In forecast mode (using 1992-2011 data), for 2012, my method still predicts 3.83, essentially the same as in hindcast mode (3.82). The final extent was 3.62, so a reasonable prediction.
In forecast mode (using 1992-2006 data), for 2007, my method predicts 5.06, slightly higher than the 4.77 prediction in hindcast mode, but still predicting a record low. The final was 4.29 which is quite a bit lower than either forecast or hindcast mode predictions.
So, the first (for 2007) showed .48 lower actual, the second (2012) .20 lower actual. Let's assume your forecasts aren't including
methane and some other positive feedbacks with exponential forcing character for melting that should be included now rather than in 5 years, like with the IPCC, I think it's fair to say your prediction for 2017 could be about ~.8 off actual.
Taking that on board, for 2017, you predict 3.85 M km^2 NSIDC 'area' in September. I say that may well end up being no more than 3 M km^2 area in September.
Look at global SST's again, using the newer more accurate satellites, for July 23rd in 2016;
and 2017:
Note how the El Niño effects have barely, if at all, "disappeared". If anything, sea surface temps have gone more dense in heat overall, in the Mediterranean too. Looking at this global scale doesn't make it really clear, but heat has moved about 100km polewards, on both sides. This heat will exchange with surface air, move around, and ultimately melt ice.
Thickness and volume will, in my view, derived from first-hand work near Шпицберген 2 months ago, reach an absolute record low this year. And, like I wrote earlier, my 51 years on this planet tell me we've seen Peak Ice globally. For trustworthy predictions, I think we need to look at total energy in atmo. The energy has to be and go somewhere, when it's not radiated out to space. (Adding
4 Hiroshima bombs per second is not something to ignore.)