Since AR4, there have been a series of new multiproxy studies, several of which were cited in AR5 (Mann et al 2008; Ljungqvist et al 2010; Christiansen and Ljungqvist 2012; Shi et al 2013). A distinctive feature of these and other recent multiproxy studies is the incorporation of varve thickness and near-equivalent mass accumulation rate (MAR) series, in which varve thickness (positively oriented) is interpreted as a direct proxy for temperature. The following table shows the usage of varve thickness and near-equivalent mass accumulation rate (MAR) series in post-AR4 multiproxy studies (“long” series shown below). It is evident that the varve thickness data in multiproxy studies is anything but “independent”.
Table 1. Varve thickness and MAR (mass accumulation rate) series used in multiproxy studies which are both “long” (including the medieval period) and which have not been truncated in the modern period. Both logged and unlogged versions are used. In a couple of cases, the mass accumulation rate is limited to organics (“dark”). I’ve also included the Igaliku pollen accumulation rate series, because it appears to me to be closely related to MAR series. XRD (Xray density not included).
One of the most obvious features of the above table is the repeated use of a small number of varve thickness series used in Kaufman et al 2009: Big Round, Blue, C2, Donard, Iceberg and Lower Murray Lakes. Five of the six series were used in Shi et al 2013. In my recent discussion of Shi et al 2013, I observed that a composite of the five series (and the same is true for all six) had something of an HS-shape, though the series otherwise had negligible common “signal” (as demonstrated clearly by their eigenvalues). Further, several of the series (especially Iceberg which had been discussed in prior CA posts) had serious problems, compromising or potentially compromising any potential utility as a temperature proxy. This certainly suggested to me that the somewhat HS-ness of the varve thickness composite was more likely to be an artifact of selection from a noisy network rather than actual scientific knowledge. Skeptic blogs have long discussed this phenomenon, but it is one to which academic literature in the field has been wilfully obtuse.
Blog discussion has been mostly based on red noise examples. So I think that readers may be interested in seeing the phenomenon at work with actual data.
In the course of examining literature on varves, it quickly became evident that specialist literature prior to the relatively recent multiproxy articles had regarded thick varves as evidence of glacier advance (rather than “warmth”). Readers (and myself) wondered how the prior consensus (so to speak) that thick varves were related to glacier advance (and vice versa) had been replaced by a model in which thick varves were now interpreted as evidence of warmer temperatures. This proved to be an interesting backstory. I’ll also contrast the varve thickness series from Iceberg Lake, a canonical series in Kaufman et al 2009 and subsequent multiproxy studies, with “non-canonical” varve thickness series from Silvaplana, Switzerland and Hvitarvatn, Iceland, where thin varves are interpreted as evidence of warmth.
Overpeck, Bradley and Kaufman
The scientists most “responsible” for the introduction of varve thickness series as temperature proxies in multiproxy studies are arguably Jonathan Overpeck, Raymond Bradley and Darrell Kaufman, the latter a protégé and frequent co-author of Overpeck’s.
Bradley and Overpeck are both familiar to Climate Audit readers, but their connection to varve thickness data has not (to my recollection) been previously discussed here.
In the late 1980s and early 1990s, Overpeck and Bradley both received numerous NSF grants for the development of varve data in the eastern Canadian Arctic.
In various grant applications, Overpeck undertook to take sediment cores from 6-11 lakes (the precise number depends on overlap between proposals) and to create a large public database. The number of published Overpeck cores seems noticeably fewer than the number that were promised: I have thus far located only two varve thickness series published by Overpeck or his students: Donard and Upper Soper Lakes in Baffin Island (Hughen et al 2000; Moore et al 2001), both sites having been drilled in the early 1990s. In the early 2000s, Hughen was separately funded to extend the Donard record, but thus far no results appear to have been published. In 2009, Hughen made a partial archive of data from Ogac Lake, unpublished and unarchived since the early 1990s. Winton Bay Lake and other Overpeck sites (if actually drilled) remain unpublished and unarchived.
At the same time that Overpeck was receiving NSF funds for Baffin Island, NSF was funding Bradley for sediment cores in Ellesmere Island. This resulted in two short varve series (Tuborg – Smith et al 2004; C3 – unpublished) and one longer series (C2). Bradley and students (Hardy et al 1996; Lamoureux and Bradley 1996) reported a positive correlation between local temperatures and C2 varve thickness in the (short) instrumental period, a correlation that was widely cited in later varve literature purporting to justify temperature reconstructions from varve thickness.
In 1997, Overpeck was lead author of the first multiproxy study (Overpeck et al 1997) to incorporate varve thickness data into a multiproxy composite (four centuries from 1600AD on). Included in the 29 Arctic proxy series of Overpeck et al 1997 were two of his own series (the then unpublished Donard and Upper Soper series), three Bradley series (C2 and the unpublished Tuborg and C3) plus a very short Arctic island series from Gajewski (DV09).
In 2005, NSF again funded Overpeck (including Kaufman, who by then had joined Overpeck at the University of Arizona) to collate Arctic lake sediment data. This resulted in a special 2009 volume of the Journal of Paleolimnology, in which several now “canonical” series were published: Iceberg Lake, Alaska – Loso 2009; Big Round Lake, Baffin Island – Thomas and Briner 2009; Blue Lake, Alaska – Bird et al 2009; and a new series from Bradley and his students: Lower Murray Lake, Ellesmere Island – Cook et al 2009.
In the same volume, Kaufman et al 2009 (lead author Kaufman, senior author Overpeck, co-author Bradley) introduced a new multiproxy reconstruction, in which lake sediment series (and especially varve thickness series) played a much more prominent role. Kaufman et al 2009 had 23 series, of which the majority (12) were lake sediment series. It also included 4 tree ring series (including a Yamal superstick) and 7 ice core series. The ice core data is fairly nondescript in the period. Thus, any HS-ness in the data results from the tree ring data (especially Yamal) and the varve data.
In addition to the six “canonical” varve thickness series (Big Round, Blue, C2, Donard, Iceberg and Lower Murray), Kaufman et al 2009 also used three Finnish lake sediment series, which attracted much commentary, the usage of which I will briefly parse in an Appendix below. In both Kaufman et al 2009 and nearly all subsequent usage, these series were used in a truncated form, thereby avoiding the contamination that became controversial as a result of Mann et al 2008. The one exception was Tingley and Huybers 2013 (Nature), which also used the contaminated portion of the Korttajarvi data in a very superstick form. PAGES2K’s use of the Igaliku data was a similar error discussed at CA in the past here ^.
Combinations of the “canonical” varve thickness series were used in all subsequent multiproxy studies (Ljungqvist 2010; Christensen and Ljungqvist 2011; Christensen and Ljungqvist 2012; Kjungqvist et al 2012; Shi et al 2013; Tingley and Huybers 2013; PAGES2k Arctic 2013.) Sometimes, the series are used after taking logarithms; sometimes not. Some studies also include shorter varve series (e.g. DV09, Upper Soper).
The most recent studies (Tingley and Huybers 2013; PAGES2K) include “new” varve thickness series from Ogac Lake and Hvitarvatn respectively. However, Ogac Lake is not really “new” – while it was only archived in 2009, it was collected by Overpeck in 1992-93 at the same time as Upper Soper Lake and the still unarchived Winton Bay series.
Numerous varve thickness series, that have been reported in specialist literature, remain unarchived, including Hector Lake (Leonard); Cheakamus Lake (Menounus) and Silvaplana (various authors).
Comparing Iceberg to Silvaplana and Hvitarvatn
The most HS-shaped series of the six canonical varve thickness is the highly problematic Iceberg Lake AK series (Loso 2009), discussed on several occasions at CA (see tag) and used in ALL multiproxy studies subsequent to Kaufman et al 2009. I will re-visit it today, contrasting it with two “non-canonical” series – Hvitarvatn and Silvaplana. Despite its repeated use, the series is defective and should never have been incorporated into these studies. Warnings against its use were sounded at CA even before Kaufman et al 2009 – see many thoughtful comments in a 2007 CA thread here.
Iceberg Lake is dammed by an unstable moraine and, over the past 200 years or so, Iceberg Lake has experienced a series of dramatic changes in lake level as the channel changed path. The most recent change occurred in 1957, when the shoreline was lowered by 25.7 meters (from 959.0 to 933.3 meters). Varve thickness is dramatically affected by closeness to the inlet. The 1957 event at Iceberg Lake resulted in extremely thick layers at the time of the event as well as ongoing thicker varves – see top panel of Figure 1 below.
Loso was aware of the 1957 event (which is clearly disclosed in his article) and excluded 1957 values from his composite – a limited hangout that ignored the post-1957 inhomogeneity. Loso 2009 acknowledged the inhomogeneity resulting from change in inlet location, but his remedy completely missed the point: Loso merely used log values in his temperature reconstruction, a device that was irrelevant to the inhomogeneity (as pointed out at CA at the time). In Loso’s most recent publication (Dietrich and Loso 2012), Loso acknowledged inhomogeneity in the Iceberg Lake series, but did not withdraw or amend the 2009 version of the series, which remains in near universal use in multiproxy studies.
In order to highlight these issues, I’ve plotted Silvaplana (Switzerland) varve data in the middle panel, showing 1957 as a dotted red line. At Silvaplana, 20th century varve thicknesses decline dramatically from Little Ice Age values, the exact opposite of what one expects under the Overpeck-Bradley-Kaufman linear model. In related articles, Silvaplana authors report a close association on a decadal-centennial scale between varve thickness and glacier advance, an association matching the common sense understanding of the Little Ice Age.
Curiously, despite this dramatic decline, within a 20th century “calibration period”, Silvaplana varve thicknesses actually have a positive correlation to temperature (as do, at least over short calibration periods, varve thicknesses in other canonical locations. The authors of the Silvaplana data (Trachsel et al 2010) noted the inconsistency as follows:
We found a significant positive correlation between local summer air temperature (May-September) and the annual sediment mass accumulation rate (MAR) in Lake Silvaplana (46°N, 9°E, 1800 m a.s.l.) during the twentieth century (r = 0.69, p < 0.001 for decadal smoothed series). Sediment trap data (2001-2005) confirm this relation with exceptionally high particle yields during the hottest summer of the last 140 years in 2003. On this base we developed a decadal-scale summer temperature reconstruction back to AD 1580. Surprisingly, the comparison of our reconstruction with two other independent regional summer temperature reconstructions (based on tree-rings and documentary data) revealed a significant negative correlation for the pre-1900 data (ie, late ‘Little Ice Age’). This demonstrates that the correlation between MAR and summer temperature is not stable in time and the actualistic principle does not apply in this case. We suggest that different climatic regimes (modern/‘Little Ice Age’) lead to changing state conditions in the catchment and thus to considerably different sediment transport mechanisms. Therefore, we calibrated our MAR data with gridded early instrumental temperature series from AD 1760-1880 (r = -0.48, p < 0.01 for decadal smoothed series) to properly reconstruct the late LIA climatic conditions.
It is, of course, ludicrous for them to suggest that the “actualistic principle” fails to apply merely because correlations change sign under different circumstances. What’s happening here is that there are multiple factors affecting varve thickness, of which current temperature is only one. Glacier advance/retreat is another important factor (as was well known in the pre-Overpeck pre-Bradley literature.)
In the third panel of the above Figure, I’ve shown the varve thickness series from Hvitarvatn, Iceland that I’ve recently discussed, since its authors (Miller et al 2012) also associated varve thickness on a centennial scale with glacier advance rather than warmer temperatures. Although its interpretation matches that of the authors of Silvaplana, the timing of the series is definitely not synchronous with Silvaplana. Whereas varve thicknesses at Silvaplana have declined in the 20th century, 20th century varve thicknesses at Hvitarvatn are among the thickest in the entire record.
Despite the apparent inconsistency, I think that it may be possible to extract a consistent story from this data, incoherent as it seems. I’ve flipped the Silvaplana and Hvitarvatn series in the figure below so that “warm” is up. Thus oriented, the Silvaplana varves provide a convincing record of recovery from the Little Ice Age, with the modern thin varves standing as evidence of glacier retreat.
In contrast, the Hvitarvatn record suggests to me that the Little Ice Age in Iceland was (unsurprisingly) much more intense than in Switzerland and that recovery remains incomplete. It seems virtually certain to me that Hvitarvatn varves will become thinner over the 21st century as the glacier recedes. That present varves are thicker than medieval varves indicates to me that present glacier retreat, rapid as it may be, has still not reached medieval (and Holocene) retreat stages. In picturing glaciers this way, I am (in my own mind) picturing glaciers as a sort of Last In-First Out inventory – a form of non-linearity that is intuitive in business applications and which seems completely applicable here. Under this concept, Switzerland would have entered the LIA later than Iceland and left sooner.
I mention the above concept with the following caveat: defects in my interpretation of the varve thickness series are irrelevant to the garbage-ness of their use as linear proxies for temperature in multiproxy studies. I see no redeeming qualities in the Overpeck-Bradley-Kaufman interpretation of this data.
Even if modern temperatures are warmer than medieval temperatures (a distinct possibility), it is my view that any HS-ness in the composite of varve thickness series used in recent multiproxy studies is entirely artifact and does not represent an authentic HS signal.
APPENDIX: Note on Finnish Varve Series
In the preparation of this note, I closely examined the versions of the various Finnish series (including upside-down Tiljander) used in the multiproxy studies, which I’ll place here as a sort of memo to file.
Mann used four Korttajarvi series: three varve thicknesses series: light thickness, dark thickness and combined thickness; and X-ray density (XRD). The enormous supersticks were in the thickness series, rather than XRD, which had elevated 20th century levels, but not crazy levels (like the other series). Tiljander oriented light thickness, combined thickness and XRD negatively; thus, for these series, in addition to using contaminated data, Mann used them upside-down to the interpretation of the original authors (an interpretation which some Mann defenders contest). Dark laminae thickness is less common in this field, but occasionally used by others e.g. Overpeck at Upper Soper Lake, Baffin Island.
Kaufman used the Korttajarvi XRD series (not thickness), but cut off at AD1800 because of modern contamination. Kaufman originally used this series upside down to the orientation of the original authors, but reversed the orientation in their corrigendum. Since they had already truncated the series at 1800AD, the flip had little effect on their final result. It is useful to note that Kaufman had no difficulty in understanding what it meant to use this series “upside down” – though many commenters e.g. William Connolley have pretended that no climate scientist could reasonably be expected to understand this criticism without much greater elaboration.
The Korttajarvi series was avoided in Ljungqvist studies, other than Ljungqvist et al 2012, where the XRD series truncated to 1899 was used in negative orientation. However, the Korttajarvi superstick (bizarrely) re-emerged in Tingley and Huybers 2013, who used the supersticked varve thickness (dark laminae) without truncation or apology. PAGES2K used a truncated version of the XRD series in negative orientation.
Kaufman also introduced the Lehmilampi varve thickness series. Kaufman originally used it in positive orientation, slightly truncated back to 1945. It had elevated closing values, but not supersticked. In their corrigendum, Kaufman inverted the Lehmilampi varve thickness series to match the interpretation of the original authors – thin varves now denoting warmth – retaining the 1945 truncation. Lehmilampi was not used in the Ljungqvist studies except for Ljungqvist et al 2012, which used the XRD data in negative orientation (as with the other Finnish sites). PAGES2K returned to use of varve thickness, again inverted and truncated to 1800AD.
Nautajarvi, the third Finnish series used by Kaufman, was another varve thickness series (dark laminae thickness used in positive orientation) which was truncated in AD1800. It was left unchanged in the Kaufman corrigendum. As with the other Finnish series, in the Ljungqvist studies, the site was only used in Ljungqvist et al 2012, where the XRD data was used in negative orientation (consistent with Ljungqvist et al 2012 policy on the other two sites). PAGES2K reverted to the use of dark laminae thickness in positive orientation, truncated to 1800AD.