Last time I promise...
ugh, I now realize that I can't use whole-Arctic values for linear regression because as peripheral seas melt out they can no longer contribute to the ice loss rate. Thus a trend based on earlier years when those seas contributed to year-to-year losses is invalid because doing so contributes fictional negative volumes into the trend.
The correction is to focus on the CAB trend only. But even that combines earlier melting of the peripheral CAB areas with the areas where the final 1M km2 ice extent (the consensus threshold for BOE) is likely to occur. The best approach I can think of is to partially correct for that by ignoring CAB data from 1979-1999 or 1979-2004 to discount the earlier years when the peripheral CAB areas still had September ice, and thus base the trend on losses after those areas were no longer contributing to the year-to-year losses.
I looked at CAB Sept. average volume linear, polynomial and logarithmic trends for 2000-2019 and 2005-2019. Starting in 2005 gives a flatter slope, and a lower R2 than starting in 2000. Defining BOE threshold as 0.8K km3, the estimated date for when there is a 50% cumulative chance for having had a first year of BOE is 3.5 years earlier when using the 2000-2019 data. The estimate for 95% cumulative chance of BOE is 5.3 years earlier using the 2000-2019 data. And the date for when each Single year has a 50% chance of being <BOE threshold is 6.2 years earlier when using the 2000-2019 data.
I also looked at CAA as a potential last refuge area, which was interesting because it brought to my attention the fact that in 2012 the CAA Sept. volume was almost down to zero. But then I found this map of the where the last 1M km2 ASI extent is expected posted by Tor Bejnar at
https://forum.arctic-sea-ice.net/index.php/topic,417.msg122285.html#msg122285 Comparing that "final ice extent map" to the NSIDC Arctic seas boundary lines at
https://nsidc.org/data/masie/browse_regions shows that the final ice refuge is almost (perhaps entirely) within the NSIDC defined boundaries of the CAB. In addition, the CAA trend has it blinking out shockingly soon. So the estimates below are derived from the 2005-2019 CAB-only volume trend. The CAB volume data are from PIOMAS, which defines the CAB boundary somewhat differently than NSIDC, but that should not affect the resulting date estimates.
I am using 0.8K km3 as the volume threshold proxy for BOE. If you don't like that, we all at least agree that 0 volume means zero extent. This is the most demanding possible definition for BOE. Using a 0 volume threshold only delays the 50% and 95% cumulative chance dates, and the 50% Single year date, by 3.0 to 3.3 years. Thus, whether we define BOE as 0.8K km3 or Zero km3 doesn't make much difference.
Enough fussbudgetry, here are the numbers for my latest and last attempt to use linear regression to estimate BOE dates.
*************************
DATA: PIOMAS Central Arctic Basin (CAB) September average volume estimates based on linear trend from 2005-2019. Threshold: 0.8K km3 as an arbitrary definition of BOE based on volume.
Values rounded to the nearest whole percent.
Year Chance of that single year Chance of at least one year since 2019
going below 0.8K km3 having gone below 0.8 km3
2020 0% 0%
2021 0% 1%
2022 1% 1%
2023 1% 2%
2024 2% 4%
2025 3% 6%
2026 4% 10%
2027 6% 15%
2028 8% 22%
2029 12% 31%
2030 16% 42%
2031 21% 55% First BOE more likely than not by 2031
2032 28% 67%
2033 34% 78%
2034 42% 87%
2035 49% 94%
After 2035, each single year has >50% chance of BOE
2036 57% 97%
2037 65% 99%
2038 72% 100%
2039 78%
2040 83%
Finally, several points that seem worth repeating.
1) Some of the difference in intuitive impression for when the ASI will reach BOE status is due to the difference between the
cumulative chance of a first year BOE (almost certainly to be followed by subsequent non-BOE years), and when each
individual year has >50% of going below a BOE threshold.
2) As experienced ASIF members have commented, melt season weather will determine when the first BOE occurs, and the Arctic may have already reached a state where BOE is within reach of an extremely warm and significant cyclone year. But the trend analysis shown above estimates that ia first BOE year s unlikely until after 2030.
3) Scientists who study ASI decline say that any such estimates are subject to +/- 20 years error due to high variability in the ASI! (In other words "You don't know, and we don't know either!").
4) BOE is an arbitrary marker, and the world won't end when there is a first BOE. But it serves as an important observable and easily understood marker with visual impact to communicate climate change progression. As the late-summer ASI ice decline continues and eats into August (which trails September by only a few years) and into July, the effects on albedo, weather, ecosystems etc., and the consequent impacts on human civilization, will increase.