99, the load-bearing proxies are very problematic and, as previously with the Graybill strip bark chronologies (which remain in use), known defects are ignored using contrived excuses.
3. In this case, Mann notes caveats by the original authors (e.g. Lake Kortajarvi), but, once again, does a typical Mannian calculation, purporting to show that he can get a similar answer without the flawed data, begging the question of why the flawed data was used at all, if it doesn’t “matter”. It’s not as though the selection criteria for this grabbag network are clearly expressed. In this case, pardon me if I assume for now that the flawed data does “matter” as we found out with MBH98, where any attempt to do a sensitivity analysis without the
4. Remember that even Mannian RegEM still results in linear weights for the underlying proxies. So if Mannian EIV, whatever it proves to be, ends up being more-HS like than Mannian CPS in the non-dendro network, as seems to be the case, this merely means that Mannian EIV concentrates the load on the HS-shaped proxies and thus “gets” a more HS-shaped result. Exactly how this is done within the algorithm is unclear right now, but, based on present information, we already know that this happens.
This is precisely the sort of issue that was encountered with Mannian PCA/regression. If more load was placed on the bristlecones, he could “get” a HS. Given the heavy load borne by the bristlecones, any sensible statistical program would then assess whether they were a good proxy. Instead the Team used contrived arguments that given heavy weights to bristlecones was statistically “right” and “proper”.
This is going to go down the same road. The flash figure below (presented in an earlier post) shows standardized versions of the 41 non-dendro series starting on or before 1010 that contribute to the early reconstruction.
You can see that the following series contribute most of the HS-ness: the 4 Kortajarrvi lake sediment series (all from the same study); 3 Thompson ice core series – Dunde, Huascaran and especially Dasuopu; the Agassix melt series (also an important contributor to Moberg); an unpublished Socotra Island, Yemen speleothen dC13 series and a Scottish speleothem.
If these 11 series receive more weight, you get more of an HS and conversely.
We already have a thread on the Kortajarvi sediments containing observations by two Finns. As noted above, the original paper noted anthropogenic disturbance to the lake sediments – something that is quite common.
we also examined whether or not potential problems noted for several records (see Dataset S1 for details) might compromise the reconstructions. These records include the four Tijander et al. (12) series used (see Fig. S9) for which the original authors note that human effects over the past few centuries unrelated to climate might impact records (the original paper states ‘‘Natural variability in the sediment record was disrupted by increased human impact in the catchment area at A.D. 1720.’ and later, ‘‘In the case of Lake Korttajarvi it is a demanding task to calibrate the physical varve data we have collected against meteorological data, because human impacts have distorted the natural signal to varying extents’). These issues are particularly significant because there are few proxy records, particularly in the temperature-screened dataset (see Fig. S9), available back through the 9th century. The Tijander et al. series constitute 4 of the 15 available Northern Hemisphere records before that point.
In addition there are three other records in our database with potential data quality problems, as noted in the database notes: Benson et al. (13) (Mono Lake): ‘‘Data after 1940 no good— water exported to CA;’ Isdale (14) (fluorescence): ‘‘anthropogenic influence after 1870;’ and McCulloch (15) (Ba/Ca): ‘‘anthropogenic influence after 1870′. We therefore performed additional analyses as in Fig. S7, but instead compaired the reconstructions both with and without the above seven potentially problematic series, as shown in Fig. S8.
Mann’s SI Figure 8 shows no difference with/without these 7 series in the CPS version, but the EIV version is affected. This figure appears to be a splice which hides the impact on the reconstruction using the MWP proxies, which, after all, is the case of interest to readers. SI Figure 8 should be shown using only the MWP proxies, as this is all that are relevant for the MWP-modern estimates.
Secondly, this figure appears to include both dendro and non-dendro proxies. Now much attention has been paid to problems with tree ring proxies, but there is a special issue with the validity of the Graybill bristlecone chronologies, which continue large as life in this study and carry lots of weight in the SI Figure 8 comparison. We know that Ababneh didn’t replicate Graybill’s results at Sheep Mountain, but, instead of using this up-to-date, Mann and Hughes have continued to use the problematic Graybill strip bark series. Even though this issue was raised by the NAS panel, flaccid reviewing failed to pick up even such an obvious point.
If there is a little diversification among the flawed “proxies”, then a few of them can be removed, with others picking up the slack.
In being critical of these particular proxies, I’m not objecting to them merely because they are HS-shaped. What I’m doing is simply isolating the HS proxies for closer scrutiny – a proper data analysis procedure – and checking the underlying literature and rationales. Clearly Kortajarvi and Graybill strip bark are not great great foundations.
I’ve observed that Mt Logan in the Yukon and Law Dome in Antarctica do not have a HS pattern. Though long versions of both these series are available, Mann doesn’t use them: why? Or for that matter, why aren’t Puruogangri and Bona-Churchill used? Oh that’s right, Thompson hasn’t archived anything for Bona-Churchill yet. I’ve commented on these series in other threads – see the Thompson category and will not do further right now, other than to note that these series are very important in a non-dendro HS through the MWP, especially when the ridiculous Kortajarvi data is removed.
The Socotra Island speleothem results are unpublished for this period. Mann’s reference is to a study that does not discuss or present this information. The data is from Bradley’s group and is not archived. I have seen no evidence that dC13 speleothem values can be relied on as a climatic proxy; there are some caveats in specialist literature against trying to use dC13 for climate. Fleitmann, one of the Socotra Island authors (within Bradley’s group) stated:
d13C values of speleothems are difficult to interpret in terms of climate variability and only a small number of samples show AGLs. Consequently, most of our paleoclimate records in Oman and Yemen are based on d18O alone.
and goes on to say:
Calibration against meteorological observations: A handful of meteorological record is available for Oman and Yemen. Unfortunately, all of these records are geographically widely distributed, generally shorter than 40 years and fragmentary. Due to the lack of meteorological observations, a calibration of our speleothem records is not possible.
This caveat wasn’t a problem for Mann who reported correlations of 0.51 (1850-1996) and 0.71 (1850-1949) for the Socotra Island dC13 record. Mann went on to say:
One advantage of the non-tree-ring proxy series used is that, in most cases, there is little reason to believe a priori that there are any problems with the series that are likely to eliminate the reliability of multicentury to millennial time scale information.
Once again, an assertions made out of thin air. You can’t just say things – they have to be shown with technical reports.
The last one that I’ll comment on briefly today is a Scottish speleothem. Here there seems to be a long-term within the record. I’ve noticed that speleothem data sets sometimes remove such trends as sometimes being due to physical causes – is something like that going on here? This is the sort of thing that Mann et al need to report on.
This is a pretty ragtag collection of proxies upon which to found claims that they can “get” “skilful” reconstructions without Graybill bristlecone chronologies. And oh yes, wouldn’t it be nice if Thompson archived a complete data set – all samples and all measurements – so that a proper analysis of the Thompson Three could be carried out.