Herring Scrap 26: Unfished Biomass

In the last few weeks, Alaska Department of Fish and Game published an extraordinary piece of statistical storytelling about herring in Sitka.

It's called:

[fanfare please]

A Simulation Study to Estimate the Unfished Biomass of Sitka Sound Pacific Herring

[disappointed groans]

I'd like to tell you about it.

(Average) Unfished Biomass is a modern fisheries idea used as a reference point and baseline to represent what an average abundance of something would be absent commercial fishing pressure. One figure to serve as a notional average of pristine abundance for all years. If there is no data record for such conditions because all data has been collected in the context of ongoing fishing pressure, then a simulation must be performed in order to determine what such a level might be.

This new study - you can snag a fresh pdf copy of it here - marks the second time that Alaska Dept of Fish and Game (ADF&G) has calculated a figure to serve as the Unfished Biomass for Sitka herring, to describe what an average year herring population size in the Sitka herring management area would be in conditions absent fishing pressure. The first study was done back in 1997 using similar methods.

Why?

In terms of the numeric mechanisms that guide the implementation of the fishery for ADF&G, the way that the Average Unfished Biomass comes into play is (currently, at least) sort of subtle. The Unfished Biomass is associated with what's called the threshold level - the level below which fishing will not be allowed. When the Average Unfished Biomass study was conducted in '97, ADF&G's suggestion was that the no-fishing threshold should be set at 25% of the AUB. As a slight conservation measure, the Board of Fisheries that year set the threshold level higher than that 25% recommendation (at 20,000 tons) and later it was incrementally increased again (to 25,000 tons). Since, biomass assessments have perennially far exceeded that threshold level and thus no fisheries have been halted by such a mechanism.

That threshold is further implicated in the doings of the fishery because the exact percentage of the biomass which is allowed to be harvested in a given year will fluctuate based on the degree to which the biomass is thought to exceed the threshold. And so, the AUB is conceptually linked but not mathematically tethered to the threshold level, and, by extension, the allowed Guideline Harvest Level for the commercial fishery. But, in all of those formulae, AUB is invisible, and its linked but ultimately arbitrary proxy, the threshold level, is doing all of the work. From it's direct absence in any of the math involved in the fishery, we can understand that the Unfished Biomass primarily exists as a storytelling device.

OK, so how?

In order to fabricate a figure to serve as the Unfished Biomass, ADF&G fed their data from population assessments done from 1980 to 2022 to a computer, and then resampled and remixed that data 1000 different ways in 30,000 year simulations - and found the average. What they came up with is that the Unfished Biomass of herring in Sitka Sound is 85,576 tons. That figure will replace the old Average Unfished Biomass (67,036 tons) calculated in 1997.

One of the reasons that they described for reassessing the Average Unfished Biomass was their observation that "for 14 of the 26 Sitka Sound herring sac roe seasons since the Alaska BOF decision in 1997", "The hindcasted estimates for spawning stock size from the current 2023-forecast ASA model exceed the average Unfished Spawning Biomass estimate" from 1998. The report continues, "While fished biomass can exceed an average unfished spawning biomass in one or more years and not indicate that the carrying capacity of the ocean has changed, if estimates of annual fished biomass persistently exceed average unfished spawning biomass over many years, then an analysis with updated data like that provided here is valuable to assess whether the carrying capacity of the ocean has changed."

And so they went ahead and updated their methodology for the study and came up with this new Unfished Biomass figure - 85, 576 tons (and they aren't calling it Average Unfished Biomass anymore - it's just Unfished Biomass now). The thing is, this new Unfished Biomass is still lower than ADF&G biomass assessments for 13 of the last 20 years. Going into 2024, ADF&G forecast a spawning biomass of 406,228 tons - nearly 5 times more than their new Unfished Biomass figure. Put another way, ADF&G statistics tell a story of a herring abundance in Sitka Sound that has exceeded an Edenic average for most of the 21st century so far – despite a rise in commercial fishing pressure in the same timeframe.

To demonstrate the degree to which this is the story being told, I want to share one of the figures from the new Simulation Study. Going into 2024, ADF&G was forecasting a Sitka spawning herring biomass of 406,228 tons. If you compare that with the simulation used to determine Unfished Biomass, you can see how rare an event this abundance supposedly is; there are only 5 or 6 occassions in the 30,000 year simulation which exceed this year's supposed biomass:

And I think, faced with this scenario, you can prefer one (or two) of three initial hypotheses at this point: a) the Unfished Biomass figure is artificially low, b) contemporary assessments of abundance are artificially high, or c) these are extraordinary once-in-10,000-year times for Sitka herring abundance.

Given that I 1) trust and believe the strong observational voice in the public record that abundance has not felt that good in recent decades; 2) trust and believe the common sense that killing 10-20% of spawning herring every year for decades will not regularly provide pristine-like conditions; 3) have not noticed anything about herring presence in Sitka that seems qualitatively exceptional compared to various descriptions from other times (including in the last half-century); 4) know that the Sitka Tribe of Alaska did an unfished biomass using methodology faithful to the original 1997 study for the 2022 Board of Fisheries cycle and came up with an Unfished Biomass much higher than the ADF&G estimate; and 5) recognize Sitka as a place that has been known for spectacular herring abundance since time immemorial... I reject conclusion C, reserve judgement on conclusion B, and select conclusion A: the Unfished Biomass figure is artificially low.

Meanwhile, ADF&G seems inclined towards conclusion C.

ADF&G and I see the same weird feature - gosh, this AUB (whether the new one or the old one) is lower than the herring population in most years nowadays, how exceptional - but we reach different conclusions about what that means. In explanation, ADF&G offers that perhaps ocean conditions have been more productive for Sitka herring than ever before in the last quarter century. I offer that ADF&G has a poor grasp of history. I really think it's a superior explanation!

But before I can even begin to defend that position, we need to take a detour.

How to Count 27,000,000,000,000 Eggs Underwater

I've just realized that I haven't yet described in any of these herring scraps how ADF&G comes up with their herring biomass estimates every year; it isn't actually by counting herring; it's by counting their eggs.

Here’s how they do it: throughout the season, daily so long as conditions allow, the Department flies over area shoreline - a really quite substantial area, a couple-few hundred miles of coast - and marks the sections of shoreline receiving spawning activity, apparent from the sky as the water turns a milky color from the milt. Then, 10 days or so later, a team of divers goes out and dives transects on some of those places: at increasing depths along each transect line, they estimate how many eggs there are in a 10x10cm area, extrapolate that to the square meter around it, and move on down the transect that way, and then on to the next transect. The collective findings from all of the survey transects are then averaged and applied to the full length of the spawning grounds, providing an estimate of the total number of eggs in the Sitka herring fishing area. In 2021, for example, they figured 27.302 trillion eggs were in the survey area - their highest such estimate ever.

That number of eggs is then compared to the herring samples taken in the midst of the fishery (assessing the proportions represented by each age class, and how heavy each age class is on average), and then using a nearly-arbitrary fecundity-at-weight relationship to guess how many eggs are contributed by fish of various sizes, and from there, how much they collectively weigh. After applying some assumptions about survival and recruitment between fishing seasons, a forecast biomass for the next year is produced. From the biomass, the Guideline Harvest Level can be derived, with up to 20% of the biomass allotted to the seine fleet each spring.

It’s an adventurous, meticulous, gas-guzzling, expensively efficient performance of regulatory science. In a lot of ways it’s a pretty reasonable and (now) well-developed way to attempt a guess at how many herring are spawning in a given area, if that’s what you want to know.

But it's important though to recognize how effort- and technology- and HR- and fuel-budget dependent that method is. Areas where spawning occurs but is not observed will not be counted, and thus the regularity and range of aerial surveys, for example, matters. The experience, confidence of the divers and the degree to which they are resourced matters. The goals that they conduct their work with matter. Where they dive matters. What the computer model does with that data matters. What shortcuts are taken along the way matters. And advancing technology sure helps, every step of the way. I'll point out some of those details in a follow-up, and others in others. But ADF&G doesn't mark or adjust for all of those reasons and others that the biological surveys of this nature are likely more thorough now than they used to be.

As a consequence, the historical data in the 1980-2022 time series which informs the model's 30,000 year simulation offers a distorted fun-house view of real biomass over that time span. As I've posited elsewhere, and will describe more fully as I go, the historic time series about herring abundance is contaminated by the growing intent and efficacy of scientific effort, among other problems. The exact degree to which the numbers are distorted for any given year is very hard to pinpoint, but it is an obvious problem that ADF&G pretends doesn't exist and doesn't adjust for.

Ok, you want proof. Here's where we can start, with something that, as I understand it, John Littlefield was talking about in Sitka back in the early 2000s (Patrick et al 2009): From the 1980's into the 1990's, biomass was assumed at an arbitrary 500 tons per nautical mile of spawn.

Conveniently, for those of us who like simple math, it was further assumed that 100,000,000 eggs represents one ton (2,000 lbs) of spawners.

This means that both the biomass and and egg-deposition estimates are derived, for those years, from nautical miles of spawn (aerial surveys) using simple linear relationships.

By extension, every nautical mile of spawn for those years is assumed to be made up of 50 billion eggs, and twenty miles of spawn means one trillion eggs. It works out that if you divide nautical miles by 20, then you get the same number as the estimated number of eggs (in trillions) for several years early in the time series. Here's what ADF&G's data on the scaled nautical miles of spawn (divided by 20) looks like charted alongside their egg deposition estimates (in trillions):

In the course of that chart, we can see the egg deposition estimates - which are a proxy for biomass - wriggle free of a bad mathematical relationship over time.

The relationship loosens and changes in 1986 and then some more around 1992. For the years up to that point, at least, two conclusions are inevitable: 1) there is not quality egg deposition data for those years; 2) The actual scientific/observational content of those years is minimal, and those biomass estimates were wild guesses. Wildly conservative guesses, if egg deposition estimates since are any indication.

And it's those wildly conservative guesses from outdated research that are (in part) dragging down the Unfished Biomass in this wacky study that ADF&G has accomplished. In the next scrap, I'll pick up on why it matters.

Thanks for reading!

Peter

P.S. I picked up that scale-by-20 trick in an unpublished report for STA from 2009 by Patrick, Brown, Kiefer, Meuret-Woody. They were trying to figure out this whole mess back then. In that report, they actually divide by 17.7, which works for some of the late-80's ye for reasons that I don't want to dig into here. Anyhow, the authors of that report credited John Littlefield of Sitka for insisting back in the early 2000s that this was an important detail in understanding the precarity of ADF&G's historic claims, and so I want to extend that credit here. I feel sure he was right.