DMI puts it in mid-August, while PIOMAS puts it in September! Do we have clues from submarine data or other sources as to which one is right and which is wrong?
During what period (years) does DMI show this behaviour?
The main process at work in summer is melt, heat flux into the ice/ocean system is positive, therefore there is no net growth of ice and no net increase in volume from mid August...
At least in recent years!
It is feasible that in the old Arctic sea ice system there might be some late season growth within the core of the Central Arctic, as the sun's angle drops near the horizon and the large expanse of ice acts to cool the atmopshere over the ice, with low insolation, low angle of incidence it is just feasible that some growth might begin in August. But I doubt it. NCEP/NCAR shows temperatures for 20 to 30 August dropping to just below zero in the early years i have checked, and a bit warmer in recent years, although 2014 had around or below zero temperatures for that period. This could cause some ice growth, but whether this would offset peripheral melt - that seems unlikely to me. Taking freezing point of sea water as -1.8degC, with say a -3.8degC air temperature, that would give a 2degC temperature difference. This would not drive much ice growth.
Note that PIOMAS Central Arctic region has a negative August September difference every year from 1978 to 2014, so it never shows this behaviour of an August start to melt, even in the Central Arctic, the same goes for the CAA. Looking at daily PIOMAS data for the Central Arctic volume, which only goes back to 2000, in 2000 and 2001 the turnaround to growth starts mid September, not august.
None of the submarine based papers I have read indicates the turn around of the seasonal cycle. I doubt if there is enough data to addres this using such data. Which is why I started in my first paragraph with the physics.
