One of the reasons why scientists have been so quick to use tree ring information despite all the problems is that, for the most part, there is excellent dating control on tree ring chronologies, something which can be problematic in other proxies.
Today I want to document some notes on dating the Arabian Sea G Bulloides. In this case, although Moberg (and Juckes) present one series in their data sets, this series actually results from a splice of information from two different cores – a splice not actually made by the original authors (although one of their figures is suggestive), but by Moberg. But how legitimate is it to splice the two cores?
Today I’ll look at some potential problems with homogeneity of the splice and also with even dating the cores.
Here once again is the figure from Anderson et al 2002 comparing the G Bulloides series to the Hockey Stick – thereby placing this data squarely on Team radar (and thus Moberg used it in his influential reconstruction even though there was no direct relationship of the proxy to local temperature – actually the opposite: the proxy was inversely related to local SST.
Another thing to observe here: note that the original authors do not use the percentage G Bulloides directly, but instead use the square root of the difference in composite G. bulloides abundance with respect to the 1975 average (thick line). This reduces the very great non-normality of the G Bulloides percentage data – a precaution abandoned by the Team.
Anderson, Overpeck et al 2002. Fig. 2. (B) Time series of Northern Hemisphere temperature variations from (28 – MBH) (thin line) superimposed on the index linearly related to monsoon wind speed, the square root of the difference in composite G. bulloides abundance with respect to the 1975 average (thick line).
Now here’s a figure of Moberg’s version of this proxy. A couple of important differences: the proxy goes longer than in the Anderson et al 2002 article (this is not a problem in itself as Moberg refers to a later version in Gupta et al 2003); perhaps more important is the abandonment of even a token attempt to normalize the data. Moberg’s SI described the splicing as follows:
A combination of two marine sediment records from the Arabian Sea14-15 in which the percentage of the foraminifera Globigerina bulloides reflects the extent of ocean up-welling, which is determined by the strength of monsoons, which in turn indirectly reflect both summer and winter large-scale temperature changes through the differential seasonal heating and cooling of the Asian continent and surrounding oceans16. We used data from Core 723A15 for the early years up to 1390 A.D. and data from Core RC273014 from 1391 to 1986 A.D. (c.f. Fig. 3c in ref. 15). Although this record reflects temperatures only indirectly, it was included to improve the balance in the geographical distribution of proxy sites.
This latter point seems somewhat of a bait-and-switch to me and rather misleading: actually it’s untrue that the “record reflects temperatures only indirectly”. The proxy actually records temperature rather directly – the negative correlation between percentage G Bulloides and water temperature is actually rather strong relative to other proxies in the network.
The only thing that’s “indirect” is that colder water offshore Oman is held, I guess, to be indirect evidence of global warming, while of course warmer water offshore Oman in the CRU data is held to be direct evidence of global warming. I guess this makes sense to real climate scientists.
Moberg et al 2005 Version of the Arabian Sea G Bulloides series
Next here is Gupta et al 2003 Figure 3 which illustrates G Bulloides values from two different cores: RC2730 and 723A. An overlap between the two cores is shown from about 900-500BP. A horizontal flip of the Moberg version is shown to evidence that these are the same plots.
Gupta et al 2003. Figure 3. G. bulloides percentage in Hole 723A (filled circles) and box core RC2730 (open circles), showing also dated depths in 723A (plus signs) and RC2730 (crosses). The North Atlantic Medieval Warm Period (MWP) and Little Ice Age (LIA) extents are based on the records shown here.
There are a few interesting points to this graphic:
1) the most modern values shown here are under 30% G Bulloides, while the most recent values in the Moberg version are over 30%. The Moberg version is consistent with RC2730 data archived at WDCP, – my graph of WDCP data here – but the published illustration omits the two latest values in the archive – both of which are high. Why is that?
2) the general tenor of G Bulloides percentages in this Figure prior to the most recent values are under 15% and most are under 10%. These low values are associated with water of at least 26 deg C i.e. no upwelling. As noted in my previous post, high G Bulloides percentages are associated with water of about 22 deg C. Taken at face value, these G Bulloides values indicate very low monsoon strength prior to the modern period.
These core values seem to prove too much – the monsoon strengths indicated here seem too weak and I thought it would be a good idea to take a look at the splicing and the dating. Here are age-depth plots for RC2730 and 723A on the same scale with radiocarbon points marked with solid circles. Points at which there are G Bulloides measurements are marked with red + (RC2730) and o (723A).
Again a couple of quick observations:
1) the implied deposition rate of 723A (31 cm/ kyr) is over three times more rapid than the implied deposition rate of RC2730 (10 cm/kyr). This seems to me to be a big inhomogeneity: could this be related to G Bulloides levels? I don’t know, but it seems odd to splice two series of percentage G Bulloides from cores with such different deposition rates. (Update: Richard T observes below that the different deposition rates could result from different coring methods, with one core mixing more mud with the core. I don’t know how likely this explanation is, or whether it’s merely an idea.)
2) Only part of RC2730 has been reported. Anderson et al 2002 report that the core was 51 cm long, while only the top 10 cm has been reported. It would be interested to see whether the proposed correlation holds up over the rest of the core. I’m not suggesting that it won’t – it just seems odd that the job would have been left half-done.
3) Notice that the radiocarbon points for the two cores do not overlap. The oldest radiocarbon measurement for RC2730 is younger than 723A. (BTW there appears to be an older portion of RC2730, but no analysis has ever been reported.) The top part of 723A is said to be about 500BP but this is not based on any measurement, but, as far as I can tell, on nothing more than extrapolation and/or wiggle matching. I wonder why they wouldn’t have tried to date the core top. If the failure to report a top date for 723A seems surprising, here’s what they said in the Nature SI:
To determine the age of the uppermost sample, we correlated the top of our record with a well-dated Oman margin box core (RC2730), finding an age of ~560 yr, suggesting that the top few centimetres were lost at the time of drilling.
Fig. 1s. Hole 723A calendar age versus depth. Ages were corrected using the Calib program HTML version 4.2 (http://depts.washington.edu/qil/calib/) and a reservoir correction (delta R) of 207 years. The one-sigma uncertainty is shown by error bars. The core top age (circle) was estimated by stratigraphic correlation with well-dated box core RC2730, and one depth (2.18m) was not used in the age model.
I’m not sure what “stratigraphic correlation” means in operational terms in this case. I don’t think that it means “stratigraphic correlation” as practised by geologists, but merely the wiggle matching of Gupta et al 2003 whon above.
Radiocarbon procedures for core RC2730 were explained in Anderson et al 2002 SI as follows:
We sampled two Soutar box subcores at continuous 2-mm spacing using the method developed for Cariaco Basin sediments (22) (supporting online text). Nine AMS (accelerator mass spectrometry) 14C dates on planktonic foraminifers from core RC2730 provided the age control (fig. S1). When corrected for the average reservoir age of the Arabian Sea (23, 24), the uppermost four samples, including the replicated 0- to 2-mm sample, were too young (less than zero age), which can be explained by the presence of bomb radiocarbon that has mixed downward over the upper 10 mm (table S1). Each of the box core sediment surfaces was observed to be pristine during core recovery. This observation, combined with the presence of bomb radiocarbon in the upper samples, supports our assumption that the top of the box core corresponds to the date of collection (1986). We constructed an age model by fitting a line through 1986 at the core top and through the remaining reservoir-corrected, nonzero 14C dates [the corrected age at zero depth (black symbol) in fig. S1 is the inferred age of “36 years (with respect to 1950)]. The date for 40 mm is too old to fit the linear sedimentation rate model, and we discounted this date because it comes from the minimum in G. bulloides abundance, where mixing could most easily bias the age. All the 14C dates were determined on the core with the highest sedimentation rate (RC2730),
The reservoir age was not reported here, but I’ve seen a figure of 604 years elsewhere:
An average reservoir age of 604 years was calculated for the western Arabian Sea, based on 14C measurements of known age mollusc, gastropod and coral samples from 7 sites in the region (Southon et al., 2002). However, a detailed study determined that the reservoir age of the Northern Arabian Sea has varied from 600-1200 years during the Holocene (Staubwasser et al., 2002). This large range in “possible” reservoir results in adding significant uncertainty to any Arabian Sea chronology.
I’ve also been keeping my eye out for any other information on G Bulloides percentages in the Arabian Sea (as I’ve reported eslewhere, percentage G Bulloides at Cariaco, Venezuela is used as a wind speed proxy by Black et al – who condemned its use by Soon and Baliunas as MWP evidence.
(On a previous occasion, David Black showed up briefly here, but when challenged to be consistent i.e. either condemn Moberg’s use of G Bulloides as a temperature proxy or withdraw his criticism of Soon and Baliunas, David Black quickly, and all too typically, disappeared.)
Anyway here’s a graphic showing G Bulloides percentages for 723B, RC2735 as well as 723A and RC2730 on three different scales: left long (300,000 years; middle -the Holocene 10,000 years; right – recent 1500 years.) 723B was sampled at coarse intervals, but even this coarse sampling indicates that G Bulloides percentages above 30% (as in Moberg) are far from unusual on a long-term basis. Do the high G Bulloides values in ice age periods indicate high monsoon levels or simply cold water. A question for another day.
Anil K. Gupta, David M. Anderson & Jonathan T. Overpeck, 2003. Abrupt changes in the Asian southwest monsoon during the
Holocene and their links to the North Atlantic Ocean, Nature.
David M. Anderson, Jonathan T. Overpeck, Anil K. Gupta, 2002. Increase in the Asian Southwest Monsoon During the
Past Four Centuries, Science.