The "extent vs volume" problem keeps coming back. I thought we could use a thread that looks at that problem from an empirical, quantitative angle. There are other angles (which metric is "better" or "more important" or whatever) but I won't address those much here.
In this post I'll use monthly mean data on volume (PIOMAS) and extent (NSIDC). It will be a very long post.
First, let's look at the relationship between June and September monthly means, and use that relationship to forecast the 2017 September mean for each metric:
Fig 1.
Fig 2.
(If I'd waited a week to do this, we could have a more accurate forecast using July instead of June numbers, but for now let's go with this.)
To show them on the same graph, we can scale them to their respective 1980-1989 baselines. In other words, each data point here shows the September monthly volume or extent in year X, as a fraction of the average September monthly volume or extent during the 1980s:
Fig 3.
Clearly the two series seem to be diverging. This is a problem because when volume reaches zero, extent should also be zero.
The trend lines in Fig 3 are simple quadratic extrapolations. Note the following:
- Both the linear and quadratic terms are highly significant for both the volume and extent models.
- One could create alternative versions of this post using other models for extrapolation -- linear, Gompertz, etc. -- but for now I will just use quadratic models in all cases, for consistency.
Each year, when a new September data point is added, those extrapolations change -- sometimes a little, sometimes a lot. Let's look at the history of those changing extrapolations:
Fig 4.
In Fig 4, the left-most data point in each series (2002) represents the extrapolated year in which September ice volume or extent first reaches zero, based only on data through Sept. 2002. The next data point represents the same forecast using data through Sept. 2003, and so on. The end of the solid line is the most recent "real" forecast (using data through Sept. 2016) and the dashed line gives the estimate of what it would look like when 2017 is added.
Clearly, the two series, volume and extent, are giving us very different stories about what the future looks like:
Using PIOMAS volume, a naive quadratic extrapolation suggests that we will have an ice-free September around 2022. There are two important points to remember:
- This is the monthly average, not the daily minimum.
- This is actually zero ice, not some other nominal threshold.
So that is a forecast of literally no ice for the entire month of September, in only five years from now. (It's worth noting that as recently as 2011 and 2012, this PIOMAS-based extrapolation was forecasting an ice-free September in
2017, i.e., this year.).
In contrast, using NSIDC extent, the same quadratic extrapolation suggests that a totally ice-free September won't happen until approximately 2041.
Worse yet, over the past decade the gap between volume and extent has been widening, not narrowing:
Fig 5.
In 2008 there was only an eight-year gap between the two forecasted "zero years", but now that gap is up to 19 years.
Figure 6 summarizes the recent years' forecasts for ice-free conditions in September, with the (expected) 2017 forecasts in solid gold (volume) and solid green (extent):
Fig 6.
It's of course possible that the PIOMAS-based forecast could be completely correct, in which case extent decline would have to speed up dramatically over the next five years.
Alternatively, it's also possible that the NSIDC-based forecast could be completely correct, in which case the volume decline would have to slow down by a lot.
Another possibility is that both series converge on an intermediate date for the first ice-free September. That could happen many ways; two examples are shown in Fig 6. The solid blue dots show weighted averages of the 2017 forecasts -- one using equal weights (i.e., a simple average of the 2017 forecasts from PIOMAS and NSIDC) and the other weighting each series based on its historical variance in forecast dates. Because PIOMAS has less variance in its recent past, it is weighted more heavily, and the "convergence" date is thus closer to 2022 than 2041.
I'm not going to get into all the various arguments in favor of one or the other
(PIOMAS is just a model, not a measurement! Extent neglects the third dimension! etc. etc. etc. ad nauseam). I am personally inclined to doubt the PIOMAS-based "ice-free September in 2021" forecast, for the following reasons:
- Just a few years ago it was predicting that we'd have an ice-free September this year. That seems unlikely to happen. Its prediction for an ice-free September has been moving into the future by one year per year.
- It seems probable that there would be individual ice-free days long before the first ice-free month, and we have not yet come even remotely close to a single ice-free day.
- It seems to me that the volume/extent gap would likely begin narrowing long before we reach "zero ice", but instead that gap has been widening.
Aside from not believing the "ice-free September in 2021" forecast, I'm basically agnostic about the relative credibility of PIOMAS volume vs NSIDC extent.