Updated dogfish assessment

Quang Huynh

February 6, 2024

1

Summary

• Dogfish are long-lived, late-maturing, low fecundity species
• Fit a 2-sex, age-structured model (Stock Synthesis 3) to indices and length composition for the outside stock in BC
• Large fishery in the 1950s
• Notable recent declines in indices despite little retention/discards today
• Models with harvest-only effects on the stock do not fit the indices
• Explore alternative hypotheses to explain stock dynamics

​

​

2

Data�Catch

3

Catch is organized into 5 fleets by gear/retention:

​

• Bottom Trawl Landings
• Bottom Trawl Discards
• Midwater Trawl (no readily apparent difference in size composition between landings and discards)
• Hook Line Landings
• Hook Line Discards

​

​

Very apparent that the high catch (landings) pre-1950 have never been repeated

​

Current TAC = 10k tonnes outside, 5k inside

Data - Fishery Lengths

• Update the fleet structure in the SS3 model
• Length composition show retained dogfish are mostly large females while discards are smaller with balanced sex ratio (pattern can’t be seen by unsorted samples because discarding behavior changes over time)

​

4

Data�Surveys

• Dogfish are caught in all three major surveys
• Rapid declines in the HBLL and SYN indices

​

5

Data

• Coefficients for Julian day (HBLL survey timing) and depth (all indices) are significant, dogfish are moving deeper over time but year effects are much stronger
• Fish are moving deeper in outside waters, but on its own does not explain decline

​

6

Data�Survey Lengths

• Lengths are not measured in the HBLL OUT survey

7

Data�Survey Lengths

• Adjusted IPHC lengths upwards from pre-caudal to total lengths

​

• Discussion on whether to mirror HBLL OUT selectivity to SYN or IPHC survey. Ongoing discussion based on HBLL INS and IPHC sets in 4B

Proceed to mirror to SYN

8

Growth

Updated with NWFSC samples (Model 2) to get better estimate of t0

• Sensitivity analyses removing 4B samples and large females (age is underestimated) (Models 3-4)

Let’s proceed with Model 2

​

9

Growth

Linf < 100 cm is smaller than previous estimates (Macfarlane, Ketchen)

- Large females are explained by high residual SD

10

Maturity

11

​

Maturity at length:

Code 55 = Females with mature gonads

Code 77 = Females carrying pups

​

Use code 77 ogive to calculate spawning output in the model and stock-recruit parameters

SS3 conversion to age ogive is dependent on growth curve:

- Ogive maximum of 0.5 is consistent with life history (2 year gestation period)

�Fecundity

12

​

Fecundity (pups/pregnant female) is a linear function of size (Ketchen 1972):

SS3 converts to age-based values:

Natural mortality

Maximum observed age:

73 for females, 70 for males (4B) M = 0.074 (per Hamel and Cope 2022)

54 for females, 53 for males (outside waters)

80? M = 0.068

​

Proceed with M = 0.074 for both sexes

​

​

​

​

13

Questions

14

Stock recruit relationship

Use Ian Taylor’s stock-recruit relationship where recruitment explicitly can not exceed spawning output

�Illustrative example:

• Unfished population N = 100 adults, which produce 10 pups/adult, that's 1,000 pups (B0)
• Density-dependent unfished pup survival is 0.6, then we get 600 recruits (R0)

​

• A population of N = 20 produces 200 pups (0.2 B0). If steepness is 1 in a closed population, we'd have to conjure 600 recruits from 200 pups. Thus, steepness is capped at 0.33 (200/600)

​

15

Stock recruit relationship

16

Stock recruit relationship

More interesting dynamics:

• Both spawning output and recruitment are in units of abundance. Therefore, the unfished replacement line cannot exceed one
• Given a fecundity schedule, there is an upper bound on possible natural mortality values
• If M is too high, then the spawning output is too low for the lifetime of a female to replace itself
• I can’t bound the unfished replacement line below 1
• Affects MCMC if M is sampled with lognormal prior. Can either sample with upper bound (right figure)
• Proceed with fixed M right now
• Estimate the zfrac parameter (steepness analogue, 0-1 with Beta(1.01, 1.01) prior, almost uniform but avoids bounds)

17

*Assume 50% female pups

Hypotheses for stock history

18

Recent observations: Declining stock, three different index trends in the 21^st century

- Each index as a separate hypothesis for stock trend

​

​

Historical observation: Stock was fished out in the 1930/1940s and recovered by the 1960s

​

Hypotheses for recent trend

19

1. Decline due to lagged effects of fishing
2. Increase in predation
3. Reduced availability to survey (behavioral)

​

​

Hypotheses for recent trend�Set A – lagged effects of fishing�

20

• Fit to each index separately
• None of the three models are satisfactory
• With all three, the steepness parameter hits lower bound (there is no sustainable harvest level, no harvest advice unless steepness is fixed)
• With HBLL or SYN, the stock level is at zero (1e-3)

​

​

Harvest alone cannot explain recent trend in indices with density-dependence in stock dynamics

21

Hypotheses for recent trend�Set A – lagged effects of fishing�

Hypotheses for recent trend�Set B – Increased predation�

22

• Estimate a stepwise increase in natural mortality starting in 2005, progressively add indices

Hypotheses for recent trend�Set B – Increased predation�

23

• Somewhat better fit to indices, cannot explain recent increasing trend in HBLL
• Natural mortality doubles (Recent M = 0.175 in B1, 0.15 in B2-3)
• Curvature in the stock-recruit relationship (h = 0.46 in B1, 0.64 in B2-3)

Hypotheses for recent trend�Set C – Reduced availability�

24

• Model changing catchability since 2012 (random walk)
• Progressively add indices

​

• 70-80% reduction in Q
• No index has constant Q in the model. Stock is increasing but masked by declining Q
• Trying IPHC with constant Q generates Set A models (and their associated problems)

​

Retrospective

25

(A1) IPHC

​

(B2) SYN + IPHC

​

(C1) SYN

Next steps

• Remove models that do not pass first test with regards to fit
• Set A models, unless steepness is fixed

​

• MCMC to explore correlations between time-varying parameters

​

• Likelihood profiling

26

Other model observations

• Recruitment deviations have no effect on SSB. Deviations offset each other

​

• Recruitment signal seen in SYN length comps but the model can’t fit to these small fish (sigmaR = 0.8, upweight length comps)
• The model can’t pick up small lengths as a recruitment signal.

27

Other model observations

• Dome vs. flatsel survey selectivity on males does not appear to affect index trend (need to verify in current models)

28

Other model observations

• Estimated sex ratio in older age classes predominantly male. Fishery landings have targeted and fished out the females
• Can we verify with survey length data? Growth difference and selectivity (behavior?) could be confounding

29

Hypotheses for stock history

30

Historical observation: Stock was fished out in the 1930/1940s and recovered to B0 by the 1960s

​

This could not have happened in current models, constrained by low fecundity (on average 7-8 per mature female)

​

Need more than just high survival of pups to recover stock quickly, e.g.,

• More pregnant females at low stock size

​

• Lower M at low stock size (Wood 1979, can explore this in Stock Synthesis)

​

​

Hypotheses for stock history

31

Historical observation: Stock was fished out in the 1930/1940s and recovered to B0 by the 1960s

​

The incoming recruitment to age 20, 30 did recover. Presumably, these age classes would be noticeable to the fishery

​

​

​

Selectivity

32

Black line = maturity

Questions

33

Fits to lengths

34

Fits to lengths

35

Fits to lengths

36

Likelihood profile of zfrac (A1 model)

37

Notes

• Large biomass moving up annually (Dan) since 2012, stopped going into the sound. Segregating by sex below 40 fathoms. Something fundamental changing since 2012.
• Don’t inhabit hard-bottom unless there’s a large biomass
• Very migratory based on tagging (see Mcfarlane and King)

​

• Post-release mortality. What do managers assume?

​

• Sablefish survey? Deeper, autumn but traps tend to avoid dogfish

​

• Cannibalism as a mechanism for density-dependent M (Dana)

​

• Explore priors for steepness (Nick)

​

• Growth curve (Philina) and converting to fecundity at age schedule. Survey sees lots of 100+ cm females. Younger fecundity schedule as a sensitivity scenario (Nick)
• Mirror HBLL sel to IPHC, but different trends due to habitat (sand + mud, complement of rock)?
• Recreational and survey catch (low but there’s not much other sources of mortality)
• Female only model (Dana)?
• Bycatch CPUE as an index (Dana)?
• Dan to ask for fishery timing with Brian Mose and Bruce Turris

38

Data�Survey Timing

Thinking about whether survey catchability is changing due to seasonality in dogfish behavior/migration..

​

Haida Gwaii: Even years, August/September

Hecate Strait: Odd years, May/June

QCS: Odd years, July/August

WCVI: Even years, May/June

​

HBLL N: Even years, August/September

HBLL S: Odd years, August/September

​

39