When Moberg  first came out, I posted up some first comments on it. I haven’t done anything on it since then, partly because of the amount of time responding to comment on our MBH articles, partly because I got stuck on some missing data sets. Hans Erren has a really neat method for digitizing graphics from pdf’s and kindly digitized a couple of series for me. I just re-plotted them against the original data and there appears to be an extremely peculiar discrepancy in the Lauritzen stalagmite series, which also shows the irritating aspect of people using "grey" and unarchived data. See what you think.
The graphs below show the "low-frequency" series used in Moberg. Tree ring series are used by Moberg for "high frequency". Here there are no data citations and I’m still missing some data.
Before discussing Lauritzen, in passing, series #1 ends in 1966 with a downtick in my replication (there is no downtick in the Moberg version). I’ll check on this later. You will notice that this series (#1), as well as #11, is denominated in %, rather than deg C and has an obviously nonlinear relationship to the other series (as does #11). Nevertheless, it is blended with the other series in a type of weighted averaging (through wavelet methods.) I discussed series #11 previously as it is a very peculiar indicator for global warming: it measures cold water plankton and as a direct measure of SST would be turned upside-down. The hypothesis is that cold-water plankton offshore Oman shows upwelling of cold water due to strong monsoons and is thus an indicator of global warming elsewhere. This may be – but the ostensible reason for including the proxy was not as a teleconnection, but because local proxies were scarce. So it’s a strange proxy.
The downtick in #6 around 1500 is bigger in my re-plot of the data than is shown by Moberg. I haven’t gone through and exactly matched scales, so these plots are indicative.
Now look at series #8. The two versions look similar; my re-plot is smoother – but the ending dates are different.The Moberg version is to the right with respect to series #7. My re-plot is from data digitized by Hans Erren from Figure 11 of Lauritzen et al, shown below. The digitization of Figure 11 ends in 1865, while the digitization of the corresponding series #8 ends in 1937! (Update: Sept. 7, 2005. I sent an email to Anders Moberg about this. He said that he used a pers. comm. series from Lauritzen which ended in 1938 and that, if I wanted information about the difference, I should communicate with Lauritzen. So it would appear that we’re dealing with two different versions of the Lauritzen series, rather than incorrect digitization by anybody)
So checking back to Lauritzen, posted up here, Lauritzen stated:
The “ÅLittle Ice Age’ (LIA). The sample was not actively growing when collected. Since TIMS dating gave an average age of 253 years for the top 5 mm with a “Åcooling’ trend in the isotopes (Figure 7), the coldest signal (–7.12″°) here is taken as an extreme LIA signal.
Lauritzen’s Figure 11 definitely shows the series as ending in the 19th century:
Lauritzen Figure 11. The speleothem temperature history inferred for the last 2000 years compared with known “Åhistoric’ events.
Hans also provided me with a digital version of series #8 from the Nature SI. Below is an overlay of the two series – digitization of Nature black; digitization of source red.
Figure 4. Lauritzen series. Black – digitization of NAture version; red – digitization of original version
Maybe there’s something that I’m missing, but it sure looks like a cock-up. Does anyone see where I’ve gone astray here?
Does the error "matter" – a question which we hear from time to time from the Hockey Team? (I de-snided that a little.) It’s hard to tell. Moberg has not archived his methods and I ahven’t replicated his methodology yet. (I am still missing some high-frequency proxies.)
There are only 11 low-frequency proxies. This allocates one of the coldest periods of the record to the 1930s, which is one of the warmest periods in the record and presumably makes that lower than it would otherwise be.
The other question from these graphs is: 20th century warmth relative to the MWP is dependent on very few individual proxies: #1, #10 to a degree and especially #11: which as I mentioned actually measures upwelling of polar diatoms.
I haven’t quite figured out Moberg’s wavelet method. I used discrete wavelet methods in our GRL simulations, while Moberg uses continuous wavelets – an odd decision for discrete series. I’m not going to bother trying to figure out how he uses continuous wavelets, since any worthwhile results should be robust to using discrete wavelets. I’m going to try substituting a record of treeline changes (either Polar Urals or foxtails) for series #11 and fix the dating of #8 and see what happens.
#10 is a study by Yang et al., where I’ve spent some time and am stuck for a couple of references from Science in China Series D – if anyone can help by sending me a pdf, I’d appreciate it. Lou, J. Y., and C. T. A. Chen, Paleoclimatological and paleoenvironmental records since 4000 BP in the sediments of alpine lakes in Taiwan, Science in China (Series D), 40(4), 424– 431, 1997.
Lou, J. Y., C. T. A. Chen, and J. K. Wann, Paleoclimatological records of Great Ghost Lake in Taiwan, Science in China (Series D), 40(3), 284–292, 1997.