New Multistock HCR

[CamFreshwater]

My notes are a bit jumbled, but my understanding is that at the end of the last meeting we had decided to incorporate a new, more generic multistock harvest control rule that could be used to evaluate the relative benefits of more restrictive single-stock fisheries. I believe the goal was to avoid something too Fraser-centric (i.e. TAM rule), but make a comparison that was more valid than comparing fixed exploitation rates in mixed-stock fisheries to more nuanced single-stock HCRs (i.e. apples and oranges).

I also believe we'd settled on following the rough outlines prescribed by the PA document, which Carrie forwarded to me after the meeting. The generic HCR in that document is as follows:

Provisional Harvest Rule In absence of a pre-agreed harvest rule developed in the context of the precautionary approach, a provisional removal reference or fishing mortality (F) could be used to guide management and to assess harvest in relation to sustainability. The provisional harvest rule is as follows: When the stock is in the “Healthy Zone”: F < FMSY When the stock is in the “Cautious Zone”: F < FMSY x ( (Biomass – 40% BMSY ) / ( 80% BMSY − 40% BMSY) ) When the stock is in the “Critical Zone”: F = 0

For our purposes I would modify this as: When the aggregate is in the “Healthy Zone”: Ftarget = FMSY When the aggregate is in the “Cautious Zone”: Ftarget = FMSY x ( (S – Sgen ) / (Smsy − Sgen) ) When the aggregate is in the “Critical Zone”: F = 0

As before we would calculate a total TAC based on the Ftarget and, depending on the MP, allocate a fixed proportion to mixed- and single-stock fisheries. So instead of examining how single-stock fisheries performed across different target ERs, we would be testing whether more conservative single-stock fisheries out-performed a plausible (but less risk averse and less complicated) version of the TAM rule.

Questions

1. As I mentioned, my notes are less complete than I remember them being at the time! Does this sound correct to all of you? If not how would you change it?

2. This is feasible to code, but will take a few hours. Before I do so I'd like to be sure that this is what we want rather than just going with the simpler option of using the TAM rule that's already coded. If I was to speculate, this frame will result in higher fishing rates than the TAM rule and hence greater contrast between mixed- and single-stock fisheries. That plus its simplicity relative to the TAM rule are why we're going this route, correct? Or have I forgotten something?

3. Assuming 1 and 2 make this a go, the practical issue is deciding what Smsy, Sgen, and Fmsy look like at the aggregate level.

• Should the aggregate be an MU or all 4 MUs?

• Option 1: fit a SR model at the aggregate MU level to estimate alpha/beta (hence SR benchmarks and Fmsy). This is complicated by different TS lengths and may be somewhat nonsensical.

• Option 2: Use the mean Fmsy among CUs (perhaps weighted by abundance) as the aggregate Fmsy. Smsy/Sgen could be calculated as the summed abundance of these BMs within the aggregate (or MU). These could take into account Larkin BMs, in which case they would change annually (complicated) or we could only use Ricker (simpler).

• Option 3: Avoid SR BMs altogether and use percentile BMs frozen at the normative period (simplest). Fmsy would be calculated as above.

Any other ideas or suggestions? Once a few people weigh in I can code this up directly, but like I said I don't want to wander alone into this rabbit hole ;)

[carrieholt]

(1) Yes. If the PA HCR tends to provide higher target harvest rates (and is less risk averse), then this would be a good contrast to the current TAM which has emergent HR ~30%. For the PA HCR, should the min HR be revised upwards from 0 to 10% (level aligned with min plausible HR for Fraser sockeye?). Either way, we should explicitly state in what ways it differs, to help identify why results differ between PA HCR and TAM rule. (2) I think so. Kendra and Ann-Marie do you have any other recollections from the meeting? This will ground the analyses in a HCR that is meaningful to readers, and provide contrast with TAM, provided they do give different results! (3) I think option 1 makes most sense as it requires the least additional assumptions (with shortening time-series to shortest stock). We present this HCR only as a hypothetical example HCR, so the shortening shouldn’t be a problem. This HCR ignores completely sub-stock structure which is fine given it’s context in our analyses. The second option integrates single- and mixed-stock considerations, requiring summing across CUs. With this option, CU-specific spawner abundances in a given year may be a far below CU-specific benchmarks with the aggregate being above the aggregate benchmark. I find this option confusing since the objective of the mixed-stock fishery is to harvest the aggregate without consideration of individual stocks, and I don’t think the sum is a meaningful way of integrating (but see option 4). Option 3 requires additional assumptions (which OCPs to use, etc) and explanations which I think we should avoid. An alternative option 4 is to run a logistic regression between probability that all component CUs > SMSY and the aggregate abundance, to estimate the aggregate abundance associated with a high (95%) probability of SMSY being achieved for all components (as for IFR coho recovery objectives, and Steelhead recovery objectives). However, this also integrates single and mixed-stock concerns in the development of OCPs for the aggregate, mixed-stock fishery (which is confusing in our context), but avoids one of the problems of option 2 described above. If options 2, 3, or 4 are chosen, then yes, FMSY can be weighted average of components.

Thoughts @krHolt @ann-marieH ?

*Cam, Steph Peacock used this approach for “true” status of the aggregate in her report, which I forgot to highlight. Any insights from this issue/responses would be appreciated in your review. This option likely provides a minimum estimate of aggregate benchmarks/OCP compared with option 4, but may be slightly higher than option 1. At the very least these caveats/issues should be documented in the report.

[krHolt]

My responses, numbered as per above:

(1) Yes, I agree with what you suggest. My initial thought was to use 0.8Smsy and 0.4Smsy as the upper and lower benchmarks to line directly up with the 0.8Bmsy and 0.4Bmsy identified in the policy. This combo would be less risk adverse than the Smsy and Sgen combo used for Wild Salmon Policy, and would provide an even stronger contrast in harvest pressure against the TAM rule. Although, now that I think about it more .... while we could cite the PA policy as a rational for the more aggressive 0.4 and 0.8 Smsy (and hence, justify more contrast), perhaps this goes against recommendations from Carrie's initial WSP simulation studies? Carrie - did you consider 0.4, 0.8Smsy? Do you think we need to stay tied to the WSP at this aggregate level?

(2) Yes, I think we should proceed ... regardless of the outcomes of my above questions on the specific benchmarks.

(3) Maybe MU-level to be consistent with TAM??? Although, I don't have a strong opinion. Option 1 (summing all CU data for a giant aggregate SR analysis) seems like a good option for our purpose. Carrie's Option 4 is a better way for determining aggregate level benchmarks; however, I agree that it adds a whole other dimension to this paper that is likely outside of the intended scope.

[ann-marieH]

I have no notes and very little memory of this convo, but here are my thoughts, anyways:

1 & 2. generic HCR:

• a tested, generic HCR that is "WSP & PA compliant" would be super useful to have as a built-in option for the recovery tool.
• for a paper, it seems like it would be more useful to stick with generic HCR than TAM? TAM has scads of non-reproducible, non-quantified value functions that went into their selection. generic HCRs would be reproducible
• no comment re: using variations of Fmsy/Sgen/Smsy/etc., as I don't have much experience working with them - I defer to you all!

3. aggregate Qs:
   • I'd suggest focusing on 1 MU
   • the individual stock interactions are going to be difficult enough to de-tangle within one MU. Modeling 19 stocks will be interesting and useful for FR SK, but less so for any other group of fish.
   • suggest either ESum or Sum, with slight lean towards Sum b/c:
   • EStu = 1 CU only
   • Lates = LShu will drive everything and I'd hazard to say that anything to be gotten from an analysis of LShu will only be applicable to LShu. And Cultus is a bit of a hot button that it would be better to stay away from at the moment. Plus modeling it is extra problematic b/c of the hatchery influence.
   • ESum = 7 stocks (to Summer run 6 stks); shorter time series of data than most Summers. I suspect Scotch S/R curve is still developing, as the ln(R/S) vs S resulted in positive slope up until a few years ago. extra bit of explaining re: Sco-Sey as 2 stks vs 1 CU req'd
   • Summers = have both cyclic & non-cyclic; red/amber/green CUs all present; contrast in the stock abundances, but not too crazy of a contrast
   • how to aggregate - good question! (and actually, that would be a really useful thing to look at from an implementation perspective, I think. Not sure if it would make for a good primary publication?)
   • i.e., could we contrast methods of creating a generic HCR? e.g. choose an HCR that falls into each of the following categories/objectives and then compare their performance:
   • method A: ignore sub-stock structure altogether (e.g. Cam's option 1, using only the years where all stocks have S/R data, per Carrie's suggestion)
   • method B: acknowledge component stock contributions (e.g. take relative abundance of each stock into account in some way.)
   • method C: scale down to take into account the stock(s) of concern (e.g. apply the HCR that would apply to the red(dest) CU) ...will mull over actual proposed methods in a bit.

[CamFreshwater]

Okie dokie, lots to unpack here, but for now I'm going to work on developing a generic HCR based on MU level abundance with a particular focus on Summers. I'll use the lower ER as 10% (I think 0 is unrealistic given bycatch, FSC, etc.) and stick with values of Sgen and Smsy calculated from summed spawners and recruits (i.e. option 1) within the Summer MU. If we'd rather use 0.4 and 0.8 Smsy I'm happy to do that too.

I won't have the time to start implementing this for a few days, but once I do I'm sure other issues will come up and I may come begging for advice!

[carrieholt]

I do suggest using 40% and 80% of SMSY as the OCPs in the PA-compliant HCR. I meant to state this in my original comments but forgot (and thanks to Kendra for bringing it up). WSP benchmarks are meant to conservation thresholds not OCPs.