Kaufman and the PAGES2K Arctic2K group recently published a series of major corrections to their database, some of which directly respond to Climate Audit criticism. The resulting reconstruction has been substantially revised with substantially increased medieval warmth. His correction of the contaminated Igaliku series is unfortunately incomplete and other defects remain.
In several Climate Audit posts, I criticized PAGES2K for using Miller’s Hvitarvatn, Iceland varve data, upside down to the interpretation of the original author, an eminent paleo-scientist (see here; here). Kaufman stated that they “corrected temperature relation” of the Hvitarvatn series:
The interpreted temperature relation of the series from Hvítárvatn was corrected from positive to negative.
The series as now reoriented is shown at left.
This confirmation of my interpretation is of considerable interest to me, since there seems to be a close relationship between the Iceland varve series and Baffin Island varve series, a location of considerable interest from a multiproxy perspective, as discussed in a previous post here, where I showed the close relationship between the Hvitarvatn, Iceland varve thickness series and the Big Round Lake, Baffin Island varve thickness as below. PAGES2K is in the awkward position of orienting the Hvitarvatn in an opposite orientation to their orientation of Big Round Lake, though the two series are evidently responding similarly. It is surely unacceptable for assessment-quality studies to leave such inconsistencies unresolved.
Figure 1. Varve thickness (mm). Purple – Hvítárvatn from Miller et al 2012 Figure 2D; blue – Big Round Lake (NCDC sheet 7 column 4) 30-year running mean. Both shown on same mm scale. See here.
I also severely criticized PAGES2K’s use of a sediment series from Igaliku, Greenland that had been contaminated by modern agricultural runoff – an issue that one would have expected them to be attentive to, given the previous issues with Korttajarvi. For this record, they state:
removed two most recent values due to anthropogenic fertilizer influence
In my previous post on the series, I showed the following plot from original NOAA data, showing three contaminated points, showing in red a datapoint not included in the PAGES2K version. So when they say that they removed the “two most recent values due to anthropogenic fertilizer influence”, they did not remove the two “most recent” values from their previous dataset, but only one value from the PAGES2K, still leaving the series contaminated. This is the largest HS-series in the dataset and it is disquieting that it continues to be contaminated. (They also changed the dating of the series by up to 108 years in some places – not sure why.)
Kaufman’s footnotes to Table 1 state that they “restricted” four series “to temperature-sensitive section” – a post-hoc adjustment that warrants closer examination. The adjusted series are: Blue Lake, Alaska varve thickness (from 5 to 730); Forfjorddalen 2 tree rings (from 877 to 1254); Lower Lena, Russia tree rings (from 1408 to 1490); and Coppermine tree rings (from 1048 to 1288).
As an example for this sort of adjustment, they gave Korttajarvi, well known to Climate Audit readers, where agricultural contamination is readily understood. However, it seems odd that series can be determined, after the fact, to go in and out of being “temperature sensitive” in their early portions and these exclusions warrant some attention. If these factors impact the early portion, how do the specialists ensure that they do not impact the later portion.
The Blue Lake series has elevated values in the early portion, which are now removed as not “temperature sensitive”. I understand that these lake sediment series are inhomogeneous – that’s one of the large problems in using them as continuous temperature proxies – but similar or greater inhomogeneities have been observed in the Iceberg Lake series, which go the “right” way and which have been ignored (See here and prior discussion of Iceberg Lake here.)
Kaufman removed three lake sediment series as “not temperature sensitive”: East Lake, Melville Island particle size; Kepler, Alaska d18O; and Holsteinborg, Greenland Fragilariopsis cylindrus. I’ll look at these series some time.
These various lake sediment series have been the most prominent “new” thousand-year proxies in the past 7 or 8 years, but they are fraught with difficulties in terms of interpretation as temperature proxies, with post-hoc interpretation rife.
In a CA post almost immediately following PAGES2K, I observed that there was longer data available for the Lomonosovfonna, Svalbard ice core (Divine 2011); this has now been incorporated.
They incorporated a new versions for Gulf of Alaska tree rings (Wiles et al 2014) and changed their Tornetrask version from Grudd’s reconstruction with high MWP values, to CRU’s most recent rival version (Melvin) which has higher modern values. However, they continued to use the older supersticked version of Yamal, rather than the more recent much-attenuated version of Briffa et al 2013.
I haven’t been able to locate a digital version of their revised reconstruction, but the before-and-after is shown in their Figure 2 here, indicating a substantial upward revision in their medieval reconstruction and downward revision in their modern reconstruction. Jean S draws our attention to the fact that the scales of the two versions are inconsistent in the diagram: I don’t know why at present.
Original Caption: Effect of revising the PAGES Arctic 2k database on the Arctic annual temperature reconstruction published recently by the PAGES 2k Consortium … (d) Comparison between Kaufman et al.7 Arctic—wide temperature reconstruction and the revised PAGES 2k Arctic reconstruction (averaged to decadal values). Note the factor-of-two difference in the temperature scales.
They show the following diagram of changes – all in the direction of increasing MWP warmth relative to modern warmth in their reconstruction. These are large changes from seemingly simple changes in individual proxies – a longstanding CA theme.
There is much of interest both in the data and the changes. The dataset is considerably more ambiguous than PAGES2K proponents recognize, though the scope of Kaufman’s changes may change some of these perceptions. And, finally, kudos to Darrell Kaufman for chinning up to errors like upside-down Hvitarvatn and Igaliku, as he did with upside-down Tiljander. I can’t imagine that it was very easy for him to face up to these mistakes, but it’s a far better practice than trying to tough it out by sliming his critics, a la Mann.