A question for readers: which of the following proxies are used to reconstruct past Southern Hemisphere temperature in the IPCC’s graphic (Figure 5.7b) showing SH reconstructions:
1. Graybill’s California strip-bark bristlecone chronologies
2. upside down and contaminated Finnish lake sediments
3. European instrumental temperature data
4. Antarctic ice core d18O isotope data covering the medieval period
The answer, rather remarkably, is that IPCC’s SH temperature reconstructions in their signature spaghetti graph (Figure 5.7b) used California bristlecone chronologies, upside down Tiljander and European instrumental temperature data, but did not use Antarctic isotope data covering the medieval period.
This remarkable turn of events occurred because IPCC Figure 5.7 relied on SH reconstructions from Mann et al 2008, citing three SH reconstructions from Mann et al 2008: Ma08eivl, Ma08eivf and Ma08cpsl. While it may seem counter-intuitive that bristlecones and contaminated Finnish sediments would be used in a SH temperature reconstruction, anything is possible in Mannian EIV and, indeed, bristlecones and contaminated sediments turn out to contribute to the SH EIV reconstructions, as I will show below.
IPCC AR5 FIgure 5.7(b). Reconstructed … Southern Hemisphere… annual temperatures during the last 2000 years. Individual reconstructions (see Appendix 5. A. 1 for further information about each one) are shown as indicated in the legends, grouped by colour according to their spatial representation (red: land-only all latitudes; orange: land-only extra-tropical latitudes; light blue: land and sea extra-tropical latitudes; dark blue: land and sea all latitudes) and instrumental temperatures shown in black (HadCRUT4 land and sea, and CRUTEM4 land-only; Morice et al., 2012). All series represent anomalies (°C) from the 1881-1980 mean (horizontal dashed line) and have been smoothed with a filter that reduces variations on timescales less than ~50 years.
Two of the three Figure 5.7 reconstructions (Ma08eivl, Ma08eivf) are Mannian EIV reconstructions ostensibly for land (“l”) and land-and-ocean (“f”). The third reconstruction (Ma08cpsl) is a CPS reconstruction for “land”. I use “land” and “land-and-ocean” in quotation marks for reasons that will become clear later in the post.
Each EIV reconstruction is a “composite” of EIV versions calculated using different proxy networks, with each variation included only in the period in which it passes Mannian cross-validation – a murky statistical procedure not employed, to my knowledge, by any other known “scientist” (or scientist, for that matter).
Ma08eivl (the “land” reconstruction) is (as I understand it) an average of the following five slightly varied Mannian reconstructions:
1. sh_glful_cru. This variation uses the “glful” network i.e. it includes proxies from both the Northern and Southern hemisphere (GL-global) without screening (“ful” as opposed to “scr” for screened). In the AD700 step (which I’ve selected as an example), the glful network has 43 proxies, of which only 9 are in the SH.
2. sh_glful_nolut_cru – The glful_nolut network is formed by excluding the Luterbacher network from the glful network. The Luterbacher network are a gridded European reconstruction from AD1500 on, in which instrumental data is used in the recent portion. Its modern portion is not “proxy” data and inclusion of this data as supposed “proxy” data overstates statistics on proxy veracity – a point made at CA previously. The Luterbacher data does not go back to the medieval period and thus this variation is more or less identical to glful in the medieval period. (I say “more or less” because nothing can be assumed in Mannian data analysis.)
3. sh_glful_nodendr_cru. In the glful_nodendr proxy network, tree ring data is excluded. However, it still includes both upside-down Tiljander and the Luterbacher network. Without the contaminated Tiljander data, this reconstruction apparently doesn’t “validate” prior to AD1500 (according to the SI to Mann et al 2009). However, Mann did not issue a corrigendum at PNAS and the portion of this reconstruction reliant on contaminated Finnish sediments continues to contribute to the SH reconstruction.
4. sh_shful_cru. This network uses the shful network: SH proxies without Mannian screening. Nine proxies are included: Quelccaya O18 and accumulation, both in older versions (also used in MBH98); Cook’s Tasmania tree ring chronology (also used in MBH98), two spelothem series (C13, O18) from Cold Air Cave, South Africa, Thompson’s Kilimanjaro O18; Pallcacocha greyscale sediments; and a couple of Argentine tree ring series. This network should correspond to the CPS network in the same step.
5. sh_shscr_cru – This network applies Mannian screening (another murky operation) to the shful network. Only 4 proxies are eligible for the AD700 step, one of which is Quelccaya d18O, a familiar series. Mannian “validation” eliminates this variation prior to AD1300, so it doesn’t “matter” for medieval comparisons.
Thus, three of the four variations contributing to the composite in the medieval period are dominated by NH proxies. Bristlecones, contaminated sediments and Luterbacher’s European instrumental-using network all contribute in varying degrees to three of the four variations, proving my point.
SH “Land” and “Land-and-Ocean” Reconstructions
Mann’s SH land-and-ocean reconstruction uses exactly the same network as his SH land reconstruction: glful; glful_nodendr;glful_nolut; shful and shscr. The only difference arises from the target series (SH HadCRU rather than SH CRUTEM). For each proxy network used in the composite, the SH “land” and SH “land-and-ocean” reconstructions are remarkably similar. For example, the next figure compares “land” and “land-and-ocean” reconstructions from the glful network (shglfulcru and shglfulhad respectively. These reconstructions are remarkably similar not because of skill in reconstructing land versus land-and-ocean, but simply because Mann has regressed slightly different instrumental series against the same proxy network. The reconstructions have only slight differences in weights and scaling.
The look of these reconstructions is noticeably different than the original Hockey Stick. This is because this variation assigns considerable weight to Curtis’ Central American lake sediment O18 series: these have a fairly marked medieval warm period; other variations more heavily weight bristlecones and/or Korttajarvi sediments and are more Hockey Stick in appearance.
NH and SH “Land” Reconstructions
Similarly, NH and SH “land” reconstructions using the glful network are remarkably similar to one another, again not because of “skill” in differentiating one hemisphere from another but merely because the SAME network of proxies is used in both cases.
shful vs glful
In the next graphic, I’ve compared the SH “land” reconstruction from the glful network to the shful network (shglfulcru vs shshfulcru). These differ in the early portion but are more than remarkably similar since 1860: they are identical. See the right panel below where glful has been shown as a line and shful as “+”.
This phenomenon was noted in contemporary CA discussion of Mann et al 2008, but not discussed recently. The reason why the later periods are identical is that Mann’s EIV reconstructions splice instrumental temperature to proxy reconstructions. (Longtime readers will recall Mann’s notorious 2004 statement that allegations that climate scientists spliced instrumental data with proxy reconstructions was a fabrication of fossil fuel disinformation.)
EIV vs CPS
In the next graphic, I’ve compared the Mann 2008 SH CPS reconstruction to the SH EIV reconstruction using screened SH proxies: Mannian CPS methodology includes a screening step. I haven’t parsed Mannian EIV screening to verify that it is the same as Mannian CPS screening. There is considerable similarity between the variations, but also differences. These presumably arise from differences in weights, with some series being more heavily weighted in EIV than CPS. (EIV can also assign negative weights to some series i.e. flip them over.)
Also note the difference in the modern period: Mannian CPS does not splice instrumental data into the reconstruction.
Probably the most important development in Southern Hemisphere proxies since AR4 has been the publication and/or archiving of high-resolution Antarctic isotope data covering the last two millennia. Some data had been measured prior to AR4 (e.g. Tas van Ommen’s Law Dome and Ellen Mosley Thompson’s), but it had either not been published (van Ommen) or archived (Mosley-Thompson). In addition, there are important new results e.g. Steig et al 2013, commendably archived in a timely manner.
Antarctic isotope data (PAGES2K and Steig et al 2013) shows a long-term decline in isotope data. IPCC AR5 commented on this as follows:
Antarctica was likely warmer than 1971–2000 during the late 17th century, and during the period from approximately the mid-2nd century to 1250 (PAGES 2k Consortium, 2013).
Despite the importance of Antarctic isotope data, no Antarctic isotope data covering the medieval period was incorporated in the IPCC SH spaghetti graph (Figure 5.7b).
This is because Mann et al 2008 did not use any Antarctic isotope data covering the medieval period. Nearly all of the Antarctic data used in Mann et al 2008 is only from ~1760 on: it appears to come from a rather dated collation of ice core information by David Fisher in the 1990s. The earliest Antarctic ice core in Mann et al 2008 is Talos Dome, in a version which goes only to the 13th century.
Curiously, in an earlier work (Mann and Jones 2003), Mann had used an Antarctic isotope series covering the past two millennia (Law Dome), but, despite the shortage of long SH proxies, this series was not used in Mann et al 2008. The reasons for its omission were not stated. As CA readers realize, the Law Dome isotope series has a rather elevated medieval period, a feature that led to an AR4 controversy seen in Climategate emails. AR4 wanted to illustrate long SH proxies, but refused to show the Law Dome series (with its warm medieval period).
In other words, as stated in my introduction, the IPCC reconstructions in Figure 5.7 did not any Antarctic isotope data covering the medieval period.
In its running text, the IPCC stated of Southern Hemisphere reconctructions:
An increasing number of proxy records and regional reconstructions are being developed for the SH (Section 5.5), but relatively few reconstructions of SH or global mean temperatures have been published (Figure 5.7b and c).
Both statements are true, but a complete assessment would have reported that the reconstructions shown in their Figure 5.7b did not rely on these new proxy records, but, for the most part, on Northern Hemisphere proxies, including bristlecones and contaminated sediments.