Craig Loehle has sent digital versions of the 18 proxies used in Loehle 2007 which are now in the directory http://data.climateaudit.org/data/loehle listed in the following order (slightly different than the order in the previous note, shown here with url information). I’ve been able to successfully crosscheck many of these series against original provenance. I’ve got some questions yet that I’m seeking to resolve. Loehle has been very co-operative in establishing a proper data set with accurate and reproducible data citations. This compares favorably with (say) Hegerl et al 2006, where after nearly 2 years of effort and over 20 emails, I am still unable to establish the provenance of several series or Esper et al 2002, where it required the intervention of the journal to obtain data (and to this date, one series is missing,)
While it would have been better to have had this all available at the start, at this point the only thing that is “unscientific” about the data situation for Loehle 2007 is the relative promptness of the data availability and the co-operativeness and commitment of the author to ensuring an adequate archive, as compared, of course, to a realclimate scientist.
#1) GRIP borehole temperature (Dahl-Jensen et al., 1998); See Moberg Nature site supplementary material
[SM- digitization at CA/data/moberg/djgrip.txt]
#2) Conroy Lake pollen (Gajewski, 1988); ftp://ftp.ncdc.noaa.gov/pub/data/paleo/pollen/recons/liadata.txt
#3) Chesapeake Bay Mg/Ca (Cronin et al., 2003); ftp://ftp.ncdc.noaa.gov/pub/data/paleo/contributions_by_author/cronin2003/
#4) Sargasso Sea 18O (Keigwin, 1996); ftp://ftp.ncdc.noaa.gov/pub/data/paleo/contributions_by_author/keigwin1996/
#5) Caribbean Sea 18O (Nyberg et al., 2002); ftp://ftp.ncdc.noaa.gov/pub/data/paleo/contributions_by_author/nyberg2002/ converted to temperature on the Moberg article nature site suppl material
#6) Lake Tsuolbmajavri diatoms (Korhola et al., 2000); See Moberg Nature site supplementary material
[SM- Hans Erren digitization at CA/data/moberg/Korhola_fig4_points.txt]
#7) Shihua Cave layer thickness (Tan et al., 2003); ftp://ftp.ncdc.noaa.gov/pub/data/paleo/speleothem/china/shihua_tan2003.txt use col 7 temp
#8.) China composite (Yang et al., 2002) which does use tree ring width for two out of the eight series that are averaged to get the composite, or 1.4% of the total data input to the mean computed below;ftp://ftp.ncdc.noaa.gov/pub/data/paleo/contributions_by_author/yang2002/china_temp.txt
#9) Spannagel Cave (Central Alps) stalagmite oxygen isotope data (Mangini et al., 2005). ftp://ftp.ncdc.noaa.gov/pub/data/paleo/speleothem/europe/austria/spannagel2005.txtmean SST for northern Pacific site SSDP-102 (Latitude 34.9530, Longitude 128.8810) from
#10) SST variations (warm season) off West Africa (deMenocal et al., 2000); ftp://ftp.ncdc.noaa.gov/pub/data/paleo/contributions_by_author/demenocal2000/
#11)speleothem data from a South African cave (Holmgren et al., 1999); from authoremail sent for archive link
#12) SST reconstruction in the Norwegian Sea (Calvo et al., 2002);
[SM- http://doi.pangaea.de/10.1594/PANGAEA.438810?format=html matches exactly]
#13-14) SST from two cores in the western tropical Pacific (Stott et al., 2004); ftp://ftp.ncdc.noaa.gov/pub/data/paleo/contributions_by_author/stott2004/
#15) mean temperature for North America based on pollen profiles (Viau et al.,2006); http://www.lpc.uottawa.ca/data/reconstructions/index.html
#16) a phenology-based reconstruction from China (Ge et al., 2003); ftp://ftp.ncdc.noaa.gov/pub/data/paleo/historical/china/china_winter_temp.txt
#17) SST from the southeast Atlantic (Farmer et al., 2005); ftp://ftp.ncdc.noaa.gov/pub/data/paleo/contributions_by_author/farmer2005/
#18) annual mean SST for northern Pacific site SSDP-102 (Latitude 34.9530, Longitude 128.8810) from Kim et al. (2004); http://doi.pangaea.de/10.1594/PANGAEA.438838
Here is a plot of the above data sets:
Craig Loehle sent me the following plot showing all the proxies together:
He also sent in 3 diagrams showing that the impact of different smoothing intervals was negligible. The 3 figures below show 10-year, 20-year- and 30-year running averages, and the difference is obviously not substantial. I think that this would apply to gaussian smoothing or other forms of smoothing. While smoothing is an issue that always seems to provoke lots of opinions here, it’s an issue that’s very low on my list of issues.
10-year smooth
20-year smooth
30-year smooth
Climate Audit 
31 Comments
For what it may be worth, I did up a spreadsheet of all the Gajewski 1988 series, so that the Conroy Lake proxy could be examined in its original context.
The data is arranged so that it should be easy for anyone to make Excel plots, if you want to play around.
The smoothing algorithm used does no interpolation, and the scatter plots show the data density.
See here
Any comments welcomed.
A general question, why make a big deal over smoothing choices?
Why not pick a smoothing to discuss with maybe a couple small plots of other smoothings to show apple v cherries is no issue in the main body and include full size plots in an appendix for the reader to choose how far to agree with the conclusions?
DaleC,
I like that presentaion, although MA0 Y scale starts 0.5 lower than the others.
I’m not a statistician, but GRIP and kim look as though they may dominate the large MWP in this collection of data. Were they weighted down to compensate for their strong signal? I wonder what the main graph would look like without it?
DaleC,
How is it that all of your temperature curves show declines (generally) through the last millenium, but your anomalies show rising anomalies over that same period?
sonicfrog says:
And here’s one of the reasons why all data needs to be accessable: so others can replicate results or modify future studies. Doesn’t that fall under that “robustness” idea? (does the loss of one proxy drastically change the result?)
Now that you know the data sources, others can test his theories.
Different proxies, same late 20th C divergence problem (from the instrument temperature record). Looks like the divergence problem is the most robust feature in paleoclimatology. Which to me is evidence the problem isn’t in the proxies. It’s in the instrument temperature record.
The smoothings Steve M posted are after Interpolating the data linearly between the existing data points, then averaging, then smoothing. This answers the question also about a different way of treating the widely spaced data.
A good case for multi-colored spaghetti graphs! Also, many of these run offscale while there is a cluster in the center, so it would be informative to graph this both with a big enough scale to incorporate all the points, and then on a finer scale that examines the center.
Obviously, a few of them are a lot noisier than most. Any future study should select only series that have validly estimated standard errors, and then do a weighted average taking these standard errors into account. Meanwhile, it would be interesting to look at the median, which might be less sensitive to the extremes than the mean.
Maybe THAT”S what Dr. Loehle did wrong to upset JEG so. He didn’t display one of those! 🙂
I see absolutely no difference in the overall shape of any of the graphs with the different smoothing, so IMO continuing to argue that point is nothing more than a deflection of the main point: MWP was global and warmer than we currently are. And here’s a question: Will statistically analyzing this thing until we puke change that?
Considering the historical record seems to indicate a MWP, the proxy record (minus bristlecone!) seems to indicate the MWP, why can’t we agree that there is a MWP? Then we can relax, talk about the best way to get off oil (since higher temps won’t cause catastrophic changes – since its happened before) and do it in a way that won’t kill our economy. My vote – hydrogen from nuclear electrolysis – combined with waste carbon to make methanol. (Yes, off topic!) Congratulations Dr. Loehle! Nice job.
Because a strong MWP in the “offical record” will undermine all of the IPCC efforts. They worked HARD to purge it from the records so they could say we are warmer now than ever in recorded human history and it’s a result of CO2 and we need to give all our money now to reseach that won’t do a damn thing. Admitting now that a MWP period existed, was global, and warmer than we currently are would completely nullify what they’ve been saying for the last 20+ years.
Of course, they could retreat to their fallback position: “Even if we’re wrong, it’s the right thing to do….” which is why they are now careful to use the term “Climate Change” instead of “Global Warming” like they did for so long.
Smoothing aliases higher frequency data down into lower frequency bands. Extending the periods of the smooths, as indicated here would increase the aliasing frequency and tend to mitigate any problems that might arise. That being said, since smoothing has its effects on higher frequencies and the purpose of this exercise is to examine the low frequency data to dtermine the presence or absence of a prolonged MWP, the issue seems moot.
RE: #12 – It’s the same mindset as some of the more ridiculous versions of “low impact backpacking.” Some of its tennets are more based on extreme anal retentiveness than actual demonstrated benefit / risk mitigation. Another analogy is RoHS. When you get down to certain concentrations, it really no longer matters. At that point, people are just as equally impacted by naturally occurring carcinogenic mold or genetics as they are by some man made “bogeyman” substance. Which causes me to circle back to my general position in all such matters – use formalized hazard analysis, assessment and mitigation methodologie, employing fault tree analysis, FMEA, DoE, Monte Carlo and other quant methods, to actully lay out the relative values of the various risk factors. Focus on the high RPNs, mitigate, repeat until it is deemed that there is no additional value add in continuing.
When the sample spacing of data is changed (down sampled) during smoothing then the method should include a low pass cutoff to preclude mirror aliasing of high frequencies about the nyquist. If there is no changes made to sample spacing then ASAIK aliasing won’t occur just from smoothing.
Is the Sargasso plot inverted?
For computing confidence intervals, people seem to be assuming that we are dealing with annually resolved data. The data vary from annual to centennial to irregularly spaced. How do you interpolate? What does this do to the errors? Just computing ci around the points estimated by the papers I cited does not allow you to carry that through the interpolation and smoothing. What does smoothing do the the errors? What about dating error? The ref JEG gave for the “proper” way to do ci was for tree ring data (annual) with no dating error and not interpolation andno smoothing. I throw down the challenge: how do I carry uncertainty from the raw data through the temperature recon at irregular intervals into the final smoothed curve? Provide a reference please. Maybe I’m not as dumb as I look.
5, henry:
The paper addresses this, and the loss of any single proxy does not change the result. Also a random selection of 14 data sets at a time didn’t change the result.
What would be the full list of attributes necessary to declare a paper “audited”?
Here’s a start:
1. Data crosschecked against original provenance and archived
2. Analysis methodology fully described, any and all code archived
3. Internal logic and external methods validated
4. All references publically available
Re: reference. My code is available on request. I will archive it with other stuff next week. It is in Mathematica, so it would probably be easier for you R folks to replicate what I did in R as a check.
re #1
to Jim Melton, November 20th, 2007 at 8:05 am
Aaron Wells November 20th, 2007 at 8:29 am
Thanks for the comments. I have made the Y1 scale consistent on the rolled temperature charts.
The problem with the anomaly chart versus the absolute temperature charts was that I had the mean/absolute subtraction the wrong way around.
There is a new chart AnomalyChk now showing the anomaly calculation for Conroy.
Try again here
The first 8 series from Steve’s archive have data at an annual frequency. If you take just these series and extract the first principal component, the resulting weightings are 0.5343, 0.3980, 0.1691, 0.3049, 0.1375, 0.4869, 0.1579, and 0.3921. This is not a simple average. However, it gives a very similar shaped series to the one’s Loehle produced. One difference is that it does an uptick at the end of the 20th century.
Quick question: Why was Conroy Lake, ME chosen over Hells Kitchen Lake, WI? Both have a 2000 year record and it would seem that there is already an Eastern US member with the Chesapeake Mg/Ca. There is a big difference in the two series, with Conroy cooling significantly, and Hells Kitchen is pretty flat.
I’m not really grokking those statements. Aliasing is a phenomenon determined by sampling rate. Smoothing a signal gives one more “room” between the nyquist frequency (due to sampling) and the max frequency of the signal. How that “increases the aliasing frequency” is not apparent to me…maybe this is a semantic thing.
26
I think what he may be saying is that the edges of the averaging window will produce sidelobes [e.g., sine(x)/x] which add distortion. This is most serious if you have sharp peaks in the time-series and need fidelity near them. I don’t think that’s an issue here.
It can be ameliorated by using a window without sharp edges (e.g., a gaussian, but there are lots of other choices).
A filter that rings I understand, smoothing to increase nyquist I don’t. A lowpass filter will reduce aliasing…all things being equal. Like I said it probably a semantic issue.
RealClimate takes on Loehle 2007: link. The dating criticisms are pretty damning.
Nice of them to link to JEG’s personal blog for “further discussion” but not to any of the five CA threads on the matter, though.
Dear Craig,
some quick questions which are most likely already discussed somewhere on this site:
Could you either point the answers out for me or (much better) write a new answer for me in laymen terms 🙂
Thanks a lot,
LoN (engaged in a private little Hockey-stick discussion)
First:
What would be your answer to the accusation, that your smoothed proxies don’t overlap with the instrumental reprod?
Second:
How would you calculate your confidence intervals?
I am sorry for writing this, since I am pretty sure, that I have seen you answering both questions somewhere, unfortunately this was a while ago and my google-fu is not good enough to bring it up.
Re: Laws of Nature (#29), In my original paper I computed confidence intervals using bootstrap and jackknife methods. The long smoothing period is due to the widely spaced data in many proxies.
Re Laws of Nature #29,
Bender apparently still has Craig locked in his basement (inside joke 😉 ), so I’ll try to answer pending his release.
I worked with Craig to provide confidence intervals for his 2008 revision of his 2007 paper. See graphs and links at http://www.econ.ohio-state.edu/jhm/AGW/Loehle/.
The revision comes up to 1935 (based on a 29-year centered moving average of data that ends in 1950). This isn’t the present, but it overlaps the post-1850 instrumental period enough for you to splice it on to your favorite series. (Since it has been smoothed from that that extends up to 1950, it would make sense to make the link with a similarly smoothed version of the instrumental series you use.)
Great!!
Many thanks,
this should deal with that part of the argument!
The fact that it is also peerreviewed comes in handy – I totally underestimated how many good arguments could be stonewalled by ignorance..
All the best,
LoN
2 Trackbacks
[…] the last 2000 years prepared by Craig Loehle. This graph, and the discussion behind it can be found at ClimateAudit.org here. Many of these paleoclimate temperature proxies used in Craig’s graph are found behind the […]
[…] case, they cite information on Loehle that was initially made available at climateaudit. In this CA post, we discussed discussed the provenance of Loehle proxies and requested that Loehle provide his […]