I can’t hear you
Posted by Jeff Id on December 2, 2011
This email demonstrates again exactly why hockey stick temperature curves are all crap.
Bo Christiansen wrote an email (copy below) on March 16 2009, stating in no uncertain terms, what I have written here dozens of times. You can’t regress extremely noisy data on a short series (temperature range) and then project it 2000 years into the past. When the noise approaches the level of (or in the case of paleowhackology, actually dominates over) the signal, you will have a loss of variance in the historic signal in comparison to the calibration range. Loss of variance means, no big changes in historic temp. It seems pretty obvious to me that this is what happens, but in the paleo world, it is nearly completely ignored. The problem is so severe that Mann published a paper in 07 which attempts to cover it up. I use the words, ‘cover it up’ with care as it is based on a lot of experiences and it is my true opinion. Unsurprisingly, he used artificial data which conveniently had the right kind of noise and as so often happens in paleoclimate – he just barely missed detecting the problem.
What it means is that your little ice age (LIA) and your medeival warm period (MWP) in EVERY reconstruction I’ve seen are actually more extreme than shown. My estimates are that they are muted by as much as 50%. This is not a minor issue for the hockey stick poster boy of climate science or the science in general because the models follow the hockey sticks.
Now there are papers where these kinds of methods aren’t used. Ljungqvist’s work for example. The problem with these though (besides the fact that I don’t believe the data has much temperature signal in it at all) is that they use known “temperature sensitive” proxies. These are noisy proxies which have been pre-sorted by people over the years because of their correlation to temperature. It is therefore the same thing as the regression methods except it is done inadvertently by hand.
Most reconstruction papers ignore these obvious facts just as they have pretended to ignore this blog but Bo Christiansen has published what is a paleo-nightmare. What we learned from this email release is that he took the time to communicate it to many of his colleagues for which I’m sure he has gained little popularity. Think of the guts that took to publish this paper and then communicate it to the team after the Soon and Baliunas attacks.
There are no responses to this email that I can find in the released documents.
date: Mon, 16 Mar 2009 14:08:31 +0100
from: Bo Christiansen <email@example.com>
subject: Paper on climate reconstructions
to: firstname.lastname@example.org, email@example.com, Francis.Zwiers@ec.gc.ca, firstname.lastname@example.org, email@example.com, firstname.lastname@example.org, email@example.com, firstname.lastname@example.org, email@example.com, firstname.lastname@example.org, email@example.com, firstname.lastname@example.org, email@example.com, firstname.lastname@example.org, email@example.com, firstname.lastname@example.org, email@example.com, firstname.lastname@example.org, email@example.com, Juerg Luterbacher , Andrew Weaver , Alexey Kaplan , firstname.lastname@example.org, email@example.com, firstname.lastname@example.org, email@example.com, firstname.lastname@example.org, email@example.com, firstname.lastname@example.org, email@example.com, firstname.lastname@example.org, C.N.Roberts@plymouth.ac.uk, email@example.com, firstname.lastname@example.org, Peter Thejll , Torben Schmith , email@example.com, firstname.lastname@example.org
Our paper “A surrogate ensemble study of climate reconstruction methods:
Stochasticity and robustness” has now been published in J. Clim.
A VERY brief summary:
1) Seven different reconstruction methods are tested
on the same data. These methods include both direct reconstructions
of the NH mean temperature and field reconstruction methods.
The field methods include both the original method of Mann et al. 1998,
the RegEM Ridge and RegEM TTLS methods used in more recent work.
2) We use a field surrogate method to estimate the stochasticity that
is always present in regression methods.
1) All methods strongly underestimates the amplitude of
low-frequency variability and trends. This means that
it is almost impossible to conclude from reconstruction studies that
the present period is warmer than any period in the
2) There is a large element of chance in the reconstructions.
This might also explain some of the opposing results obtained
in previous studies.
In a new paper we have submitted to J. Clim. (under review) we
use the the same methodology to test sea level reconstruction methods.
The paper is here:
These methods are closely related to the temperature reconstruction methods.
Now tide-gauge measurements take the place of temperature proxies,
and satellite altimetry the place of temperature observations.
For the sea level reconstructions the situation is simpler than for
temperature reconstructions because the tide gauges are direct measurements
of sea level and do not include noise. We can therefore easier
identify the sources of the errors in the reconstructions
and stratify the errors into contributions from limited spatial coverage
of gauges and non-stationarity of the sea level field.
Best wishes -Bo
Bo is right on all counts and the hockeysticks of paleoclimate should be dead for IPCC AR5. My guess is that the HS poster boy will be front and center so everyone can be properly afraid. Bo Christiansen also took the time to post on this paper at Eduardo Zorita’s blog and allowed a guest post here.
[lalalala – insert I can’t hear you cartoons here]