Global Sea Ice Nears Record High

Global sea ice area is a hot topic these past several years. We regularly hear about polarbears drowning due to ice caps melting, ice shelves breaking off in the Antarctic. We also hear the global warming advocate blogs basically screaming about the end of the world almost universally promoting leftist government policies as the salvation. Unfortunately the mindless public follows along rarely taking time to actually look at the data. Some are suspicious yet have no idea who’s right, others simply accept scientists word. “Trust the experts”, is becoming the standard mantra in blogland on advocate sites, how can we trust if the message from guys like Hansen are so distorted.

For those who’ve been following along, we have been watching the high ice levels in the Arctic and climbing Antarctic ice for some time. This led me to suspect that we were approaching a record high ice level. This is important due to the recently released US EPA report calling CO2 pollution dangerous referencing the obvious signs of warming. The report contained dozens of items presented as fact, unfortunately many of its points were nothing but false anti-science propaganda. This is most disturbing as the propaganda is being used as the excuse to dramatically increase the cost of the very item which powers our global economies – energy.

The point of this is to say that global warming science is biased both politically and scientifically by advocates posing as scientists as well as the expected politicians. Anyway, this very important quote came from the EPA.

R Code for Ryan’s Third Post Verification of Steig et al.

Steig et al. a Successful Reproduction

Ryan O has done a brilliant job hunting down and now recreating the verification statistics in the Steig et al Antarctic reconstruction.  This article details not only the verification statistics, it clearly lays out the limitations of the verification methods used.

Absolutely top notch.

Also, my thanks to Ryan for  recognizing Dr. Steig’s cooperation.

======================================================

REPLICATING STEIG III: REGEM VS. PCA

I had mentioned in the Replicating I thread that after I finished with both the satellite portion and the ground portion, I intended to email Steig and ask about any discrepancies. Yesterday, I emailed him at about 2:45 pm with a description of what I found (including the relevant images). By 3:45 pm he had emailed me back with the explanation. After I read the explanation, the reason doing the verification statistics as described in the SI became abundantly clear.

While my actual calculations had been correct, my methodology was missing a step. Simply comparing the reconstruction to the AVHRR data (my method) does not evaluate the performance of RegEM. It only evaluates the performance of the 3 PCs.

In order to evaluate the performance of both the PCs and RegEM, Steig performed two additional reconstructions. For one, he withheld the 1982 – 1994.5 AVHRR data and performed the reconstruction using only the 1994.5 – 2006 data. For the second, he withheld the 1994.5 – 2006 data and performed the reconstruction using only the 1982 – 1994.5 data. The period where actual AVHRR data was used was the calibration period; the period where the AVHRR data was withheld was the verification period.

In other words, he deletes half of the AVHRR data, performs the reconstruction, and then compares the result to the deleted AVHRR data. Because this is a harder test to pass, some areas do actually end up with negative RE and CE values.

Within about 10 minutes of Steig’s reply, I was able to produce the following:

Fig. 1 -Replicated split calibration/verification experiments.

Compare to:

Fig. 2 -Original (Steig) split calibration/verification experiments.

Except for color scale issues, a perfect visual match.

Robustness

These reconstructions have to do with an exploration of why the Campbell and Macquarie stations were included in Steig et al Antarctic reconstruction. These stations are far from the continent and located on small islands, a properly functioning algorithm should probably completely ignore them.

Figure 1

I ran the whole time consuming RegEM mashing the data process again without the two little islands on the bottom of the graph.

Figure 2