As usual, I come to the conclusion that by far the best explanation of the behaviour of arctic ice, when it melts and how fast it melts, is found by looking at the changes in temperature ... etc."
Moving on to Notz and Stroeve 2018, it says
"The sensitivity of Arctic sea ice as described by the linear relationship between global-mean temperature and Arctic sea-ice coverage has been found to remain constant in model simulations across a wide spectrum of temperature trajectories. In particular, the linearity holds in all CMIP5 models until summer sea ice vanishes in individual simulations. Hence, the observed sensitivity can be extrapolated to directly estimate the response of the Arctic sea-ice cover to future warming."
After describing some complications, they go on to say:
"Despite these uncertainties, the different estimates result in a relatively narrow range of additional warming above present that is required to obtain a near-ice free Arctic Ocean during summer, defined as the total sea-ice coverage dropping below 1 million km2."
Then they discuss specific temperatures, and variation around those, from which I derived a median estimate of 1.68 +/- 0.25C (95% CI) above the (inferred) 1850-1900 NASA GISS global average land and ocean surface temperature. (inferred because GISSTemp doesn't start until 1880. Details to translate to 1850-1900 equivalent are trivial and don't affect the stated value).
And for those of you keeping score at home --- as of 2019 the running 5-year average GISSTemp is at +1.15 C. In earlier post, I used observed recent GISSTemp trends compared to the 1.68 +/- 0.25C to make the year estimates for when Sept gets below 1M km2 Extent.
The 1.7C estimate fits with their statement:
"As soon as the global-mean temperature has risen by slightly below 2 ◦C, the Arctic Ocean is expected to be on average nearly ice-free during September."
and slightly out of context, but still relevant:
"A possible modification of these estimates might be caused by the future evolution of anthropogenic aerosols, as they are expected to become less abundant over the next few
decades. "
"... in climate-model simulations the expected aerosol reduction causes additional ice loss..."
"This would imply that the estimates given here are too conservative."
And finally, they remind us that:
"While the observed linear relationship between sea-ice coverage and global-mean temperature allows one to estimate the long-term average future evolution of the pan-Arctic ice cover, the evolution of the real ice cover will show substantial year-to-year variability because of internal variability."
-----------------
So it seems to me that binntho and crandles are both right!