My PIPS-based June ice extent forecast minimum (dashed line) was off by 11%. Starting in mid-August, persistent winds blew from the south – pushing the ice edge back and melting out the thinner ice.
Once the winds stopped, that same region of missing ice quickly refroze, and brought the ice back to where I expected it to be.
Conclusions:
1. US Navy PIPS is an excellent indicator of ice conditions.
2. Variations in wind from year to year largely determine the relative minimum extent rankings of recent years. The rankings have little to do with the actual condition of the ice.
Arctic Ice is so highly variable making predictions is dicey.
I woudn’t call PIPS excellent just because the curves cross each other twice. But even still there doesnt seem to be a death spiral, like Tamino barks here:
http://tamino.wordpress.com/2010/09/14/death-spiral/
Since these prediction resemble a hurricane cone, is there a way to compare the “centerlines of that cone”?
For this reason SEARCH estimates the September mean, not the minimum. Maybe it would be even better to take the mid-August to mid-October average? If the extent goes up steeply during the next fortnight, according to that criterion your prognosis might even be exact.