On April 10, I was notified by NOAA that a little bit of data from Trouet was now online at NOAA WDCP. The archive is the absolute minimum; even Science would require them to archive this data if asked.
However, it is totally inadequate for replication.
However, every little bit of data helps and here’s a little progress report. It’s too bad that Esper the non-Archiver has taken an anti-archiving stance as the underlying idea of comparing the Scotland and Morocco proxies is fairly interesting. I’m still trying to understand why the authors regard this information as rebutting the MWP idea, as the proxies themselves seem to show a rather pronounced MWP/MCA.
First, the new data permits a more complete emulation of Trouet Figure S1, now that the Esper PDSI recon is available. For reference, below is the original Figure S1 with original caption.
Trouet Figure S1 Original Caption: Long-term winter proxy records from Europe. Comparison of the tree-ring based Morocco (S2) and speleothem based Scotland (S1) records with a documentary based estimate of September-June England-Wales precipitation (S10) and speleothem based estimate of winter temperature from the central Alps (S13) (A). Time series consist of 25-year averages, standardized over the common period (1075-1925). The Morocco PDSI record was inverted.
Next, here is my emulation of their Figure S1 using the speleothem version said by Andy Baker to be correct; the newly archived Esper PDSI information; my manual transcription of data from Lamb 1965; and archived Mangini information. In each case, I’ve used 50-year averages and then scaled the resulting series over 1075-1925 as in the Trouet caption, orienting the series to match the Trouet figure. As noted previously, 50-year averages match. Three of four emulations match very closely: the oddball is the Scotland speleothem, where, as noted before, Trouet has almost certainly used the detrended version, said by Andy Baker to be an incorrect version. (Andy Baker said that Trouet used the non-detrended version but I don’t think that his understanding is correct; I’ve emailed him about this.)
Figure 2. My emulation of Trouet Figure S1.
Moving along a little, here is Trouet Figure 1 from the main article. The term “30-year cubic spline” may not be familiar to statisticians and scientists from Off the Island; I think that I can explain what this means in Team dialect and will do so if necessary. Note that the two series are oriented oppositely in this plot (while the MWP has a common orientation in SI Figure 1.)
Original Caption to Trouet Fig. 1. Proxy-derived long-term NAO reconstruction. (Top) Reconstructed winter precipitation for Scotland and February-to-June Palmer Drought Severity Index (29) for Morocco. Records were normalized over the common period (1049–1995) and smoothed with the use of a 30-year cubic spline. (Bottom) Winter NAO reconstruction NAOms (black curve) is the difference of the Scotland and Morocco records. The gray area is the estimated uncertainty; yellow and red areas are the 10 and 33% highest and lowest values since 1700. The blue line represents the 30-year smoothed Lisbon-Iceland instrumental NAO index series (11).
Most CA readers hate all the smoothing of data that’s infected so many Team articles and presentations, including Trouet et al. For comparison, in the next figure, I’ve plotted raw versions of the two series in question – the Esper PDSI recon (however it was made) and the Baker non-detrended speleo, said to the preferred version, though, as noted above, unlikely to be the version used in Trouet.
Examining the “raw” versions of the data, here are my first few comments.
1. The “correct” non-detrended Scottish speleo data – whatever it shows – definitely shows a “warm” MWP on the scale of the late 20th century. The ability of the proxy to measure warm climates looks severely limited by the fact that it has a zero minimum and is bumping against the zero minimum in the MWP.
2. The eye can certainly be teased into comparing the two series. And in fairness to this perception, both series had been postulated to be connected to the NAO in independent articles (Proctor et al 2000, 2002) and Esper et al 2007, before the present joint venture article. IF both series are NAO proxies, then the big downspikes should align. Given the precision of tree ring dating, I think that there would be an excellent case for using the tree ring data to align the speleo data with its much looser dating control. I’m not exactly sure how one would express this methodology in objective statistical terms, but intuitively it makes sense to me. Perhaps Roman or UC or Hu have ideas on this.
3. Esper’s PDSI reconstruction has all sorts of ad hoc splicing, selecting and re-scaling, but it’s not possible analyse its impact without the Esper measurement data, which is under lock and key.
![Reblog this post [with Zemanta]](https://i0.wp.com/img.zemanta.com/reblog_e.png)
Climate Audit 
20 Comments
# TROUET SI FIGURE 1
trouetf=function(x,sgn=1,K=50) { #requires 2 columns one named year
names(x)[2]=”proxy”
temp= x$year>=1075 & x$year< =1925 &!is.na(x$proxy)
m0=mean(x$proxy[temp]);sd0=sd(x$proxy[temp])
x$trouet= sgn*(x$proxy-m0)/sd0
x=x[!is.na(x$proxy),]
fred=factor(floor(x$year/K))
test=tapply(x$trouet,fred,mean,na.rm=T)
y=data.frame(K*as.numeric(names(test))+25,test);names(y)=c("year","proxy")
y=y[y$year=1025,]
y$scale=scale(y$proxy)
trouetf=list(x=x,y=y)
trouetf}
Data=list()
##pdsi
url=”ftp://ftp.ncdc.noaa.gov/pub/data/paleo/treering/reconstructions/africa/morocco-pdsi2007.txt”
pdsi=read.table(url,skip=74,fill=TRUE)
names(pdsi)=c(“year”,”pdsi”) #1049 2001
Data$pdsi=trouetf(pdsi,sgn=-1)
##Scottish speleothem
url=”ftp://ftp.ncdc.noaa.gov/pub/data/paleo/speleothem/scotland/assynt96.txt”
speleo=read.table(url,skip=126,sep=”\t”,fill=TRUE,na.strings=”–“) #was 116
speleo=speleo[,c(1,2,4,6,8)]
names(speleo)=c(“bp”,”SU967″,”SU961″,”SU962″,”Avg”)
speleo$year=2000-speleo$bp
Data$scot=trouetf(speleo[,c(“year”,”SU967″)],sgn=-1)
#LAMB transcribed from journal
lamb=read.table(“http://data.climateaudit.org/data/documentary/lamb_1965.csv”,skip=3,fill=TRUE,header=TRUE)
lamb$year=apply(lamb[,1:2],1,mean)
Data$lamb=trouetf(lamb[,c(“year”,”SeptJune”)])
## Mangini
url<-"ftp://ftp.ncdc.noaa.gov/pub/data/paleo/speleothem/europe/austria/spannagel2005.txt"
#SPA 12 only
test<-read.table(url,skip=56,header=TRUE,fill=TRUE)
names(test)=c("bp","year","O18","temp")
Data$mangini=trouetf(test[,c("year","temp")])
#Winter NAO Reconstruction 1049-1995 AD.
url="ftp://ftp.ncdc.noaa.gov/pub/data/paleo/treering/reconstructions/nao-trouet2009.txt"
#April 7 archive, available Apr 10, 2009
nao=read.table(url,skip=94,fill=TRUE)#1049 1995
names(nao)=c("year","nao")
range(nao$nao) #[1] -2.6075 3.1488
##NOW DO PLOT
col0=c("red","steelblue4","cyan","green");index=c(3,2,1,4)
layout(1);par(mar=c(3,3,2,1))
plot(950:2000,.01*(950:2000),ylim=c(-3,2.15),yaxs="i",xlim=c(1050,2000),xaxs="i",type="n",ylab="",axes=FALSE)
for(i in 1:4) lines(Data[[i]]$y$year,Data[[i]]$y$scale,col=col0[i],lwd=2)
abline(h=0);mtext(side=2,line=2,"z-scores",font=2);axis(side=1);box();axis(side=2,las=1)
legend("bottomleft",fill=col0[index],legend=c("Esper PDSI","Baker 2002","Lamb 1965","Mangini")[index],cex=.8)
title("Trouet et al 2009 Figure S1")
#NATIVE DATA
nf=layout(array(1:3,dim=c(3,1)),heights=c(1.1,1,1.3))
par(mar=c(0,3,2,1))
plot(Data$pdsi$x$year,Data$pdsi$x$trouet,type="l",ylim=c(-3.6,3),xaxs="i",xlim=c(1030,2010),col="grey10",axes=FALSE)
axis(side=1,labels=FALSE);axis(side=2,las=1);box();abline(h=0,lty=2)
#lines(Data$pdsi$y$year,Data$pdsi$y$scale,col=2,lwd=2)
text(1050,2.8,font=2,pos=4,"Esper PDSI")
title("Trouet NAO Proxies")
par(mar=c(0,3,0,1))
plot(Data$scot$x$year,Data$scot$x$trouet,type="l",ylim=c(-3.6,3),xaxs="i",xlim=c(1030,2010),col="grey10",axes=FALSE)
axis(side=1,labels=FALSE);axis(side=2,las=1);box();abline(h=0,lty=2)
#lines(Data$scot$y$year,Data$scot$y$scale,col="steelblue4",lwd=2)
text(1050,2.8,font=2,pos=4,"Baker Speleo")
par(mar=c(3,3,0,1))
plot(nao$year,nao$nao,type="l",ylim=c(-2.5,3.5),xaxs="i",xlim=c(1030,2010))
text(1050,3.2,font=2,pos=4,"Trouet NAO")
Finally some math I’m familiar with, it’s nice not to need to look up parametric cubic splines to figure out what you’re talking about.
Is there a link to the article somewhere I missed so I can follow along? Sorry if it’s obvious, I’ve been busy.
PDSI (palmer drought severity index) is often elevated during hot years, so to use high PDSI to claim that the MWP wasn’t hot seems…iffy.
Re: Craig Loehle (#3), Often, yes, but not always. For the entire US, I understand it was higher in the earlier half of the century, much more than one would expect from the apparent relationship with temperature.
can you use however twice in succession like in the first paragraph, or does the second nullify the first… are there any grammarians.. that read this?
but you are right… this really doesn’t disprove any MWP, maybe they think if they can show it was localized or more pronounced in certain regions like Scotland, it would slightly discredit it? i’m guessing that’s what the morocco data is for… Keep up the good work, Steve!
Isn’t the Trouet paper supposed to be about finding a natural cause for the warm period and thence arguing that the MWP was plausibly confined to Europe? For that argument they need a proxy with a MWP. So it’s more a localization than a rebuttal.
But a climate link to the NAO has been already established for England, Armagh and Sweden though and the periodicity correlation is 7.8 years.
Click to access 445.pdf
Trouet et al. are postulating another (carrier) oscillation covering the entire MWP I think.
Maybe she didn’t notice that Mann08 had proxies showing an MWP in China and Latin America. She seems to be trying to validate Mann’s previous hand-wave, that of the EWP, but he’s already moved on to a new idea – that of the smallish, worldwide MWP.
Re: JamesG (#6),
It seems extraordinary that the current skeptical theories of Roy Spencer, Bill Gray, A. Tsonis and others, viz. that current warming is shown to be related to phases of the PDO/AMO/etc/intersections-thereof, are brushed aside by the Team without comment, and the next thing we hear is that, “ooh, let’s say the entire MWP was caused by the NAO!”
My understanding of the IPCC2007 report is that they were able to marginalise the atmospheric heating from ocean oscillations by arguing that the current positive phases of the ocean oscillations (AMO, PDO, NAO, …) could only have been caused or amplified by human-caused global warming.
Can anyone tell me what is different about the MWP? If the ocean oscillations aren’t strong enough in principle to cause this awfully massive 0.3 C temperature increase (or whatever the actual number is) since 1976 how can it be considered as the several hundred year cause of the MWP?
Also does anyone know what the standard rebuttal to Spencer’s PDO, Tsonis’s superposition theory (or whatever it’s called), and similar theories?
Which other group of calculators strives so hard to extract a convenient signal from a bundle of noise? People listening for UFOs?
Re: Geoff Sherrington (#7),
Given what I have been reading on this website for the past few years, that strikes me as a perfect analogy.
Make sense that MWP was more pronounced in Northern Latitudes — since, generally, global warming is expected to be more pronounced in Norhern Latitudes according to IPCC. Still, the subject study shows MWP WAS more pronounced than today’s warmth. Any evidence for a warm MWP discredits an interpretation that current global warming is unique (assuming that GW is happening currently).
Yet another worthwhile CA article.
Allen
“since, generally, global warming is expected to be more pronounced in Norhern Latitudes according to IPCC”
One of these 20-20 hindsight things?
Allen 63 says – “since, generally, global warming is expected to be more pronounced in Norhern Latitudes according to IPCC.”
Does anyone else think that the idea of ‘global’ warming being more pronounced in ‘Northern Latitudes’ is just silly? Or am I the only one?
(I admit, I’m skeptical about the whole notion of a global temperature which might be warmed up)
Global warming does not distribute the “heat” uniformly, or so it is written. Rather, it is anticipated that if, for example, there is 2C warming globally, there will be closer to 1C near the equator, 2C in Ohio USA, and 4C+ nearer the North Pole. This has been observed in the geological record during past warm ages (warmer poles). The forgoing is simplified and rough (for such a mathematically precise blog site).
Now, I have not proven that personally, its just what I have frequently read. All things considered, it seems physically reasonable to me.
For the record, I am an AGW skeptic. But, I would certainly welcome an additional 2C in Ohio USA (this month its a couple C colder than normal).
“Polar amplification”, i e that temperatures will rise more at high latitudes than near the Equator at times with warm climate is strongly supported by palaeoclimatological data. This is true both during past warmer interglacials and pre-glacial times essentially as far back as we have reliable palaeclimatological data. That such warming is assymetric between the northern and southern hemispheres, however is not supported by paleo data, except for short-term variations. Interglacials that were unusually warm in the north were also unusually warm in Antarctica. The last interglacial was about 5 degrees warmer than at present at Vostok and also about 5 degrees warmer on Greenland and so on.
Re: tty (#12), The “short term” probably is related to the heat capacity of the oceans, and they fact that there is more ocean in the Sothern Hemisphere. However, Antartica exhibits some odd behavior on short term time scales, where it’s habit is to just do the opposite of what therest of the world is doing. I won’t go on about why I think that is here though.
Pronounced Siberian warming will probably be “expected” in the next IPCC report. :^)
Andy Baker has replied to my email agreeing that Figure S1 uses the detrended data. However, he states that the non-detrended data was used in the NAO reconstruction. Baker:
At present, I haven’t been able to figure out exactly what sequence of smoothing and re-scaling is used to yield their NAO reconstruction and am not in a position to comment one way or another on this statement.
I’m a little bit annoyed at the haphazardness – all too typical – but I’m more appreciative of Baker at least taking the trouble to acknowledge that there was an error. I hope that they issue an erratum.
For further analysis, I’ll use the non-detrended data which Baker to his credit archived and took the trouble to annotate the archive after the above discussion.
Re: Steve McIntyre (#16),
Thank you Dr Baker for following up. However, can you now recognize the immense benefit of archiving EVERYTHING, data and code? The value of the work done by Steve M in this regard is immeasurable.
Re: Steve McIntyre (#16),
Good auditing – and a good result.
Smoothing isn’t always a bad thing. It is useful, for instance, in identifying heating events. For instance, if you treat each solar cycle as a discrete heating event and smooth both avg sunspot count and avg temps as a running average of time length equal to the solar cycle occuring at that time, you can get a very close matching of both smoothed curves.