Therefore, I experimented with instead normalizing the years to their predicted average extent from a linear regression of all the years.
Your method is skewed by the fact that the long-term extent losses are happening more rapidly in September than in January/February/March.
September extent in the last few years is about 40% lower than in the 1980s, whereas March extent has decreased by only 10%. So your "normalized" March extent has an upward trend over the last few decades, whereas the normalized September extent has a downward trend. So it's not surprising that you get a negative correlation between them. But that correlation is spurious.
A more meaningful method is to detrend the data (see e.g. for some background on detrending). It turns out that the correlation between the detrended March extent and the detrended September extent is very weak: the correlation coefficient is -0.029.
Many thanks to grixm and Steven et al. for wrestling with the numbers. It looks like the jury is still out on this one. While counter-intuitve [more Extent early leads to less 8 months later (Jan-Sept), 7 (Feb-Sept.), or 6 March-Sept.)], the insulating ice theory at least sounds plausible.
But assuming Steven's -0.029 correlation coefficient for March-September Extent is correct, that is low even for noise, and certainly not signal. I find such a low correlation equally counter-intuitive. My naive guess is that there would be some influence of March Extent on September Extent only 6 months later.
Which leaves me even more jaded about Extent as being a flukey measure for status of the Arctic sea ice. It is a real thing, and the most directly measurable, so I'm not discrediting it entirely. But this episode seems to demonstrate that caution is needed in equating the annual maximum or minimum Extent as a precise measuring stick for ASI status and trend. It is what it is, and the long-trend in Extent decline certainly shows us a real effect, but for year to year, or within-year, comparisons it seems too variable to draw short-term conclusions or predictions.
If anybody is up for graphing a detrended March vs. September average Volume, that might be more conclusive. But even that would still be subject to unknown melt season weather. Given the considerable year to year variation we see in Volume also, as shown in the recently updated Wipneus PIOMAS volume graphs --
https://sites.google.com/site/arctischepinguin/home/piomasit seems that what we need is a skilled forecast for melt season weather.
My amateur foray into estimating forecast skill at 6-9 month range for midlatitude U.S. (45 degrees N, 69W) temperature and precip found a little bit of skill beyond climatology out to 6 months for temperature, but nothing worth mentioning for precip beyond about a month. ASI melt weather seems strongly influenced by clear vs. cloudy skies. The inability to forecast precip at a more intensively monitored and presumably better understood temperate mid-latitude location beyond a month suggests that we will not have skillful multi-month Arctic melt season forecasts anytime soon. So we'll just have to wait and see what happens in 2020.
But there's still room for some alarmist notification and unmitigated voodoo. The Wipneus exponential volume trend puts the 2020 Sept. minimum more than 1 million km3 BELOW 2012. I put more faith in the straight line trend, but even that puts 2020 at matching the 2012 record low.
As for the voodoo,
1) the recent low snow cover post by Pavel
https://forum.arctic-sea-ice.net/index.php/topic,2888.msg247615.html#msg247615,
2) the speculation by El Cid about a cold Alaska winter leading to a strong melt season.
https://forum.arctic-sea-ice.net/index.php/topic,2888.msg247815.html#msg2478153) the continuing trend to thinner ice and recent obliteration of old thick ice
4) and the animated ice pack image posted above by Alumimiun - which looks to me like the Atlantic front is already retreating (though that could just be daily variation) -
have got me suspecting that 2020 could have a very active melt season. Maybe that will bring necessary attention to the larger problem.