I usually give you an advance view of my outlook contributions. There is little point doing that now it is available from SEARCH site, but by way of explanation the great Arctic Summer storm was in progress by the time it was ready to submit so Neven blog seemed very busy and the projection was out of date before it was submitted due to that summer storm. So there seemed little point adding out of date projections to an already very busy blog.
Things have quietened down a bit now and the flash-back increase in area has occured on 10 August the same day as NSIDC extent similarly saw an increase. Extent has declined since then so perhaps that is the end of the flash-back increases. So perhaps an update is now in order.
My July outlook contribution gave a prediction of 4.0 m km^2 for NSIDC average September Extent. This used the residual from a gompertz fit of 30 June area to predict the residual from gompertz fit of September average Extent.
The same technique using data to 31 July gave a prediction of 4.02 while using 10th August area data now predicts 3.75 m Km^2
This technique seems pretty good when we are some way from the minimum, however other techniques get better as we get closer to the minimum, namely trying to predict the fall from now to the minimum begins to works better.
So for the August contribution I used a weighted average of the above prediction of 4.02 m km^2 and a new technique. I calculated a weighted average of area and extent minus the extent minimum for each year. A straight line trend through these numbers gave a prediction of 4.41 m Km^2 and a weighted average of these two predictions was 4.23 m Km^2. This put me pretty near the middle of the pack for the August report using data to 31 July.
I used a straight line trend through those weightedaverage-ExtentMinimum numbers because there seemed no useful gain in trying to predict the variation from year to year. Using Area to 10 August and Extent to 12 August, it now seems a useful attempt to predict the year to year variations can be made using Extent-Area as the predictor. This gives a prediction of 3.95 m Km^2 for extent.
Rather than averaging that 3.95 with 3.75 from the first technique, I would prefer to average the 3.95 with 3.72 which has used a weighted average residual rather than the area residual. So my current prediction is 3.86.
Well I have probably thoroughly confused you all by now so I think I will leave it at that. Maybe I will add my predictions for area and volume in the comments....