[Continued discussion of evaluation of linear trend 2003-2013 based on 2013 observation] Another way to look at this is to perform distribution fit tests against the extent residuals. The residual z scores for 2003-2012 fit a normal distribution very well, and we can plot the goodness of fit versus a range of hypothetical 2013 zscores (using Kolmogorov-Smirnov P-Values as goodness of fit): https://www.dropbox.com/s/bwcpfzw29aho7jk/LinearExtentDeclineResidualsGoodnessOfFit_2003_2013.png

@Lodger: I think your discussion points were spot on but I'm cautious about #5: "a Sep 2013 SIE < 2.9 M km^2 or > 4.9 km^2 rejects the linear trend model of SIE decline at the 95% confidence level." I think the linear model over the 2003-2013 span is rejected such less strongly by such a value. We are evaluating the model with the new observation in mind, but it shouldn't have any more weight than any other year in evaluating the model. This means the question is: "What is the probability of at least 1 measurement with a residual > 2stddev in a set of 11 measurements?" This has a probability of about 0.4 Of course, you can look at 2007 and instead ask how likely it is to get 2 measurements with residuals of z >= 2; this is about 0.086. In any case, however, I think that a viable linear-decline model would assume long tails on the residuals for the measurements -- possibly drawn from distributions with indeterminate std. dev. Thus the calculated probabilities of "2 sigma" events are misleading when sigma is estimated from 10 events. Does this seem reasonable?

The data and code used for the analysis in the "Warmest Arctic Summers in 600 Years" study is available at ftp://ftp.ncdc.noaa.gov/pub/data/paleo/contributions_by_author/tingley2013/ in case anyone wants to have a go at creating some non-paywalled visualizations!

I noticed an event in Wipneus' graph of exponential trend broken out by month. The trendlines for July & November are crossing right about now, as June and January did in 2011. This sort of shift in the seasonal pattern is is notable and is clear in the data from the last few years. If the trends continue, the relative seasonal change will be quite large. I certainly think the 0 intercept for June is more plausible than for November. I hope I'm right about that, because zero volume in November by 2019 is hard to fathom. Of course, the same can be said about the entire picture conveyed by the graph -_-

There's a no-paywall copy of the 9 Oct 2012 PNAS article by Harig and Simons at Simons' princeton.edu site: http://www.princeton.edu/geosciences/people/simons/pdf/PNAS-2012.pdf

Ah, sorry, forgot the source for that quote: http://www.the-cryosphere-discuss.net/6/2931/2012/tcd-6-2931-2012.pdf

The mean September SIE (SSIE) reached this year may have particular significance. Massonnnet's work on interpreting model output for future sea ice state (http://www.elic.ucl.ac.be/repomodx/elic/index.php?id=73) has identified an extent threshold that initiates a marked acceleration in SSIE decline across models: "...we find that the CMIP5 models projected anomalies of September sea ice extent (SSIE) (with respect to their own 1979–2010 climatology) are linked in a complicated manner to the 1979–2010 characteristics of their sea ice cover, owing to an acceleration of the trends (and thus larger anomalies) in SSIE, which occurs at different times during the 21st century, but at a mean SSIE of ∼2–4 million km^2."

I've added the 2012 September minimum point to Stroeve's PIOMAS vs. CMIP5 graph: This illustration doesn't appear in the published paper ("Trends in Arctic sea ice extent from CMIP5, CMIP3 and observations, Stroeve et al 2012), but is from a presentation retrieved from http://www.cesm.ucar.edu/working_groups/Polar/presentations/2012/stroeve.pdf

Ack, closing my tag and including the link: http://ww.randcompare.net/content/dam/rand/pubs/research_memoranda/2008/RM6093.pdf The should be right after properties." [Fixed the tag, Ethan. Thanks for an interesting read. 1969? Wow, how prescient. N.]

I don't know if this paper has made the rounds here or not, but it's great reading: Rand Corporation Memorandum 6093-PR / November 1969 "Numerical Prediction of the Thermodynamic Response of Arctic Sea Ice to Environmental Changes" This memorandum's underlying purpose is to evaluate the feasibility of deliberately eliminating the sea ice through environmental modification! So it's a great set of perspectives that are fairly resistant to most skeptical objections about motivation :) Due to computational resource constraints, they had to assume a horizontal homogeneous Arctic (a much more reasonable assumption back then!), but a few key conclusions: -They conclude that to eliminate the sea ice based purely on the first-order response to increased oceanic heat flux would require a 400% increase in Atlantic->Arctic ocean advection -They conclude that the two most effective mechanisms for rapidly eliminating the central pack are albedo reduction and reduced net long wave radiation loss. Check, and Check, thanks to soot and net greenhouse forcing! And there are these two gems of quotes: "The cases discussed in this section suggest that modification of the snow or ice surface is the most effective means of large-scale ice removal. Ideally, the surface should be covered with a substance that reduces not only the albedo and Io, but evaporation and long-wave emissivity as well. However, in addition to the logistical problems involved in such a project, it may be difficult to find a material with a long-wave emissivity substantially less than one [my note: hence, block it in the atmosphere after emission!]. Furthermore, a finely distributed dark solid, like coal dust, would rapidly melt into the ice and lose its effectiveness. The ideal material would be dark, nontoxic , lighter than water, slowly soluble in water, and have low emissivity. A systematic search should be made to find a substance with an optimal combination of these properties." and, in case you thought this all sounds a little reckless: "As man's technological capabilities increase, it becomes more and more urgent that the factors influencing climate on a large scale be understood. For example, it may now be possible for man to remove the arctic ice pack; the result can only be surmised. Another possibility is that in man's increasingly effective tampering with nature, he may accidentally intervene at some critical stage in the climatic process, producing unexpected effects on a world-wide scale. Such considerations cannot be neglected in the formulation of climatic experiments." All in all, a very worthwhile read for a unique and orthogonal perspective on the sea ice from our typical discussion here :)

Re: Beaufort Sea and Yukon coastal SSTs, the NOAA/NESDIS 14KM analyzed field product for the expanded alaska region is showing temps that max out around 10C near the coast and are ~3-6C for most of the area: http://www.osdpd.noaa.gov/data/sst/fields/FS_km14sar00.gif The number of observations used for the current 48hr analysis is high: http://www.osdpd.noaa.gov/data/sst/fields/FS_km14sar01.gif and the average observation age is low: http://www.osdpd.noaa.gov/data/sst/fields/FS_km14sar02.gif These are remotely sensed products, and not in-situ observations, but seem more plausible than the 18C readings. Still really warm, though!