Recent Warming Reverses Long-Term Arctic Cooling

Below is a great example of stochastic+ reasoning (Monte Carlo simulation) gone mad, as illustrated with a global warming AAAS Science report and its rebuttal:

Recent Warming Reverses Long-Term Arctic Cooling
Vol. 325. no. 5945, pp. 1236 - 1239
DOI: 10.1126/science.1173983

The obliquity of the Earth's orbit raises and lowers the height of the analemma (i.e., the height of the figure eight) in 40K year cycles (20K rising, 20K falling). Glacial epochs over geologic history have often occurred within a few thousand years after the analemma drops to its lowest height in the Northern Hemisphere (analogous to the one-two month delay in the coldest part of winter after the autumnal equinox). Of course it's never that simple. See here.


Report:

The authors developed a 2000-year trend line of temperatures in 10-year increments using a compilation of proxy records from Arctic lakes, combined with complementary ice core and tree ring records. Again, this is northern hemisphere temperatures which should show cooling over this 2000-year period due to decreasing solar obliquity (i.e., height the sun reaches in the sky at the peak of summer, the upper right location of the sun in the image).

The article states that trend lines for cooling matched what you would expect from decreasing obliquity until you reach 20th century, "with the last half-century being the warmest of the past two millennia ... the most recent 10-year interval (1999-2008) was the warmest of the past 200 decades [authors are talking about Northern Hemisphere temperatures]. Temperatures were about 1.4°C higher than the projected value based on the linear cooling trend." [note: not absolutely 1.4°C higher]


Rebuttal:

"...the 23 proxy sites (Fig. 1) closely tracks the temperature for all of the land area north of 60° latitude, indicating that our proxy network accurately represents the Arctic-wide mean..."

Let's assume a simple average of each of the 23 sites is the arctic-wide mean temperature (a dubious proposition when you consider 2000 years of data). 23 data points for the whole of the arctic - we must be very careful in how we measure each data point since each is responsible for a lot of geographic area.

What does the average of the 4 tree rings represent? Nothing. When these tree rings thicknesses spike (illustratively meaning warmer summer temperatures), it does not mean anything for arctic-wide mean temperatures.

An ice core at one of the 7 sites could be simultaneously driving down the average of all the tree rings. So what purpose does it serve to show tree ring proxy temperatures in the Figure 3 graphs? Similarly the average of 12 lake sediment proxies or 7 ice core proxies mean nothing. There is no basis for the authors to make the outrageous claim: "from 1 C.E. to 1900 C.E. ... the cooling trend is especially clear in records from ice and lakes." - lake proxies only represent 12 out of the 23 sites.

It's only when you have data from 23 sites can you say you have "a reading." Having 22 sites means you have nothing. If you look at the mini-graph between Figure 3a and 3b you see that you only have 23 site "readings" for the years between 980 and 1800. Everything else is pure speculation (or fake data using Monte Carlo simulation which has been ostracized from most fields of study dealing with complexity).

The authors had to resort to the magic of statistics to reach their conclusions. But in reality you can't extrapolate missing data because each measurement is so important. That 23rd reading must be able to significantly affect the direction of the other 22. Otherwise it should not have been included in the 23.