The last 4-6 weeks have been exciting. First, at a time of year when SIE is normally very consistent (narrow 2SD envelope on most charts), SIE has been unseasonably low. Next, in only a week the gap (with record years at least) has largely vanished.
At the risk of damping the enthusiasm, I have a couple of points about this. First, remember no current satellite can actually measure ice area or extent. We infer those based on relative brightness across a selected set of bandwidths/polarizations (which depend on particular orbiting instrument packages and models). The brightness is measured over a moving earth footprint whose size is a function of orbital parameters and antenna geometry.
The graphs we see are a result of a model, whose design (including significant compromises) has been tuned and validated over time to the extent possible against what is known about ice conditions.
To simplify things, assume within the antenna footprint are one year ice, old ice, very thin lead ice, melt pond, and open ocean. Each of these is present in the antenna earth footprint in some proportion, and with some average patch size (relative to the antenna footprint size). I have no personal knowledge but would be astounded if any existing model is calibrated against factors for even these five surfaces. There are, of course infinite variations - thickness, roughness, wetness, and dirtiness of the ice and depth and roughness of surface pooled water are all continuous variables that may be very finely or very coarsely textured across the antenna footprint. Rather, the models depend on them to be balanced against each other in a consistent way (at least for the given time of year).
I offer that what we have likely been seeing is, to a degree, an unseasonable combination of these radiative surface proportions/patch sizes that has, over the last week or ten days, become more seasonable. A likely suspect is a combination of early ice cracking (creating very thin lead ice and eventually, open water in places and extents not typical for the season) and windblown movement. Given the weather, this seems more likely than physical changes in extent (with associated large exchanges between sensible and latent heat - even for flash freeze/ephemeral ice).
A second point to ponder is averages and standard deviations. We know (no matter what "some people" say) that there is a long term trend going on. Except in the middle year of the averaging period, this means that the expected value is NOT the average - for SIE, it will always be less than a past average. This also means that residuals will not be normally distributed, so the calculated "standard deviation" does not mean what we think it does. These combined means that the line we see on many graphs, in relation to plotted averages and SD envelopes, generate emotions that are often not warranted (based on more detailed statistics). 2015 had a minimum that was in the lowest recorded handful, but which was not markedly less than the long term trend.
While SIE has a decreasing trend every day of the year, the daily rate of change is more complicated (it is decreasing over the years on days from mid-May to mid-July, the decrease peaking near the solstice). While SIE is comparatively well behaved this time of year (low computed SD), the daily change rate is very unsettled (high SD).
I noted in another thread I had a lot of work to do before I could say anything about the dynamic behavior of SIE other than it is interesting. I will offer up one early result in closing. Without belaboring details (which I have been tweaking), I've computed residuals against the daily long term trend in SIE. Below is a plot of the fraction of the residuals (variation from expected value) for the September SIE minimum that is (statistically) explained by the residuals on preceding days of the year. NSIDC (actually, Wipneus) SIE today is still about two times the RMS residual value lower than the long term trend for the day. That said, the predictive value of today's (or that three weeks ago) SIE on the annual minimum is, historically, low.
It may or may not be that factors other than deviation from expected value do have predictive value this early in the year - albedo is one that is being examined. In the mean time, all I can say is that the minimum will become clear in weeks and months to come (but not this week or next). For the moment at least it is still safe for Vegas to run a line on the annual minimum, the specific value is still a crapshoot.