These predictions are naive in the sense that they are not based on a physical model, nor other measurements apart from the 30-odd year history of the index in question. Moreover, they are made a year in advance as winter freeze-up is just starting. The predictions are simply If ... Then statements: If trends from the recent past continue ... Then we could expect this much ice next September. Those past trends appear reasonably well characterized by Gompertz curves, fit by nonlinear least squares to the data. Error bands in Figure 1 shade a range plus or minus twice the standard deviation of the residuals (observed variation around the curve).
The resulting naive prediction for mean Arctic sea ice volume in September 2013 is 3,100 km3, plus or minus 1,900. Extrapolated beyond 2013, again as an If ... Then exercise, the curve drops below a "virtually ice free" threshold of 2,000 km3 by 2015, and approaches zero volume after 2020 -- with uncertainties that could go either way, of course.
Daily updates in Cryosphere Today sea ice area receive close attention during the melt season. In an earlier analysis I was surprised to find that the 1-day minimum area seemed no more unpredictable than the September mean. In that spirit, Figure 3 offers a prediction of CT 1-day minimum area.
The predecessor to this post, Naive Predictions of 2012 Sea Ice, offered similarly based predictions for September 2012. Observed 2012 values in every case fell below the central prediction, though still within error bounds:
Sep 2012 Sep 2012
NSIDC extent 4.3 (3.4-5.1) 3.6
NSIDC area 3.0 (2.2-5.1) 2.4
PIOMAS volume 4.0 (2.0-5.9) 3.4
CT area 1-day 2.7 (2.1-3.3) 2.2
An even earlier post from April 6, 2011, applied Gompertz curves to make naive (although surprisingly close) predictions about September 2011 ice conditions:
Sep 2011 Sep 2011
NSIDC extent 4.4 4.6
U Bremen extent 4.6 4.6
NSIDC area 3.1 3.2
PIOMAS volume 4.9 4.2
When I wrote that April 2011 post (which became my first SEARCH entry), the idea of applying Gompertz curves to sea ice had been loose in my head for a while, but I was not aware of other publications taking this approach. It seems an obvious idea, however, and I would welcome links to earlier reports.