In his New York Times “Green” blog article “Running the Numbers on Antarctic Sea Ice” Times reporter Justin Gillis generates a new index of sea ice melt that hypes the loss of Arctic sea ice relative to the gains in Antarctic sea ice. As you’ll see below, perhaps a more appropriate title would have been “Torturing the Data on Antarctic Sea Ice.”
The purpose of Gillis’ article is to downplay the observations which show that the sea ice around Antarctica has recently set a record for its all-time maximum extent (since satellite observations began in 1979).
Whenever the sea ice around the North Pole sets a new record minimum (which it did last month), the global press gets very noisy. But when a new record maximum sea ice extent is set in Antarctica, no one seems to notice.
Gillis’s article is an attempt to explain why the Arctic ice loss is so much more important that Antarctic ice gains. And to make his point stronger, he reverts to a tried and true technique of mathematical obfuscation—comparing changes in percentages, or rather ratios of differences in percentages instead just reporting on the raw numbers—in order to overinflate the apparent magnitude of the Arctic sea ice declines compared to Antarctic sea ice gains.
Consider the absurdity of comparing the trends in the summer ice minima in both hemispheres.
Figure 1 contains the plots of the sea-ice extent in both hemispheres. The North Pole is completely surrounded by water—all the way down to about 70°N. There is sea-ice to melt there. The South Pole is completely surrounded by land—all the way up to about 70°S. There is no sea-ice to melt there, except for the ring that surrounds the continent.
Figure 1. Satellite history of sea ice in the Northern Hemisphere (top) and Southern Hemisphere (bottom). (Figures from Cryosphere Today).
Each and every summer, the Antarctic sea-ice pretty much melts back to the continental margin, so it’s pretty hard to establish any trend in the minimum. Comparing the rate of the decline in summer sea-ice in the Artic to the Antarctic is essentially dividing a number by zero, which results in an infinite difference.
Recognizing this dilemma, Gillis compared two five-year periods at the beginning and end of each record. As a statistical hiccup of the data, there is a slight (nonzero) difference in Antarctica.
Another way to look at this is to look at the “anomalies”, or the departure from some long-term average in ice coverage. While the Antarctic is gaining ice, at about 0.14 million square kilometers per decade, the north is losing it at around 0.55 km2/decade. These are average rates during the entire period of record. So, overall, the Arctic is losing ice at a rate about 4 times what the Antarctic is gaining.
Gillis got a remarkable 25 times difference. Indeed, he tortured the data until it finally confessed.
First, he converted the observed sea ice extent to a percentage of the area of the total ocean over which it resides. We are not sure what this is supposed to represent, as noted before, the Arctic Ocean is largely confined by land, while the Southern Ocean is defined as being the confluence of the Pacific, Atlantic, and Indian Oceans south of 60°S. This is just an arbitrary definition of a sub-part of a vast body of global ocean. So calculating the percentage of ice cover provides an equally arbitrary number. And what’s worse as far as Gillis’s calculation is concerned is that the area of the Southern Ocean (20.3 million sq km) is about 6.2 million square kilometers greater than the area of the Arctic Ocean (14.1 million sq km). This means, that even though the annual average sea ice extent in the Arctic (11.8 million square kilometers) is a tiny bit less than the average annual sea ice extent in the Antarctic (12.0 million square miles), on a percentage of ocean coverage basis, the Arctic sea ice is greater than the Antarctic sea ice (84% vs. 59%)—which means that by Gillis’s reckoning, it has more to lose.
Next, he averages the percentage of sea ice during the first 5 years of the record in each ocean basin and does the same thing again during the most recent five years of the record.
For Arctic end-of-summer sea ice, he gets that from 1979 to 1983, the ice coverage was 51.1%, and from 2008-2012, the coverage dropped to 31.6%. The difference between the two percentages, now vs. then (31.6 minus 51.1) is -19.5%.
His similar calculation for the Antarctic end-of-summer sea ice extreme yields a coverage of 13.8% from 1979-1983 and 14.6% from 2008-2012. And the difference between the two percentages is 0.8%.
In the coup de grâce, Gillis divides the percentage change in Arctic sea ice extent (19.5%) by the percentage change in Antarctic sea ice (0.8%) and arrives that the conclusion that “So, expressed as a percentage of ocean cover, the decline going on in the Arctic is almost 25 times the increase going on in the Antarctic.”
So in a world where the actual areal loss in Arctic sea ice extent is four times that of the areal gain in Antarctic sea ice, Gillis reports that, by his physically meaningless, but hyper-sensitive measure, that Arctic sea ice minimum is declining 25 times faster than Antarctic sea ice is growing.
When the news isn’t fit to print the way it is, apparently there are ways of making it fitter. We think someone ought to complain to the UN’s High Commission on Data Torturing!