Canadian Science Advisory Secretariat (CSAS) Research Document 2022/068 Gulf Region December 2022 Assessment of the NAFO Division 4TVn southern Gulf of St. Lawrence Atlantic Herring (Clupea harengus) in 2020-2021 N. Rolland, F. Turcotte, J.L. McDermid, R.A. DeJong, L. Landry Fisheries and Oceans Canada Gulf Fisheries Centre 343 Université Avenue, P.O. Box 5030 Moncton, NB, E1C 9B6 Foreword This series documents the scientific basis for the evaluation of aquatic resources and ecosystems in Canada. As such, it addresses the issues of the day in the time frames required and the documents it contains are not intended as definitive statements on the subjects addressed but rather as progress reports on ongoing investigations. Published by: Fisheries and Oceans Canada Canadian Science Advisory Secretariat 200 Kent Street Ottawa ON K1A 0E6 http://www.dfo-mpo.gc.ca/csas-sccs/ csas-sccs@dfo-mpo.gc.ca © His Majesty the King in Right of Canada, as represented by the Minister of the Department of Fisheries and Oceans, 2022 ISSN 1919-5044 ISBN 978-0-660-45821-2 Cat. No. Fs70-5/2022-068E-PDF Correct citation for this publication: Rolland, N., Turcotte, F., McDermid, J.L., DeJong, R.A., and Landry, L. 2022. Assessment of the NAFO Division 4TVn southern Gulf of St. Lawrence Atlantic Herring (Clupea harengus) in 2020-2021. DFO Can. Sci. Advis. Sec. Res. Doc. 2022/068. xii + 142 p. Aussi disponible en français : Rolland, N., Turcotte, F., McDermid, J.L., DeJong, R.A., et Landry, L. 2022. Évaluation des stocks de Hareng Atlantique (Clupea harengus) de la zone 4TVn de l’OPANO dans le sud du golfe du Saint-Laurent en 2020-2021. Secr. can. des avis sci. du MPO. Doc. de rech. 2022/068. xiii + 148 p. http://www.dfo-mpo.gc.ca/csas-sccs/ mailto:csas-sccs@dfo-mpo.gc.ca iii TABLE OF CONTENTS ABSTRACT ................................................................................................................................. xii INTRODUCTION .......................................................................................................................... 1 DATA SOURCES .......................................................................................................................... 2 LANDINGS ............................................................................................................................... 2 Spawning stock assignment .................................................................................................. 3 TELEPHONE SURVEY ............................................................................................................ 4 FISHERY SAMPLING .............................................................................................................. 4 FISHERY-INDEPENDENT ACOUSTIC SURVEY ................................................................... 5 EXPERIMENTAL NETS ........................................................................................................... 5 SPAWNING GROUND ACOUSTIC SURVEYS ....................................................................... 5 MULTISPECIES BOTTOM-TRAWL SURVEY ......................................................................... 6 ECOSYSTEM INFORMATION ................................................................................................. 6 INPUTS AND INDICES ................................................................................................................. 6 CATCH-AT-AGE AND WEIGHT-AT-AGE MATRICES ............................................................ 6 CATCH-PER-UNIT EFFORT ................................................................................................... 7 FISHERY-INDEPENDENT ACOUSTIC SURVEY INDEX ....................................................... 9 SPAWNING GROUND ACOUSTIC SURVEYS ....................................................................... 9 EXPERIMENTAL NET INDICES .............................................................................................. 9 Relative selectivity index ....................................................................................................... 9 Catch-at-age of experimental nets ...................................................................................... 10 MULTISPECIES BOTTOM TRAWL INDEX ........................................................................... 10 MATURITY OGIVE................................................................................................................. 10 SPRING SPAWNER COMPONENT ASSESSMENT ................................................................. 10 SPRING SPAWNER MODEL ................................................................................................. 11 SPRING SPAWNER RESULTS ............................................................................................. 14 SPRING SPAWNER PROJECTIONS .................................................................................... 15 Short term projections ......................................................................................................... 16 Long term projections .......................................................................................................... 16 FALL SPAWNER COMPONENT ASSESSMENT ...................................................................... 16 FALL SPAWNER MODEL ...................................................................................................... 16 FALL SPAWNER RESULTS .................................................................................................. 18 FALL SPAWNER PROJECTIONS ......................................................................................... 20 Short term projections ......................................................................................................... 21 Long term projections .......................................................................................................... 21 PREDATOR PREY INTERACTIONS ......................................................................................... 21 DISCUSSION AND CONCLUSION ............................................................................................ 22 SPRING SPAWNING HERRING ........................................................................................... 22 FALL SPAWNING HERRING ................................................................................................. 23 iv NATURAL MORTALITY AND ECOSYSTEM INTERACTIONS: SPRING AND FALL HERRING ............................................................................................................................... 25 SOURCES OF UNCERTAINTY .................................................................................................. 26 REFERENCES ........................................................................................................................... 28 TABLES ...................................................................................................................................... 33 FIGURES .................................................................................................................................... 81 APPENDIX A. AGE READING CONSISTENCY TEST ............................................................ 127 APPENDIX B. FISHERY-INDEPENDENT ACOUSTIC SURVEY RESULTS ........................... 128 APPENDIX C. SPAWNING GROUND ACOUSTIC SURVEY RESULTS ................................. 133 APPENDIX D. MULTISPECIES BOTTOM-TRAWL SURVEY RESULTS ................................ 140 APPENDIX E. COMPARISON OF CPUE ESTIMATIONS FROM FORMER SAS CODE AND NEWLY TRANSLATED AND UPDATED R CODE ................................................................... 141 LIST OF TABLES Table 1. Landings (in tons) of 4T Herring in the spring and fall fisheries by gear (fixed and mobile) and spawning group (SS=spring spawners and FS=fall spawners). TAC allocations and target catches are also provided, as TAC is higher than the targeted catch decision due to historical shares between regions. .............................................................................................. 33 Table 2. Commercial fishery samples collected, number of fish processed (N), landings, and % TAC landed by zone in the spring (April 1-June 30) and fall (July 1-December 31). These data are used to derive the 2020 and 2021 catch and weight-at-age matrices for 4T Herring. .......... 36 Table 3. Comparison of 2020 and 2021 DMP and telephone survey results including number of respondents, mean net length (fathoms), numbers of nets set, percentage of nets of mesh size 2⅝“ in the fall fishery, and a comparative index abundance from 2020 and 2021, respectively [scale 1 (poor) to 10 (excellent)]. ................................................................................................ 38 Table 4. Spring spawner catch-at-age (thousands) for fixed gear in the 4T Herring fishery. ...... 40 Table 5. Spring spawner weight-at-age (kg) for fixed gear in the 4T Herring fishery. ................. 41 Table 6. Fall spawner catch-at-age (thousands) for fixed gear in the 4T Herring fishery, by region: a) North, b) Middle, c) South. .......................................................................................... 42 Table 7. Fall spawner weight-at-age (kg) for fixed gear in the 4T Herring fishery, by region: a) North, b) Middle, c) South. .......................................................................................................... 45 Table 8. Spring spawner catch-at-age (thousands) for mobile gear in the 4T Herring fishery. ... 48 Table 9. Spring spawner weight-at-age (kg) for mobile gear in the 4T Herring fishery. .............. 49 Table 10. Fall spawner catch-at-age (thousands) for mobile gear in the 4T Herring fishery, by region: a) North, b) Middle, c) South. .......................................................................................... 50 Table 11. Fall spawner weight-at-age (kg) for mobile gear in the 4T Herring fishery. ................ 53 Table 12. Percent of fishing days with no gillnet catch derived from the telephone survey for main fishing areas in the spring and fall fishery. ......................................................................... 54 v Table 13. Results of the multiplicative general linear model applied to the fishery catch-per-unit- effort data for each region (NAFO 4T). ....................................................................................... 54 Table 14. Spring spawner fixed gear catch-per-unit-effort values (number per net-haul) for NAFO area 4T. ............................................................................................................................ 55 Table 15. Fall spawner fixed gear catch-per-unit-effort values (number per net-haul) by region: a) North, b) Middle, and c) South. ............................................................................................... 56 Table 16. Spring spawner and fall spawner catch-at-age from the fishery-independent acoustic survey in NAFO area 4Tmno. ..................................................................................................... 59 Table 17. Relative selectivity-at-age for 2⅝” and 2¾” mesh calculated from the experimental netting survey and commercial gillnet fishery. ............................................................................ 61 Table 18. Multi-species bottom trawl survey fall spawning Herring stratified mean numbers per tow at age. .................................................................................................................................. 63 Table 19. Maximum likelihood estimates (MLEs) of January 1 spring spawner biomass (t). ...... 64 Table 20. Maximum likelihood estimates (MLEs) of January 1 spring spawner abundance (number in thousands). ............................................................................................................... 65 Table 21. Maximum likelihood estimates of the instantaneous rate of fishing mortality (F) of spring spawners by age. F6-8 is the January 1 abundance-weighted average F for ages 6 to 8 years. ....................................................................................................................................... 66 Table 22. Risk analysis table of annual catch options (between 0 and 1,250 t) for 2022 and 2024 and subsequent years until 2027, with predicted resulting SSB (kt) in 2023, 2024 and 2027, resulting probabilities (%) of SSB being greater than the LRP, resulting probabilities of increases in SSB by 5%, and resulting abundance weighted fishing mortality rate (F6-8) for the spring spawner component of Atlantic Herring from the southern Gulf of St. Lawrence. ...................... 67 Table 23. SCA maximum likelihood estimates of August 1 biomass (t) of fall spawners in the North region of the southern Gulf of St. Lawrence. ..................................................................... 68 Table 24. SCA maximum likelihood estimates of January 1 abundance (number in thousands) of fall spawners in the North region of the southern Gulf of St. Lawrence. ..................................... 69 Table 25. SCA maximum likelihood estimates of August 1 biomass (t) of fall spawners in the Middle region of the southern Gulf of St. Lawrence. ................................................................... 70 Table 26. SCA maximum likelihood estimates of January 1 abundance (number in thousands) of fall spawners in the Middle region of the southern Gulf of St. Lawrence. ................................... 71 Table 27. SCA maximum likelihood estimates of August 1 biomass (t) of fall spawners in the South region of the southern Gulf of St. Lawrence. .................................................................... 72 Table 28. SCA maximum likelihood estimates of January 1 abundance (number in thousands) of fall spawners in the South region of the southern Gulf of St. Lawrence. .................................... 73 Table 29. SCA maximum likelihood estimates of August 1 total biomass (t) of fall spawners in the southern Gulf of St. Lawrence. ............................................................................................. 74 Table 30. SCA maximum likelihood estimates of January 1 total abundance (number in thousands) of fall spawners in the southern Gulf of St. Lawrence. ............................................. 75 Table 31. SCA maximum likelihood estimates of the instantaneous rate of fishing mortality (F) of fall spawners in the North region of the southern Gulf of St. Lawrence. F5-10 is the January 1 abundance-weighted average F for ages 5 to 10 years. ............................................................. 76 vi Table 32. SCA maximum likelihood estimates of the instantaneous rate of fishing mortality (F) of fall spawners in the Middle region of the southern Gulf of St. Lawrence. F5-10 is the January 1 abundance-weighted average F for ages 5 to 10 years. ............................................................. 77 Table 33. SCA maximum likelihood estimates of the instantaneous rate of fishing mortality (F) of fall spawners in the South region of the southern Gulf of St. Lawrence. F5-10 is the January 1 abundance-weighted average F for ages 5 to 10 years. ............................................................. 78 Table 34. SCA maximum likelihood estimates of the instantaneous rate of fishing mortality (F) of fall spawners in the southern Gulf of St. Lawrence. F5-10 is the January 1 abundance-weighted average F for ages 5 to 10 years. ............................................................................................... 79 Table 35. Risk analysis table from the SCA model of annual catch options (between 2,000 and 18,000 t) for 2022 and 2023 and subsequent years until 2027, with predicted resulting SSB (kt) in 2023, 2024 and 2027, resulting probabilities (%) of SSB being lower than the LRP, resulting probabilities of increases in SSB by 5%, and resulting fully-recruited fishing mortality rate (F5-10) for the fall spawner component of Atlantic Herring from the southern Gulf of St. Lawrence....... 80 LIST OF FIGURES Figure 1. Southern Gulf of St. Lawrence Herring fishery management zones (upper panel, a), Northwest Atlantic Fisheries Organization (NAFO) Divisions 4T and 4Vn, where purple represents the North region, blue = Middle region, and green = South region (middle panel, b), and geographic areas used in the telephone survey of the Herring gillnet fishery (lower panel, c). ................................................................................................................................................ 81 Figure 2. Reported landings (tonnes) of southern Gulf of St. Lawrence Atlantic Herring (spring and fall spawners combined) by NAFO Division (upper panel, a), by gear fleet (middle panel, b), and by fishing season (lower panel, c), 1978 to 2021. In all panels, the corresponding annual TAC (tonnes) is shown. For landings by season, the landings in Div. 4Vn were attributed to the fall fishing season. Data for 2020 and 2021 are preliminary. ...................................................... 82 Figure 3. Estimated landings (tonnes) of the spring spawner component (left) and fall spawner component (right) of Atlantic Herring from the southern Gulf of St. Lawrence, 1978 to 2021. Panel a and d shows the estimated landings by gear type and the proportion of the landings attributed to the fixed gear fleet and the TAC for the spawner component (red symbols) for 1991 to 2021. Panels b and e shows the estimated landings of Herring in the fixed gear fleet that occurred in the spring fishery season and the fall fishery season as well as the proportion of Herring landed in the matching fishing season. Panels c and f shows the estimated landings of Herring in the mobile gear fleet that occurred in the spring fishery season and the fall fishery season as well as the proportion of Herring landed in the matching fishing season. For landings by season, the landings in NAFO Division 4Vn were attributed to the fall fishing season. Data for 2018 and 2021 are preliminary. .................................................................................................. 83 Figure 4. Catch-at-age of the spring spawner component from the fishery, all gears combined, 1978 to 2021. Size of the bubble is proportional to the catch numbers by age and year. The diagonal line represents the most recent strong year-class (1991). The values indicated at age 11 represent catches for ages 11 years and older. .............................................................. 84 Figure 5. Bubble plots of fishery catch-at-age (number) by region for both mobile and fixed gear combined, 1978 to 2021. The size of the bubble is proportional to the number of fish in the catch by age and year. The values indicated at age 11 represent catches for ages 11 years and older. .................................................................................................................................................... 85 vii Figure 6. Mean weight (kg) of Atlantic Herring for ages 4, 6, 8, and 10 of spring spawners (left panels) sampled from catches in the spring season and fall spawners (right panels) sampled from catches in the fall season from mobile (upper panels) and fixed (lower panels) commercial gears, in NAFO Div. 4T for 1978 to 2021. ................................................................................... 86 Figure 7. Bubble plot of spring spawner Herring fixed gear catch-per-unit-effort values (number per net-haul per trip) at age, 1990 to 2021. The size of the bubble is proportional to the maximum CPUE index value. ..................................................................................................... 86 Figure 8. Fall spawner (FS) fixed gear age-disaggregated catch-per-unit-effort values (number per net-haul per trip) by region (upper panel North, middle panel Middle, and lower panel South), 1986 to 2021. The size of the bubble is proportional to the CPUE index value. ............ 87 Figure 9. Bubble plot of abundance-at-age (number) from the fisheries-independent acoustic survey for spring spawners (upper panel a); ages 4 to 8) and fall spawners (lower panel b); ages 2 to 3) from 1994 to 2021. .................................................................................................. 88 Figure 10. FSCP acoustic biomass indices of NAFO Division 4T fall spawning Atlantic herring in the North, Middle and South regions between 2015 and 2021. Points are average and vertical lines are 95% confidence intervals. ............................................................................................ 89 Figure 11. Bubble plots of catch-at-age indices (number) of fall spawners from the experimental netting survey by region (upper panel North, middle panel Middle, and lower panel South) from 2002 to 2021. The size of the bubble is proportional to the index value. .................................... 90 Figure 12. Variations in the proportions of gillnets with mesh sizes 2 5/8 inches by region, 1986 to 2021. It is assumed that all other nets used were of mesh size 2 ¾. ..................................... 91 Figure 13. Multispecies bottom trawl survey abundance index (number of fish per standardized tow) for fall spawning Herring ages 4 to 6 years, 1994 to 2021. ................................................. 91 Figure 14. Residuals in PAA (observed – predicted indices) for the population model of spring spawners in the southern Gulf of St. Lawrence. The upper panel shows residuals for the CPUE index and the bottom panel shows residuals for the acoustic index. Rows are for ages and columns for years. Circle radius is proportional to the absolute value of residuals. Black circles indicate negative residuals (i.e., observed < predicted). ............................................................. 92 Figure 15. Observed (circles) and predicted (lines and shading) age-aggregated CPUE (upper panels) and acoustic (lower panels) indices (kg) for the population model of spring spawners in the southern Gulf of St. Lawrence. The lines show the median predicted indices and the shading the 95% confidence intervals of the predictions based on MCMC sampling. ............................. 93 Figure 16. Retrospective patterns in estimated spawning stock biomass (SSB) of ages 4 to 10 and years 2021 to 2015 for spring spawners in the southern Gulf of St. Lawrence. Lined colors correspond to peels between years 2015 and 2021. .................................................................. 94 Figure 17. Estimated fully-recruited catchability to the CPUE index (q) from the spring spawners population model. Lines show the median estimates and shading their 50% (dark shading) and 95% (light shading) confidence interval based on MCMC sampling. .......................................... 94 Figure 18. Fully-recruited catchability to the CPUE gillnet fishery (q) in function of SSB (kilotons) for spring spawning Herring between 1990 and 2021. ............................................................... 95 Figure 19. Estimated instantaneous natural mortality rate (left axis) and annual mortality (%, right axis) of spring spawning Atlantic Herring from the population model, for ages 2 to 6 (upper panel) and 7 to 11+ (lower panel). Lines show the median estimates and shading their 95% confidence interval based on MCMC sampling. .......................................................................... 96 viii Figure 20. Estimated beginning of the fishing season (April 1) SSB of the spring spawner component of Atlantic Herring in the southern Gulf of St. Lawrence, 1978 to 2021. The solid line is the median MCMC estimate and shading its 50% (dark shading) and 95% (light shading) confidence intervals. The red dashed horizontal line is the Limit Reference Point (LRP) (46,340 t of SSB). ....................................................................................................................................... 97 Figure 21. Estimated January 1 abundance of 2 year old Herring (blue bars), and Herring 4 years and older (black line) of the spring spawner component in the southern Gulf of St. Lawrence. Black line show the median MCMC estimate and vertical lines and shading show 95% confidence interval. ............................................................................................................. 97 Figure 22. Estimated January 1 abundance of 4 year old Herring (blue bars), and Herring 4 years and older (black line) of the spring spawner component in the southern Gulf of St. Lawrence. Black line show the median MCMC estimate and vertical lines and shading show 95% confidence interval. ............................................................................................................. 98 Figure 23. Recruitment rates for age 2 recruits for the 1978 to 2019 cohorts of spring spawning Atlantic Herring in NAFO Div. 4T. Vertical lines indicate 95% confidence intervals.................... 98 Figure 24. Estimated January 1 abundance weighted age 6 to 8 fishing mortality (F6-8, left axis; annual exploitation rate, right axis) of spring spawning Herring in the southern Gulf of St. Lawrence. Circles are the median estimates and vertical lines their 95% confidence intervals. 99 Figure 25. The southern Gulf of St. Lawrence Atlantic Herring spring spawner component trajectory in relation to SSB (kt = thousand t) and abundance weighted fishing mortality rates for ages 6 to 8 years. The red vertical line is the LRP and the green dashed vertical line is the Upper Stock Reference (USR). The orange solid horizontal line is the removal rate reference value (F0.1 = 0.35) in the Healthy Zone and orange dashed line is the provisional harvest decision rule of the Precautionary Approach Framework in the Cautious and Critical Zones. Point labels are years (83 = 1983, 0 = 2000). ........................................................................... 100 Figure 26. Projected April 1 SSB (in kt) of spring spawning Atlantic Herring from the southern Gulf of St. Lawrence under a recent 5 years average recruitment level and 2 years average natural mortality level at various catch levels in 2022 and 2023. Lines show the median estimates of the April 1 SSB, dark shading the 75% confidence interval and light shading the 95% confidence intervals of these estimates (based on MCMC sampling). Black and grey indicate the historical period and blue the projection period. The red horizontal line is the LRP. .................................................................................................................................................. 101 Figure 27. Projected ages 6 to 8 fishing mortality rate (F) of spring spawner Atlantic Herring from the southern Gulf of St. Lawrence at various catch levels in 2022 and 2023. Lines show the median estimates of fishing mortality, dark shading the 75% confidence interval and light shading the 95% confidence intervals of these estimates (based on MCMC sampling). Black and grey indicate the historical period and blue the projection period. ..................................... 102 Figure 28. Projected April 1 SSB (in kt) of spring spawner Atlantic Herring from the southern Gulf of St. Lawrence under a recent 5 years average recruitment level and 2 years average natural mortality level at various catch levels in all years between 2022 and 2027. Lines show the median estimates of the April 1 SSB, dark shading the 75% confidence interval and light shading the 95% confidence intervals of these estimates (based on MCMC sampling). The red horizontal line is the LRP. ......................................................................................................... 103 Figure 29. Fishery catch PAA residuals by region (North, Middle and South) from the SCA population model of fall spawning Herring from the southern Gulf of St. Lawrence. Rows are for ix ages and columns are years. The circle radius is proportional to the absolute value of residuals. Black circles indicate negative residuals (i.e., observed < predicted). ...................................... 104 Figure 30. CPUE index PAA residuals by region (North, Middle and South) from the SCA population model of fall spawning Herring from the southern Gulf of St. Lawrence. Rows are for ages and columns are years. The circle radius is proportional to the absolute value of residuals. Black circles indicate negative residuals (i.e., observed < predicted). ...................................... 105 Figure 31. Experimental nets index PAA residuals by region (North, Middle and South) from the SCA population model of fall spawning Herring from the southern Gulf of St. Lawrence. Rows are for ages and columns are years. The circle radius is proportional to the absolute value of residuals. Black circles indicate negative residuals (i.e., observed < predicted). Results are only provided for the years during which the acoustic survey was conducted. ................................ 106 Figure 32. RV survey index (top) and Acoustic survey index (AC, bottom) PAA residuals from the SCA population model of fall spawning Herring from the southern Gulf of St. Lawrence. Rows are for ages and columns are years. The circle radius is proportional to the absolute value of residuals. Black circles indicate negative residuals (i.e., observed < predicted). ................. 107 Figure 33. Observed (circles) and predicted (lines and shading) age-aggregated commercial gillnet CPUE indices by region (CPUE North, CPUE Middle, CPUE South) from the SCA population model for fall spawners from the southern Gulf of St. Lawrence. The lines show the median predicted indices and the shading the 95% confidence intervals of the predictions based on MCMC sampling. ................................................................................................................. 108 Figure 34. Observed (circles) and predicted (lines and shading) age-aggregated RV indices (RV, all regions combined) and acoustic indices (AC, all regions combined) from the SCA population model for fall spawners from the southern Gulf of St. Lawrence. The lines show the median predicted indices and the shading the 95% confidence intervals of the predictions based on MCMC sampling. ................................................................................................................. 109 Figure 35. Observed (circles) and predicted (lines and shading) age-aggregated FSCP Acoustic Biomass Index from the SCA population model for fall spawners from the southern Gulf of St. Lawrence. The lines show the median predicted indices and the shading the 95% confidence intervals of the predictions based on MCMC sampling. ............................................................ 110 Figure 36. Retrospective patterns in SSB and Mohn’s rho of fall spawners within the three regions (North, Middle, South) for the SCA population model of Atlantic Herring of the southern Gulf of St. Lawrence. Colored lines shows retrospective peels between 2017 and 2021......... 111 Figure 37. Estimated fully-recruited catchability for the commercial gillnet CPUE index by region (North, Middle, South), from the SCA population model of fall spawning Atlantic Herring in the southern Gulf of St. Lawrence. Lines show the median estimates and shading their 95% confidence intervals based on MCMC sampling. ...................................................................... 112 Figure 38. Estimated fully-recruited catchability for the commercial gillnet CPUE index in relation to SSB by region (North, Middle, South), from the SCA population model of fall spawning Atlantic Herring in the southern Gulf of St. Lawrence. .............................................................. 113 Figure 39. Estimated instantaneous natural mortality rate (left axis) and annual mortality (%, right axis) of fall spawning Atlantic Herring for three regions of the sGSL (North, Middle, South) from the SCA population model, for ages 2 to 6 (blue) and 7 to 11+ (red). Lines show the median estimates and shading their 95% confidence interval based on MCMC sampling. ...... 114 Figure 40. Estimated beginning of fishing season (August 1) SSB of fall spawning Herring by region and overall (Total) for the southern Gulf of St. Lawrence from the SCA population model. The black line shows the median estimates of the MCMC sampling and the shading their 95% x confidence intervals. In the bottom right panel for Total, the solid and dashed yellow horizontal lines represent the USR level and the red horizontal line is the LRP. SSB, USR and LRP values are adjusted to August 1st using natural mortality estimates at age for 7 months. ................... 115 Figure 41. Estimated January 1 abundance of 2 year old Herring (blue bars), and Herring 4 years and older (black line) of the fall spawner component in three regions (North, Middle, South) in the southern Gulf of St. Lawrence from the SCA population model. Black line show the median MCMC estimate and vertical lines show 95% confidence interval. .............................. 116 Figure 42. Estimated January 1 abundance of 4 year old Herring (blue bars), and Herring 4 years and older (black line) of the fall spawner component in three regions (North, Middle, South) in the southern Gulf of St. Lawrence from the SCA population model. Black line show the median MCMC estimate and vertical lines show 95% confidence interval. .............................. 117 Figure 43. Estimated recruitment rate (recruits per kg of SSB) at age 2 (circles) of fall spawners in the three regions (North, Middle, South) and summed over regions (Total) of the southern Gulf of St. Lawrence, from the SCA population model. Bars show the median estimates and vertical lines show the 95% confidence intervals. ..................................................................... 118 Figure 44. Estimated fishery (top row), CPUE (Middle row) and experimental nets (bottom row) selectivity for three populations of the southern Gulf of St. Lawrence (North in the left column, Middle in the Middle column and South in the right column), from the SCA population model. Lines show the maximum likelihood estimates for years or time-periods identified in respective Figure legends. ......................................................................................................................... 119 Figure 45. Estimated beginning-of-the-year abundance averaged age 5 to 10 fishing mortality (F5-10, left axis; annual exploitation rate, right axis) of fall spawning Herring by region and averaged over regions (weighted by region-specific abundance at ages 5-10 years) in the southern Gulf of St Lawrence from the SCA model. Lines show the median estimates and shading their 95% confidence intervals. ................................................................................... 120 Figure 46. Southern Gulf of St. Lawrence Atlantic Herring fall spawner component trajectory in relation to SSB/USB and fishing mortality rates for ages 5 to 10 years from the SCA population model. The red vertical line is the LRP and the green vertical line is the USR. The orange dashed line is the provisional removal reference of the Precautionary Approach Framework. 121 Figure 47. Projected SSB (in kt) of fall spawning Atlantic Herring from the southern Gulf of St. Lawrence at various catch levels in 2022 and 2023, under a 5 recent years average recruitment and 2 recent years average natural mortality scenario. Lines show the median estimates of August 1 SSB, dark shading the 95% confidence intervals and light shading the 50% confidence interval (based on MCMC sampling). Black and grey indicate the historical period and blue the projection period. The red horizontal line is the LRP. ............................................................... 122 Figure 48. Projected average fishing mortality (F5-10) of fall spawning Atlantic Herring from the southern Gulf of St. Lawrence at various catch levels in 2022 and 2023, under a 5 recent years average recruitment and 2 recent years average natural mortality scenario. Lines show the median estimates of fishing mortality, dark shading the 95% confidence intervals and light shading the 50% confidence interval (based on MCMC sampling). Black and grey indicate the historical period and blue the projection period. ....................................................................... 123 Figure 49. Six years projections of SSB (in kt) of fall spawning Atlantic Herring from the southern Gulf of St. Lawrence at various catch levels from the SCA population model, under a 5 recent years average recruitment and 2 recent years average natural mortality scenario. Lines show the median estimates of August 1 SSB, light shading shows the 95% and dark shading xi shows the 50% confidence intervals (based on MCMC sampling). The green and red horizontal lines are the USR and LRP, respectively. ................................................................................. 124 Figure 50. Scaled relative abundance indices for Herring major predators (Atlantic cod, White Hake, Grey seal, Atlantic Bluefin Tuna, Northern Gannet) between 1970-2021 alongside with natural mortality (M) estimates for age groups 2-6 (M2-6) and 7-11+ (M7-11) from the SCA spring and fall herring stock models. ........................................................................................ 125 Figure 51. Correlation matrix between the scaled relative abundance indices for Herring major predators (Atlantic cod, White Hake, Grey seal, Atlantic Bluefin Tuna, Northern Gannet) between 1970-2021 alongside with natural mortality estimates for age groups 2-6 (m1) and 7- 11+ (m2) from the spring and fall herring stock models. ........................................................... 126 xii ABSTRACT Atlantic Herring (Clupea harengus) in Northwest Atlantic Fisheries Organization (NAFO) Division 4T, referred to as the southern Gulf of St. Lawrence (sGSL), consists of two spawning components, spring spawners and fall spawners. This document presents the most recent information on trends in abundance, distribution, and harvest for the spring and fall spawning Herring components in NAFO Division 4T. This includes catch-at-age and catch-per-unit-effort (CPUE) indices, fisheries-independent acoustic indices, experimental gillnet survey indices, mesh selectivity, fishery-dependent acoustic indices and catches in the multi-species bottom trawl survey of the sGSL. The data and indices are reported for the sGSL for the spring spawners, and regionally-disaggregated (North, Middle, and South regions) for the fall spawners where applicable. Spring spawners were assessed using a statistical catch at age (SCA) model that allowed for time-varying catchability to the gillnet fishery and time-varying natural mortality. The model estimated that spawning stock biomass (SSB) has been in the critical zone of the Precautionary Approach framework since 2002. The SSB median estimate in April 1 2022 is estimated to be 28,835 tons (t); 62% of the limit reference point (LRP = 46,340 t). Under current low recruitment and high natural mortality conditions, this stock is not expected to recover in the short or the long term. Reducing fishing mortality will have marginal effects on the projected SSB trends. By 2027, the probability of exceeding the LRP was not more than 20% at all catch levels, with SSB values ranging between 32,500 and 35,400 t. Fall spawners were assessed as regionally-disaggregated populations using a SCA model that allowed for time-varying catchability to the gillnet fishery and time-varying natural mortality. Estimated SSB has been declining in all three regions in recent years and is currently in the Cautious Zone of the Precautionary Approach framework. At the target catch level in 2021 (~12,000 t), the probabilities of a 5% increase in SSB by 2024 are all under 40%. Long-term projections show a continuous decline of SSB, however the probability of moving into the Critical Zone (under the LRP) by 2027 was 0% at all catch levels. As a consequence of low productivity and high natural mortality, exploitation of this stock should assert caution until high recruitment is observed for consecutive years. 1 INTRODUCTION Atlantic Herring in the southern Gulf of St. Lawrence (sGSL) are found in the area extending from the north shore of the Gaspé Peninsula to the northern tip of Cape Breton Island, including the Magdalen Islands. Adults overwinter off the north and east coast of Cape Breton in the Northwest Atlantic Fisheries Organization (NAFO) Divisions 4T and 4Vn (Claytor 2001; Simon and Stobo 1983; Figure 1). Studies in the early 1970s indicated that southern Gulf Herring also overwintered off the south coast of Newfoundland, but an exploratory fishery in 2006 did not detect any concentrations (Wheeler et al. 2006). Herring is a pelagic species that schools particularly during feeding, spawning periods, and annual migrations. Eggs are attached to the sea floor and large females can produce up to 360,000 eggs (Messieh 1988). First spawning behavior typically occurs at four years of age. Herring in the sGSL are managed across seven Herring Fishing Areas (HFA) (16A-16G; Figure 1a). These HFAs cover the same region as NAFO Division 4T (Figure 1b). The Herring population in the sGSL consists of two spawning components: spring spawners and fall spawners. Spring spawning occurs primarily in April-May but extends to June 30 at depths < 10 m. Fall spawning occurs from mid-August to mid-October at depths of 5 to 20 m, but can occur as early as July 1. Both spawning behaviors are explained by the genetic differentiation between these stocks (Lamichhaney et al. 2017). Spring and fall herring spawners within 4T are therefore considered distinct stocks and are assessed separately. Herring also show high spawning site fidelity (Winters and Wheeler 1985; McQuinn 1997; Brophy et al. 2006) and local stocks are targeted by the gillnet fishery which takes place on the spawning grounds. Fall spawning Herring in the sGSL are therefore assessed using regionally-disaggregated assessment models (North, Middle, South regions; Figure 1b). The sGSL Herring are harvested by a gillnet fleet (referred to as “fixed” gear fleet) and a purse seine fleet (“mobile” gear fleet). The mobile gear fleet mainly consist of large vessels (> 19.8 m), but some small seiners (< 19.8 m) can also participate in the inshore fishery as part of the gillnet fleet. The fixed gear fishery is focused in NAFO Division 4T, whereas the mobile gear fishery occurs in 4T and historically, occasionally in 4Vn (Figure 1b). During the spring and fall fishing seasons, the mobile fleet are prohibited from fishing in areas set aside exclusively for the fixed gear fleet (Claytor et al. 1998). In the spring fishing season, mobile gear fleets fish along the northern boundary of NAFO region 4Tf, which is referred to as the “Edge” fishery. In the fall fishing season, mobile gear fleets fish in the Baie-des-Chaleurs area. Both spring and fall spawning Herring are harvested in the spring and fall fishing seasons and must therefore be separated into the appropriate groups for assessment purposes. Prior to 1967, sGSL Herring was mainly exploited by fixed gear and average landings from 1935 to 1966 were 34,000 tons (t). In the mid-1960s, a mobile gear fishery was introduced and average landings by both fleets were 166,000 t from 1967 to 1972. Since 1981, fishing effort was reduced in the mobile gear fleets and the fixed gear fleet has accounted for most of the catch of spring and fall spawners (McDermid et al. 2018). A global allocation or Total Allowable Catch (TAC) was introduced in 1972 at 166,000 t, and reduced to 40,000 t in 1973. Separate TAC for the spring and fall spawners components began in 1985. The TAC were first allotted by fishing season (spring and fall) and later attributed to spring or fall spawners landings based on biological samples taken during the fishery. The percentage of spring and fall spawners in the catch varies according to season and gear type. As a result, landings during the spring and fall fishing seasons must be separated into the appropriate spring and fall spawners groups to determine if the TAC for these groups has been attained. 2 For this assessment, the population modelling is conducted for spring and fall spawning Herring to the end of 2021, with projections for 2022, 2023, and 2027. DATA SOURCES For the spring spawning Herring assessment, data collected in NAFO Div. 4T is used to model the population at the scale of the sGSL. The spatial distribution of the data collected during the spring fishery does not permit, for now, the use of a regionally-disaggregated model as for the fall spawning stock. For the fall spawning Herring assessment, a regionally-disaggregated model is used to evaluate the population in three regions (North, Middle, and South) that encompass the entire NAFO Div. 4T. The regions are defined on the basis of traditional Herring spawning beds and fishing areas (Figure 1): • North (Gaspé and Miscou; 4Tmnopq), • Middle (Escuminac-Richibucto and west Prince Edward Island; 4Tkl), and • South (east Prince Edward Island and Pictou; 4Tfghj). The choice of three regions was dictated by geographic proximity of spawning beds and is the finest level of disaggregation that can presently be supported by the available data. The regionally-disaggregated models include inputs that are region-specific (e.g., catch-at-age, catch-per-unit-effort, experimental nets proportions-at-age (PAA), selectivity-at-age, biomass indices from hydroacoustic surveys on spawning grounds) and inputs that are common to the entire area (e.g., acoustic survey index, RV survey index). LANDINGS Catch data were extracted from purchase slips and ZIFF (Zonal Interchange File Format) files collected by the Statistics Branch of Fisheries and Oceans Canada (DFO). Catch data to 1985 are available by fishery (fixed and mobile) and by fishing area. Beginning in 1986, the catch data are further reported by vessel and trip. The ZIFF files are based on information collected by the Dockside Monitoring Program (DMP). This program provides accurate, timely, and independent third-party verification of fish landings. Contracted companies are hired by the fishing industry to observe the offloading of fish and to record and report the landings information to DFO. The fishery TACs within NAFO Div. 4T are set for the sGSL spring spawners and fall spawners components, separately. In 2020 and 2021, the TACs were set at 500 t for the spring spawners and 12,000 t for the fall spawners, for a total of 12,500 t (Table 1; Figure 2). Bait removals were not counted against the TAC. Seventy-seven percent of the TAC for each spawning component was allocated to the fixed gear fleet with the remaining 23% for the mobile gear fleet (Table 1). The preliminary estimated landings of spring spawning Herring in both the spring and fall fishing season were 603 t and 403 t for 2020 and 2021, respectively (Table 1; Figure 3). Most of the spring spawning Herring were estimated to have been landed in the fixed gear fleet over the 1981 to 2021 period. In 2020 and 2021, the fixed gear fleet was estimated to have landed 59% and 98%, respectively, of the total harvests of spring spawning Herring (Table 1; Figure 3a). The 2021 value was exceptionally high as the mobile fleet had a fairly limited activity that year. For 2020 and 2021, more than 95% of the spring spawning Herring landed by the fixed gear fleet was landed during the spring fishing season, whereas 100% of the spring spawning Herring landed by the mobile fleet was landed in the fall season (Table 1). Historically and on average, more than 80% of the spring spawning Herring landed by the fixed gear fleet has been 3 landed during the spring fishing season, whereas more than 80% of the spring spawning Herring landed by the mobile fleet has been landed in the fall season (Figure 3b, c). The preliminary landings of fall spawners in 2020 and 2021 were 10,065 t and 10,834 t, respectively (Table 1; Figure 3d). Over the 1978 to 2021 period, most of the fall spawning Herring have been landed in the fixed gear fleet. In 2020 and 2021, the fixed gear fleet was estimated to have landed 97% and 99.9%, of the total harvests of fall spawning Herring, respectively (Figure 3). The majority (nearly 100%) of the fall spawning Herring captured in the fixed gear fishery are landed during the fall fishing season (Figure 3e). Of all the fall spawners landed by the mobile fleet, 100% were landed in the fall fishing season in 2020 and 2021 (Figure 3f). The recent 2017 to 2021 mean proportion of the total catch caught by fixed gear was 74% of the spring spawners and 96% of the fall spawners (Table 1). Over 37% and 29% of the 2020 and 2021 spring fishery fixed gear catches occurred in Herring areas 4Th (South) and 4Tmn (North), respectively (Table 2). Meanwhile, 55% of the 2020-2021 fall fishery fixed gear catches occurred in Herring area 4Tmn (North; Figure 1; Table 2). The mobile gear (Edge) spring fishery was not active in both 2020 and 2021. However the fall fishery 2020 and 2021 mobile gear catches were 646.2 t and 13.8 t, respectively, and both from 4Tmn (North; Figure 1; Table 2). In 2020, 120.6% of the spring spawners TAC was attained compared to 80.6% in 2021 (Table 1). For fall spawners, 83.9% and 90.3% of the TAC was attained in 2020 and 2021, respectively (Table 1). Herring fishing area landings information can be found in Table 2. A rebuilding plan was introduced for the spring spawners in 2010. This plan includes: • fishing closure on some spawning areas in all HFA except 16A and 16F, • weekly landing limits of 10,206 kg in all HFA except 16A, 16D, and 16F, where no restrictions apply, and • no nets or Herring allowed on board during a fishing trip between 18:00 and 04:00 (ADT) in 16C-G and between 22:00 and 03:00 (ADT) in 16A and 16B. Spawning stock assignment Gulf Region Science uses three methods to assign Herring samples to either spring or fall spawners based on gonad maturity stages (Cleary et al. 1982): 1. For immature Herring of maturity stages 1 and 2 (juveniles), the season of hatching is based on the size at capture and visual examination of otolith characteristics (Messieh 1972). The spawning component assignment for juvenile Herring is its hatching season (Cleary et al. 1982). Juveniles represent a small percentage of commercial catch, but are a higher proportion in the research survey samples. 2. Adult Herring with ripe or spent gonads are assigned their maturity stage by macroscopic laboratory examination of the gonads. The fish are assumed to belong to the spawning component of the season in which they were caught. These represent over 90% of the gillnet catches and 75% of the total yearly landings. 3. Adult Herring with unripe gonads are assigned their maturity stage by using a gonadosomatic index (GSI) based on a discriminant function model. The GSI is based on the length of the fish and its gonad weight (McQuinn 1989). Once the maturity stage is determined by GSI, the spawning component is assigned by using a maturity schedule decision rule (a table cross-referencing maturity stage assigned by GSI and the date of capture to assign a spawning component) (Cleary et al. 1982). 4 For the month of June, the GSI and macroscopic examination methods historically resulted in different assignment of samples to spawning components. In particular, the 2012 and 2013 Cabot Strait Edge fishery samples were not well classified by the GSI method. The macroscopic examination identified at least 95% of the gonads as developing gonads therefore classifying them as fall spawners. The GSI discriminant function reclassified at least 20% of these developing gonads as spent gonads resulting in a classification of spring spawners. A change was made to the decision rules for the GSI method such that a “spent” gonad in June is classified as a fall spawner. TELEPHONE SURVEY A telephone survey has been conducted annually since 1986 to collect information on the fixed gear fishery and opinions on abundance trends (details in LeBlanc and LeBlanc 1996). The sGSL was divided into eight telephone survey areas corresponding to the areas where the major fisheries occur (Figure 1c). Active commercial licence holders were asked a series of questions concerning the number, dimensions, and mesh size of nets used, the frequency of fishing and how the abundance in the current year compared to the previous year and the medium-term trend. A 2008 review of the consistency of the abundance relationship among years concluded that this index should not be used as a biomass index in the population model. The telephone survey responses inform the fishing effort calculation for the CPUE in the gillnet fishery. The 2020 fixed gear telephone survey contacted 251 fishermen randomly selected out of approximately 421 active commercial licence holders in both seasons combined. A total of 37 fishermen responded to the spring fishing season survey and 139 fishermen responded to the fall fishing season survey for a total of 176. The 2021 fixed gear telephone survey contacted 269 fishermen randomly selected out of approximately 452 active commercial licence holders in both seasons combined. A total of 55 fishermen responded to the spring fishery survey and 130 fishermen responded to the fall fishery survey for a total of 185. The distribution of respondents across the 8 telephone survey areas, mean net hauls, net lengths, and trend in the abundance from the previous year are shown in Table 3. Overall, fishermen felt that abundances in the 2021 spring fishery were slightly higher than those of 2020 and to those in the previous years. For the fall fishery there was a sense that the 2020 abundance in both the North and Middle regions has increased slightly compared to 2019, and decreased in the South. When comparing 2021 to 2020 in the fall fishing season, the North region respondents indicated a status quo, the Middle region a slight decrease and a strong increase in the South (Table 3). In each year, the data source (either DMP or phone survey) with the greater number of responses was used to calculate the fixed gear CPUE abundance index. In the spring fishery, mesh sizes of gillnets has been relatively constant at 2½”. In the fall fishery, 2⅝” mesh is the most common. However, many fishers started using bigger mesh sizes (2¾”) in 1992. By 2002, the proportion of 2⅝” mesh reverted to pre-1992 numbers. The proportion of 2⅝” mesh in 2020 and 2021 was 100% (Table 3). FISHERY SAMPLING Commercial fishery catches are sampled dockside by DFO scientific personnel for the fixed and mobile fisheries, and at sea by fisheries observers in the mobile fishery. Sampling procedures are designed to obtain samples that are spatially and temporally representative of landings. The landings and samples by area used to calculate catch-at-age are shown in Table 2. The samples are used to determine the size, age, and spawning component (spring spawners or fall spawners) composition of the catch. Yearly age reading consistency tests are done in order to evaluate and ensure the consistency of age reading over time (Appendix A). 5 FISHERY-INDEPENDENT ACOUSTIC SURVEY Since 1991, an annual fishery-independent acoustic survey of early fall (September-October) concentrations of Herring has been conducted in the sGSL. The standard annual survey area occurs in the 4Tmno areas where both NAFO Div. 4T Herring spawning components aggregate in the fall. The survey uses a random stratified design of parallel transects within predefined strata. Surveys are conducted at night and use two vessels: an acoustic vessel to quantify the fish schools biomass using a hull-mounted 120 KHz split-beam transducer, and a fishing vessel to sample aggregates of fish with a pelagic trawl (details in LeBlanc et al. 2015; see also LeBlanc and Dale 1996). The acoustic survey covered a total transect distance of 886 km and 1,022 km in 2020 and 2021, respectively (Appendix B Figure B1). All strata were covered in 2021, but in 2020 the northern strata along the coast of New-Richmond, with historically low abundance of fish, were skipped due to time constrain and vessel availability. The trawl samples are used to separate the estimated biomass by spawning component and age, determine species composition, and size distribution for the estimation of the target strength (LeBlanc and Dale 1996; LeBlanc et al. 2015). EXPERIMENTAL NETS Part of an industry partnership project between DFO and fishery associations, experimental gillnets consisting of multiple panels of varying mesh size were weekly deployed by fishermen during the fall fishing season. These modified gillnets catch a wider range of fish sizes and provide information on the relative selectivity of various mesh sizes. Each experimental gillnet had five panels, each with a different mesh size, from a set of seven possible mesh sizes, ranging from 2” to 2¾” in ⅛” increments. All gillnets had panels with mesh sizes of 2½”, 2⅝”, and 2¾”, plus two smaller mesh sizes that varied among fishermen. Harvesters in the fall fishing season participated in the study on the following spawning grounds (Figure 1a): Miscou Bank (North region; 16B), Gaspé (North; 16B), Escuminac (Middle; 16C), West PEI (Middle; 16E), Fisherman’s Bank (South; 16G), and Pictou (South; 16F). The target fishing procedure was a one hour soak and nets were set on the fishing grounds during the commercial fishery. Data from Pictou prior to 2015 were corrected for gillnet depth as nets in this region were 5 m (17 ft) deep compared with the standard 2.4 m (8 ft) used on other spawning grounds. A correction factor of 8/17 (in ft) was applied to the Pictou nets to address the difference in net depth size. Catches from the experimental nets has been used to estimate the relative size-selectivity of gillnets of different mesh sizes (details in Surette et al. 2016) and to produce PAA. Both are inputs to the fall spawners assessment model. SPAWNING GROUND ACOUSTIC SURVEYS In 2015, a spawning ground acoustic survey that follows the design of the fishery-independent acoustic survey described above was initiated. This survey is the result of a partnership between DFO and fishery associations. The survey design uses random parallel transects within predefined strata that cover the same spawning grounds as the experimental nets (Appendix C). Surveys are conducted by fishermen in the fall fishing season according to protocols developed by DFO. The survey is conducted at night, during the weekend fishery closures except in Herring fishing area 16C and 16E in 2015 to 2017 (Middle; Figure 1a), where this region didn’t have weekend closures. The spawning ground acoustic survey is meant to provide a nightly estimate of spawning biomass among regions. It is analyzed in the same manner as the fishery-independent acoustic survey. The catches from the experimental nets are used to calibrate the spawning group specific target strength in order to obtain the nightly estimates of spawning biomass. 6 In this assessment, this biomass index has been incorporated into the fall population model for the first time. The detailed results of the 2020-2021 surveys are available in Appendix C. MULTISPECIES BOTTOM-TRAWL SURVEY The annual multi-species bottom trawl survey, conducted each September since 1971, provides information on the abundance and distribution of NAFO Div. 4T Herring throughout the sGSL in September (Savoie 2014). Total catch weights and numbers, representative length frequency and representative individual length-weight data has been recorded for each fish species in each survey set since 1971. Since 1994, additional sampling of Herring catches has been undertaken to disaggregate catches by spawning group and age (additional details in Hurlbut and Clay 1990). Herring were primarily caught near shore in waters < 30 fathoms, mostly off northeast P.E.I., west of Cape Breton, as well as in the Northumberland Strait, and Baie-des- Chaleurs (Appendix D Figure D1). ECOSYSTEM INFORMATION The abundance of major predators of Herring has changed over the time-series of the assessment. Abundance information for age 5+ Atlantic Cod and for Grey Seals was obtained from Neuenhoff et al. 2019. Atlantic Bluefin Tuna abundance information specific to the sGSL was obtained from the rod and reel CPUE index in ICCAT 2020. White Hake abundance data was obtained from Rolland et al. (2022), and mature Northern Gannet abundance data was obtained from (Rail 2021). Missing values in northern Gannet time series were obtained using linear interpolation (zoo R package, Zeileis et al. 2021). The proportion of immature Gannets, who also consume Herring, in the population was estimated to be 28% (J-F Rail, personal communication). Annual Gannet abundance was calculated by adding the equivalent of 28% of the mature Gannet population to the yearly abundance estimate (Benoît and Rail 2016). As predator data were in different units, values of abundance indices for each predator and natural mortality estimates were scaled by subtracting the mean and dividing by the standard deviation of the individual data vector. Correlation between variables was assessed using a correlation matrix, bivariate scatterplots and the Pearson’s correlation coefficient. Environmental effects on spring spawning Herring recruitment, stock-recruit relationships and population projections were assessed in Turcotte 2022. The GSLEA R package (Duplisea et al. 2020) was used to obtain a matrix of environmental variables for the ecosystem approach region 5 (Magdalen Shallows). For the assessment, time-series of zooplankton abundance data was only available for years 2001 to 2019. Recruitment results interpretation for spring spawning Herring were based from Turcotte 2022. Environmental effects on fall spawning Herring was assessed in a qualitative manner, using recent literature to infer the relationship between recent estimates of recruitment and recruitment drivers. INPUTS AND INDICES CATCH-AT-AGE AND WEIGHT-AT-AGE MATRICES Catch-at-age and weight-at-age matrices for 4T Herring spring spawners and fall spawners include catches from both fixed and mobile gear fleets. These were calculated using age-length keys and length-weight relationships for each spawning component, gear type, and fishing season (Table 2). For missing length cells, the age-length keys were completed by assigning a distribution of probability of an age based on data available for each season in a defined strata. Historically, when fewer than 30 fish were sampled for detailed analysis, the overall length- weight relationship and age-length key most similar and adjacent in gear, geography, and time were used to estimate the catch-at-age. However for both 2020 and 2021, that threshold was 7 decreased to 25 to compensate for the lack of specimens in some samples. Catch-at-age and weights-at-age are presented for fixed gear (spring spawners: Table 4-Table 5, fall spawners: Table 6-Table 7) and mobile gear (spring spawners: Table 8-Table 9, fall spawners: Table 10- Table 11). The dominant age in the 2020 spring spawners catch was age 7 belonging to the 2013 year- class. In 2021 the dominant age was the same year-class, now age 8 (Table 4 and Table 8; Figure 4). For fall spawners, the dominant age was 7 (2020) and 8 (2021) in the North (2013- 2014 cohorts), age 8 in the Middle for both years (2011-2012 cohorts), ages 7 to 8 in 2020 (2013 to 2012 cohorts) and age 8 in 2021 (2013 cohort) in the South (Table 6 and Table 10; Figure 5). Beginning of year weights-at-age are calculated from the weight-at-age for fixed and mobile gear combined. For age a at the beginning of year t, it is the geometric mean of weight-at-age a- 1 in the fishery in year t-1 and the weight-at-age a in the fishery in year t. Mean weight-at-age of the spring spawners caught in the mobile and fixed gears in the spring season have declined since the 1990s for mobile gears, and since the mid-1980s for the fixed gears (Table 5 and Table 9; Figure 6). The average weight-at-age declined by 39.6% between 1978 and 2021. Mean weight-at-age of fall spawning Herring from fixed and mobile gears has declined almost continuously over the time period 1978 to 2015 and has then stabilized until 2021 (Table 7 and Table 11; Figure 6). The mean weight-at-age declined by 30.2% between 1978 and 2021. Mean weight-at-age is an indication of stock status, affecting stock biomass for a given stock abundance. Similar to the previous assessment, seiner catch from 4vn was re-distributed to the North, Middle and South regions in proportion to the region’s fixed gear landing. Historically, re- distribution was based on seiner landings in each region, resulting in regions without seiner landings receiving no catch redistribution from 4Vn seiner landings. Similarly, seiner catch from the edge fishery was re-distributed to North, Middle and South regions in proportion to their fixed gear landings. Prior to the last assessment, these landings were all attributed to the South region. CATCH-PER-UNIT EFFORT The fixed gear fisheries occur on the spawning grounds. Landings from this fishery in 2020 account for approximately 59% of the spring spawners catch and 97% of the fall spawners catch. In 2021, this fishery account for more than 98% for both spawning groups. Fixed gear catch and effort data were used to construct CPUE abundance indices for spring and fall spawners. The fixed gear CPUE indices are defined as catches in kg/net-haul/day (or kg/net- haul/trip). Before 2014, a default 15 fathoms (27.4 m) net length was used when the information was not recorded, while starting in 2014 a value of 14 fathoms (25.6 m) is used. For all years all net length have been standardized to 14 fathoms. Total CPUE indices and PAA for ages 4-10 are used in the assessments for both stocks. Catch data were taken from the landings data. From 1990 to 2021, spring fishing season use landing data from DMP and complete the missing statistical districts with landings from purchase slips and ZIFF (Claytor et al. 1998, LeBlanc et al. 2002). Since 1978, fall fishing season use landing data from purchase slip and ZIFF. Fishing effort was calculated as the average number of gillnets deployed by season and area for the sGSL since 1978. From 1978 to 1985, the average number of nets used was collected by questionnaires done on wharves and by mail (Clay and Chouinard 1986). Since 1986, the fishing effort was calculated as the number of trips (purchase slips) multiplied by the estimated number of standard net hauls, which were determined from the DMP records (since 1990, see LeBlanc et al. 2008) and the annual telephone survey depending on which has the most data (Table 3). The number of hauls, 8 available since 1986, is used only for the fall fishing season (Claytor et al. 1998; LeBlanc et al. 2009). The percent of fixed gear fishing days with no catch has been recorded since 2006 based on responses to the telephone survey (Table 12). The percentage of days without catch in spring for both 2020 and 2021 was 24.3%, which is below the average of 32.7%. In the fall, the days without catch are still among the highest in the time series for the both years of the fall fixed gear fishery at 37.3% while the average is at 29.3%. As this information is only available for the most recent period, it is not yet included in the calculation of fishing effort. A multiplicative model (GLM) was used to calculate the standardized CPUE indices, based on the following formulation: 𝑙𝑙𝑙𝑙�𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑖𝑖𝑖𝑖𝑖𝑖� = 𝛼𝛼 + 𝛽𝛽1𝐼𝐼 + 𝛽𝛽2𝐽𝐽 + 𝛽𝛽3𝐾𝐾 + 𝜖𝜖 where 𝐼𝐼 indexes year, 𝐽𝐽 indexes herring management area by province, 𝐾𝐾 indexes week and ∈ is the residual error. For spring, data was aggregated by day and area and weighted by the catch for that area. For fall, data was aggregated by week. For the spring spawners, the model was applied to the data for the whole stock area. For the fall spawners, GLMs were run by region (North, Middle, and South) and did not include the area term. The spring spawner analysis was limited to weeks 11 to 22, whereas the fall spawner analysis was restricted to weeks 27 to 43 (see table 19 in LeBlanc et al. (2012)). Days by area (for spring) or by region (for fall) with less than 5 trips were also removed from the analysis. In order to improve the year to year repeatability of the CPUE estimations used by the population model, the historical method using SAS (Statistical Analysis Software) was translated into R (R Foundation for Statistical Computing Platform) programming language. The similarities between both methods are presented in Appendix E and the new method is proposed to become the standard for future assessments. This assessment uses the historical method. The models explained 40% of the variance in the spring data and the factors for year, week, and area were statistically significant. For the fall data, models explained between 51% and 69% of the variance in the data and the factors for year and week were statistically significant (Table 13). Age-specific CPUE indices for ages 4 to 10 was derived by dividing the gillnet catch- at-age by the standardized effort (CPUE) from the multiplicative GLM model. The CPUE age- specific abundance index included the years 1990 to 2021 for spring spawners and 1986 to 2021 for fall spawners. The indices presented in Table 14-Table 15 and Figure 7-Figure 8 account only for catch and effort, and do not account for possible changes in selectivity or catchability, which are addressed as part of the population modelling. The CPUE index for spring and fall spawners shows internal consistency as the abundance of cohorts is correlated between years (Figure 7- Figure 8). Fixed gear catches of spring spawners were composed mostly of ages 5 to 7 for 2020 and ages 6 to 8 for 2021 (Table 4). The CPUE of spring spawners in 2020 and 2021 has increased compared to the low values of 2018-2019 and for ages 7 and 8 has returned to the higher values observed in 2017. For 2021 the dominant ages were 7 and 8 (2013-2014 cohorts, Table 14; Figure 7). In the North region, catches of fall spawners in 2020 were dominated by ages 7 to 9 (2011 to 2013 cohorts), while in 2021 age 8 (2013 cohort) was the most abundant. In the Middle region, catches of fall spawners in 2020 were dominated by ages 7-9 (2011-2013 cohorts), and in 2021 age 8 (cohort 2013) was the most abundant. In the South region, catches of fall spawners in 2020 and 2021 were dominated by ages 7 to 8 and 6 to 9, respectively (2011 to 2015 cohorts; Table 6). Except for the South region in 2021, overall catch in all three regions is much lower compared to the last assessment period of 2018-2019. The CPUE of fall spawning Herring increased in 2020 for both the North and Middle regions but decrease in the South. In 2021, the CPUE decreased in the North, but increased in the Middle and South 9 regions (Figure E2, SAS method). Across regions, the CPUE of fall spawning younger fish (ages 4 and 5) has remained low since 2011, although the values are slightly higher for both North and Middle regions compared to 2018-2019 (Table 15; Figure 8). FISHERY-INDEPENDENT ACOUSTIC SURVEY INDEX A second standardized abundance index is generated from the annual fishery-independent acoustic survey. This index includes catch-at-age data from NAFO areas 4Tmno which has been surveyed yearly since 1994. The age-disaggregated acoustic abundance index for ages 2 to 10 for spring spawners and fall spawners is presented in Table 16. The 2020 and 2021 acoustic biomass index of the 4Tmno areas for spring and fall spawners combined were 30,081.8 t, and 37,953.1 t, respectively. In 2020, the biomass was composed of 30% spring spawners and 70% fall spawners. In 2021, the biomass was composed of 37% spring spawners and 63% fall spawners. A summary of the acoustic survey results is available in Appendix B. The spring spawner assessment model uses results for ages 4-8. For 2020 and 2021, the acoustic survey estimated that catch rates (in numbers) of spring spawners ages 4 to 8 were overall slightly higher than those observed in 2018 and 2019. The catch was dominated by ages 4 and 6 in 2019, ages 5 and 7 in 2020 and age 4 in 2021, indicating the 2013 cohort was relatively strong, as also seen in the CPUE index but also that the 2017 cohort appears to be stronger than expected. The observed trend is consistent with the low numbers experienced since the early 2000s (Table 16; Figure 9). For the fall spawner assessment model, the acoustic survey provides an abundance index of recruiting Herring (ages 2 and 3; LeBlanc et al. 2015). It is not thought to provide a useful abundance index for older ages given that the survey is limited to a restricted portion of the sGSL at a time when older Herring are spawning in areas throughout the sGSL. The acoustic abundance of ages 2 and 3 were much higher in both 2020 and 2021 than those of 2019, with the most abundant being age 3 (2017-2018 cohorts) in both years (Table 16; Figure 9). SPAWNING GROUND ACOUSTIC SURVEYS The sampling effort varied between regions and years, generating data with missing values, which can create biased biomass estimates when the mean annual value is calculated. To account for missing samples, a predictive model of nightly Herring biomass by year, region and Julian day was used to obtain a complete data grid and produce unbiased biomass indices (Turcotte et al. 2022). The average North region nightly biomass showed a general decline through the time series, from a peak of 7,667 t in 2016 to 600 t in 2021. The Middle region has seen a slower decline than the North region, with more interannual variation. Average nightly biomass declined from 3,175 t in 2015 to 1,036 t in 2021. The South region average nightly biomass declined between 2015 (3,563 t) and 2018 (335 t), but then increased until 2021 to reach a value of 2,816 t (Figure 10). EXPERIMENTAL NET INDICES Relative selectivity index A relative selectivity index was developed to account for changes in the proportion of 2⅝”, and 2¾” meshes used by commercial fishermen (Figure 12), as well as changes in mean length-at- age which have generally decreased over time. Selectivity-at-age (Table 17) and selectivity- adjusted CPUE calculations are described in the fall spawner model below. 10 Catch-at-age of experimental nets Similar to the previous assessment, the observed catch-at-length of each mesh size was summed per day and per region, and then the mean catch-at-length per region and per year was calculated. The catch-at-age data was then constructed using age-length keys as described above. The selectivity of the different mesh sizes was dealt with within the model (see fall spawner model). The experimental net index catch-at-age shows a greater proportion of fish ages 3 to 4 until 2009, after which the numbers decline. No major trend was observed in older Herring over the time series. No data was available for the North region in 2021 and in 2020 the proportion in the catch-at-age was much less than what was observed in 2018 and 2019. For both middle and south regions, proportions in the catch-at-age show greater catches of fish ages 5 to 8 (Figure 11). The fall spawning Herring population model uses proportions-at-age from the catch- at-age in experimental nets as a data input for years where the spawning ground acoustic survey are available (2015 to 2021). MULTISPECIES BOTTOM TRAWL INDEX This index consists of an age-disaggregated index using data from 1994-2021 for the fall spawners only (Table 18; Figure 13). Since the last assessment, the diel adjustment factor was not used to calculate the bottom-trawl survey index (see Turcotte et al. 2021b for details). The spatial distribution since 1971 is provided in Appendix D. The annual stratified mean catch-at-age values (standardized for tow distance) from the survey were used to produce an index of abundance. The results suggest an increase to relatively high abundance of ages 4-6 in 2010-2014 followed by a steady decline to very low abundance of these ages down to 2020 and an increase in 2021 to values previously observed in 2017 (Figure 13). MATURITY OGIVE For the purposes of the assessment, Herring are assumed to follow a knife-edged maturity schedule, with 100% maturation occurring between the ages of 3 and 4. SPRING SPAWNER COMPONENT ASSESSMENT Similar to last assessment, a SCA model with time-varying parameters was used. Such SCA model 1) assumes that there is observation error in the PAA in the fishery catches, 2) fits to the age-aggregated biomass indices and to the PAA in the fishery and survey catches; which accounts for the lack of independence between catches at different ages in the same year, and 3) is forward projecting from abundance-at-age in the first year and at the first age in all years. The model allows fishery catchability and natural mortality to vary over time which ensure the best fit to indices, minimized the residuals and showed no retrospective pattern in SSB estimates (Turcotte et al. 2021a). Fisheries stock assessment is often based on the assumption that natural mortality is constant through time, yet numerous examples show that predator-prey interactions are dynamic. Failure to account for increases in natural mortality due to changes in predator-prey interactions in stock assessment can result in biased estimates of population parameters and vital rates. Natural mortality also includes mortality from disease and unreported catches, including the bait fishery removals, for which no information is available. This component of the fishery has raised many questions over the year and is included in the assessment, although its effect cannot be distinguished from other sources of mortality. Disease 11 mortality is expected to be a low fraction of total natural mortality, as no mortality event due to disease were recorded during the time series. SPRING SPAWNER MODEL The SCA model of the spring spawners component was implemented using AD Model Builder (Fournier et al. 2012). Data inputs to the model included: • total fishery catches, and catches-at-ages 2 to 11+ from 1978 to 2021 in PAA; • catch-per-unit-effort (CPUE) index PAA and age-aggregated biomass index from 1990 to 2021 (ages 4 to 10); • fishery-independent acoustic survey index PAA and age-aggregated biomass index from 1994 to 2021 (ages 4 to 8). For yearly PAA in all data sources, where PAA was smaller than 0.01, plus or minus groups were created with adjacent ages until PAA was greater than 0.01. Estimated model parameters included the numbers-at-age in the initial year (1978), yearly recruitment (average recruitment and yearly recruitment deviations in numbers of age 2 fish), selectivity parameters in three time blocks to account for changes in selectivity and gear proportion in the catch, initial fishing mortality prior to 1978, CPUE and acoustic survey q and yearly q deviations for the CPUE index, initial M and yearly M deviations for two age groups (2-6 and 7-11+) and the observation error to the indices. All parameters were estimated on the log scale. Independent time-series of M for two age groups were estimated: ages 2-6 (𝑗𝑗 = 1) and 7-11+ (𝑗𝑗 = 2). These time series were estimated on the log scale as random walks: log�𝑀𝑀𝑖𝑖,𝑡𝑡� = log𝑀𝑀𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑡𝑡 where 𝑡𝑡 = 1978 log�𝑀𝑀𝑖𝑖,𝑡𝑡� = log�𝑀𝑀𝑖𝑖,𝑡𝑡−1� + 𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑖𝑖,𝑡𝑡 where 𝑡𝑡 > 1978 𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑖𝑖,𝑡𝑡~𝑁𝑁𝑁𝑁𝑁𝑁𝑁𝑁𝑁𝑁𝑙𝑙(0,𝜎𝜎𝑖𝑖𝑀𝑀) where log (𝑀𝑀𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑡𝑡) and 𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑖𝑖,𝑡𝑡 are parameters estimated by the model. The M deviations (𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑖𝑖,𝑡𝑡) were assumed to be normally distributed with a mean of 0 and standard deviation 𝜎𝜎𝑖𝑖𝑀𝑀 fixed at 0.075 for all 𝑗𝑗. The random walk started in 1979. Priors were supplied for 𝑀𝑀𝑖𝑖𝑖𝑖𝑖𝑖𝑡𝑡. These priors were normally distributed with means of 0.2 and standard deviations of 0.1 for both age groups (i.e., 𝑀𝑀𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑡𝑡~𝑁𝑁(0. 2, 0. 1)). The model likelihood included penalty terms due to the priors on M: 0. 5� (𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑖𝑖,𝑡𝑡 2 𝑖𝑖,𝑦𝑦 )/(𝜎𝜎𝑖𝑖𝑀𝑀)2 + 0. 5� exp (log�𝑀𝑀𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑡𝑡� − 0. 2)2/0. 12 𝐽𝐽 The model allowed for process error in fully-recruited catchability (q) to the fixed gear fishery. The initial value of q in 1990 (the first year with CPUE data) was a model parameter and the subsequent values of q were estimated as a random walk: 𝑞𝑞𝑡𝑡 = exp (log 𝑞𝑞) where 𝑡𝑡 = 1990 𝑞𝑞𝑡𝑡 = 𝑞𝑞𝑡𝑡−1 ∗ exp (𝑞𝑞𝑀𝑀𝑀𝑀𝑀𝑀𝑡𝑡) where 𝑡𝑡 > 1990 𝑞𝑞𝑀𝑀𝑀𝑀𝑀𝑀𝑡𝑡~𝑁𝑁𝑁𝑁𝑁𝑁𝑁𝑁𝑁𝑁𝑙𝑙(0,𝜎𝜎𝑞𝑞) 12 where log(𝑞𝑞𝑡𝑡) and 𝑞𝑞𝑀𝑀𝑀𝑀𝑀𝑀𝑡𝑡are parameters estimated by the model. The q deviations (𝑞𝑞𝑀𝑀𝑀𝑀𝑀𝑀𝑡𝑡) were assumed to be normally distributed with a mean of 0 and a standard deviation 𝜎𝜎𝑞𝑞 fixed at 0.1. The model likelihood included a penalty term due to the prior on the q deviations: 0. 5� (𝑞𝑞𝑀𝑀𝑀𝑀𝑀𝑀𝑡𝑡2 𝑡𝑡 )/(𝜎𝜎𝑞𝑞)2 Selectivity 𝑆𝑆𝑔𝑔,𝑎𝑎,𝑡𝑡 was indexed by catch source g, age a and year t. Fishery selectivity (g =1), selectivity to the CPUE in the gillnet fishery (g =2) and to the acoustic survey (g =3) were assumed to be logistic functions of age. It could be argued that selectivity to the CPUE index and to the fishery may be dome shaped due to the use of gillnets. Selectivity models that allowed for a dome shape (e.g., double logistic, gamma, exponential logistic) were also examined and they did estimate that selectivity was dome shaped. The descending limb of the dome was steeper and declined to a lower level in the 2005-2017 period than in the 1990-2004 period. For example, using the above three selectivity models, selectivity-at-age 10 in the gillnet fishery was estimated to be about 0.5, 0.8 or 0.9 in 1990-2004 respectively and 0.2, 0.2 and 0.8 in 2005 to 2017 (see Turcotte et al. 2021a Appendix 2 for details). However, size-at-age of herring has been declining since the mid-1980s (Figure 6). If selectivity was dome-shaped, old herring (e.g., age-10) would be on the descending limb. Consequently, decreases in size-at-age would increase their selectivity to the gillnet gear, not decrease it. Independent estimates of relative selectivity-at-age of fall spawners confirms that their selectivity at older ages has increased, not decreased, as their size-at-age has declined. Declining abundance at old ages that is not accounted for by fishery catches and estimated natural mortality can be spuriously accounted for by estimating declining selectivity at old ages. Consequently, these estimates of declining selectivity for older herring in recent years were judged to be spurious and the decision was made to use logistic selectivity models. For the commercial fishery and the CPUE index, separate selectivity functions were fit to three time periods: 1. 1978 to 1989 (p =1), 2. 1990 to 2004 (p =2), and 3. 2005 to 2021 (p =3) (i.e 𝑆𝑆1,𝑝𝑝 = 𝑓𝑓�𝑠𝑠1,𝑎𝑎,𝑡𝑡� and t ∈ 1978,1979, ... ,1989 for p =1, etc.). These time periods were chosen based on an examination of the yearly fixed/mobile gear proportions in the commercial fishery. Population abundance-at-age 2 (recruitment) in year t was estimated based on log average recruitment (𝑅𝑅�) and annual recruitment deviations 𝑅𝑅𝑀𝑀𝑀𝑀𝑀𝑀𝑡𝑡: 𝑅𝑅𝑡𝑡 = exp (𝑅𝑅� + 𝑅𝑅𝑀𝑀𝑀𝑀𝑀𝑀𝑡𝑡) 𝑅𝑅𝑀𝑀𝑀𝑀𝑀𝑀𝑡𝑡~𝑁𝑁𝑁𝑁𝑁𝑁𝑁𝑁𝑁𝑁𝑙𝑙(0,𝜎𝜎𝑅𝑅) where 𝑅𝑅� and 𝑅𝑅𝑀𝑀𝑀𝑀𝑀𝑀𝑡𝑡 are parameters estimated by the model. The recruitment deviations (𝑅𝑅𝑀𝑀𝑀𝑀𝑀𝑀𝑡𝑡) were assumed to be normally distributed with a mean of 0 and standard deviation 𝜎𝜎𝑅𝑅 fixed at 0.5. For older ages a (a ∈ 3, 4, ... 11+) in year 1, population abundance was estimated by projecting cohorts forward from age 2 in year 1 minus (a-2) to their age in year 1, as follows. For abundance-at-age a ∈ 3,4, ... A-1 in year 1, where A is the last age (11+): 𝑁𝑁𝑎𝑎,1 = exp (𝑅𝑅� + 𝑅𝑅𝑀𝑀𝑀𝑀𝑀𝑀𝑎𝑎𝑟𝑟1 − � �𝑠𝑠𝑏𝑏,1𝐹𝐹𝐹𝐹 + 𝑀𝑀𝑏𝑏,1� 𝑏𝑏=𝑎𝑎−1 𝑏𝑏=2 ) 13 For abundance-at-age A in year 1: 𝑁𝑁𝐴𝐴,1 = exp (𝑅𝑅� + 𝑅𝑅𝑀𝑀𝑀𝑀𝑀𝑀𝐴𝐴𝑟𝑟1 − ∑ �𝑠𝑠𝑏𝑏,1𝐹𝐹𝐹𝐹 + 𝑀𝑀𝑏𝑏,1�𝑏𝑏=𝐴𝐴−1 𝑏𝑏=2 ) 1 − exp (−�𝑠𝑠𝐴𝐴,1𝐹𝐹𝐹𝐹 + 𝑀𝑀𝐴𝐴,1�) where 𝑁𝑁𝑎𝑎,1 is abundance-at-age a in year 1, 𝑅𝑅𝑀𝑀𝑀𝑀𝑀𝑀𝑎𝑎𝑟𝑟1 are recruitment deviations used to initialize abundance-at-age a in year 1, 𝑠𝑠𝑏𝑏,1 is fishery selectivity-at-age b in year 1, 𝐹𝐹𝐹𝐹 is fully-recruited fishing mortality for initializing abundance-at-age in year 1, 𝑀𝑀𝑏𝑏,1 is natural mortality-at-age b in year 1, and b indexes age in the summations. The model likelihood included penalty terms due to the priors on the recruitment deviations used to initialize abundance-at-age 2 in all years and at older ages in year 1: 0. 5� (𝑅𝑅𝑀𝑀𝑀𝑀𝑀𝑀𝑡𝑡2 𝑡𝑡 )/(𝜎𝜎𝑅𝑅)2 + 0. 5 � (𝑅𝑅𝑀𝑀𝑀𝑀𝑀𝑀𝑎𝑎𝑟𝑟𝑖𝑖)2/(𝑠𝑠𝑅𝑅)2 𝑎𝑎 After recruitment to age 2, cohorts were projected forward in the usual manner: 𝑁𝑁𝑎𝑎,𝑡𝑡 = 𝑁𝑁𝑎𝑎−1,𝑡𝑡−1 × exp (−𝑍𝑍𝑎𝑎−1,𝑡𝑡−1) 𝑍𝑍𝑎𝑎,𝑡𝑡 = 𝑠𝑠1,𝑎𝑎,𝑡𝑡 × 𝐹𝐹𝑡𝑡 + 𝑀𝑀𝑎𝑎,𝑡𝑡 where a and t index age and year, 𝑁𝑁 denotes abundance, 𝑍𝑍 is total mortality, 𝑀𝑀 denotes natural mortality, 𝐹𝐹 is fully-recruited fishing mortality and 𝑠𝑠1,𝑎𝑎,𝑡𝑡is selectivity-at-age a in year t in the fishery. The objective function for the model included the following components: • discrepancies between observed and predicted values of the age-aggregated biomass indices for the CPUE in the gillnet fishery and acoustic survey. Indices were assumed to be lognormally distributed with standard deviations estimated by the model. The model allowed for weighing of the biomass indices likelihood, • discrepancies between observed and predicted PAA in the fishery, CPUE and acoustic survey catches. The PAA were assumed to follow a multivariate logistic distribution, which estimates data variances, • a normal prior for the log M deviations, • a normal prior for the initial values of log M, • a normal prior for the log q deviations, • a normal prior for the log recruitment deviations in years 1979 to 2021 and • a normal prior for the log recruitment deviations used to calculate abundance-at-age in 1978. Based on preliminary analysis of model fit to the age-aggregated indices, the CPUE biomass index likelihood was given a weight of one, while the acoustic biomass index likelihood was given a weight of three. Approximate 95% credible intervals were obtained for quantities estimated by the model based on 210,000 Markov chain Monte Carlo (MCMC) samples with the first 10,000 samples discarded and every 40th of the subsequent samples saved. Population estimates are posterior medians based on the MCMC sampling. Goodness-of-fit to indices was assessed by visual examination of estimated and observed aggregated biomass plots. Discrepancies between predicted and observed PAA were assessed by plotting the residuals by year and age, and looking for “blocking” through ages or years. Residuals were calculated in log space as observed values minus predicted values, minus the average difference by year. The 14 sum of squares of the residuals were calculated for each index of abundance. Retrospective patterns in SSB estimates were assessed by plotting SSB time-series estimated by sequentially removing the terminal year of data, for 4 years (2018 to 2021). SPRING SPAWNER RESULTS Residual patterns indicated an acceptable fit of the model to the age-disaggregated CPUE and acoustic indices, without apparent blocking (Figure 14). Fits to the age-aggregated indices are good for both the CPUE and acoustic indices (Figure 15). The SSB retrospective pattern analysis doesn’t show any progressive changes in a consistent direction as additional data are added to the model for the recent past (Figure 16). Catchability to the CPUE index averaged about 0.0019 in the early 1990s, increasing to a peak of approximately 0.0062 in 2007-2008, and stabilizing at 0.0056 on average between 2017 and 2021 (Figure 17). Estimated CPUE index catchability increased as the SSB declined (Figure 18). Natural mortality estimates for the age group 2-6 varied between 0.24 and 0.51 (between 21% and 40% annual mortality) over the time series (Figure 19). Estimates decreased slightly from 1978 to 1988, values were then relatively stable until 1995 when M increased to reach its highest values between 2000 and 2011. M decreased from 0.51 in 2009 to 0.27 in 2017, and has stayed at that level up to 2021. For the age group 7-11+, M increased gradually from 0.30 to 0.56 (between 26% and 43% annual mortality) between 1978 and 2006, before decreasing down to 0.47 (37% annual mortality) in 2009 (Figure 19). Starting in 2010, estimates sharply increased to reach a maximum of 1.05 (65% annual mortality) in 2018 before decreasing down to a mean value of 0.9 (59% annual mortality) in 2020 and 2021. Before the last assessment, models used to show estimates to the beginning of the year (January 1) while assuming a constant natural mortality of 0.2 (18% annually), meaning that SSB declined by only 5% between January 1 and April 1 (when the spring herring fishery started). Since the last assessment, the model uses time-varying natural mortality estimates, which has been very high in recent years. It is therefore important to account for the timing of the fishery in the estimates of stock status. Since the fixed-gear fishery is restricted to a limited period of the year, and M is estimated to be very high in some years for some ages of Herring, April 1 was used to estimate SSB, calculate the reference points, and to make projections. The limit reference point (LRP) in 4T Herring is Brecover, which is the lowest biomass from which the stock has been observed to readily recover. It is calculated as the average of the 4 lowest SSB estimates in the early 1980s (i.e., 1979-1982). Consequently, this value is model dependent. If the model changes, stock biomass may be re-scaled upwards or downwards. For this assessment, the LRP was estimated to be 46,340 t which is ~1.9% lower than the 47,250 t presented in the last assessment (Turcotte et al. 2021b). The upper stock reference (USR) was determined in 2005 as an interim reference point (Chouinard et al. 2005). Calculations used a yield per recruit analysis assuming M = 0.2 and specific partial recruitment vectors to the fishery that would not apply for the current model and SSB estimates based on time varying M. Consequently, since the last assessment, the USR was scaled upwards by the same proportion as the LRP. The historical USR was 54,000 t of SSB, and the re-scaled USR is 129,994 t. The LRP and USR were calculated to April 1 to account for three months of natural mortality for both age groups. The fishing removal reference in the Healthy Zone was defined as F0.1 and this assessment used the same value of 0.35 as used in previous assessments. 15 Estimated SSB increased from low levels in the early 1980s to highest levels in the mid-1980s to mid-1990s. SSB declined in the mid-1990s to reach the Critical Zone in 2002. SSB increased slightly until 2010, still in the Critical Zone, but then declined again and fluctuated around a mean value of 39,550 t until 2021. The MCMC estimates of April 1 SSB in 2020 and 2021 were 38,402 t (95% confidence interval: 23,771 – 69,893) and 35,626 t (95% CI: 22,012 – 66,950), respectively. The estimate for 2021 is 77% of the LRP. The probabilities that April 1 SSB was under the LRP (in the critical zone of the Precautionary Approach) were 23% in 2020 and 30% in 2021 (Figure 20). SSB has been declining since 2018. Estimated recruitment (number of age 2 fish) was highest in the early 1980s, 1990 and 1993 (Figure 21). Recruitment has been relatively stable at lower values since 1993, with slightly higher values between 2006 and 2008. Recruitment declined to lowest values of the time-series after 2008 up to 2020, except a small peak in 2015. Recruitment rate (number of age-2 fish per kg SSB) was highest in the early 1980s and around 2005, and at its lowest between 1992 and 2000. Since 2006, recruitment rates have declined to low values except for a small peak in 2013 and another in 2019 (Figure 23). Estimated abundances of recruits to the fishery (age-4 fish) were highest in the mid-1980s, 1992 and 1995 (Figure 22). The number of fishery recruits declined from 1995 to the lowest level observed in 2004 and has remained at a very low level since then (average 102.8 million Herring, Figure 22; Table 20). The 2020 MCMC median spawner (4+) abundance estimate is 284.5 million Herring (95% CI: 175.5 – 515.3), while the 2021 MCMC median is 250.2 million Herring (95% CI: 155.5 – 469.5) about 34.2% of the average spawner abundance in 1985 to 1995. Estimated fishing mortality (abundance weighted F6-8) was high in 1979-1980, decreased until 1984 and then increased steadily to 0.59 in 2004. F then decreased rapidly to a low value (0.03) in 2012 and has since remained at this low value. The lowest value was observed in 2021 (<0.02) (Figure 24; F values in Table 21). Fully recruited F6-8 median MCMC estimate was 0.025 (95% CI: 0.013 – 0.041) and 0.018 (95% CI: 0.009 – 0.030) in 2020 and 2021, respectively (annual mortality of 2.5% and 1.8%). The spring spawning Herring population trajectory with respect to SSB and fishing mortality levels is shown in Figure 25. The figure shows the Healthy, Cautious and Critical Zones of the Precautionary Approach. The removal reference in the Healthy zone for the spring spawning Herring stock is F0.1 = 0.35. There are no harvest control rules in the cautious and critical zone for this stock. The provisional Precautionary Approach removal reference is thus provided but may not be as restrictive as formally developed harvest control rules. Fishing mortality exceeded the removal reference level in 28 of the 44 years of the time series. Fishing mortality exceeded the Precautionary Approach removal reference in all years after 1998 and was especially high during and soon after the SSB decline, between 1999 and 2007. SPRING SPAWNER PROJECTIONS The population model was projected forward to 2023, 2024 and 6 years forward to 2027 during the MCMC sampling of the joint posterior distribution of the parameters. This takes into account uncertainties in the parameter estimates. Projections were conducted at several levels of annual catch (0, 250, 500 and 1,250 t). Recruitment has been stable at low values in recent years, projections were thus conducted using random recruitment values of the last five years (2017- 2021). Natural mortality for age group 2-6 has been stable for the last 5 years. For age group 7- 11+, natural mortality increased in the last decade to highest values in 2018 and 2019 and slightly decrease in 2020 and 2021 (Figure 19). Projections were thus conducted using the average of the 2017-2021 M values for each age groups. Two year projections of SSB to April 1 16 and abundance weighted fishing mortality for ages 6 to 8 are shown in Figure 26 and Figure 27, and the probabilities of meeting various objectives are given in Table 22 for each catch level, for six years. Six year SSB projections are shown in Figure 28. Projected April 1 2022 SSB is 28,835 t (95% CI: 17,255 – 55,772), keeping the stock in the Critical Zone of the Precautionary Approach. Short term projections At annual catches of 0, 250, 500 or 1,250 t in 2022 and 2023, SSB was expected to increase slightly from 2022 to 2023, and to remain stable from 2023 to 2024 (Figure 26, Table 22). The probability of an increase in SSB between April 1 2022 and April 1 2023 was between 64.5 and 68.5% at all catch levels. The probability of a greater than 5% increase in SSB between April 1 2023 and April 1 2024 was between 42.3% and 44.3% at all catch levels. For the short term projections, all catch levels (including no catch) resulted in under a 20% probability that SSB would exceed the LRP to reach the Cautious Zone in 2024. In the short term, there is no chance that the population would reach the USR by 2024. Catches of 250 t would result in abundance-weighted ages 6 to 8 fishing mortality (F) values of 0.017 in 2022 (1.7% annual mortality) and 0.016 in 2023 (1.6% annual mortality), which correspond to lower values than F in recent years. Catches of 500 t would result in F values of 0.034 in 2022 (3.3% annual mortality) and 0.032 in 2023 (3.1% annual mortality), values similar to recent F. Catches of 1,250 t would result in an increase in F from recent years, with values of 0.085 in 2022 (8.1% annual mortality) and 0.083 in 2023 (8.0% annual mortality) (Figure 27, Table 22). Long term projections Six years projections in SSB show no changes from 2022 to 2027. By 2027, the probability of exceeding the LRP was between 15.8 and 20.4% at all catch levels, with SSB values ranging between 32,477 and 35,445 t (Figure 28, Table 22). FALL SPAWNER COMPONENT ASSESSMENT FALL SPAWNER MODEL The fall spawning Herring component was assessed using a SCA model implemented using AD Model Builder (Fournier et al. 2012). This model estimates time varying CPUE catchability (q) and natural mortality (M) (Turcotte et al. 2021a). Data inputs to the models included: • fishery catches-at-ages 2 to 11+ by region from 1978 to 2021, in PAA, • catch-per-unit-effort (CPUE) PAA index and age-aggregated CPUE biomass index by region from 1986 to 2021 (ages 4 to 10), • PAA in experimental nets and the average nightly biomass from the spawning grounds acoustic survey by region from 2015 to 2021 (ages 3 to 9), • fishery-independent acoustic survey PAA and age-aggregated biomass index from 1994- 2021 (ages 2 and 3), • multispecies bottom trawl survey (RV survey) PAA index and age-aggregated biomass index across the sGSL from 1994 to 2021 (ages 4 to 6), 17 • the proportion of gillnets with 2 ⅝ inch mesh and the relat