Proxy spikes: The missed message in Marcott et al

I got a bunch of replies and will answer a few of them:
Benfrommo: “But that is the way science works. Science ASSUMES that the null hypothesis is true until it is PROVEN false. ”
In this case it is proven false. Our CO2-emissions are clear as is the rapidly rising concentration in the atmosphere. Your “null hypothesis” would have to pretend industrialization never happened or that CO2 isn’t a greenhouse gas. You should be careful about what you call the “null hypothesis” as that is usually far from obvious.
Jerome: “If you see planets around our perfectly normal sun, then it is fairly safe to assume that we are NOT the anomaly”
Then you have to know that our sun is “perfectly normal” which you can’t do unless you have knowledge about how common planets are. Assume only one sun in a million has inhabitable planets. What are the odds our sun has it? Still 100%. I’m a perfectly ordinary human. I was born in Sweden. Does that mean that everyone is born in Sweden? Statistics from a single sample is risky.
markx “So, how do we really know whether the 20th century spike is ‘special’ or not?”
We can’t, there is always the possibility that there has been rapid change that just is too small or too rapid for us to be able to see it in the proxies (God in the gaps). However, that was not what I argued against. Nancy Green claimed the much stronger “Rather than being an anomaly, the 20th century spike should appear in many places as we improve the resolution of the paleo temperature series”. She used a study that didn’t find any spikes before the current one as support for the claim that there ought to be others.

Anthony, this is a figure I was thinking of generating myself, but it bothers me when the caption states that the “vertical scales don’t match”. Do the ORIGINS not match (naturally) or is it the SCALE that does not match? This is a crucial question. The point can be made mathematically precisely — the annual variance (really, more than just variance, several cumulants/moments) about a running mean temperature is known to reasonable accuracy from the data from the thermometric era, and it is bone simple to generate colored noise with the same distribution to add to the oversmoothed data from Marcott et. al.
I would argue that if the SCALES of the curves match up, this figure is meaningful even “as is” above, but it would be much more meaningful if one generated a half dozen variations of Marcott plus annual noise that matches the observed natural variability of the climate about 100 year smoothed trends or 50 year smoothed trends.
In any event the term “scales don’t match” needs to be clarified — if the size of the degrees match they have the same scale (but are overlaid with different vertical origins to have roughly the same mean). To REALLY get the right feel for things, the noise should be applied only to a segment of Marcott with the same length as the added sequence of noise — indeed, it should decorate the curve so the first 10,800 years match the noise apparent in the last 200 years in their actual figure.
rgb

Leif Svalgaard says:
April 3, 2013 at 7:37 pm
You can infer the long-term trend among other things. Don’t overstate your case.
=========
Dr Svalgaard, thank you for taking the time to review my paper. I considered your argument when I wrote the sentence in question and in the general sense I believe you are correct.
I used the word “reliable” in the sense that as you increase the resolution the trend may change, depending upon the error in the original signal. What I was trying to convey is that Mother Nature tends to be full of unexpected surprises and we should not over estimate the reliability of our findings. Do we really known that the trend over 8000 years is for example 1.1C? Might not better data change this to 1.3C?

I have repeated Tamino’s work on peak detection in the Marcott data generating three 200-year long triangular peaks of amplitude 0.9C at dates 7000BC, 3000BC and 1000BC. I did exactly the same analysis but using the individual proxy measurements binned into 50 year intervals. I used the Hadley algorithms for generating the global averaging of the 5×5 degree grid. I found that it is indeed correct that the peaks would be detected. The signal however is more smeared out than that of Tamino and I suspect would be further reduced by ~30% folding in measurement error.
What is even more interesting is that the underlying data actually does show a few slightly smaller peaks similar to the generated ones. One of these coincides with the medieval warming period !
see: http://clivebest.com/blog/?p=4833

Look at this series of numbers;
2,2,3,2,3 mean = 2.4
4,1,1,2,4 mean = 2.4
6,1,2,1,2 mean = 2.4
1,2,2,3,4 mean = 2.4
Consider the first 3 sets to be proxy measurements that have been averaged to give, say, 200 years resolution. All the “high frequency” variability is lost.
The last set is instrumental data. If it is treated in the same way as the previous 3 sets, it gives the same mean.
However if it is plotted as its individual values, preserving the high frequency variability, it gives the appearance of a rapidly rising (temperature) signal.
This is why you cannot simply “splice” proxy and instrumental data, of greatly different resolutions and expect to get a meaningful result.

@Leif Svaalgard
Leif I will give you this, you state more personal opinions than almost anyone. I have to totally disagree with you about using low resolution models to spot long term trends…and speaking of long term trends, the current trend is thus
GAT is still above CAT, and average atmospheric co2 is also below the running average. That you and the alarminati choose to study co2 in the context of human history is just a biased opinion. The running average is the more objective number to use. And what you have now is a damaged science because the alarminati continue to produce deceitful graphics in which they show current temperatures running warmer than the Roman and Medieval warm periods which is a lie.

Thomas says:
April 3, 2013 at 10:55 pm
This analogy only works if you start by assuming that there is nothing special about the 20th century
==========
Marcott ends at 1940. The population explosion and global Industrialization based on fossil fuels are almost exclusively post WWII. Therefore, the spike in Marcott cannot be attributed to AGW except in a minor sense. The majority of the spike must be due to natural causes or error in the underlying work.

Alkenone series suffer from the same limitations as tree rings. Any subset of measurements could be disturbed by bioturbidity, sediment slumping, changes in currents, etc., so you have to average a lot of series together. A high resolution series is a single core, but with lots of uncertainty. This leaves lots of room for statistical interpretation.
Mark T. and Leif Svalgaard are both mostly right. You should be able to ascertain a trend from low resolution data as long as you can safely rule out systematic error like aliasing.
Another point should be made. Marcott et al. is geting a lot of press as being the “first global study that spans the Holocene.” Folks need to remember that there are lots of proxies that cover the Holocene, and they tend to show the Holocene Thermal Maximum as much warmer than today. But you have to go back to studies written on paper not bytes, and modern climate scientists seem unwilling to do that.

Apologize in advance for constantly flogging this point regarding the centrality Tennessee, Oak Ridge National Laboratory, and politcal climate science. Marcott, Shakun, et al ran their computations through it’s custom made hockey stick generating Jaguar supercomputer. ORNL is a creature of the nuclear power industry and perfectly represents our former Senator Al Gore’s interest in cutting edge technology and edgy environmentalism. If you see ORNL on the list of ingredients you may want to consider the contents carefully.

Historical temp spikes near the poles are likely to be larger than temp spikes in temperate areas, just as shift in temps today are enhanced nearer the poles. So Vostok can make the case about spikes, but may overstate it.
Matt Ridley had a post on higher temps during the Holocene Optimum (about 6 to 8 thousand years ago):
http://www.rationaloptimist.com/blog/unprecedented-warming.aspx
As part of that post, he linked to a site with several tens of records of high latitude speleothems, which apparently nobody has put together into an article:
http://www.ncdc.noaa.gov/paleo/speleothem.html
Maybe somebody who is very good with numbers and is familiar with these issues might do a blog post, followed by a journal article, showing what these temp records look like? We could see the extent of temp spikes, then, not just at the poles but also nearer the equator.

The reason for the spikes are the distorted scales especially the Y axis. If sensible scales were chosen then there would be no spikes just minor bumps and hollows. The proxies are treated as absolutes when in fact they are just educated guesses?
The quasi religious debate over global warming will only end when scientists with no axe to grind are formed into a group and are tasked to look objectively at the scientific evidence. Unfortunately hell will freeze over before that happens?