Some happy amateur statistics points to the discussion about ice melt projection functions:
1) While area and extent data are more available and also more visual, they say less than volume about how much melt went on in that particular season. True, the 'target' is zero for all data groups, but the physical (3D) reality is best represented by 3D volume numbers. (This would of course change a bit when/if all areas become thin 1–2 m ice.) Melting work – extremely energy requiring – has to be done on the 3D ice, although the area where work can be done (from above and below) is roughly the same for 2D and 3D ice.
2) Demands and requests for elaborate physical models to predict the first Zero Ice Day (ZID?) are frequent, but are we sure they are a) possible or even b) better than a purely mathematical function, at describing the current and future ice melt trend? I would lean heavily towards a 'no' on both accounts. Recent developments in nearby Iceland as well as the unpredictability of the ENSO and their aftermaths, suggest that a fully accounting physical model for the Arctic Ocean melt trend is next to impossible. At best, such a formidable and laborious project would also have to adjust its projections following yearly major events, just as the mathematical whole–system functions will have to take new data points into account for updated projections of the elusive ZID.
3) Already mentioned by others, but a purely mathematical function with its inherent confidence intervals is not proven wrong by one or two data points beyond those confidence intervals. The function doesn't say no data can deviate from the projection, it merely says this is the area where you'll most likely find future data points. Also, any model or function, like my own 3rd Degree Polynomial function, must be expected not to be correct at first run, say, in August 2014, but should rather be expected to be the best available projection for an August or September ZID at the point of publication, with further updates and adjustments along the way. The fact it allows for easy inclusion of new data points that can then quickly and with no doubt or hesitation whatsoever be drawn upon to produce *even better* projections, is in fact a strength, not a weakness. The fact that a 9 year old can run the model off of a future web page and get updated and accurate ZID projections also suggests robustness. Simplicity and whole–system nature means fewer things can go wrong. (If Iceland goes BANG, it will be a new datapoint, or several, yet it's in no way a crisis for this model. It's by definition adjustable to unforeseen events.)
4) September 2012 only contributed 1.1% of the total ice mass — or 259 km³ – to the 2012 yearly melt, or minimum. 259 km³ melted in half a day in June of 2007, in less than 12 hours. 99% of the 2012 melt work was already done by the last day of August. This suggests that the first ZID event may well come in August, following an exceptionally successful start of the melting season in a way hotter year than 2012, as these outlying years go. If the overall meltdown % is raised all the way from 2012's 84% of the Winter Maximum up to 99 or even 100% — in such an extraordinary year — it is also highly likely that no more work is needed on the ice by September 1st. My simple math model suggests August & September will crash in the same year, or two adjacent years. I think August will say 'Oi! Join the club!' to September.
Your thoughts on all of these, please!
