5. Assessment of Greenland Turbot (Reinhardtius hippoglossoides) in the Bering Sea and Aleutian Islands Steven J. Barbeaux, James Ianelli, Dan Nichols and Jerry Hoff U.S. Department of Commerce National Oceanic and Atmospheric Administration National Marine Fisheries Service Alaska Fisheries Science Center 7600 Sand Point Way NE, Seattle, WA 98115-6349 Executive Summary Relative to last year’s assessment, the following changes have been made in the current assessment. Summary of Changes in Assessment Inputs Change in weight at length relationship A new weight at length relationship has been developed using the combined weight and length data from all bottom trawl surveys conducted by the Alaska Fisheries Science Center in the Bering Sea and Aleutian Islands from 1983 to 2011. The analysis presented at the September 2012 Plan team and October 2012 SSC meeting (Appendix 5.1) showed a decrease in estimated abundance due to this change of ~20%. Removal of the pre-2002 Slope survey biomass estimates Slope survey abundance index values for surveys conducted prior to 2002 were not included in this year’s model. These data were removed after discussions with the current Chief Scientist for the slope survey, Dr. Jerry Hoff. The earlier surveys differed in vessel power, in gear used, and in the ability to determine whether the gear was in contact with the bottom. Therefore the older Slope survey data were considered not comparable to the more recent surveys. Addition of new fishery and survey data There were new Slope, Shelf, and Auke Bay Laboratory (ABL) longline surveys in 2012. The abundance estimate (or RPN for the ABL longline survey) and length data were added to this assessment. Fishery catch and length frequency data were updated to the 2012 numbers. The 2009 through 2012 ABL longline survey length data have become available and added to the assessment. Changes in length and age composition data Fishery length composition data were treated differently this year than in previous years. The raw Trawl and Longline fishery length composition data were proportioned to catch numbers by haul to obtain a more accurate representation of the catch composition. The proportion (P) of fish for a particular length bin (l) and year (y) was calculated as 𝑛𝑦𝑙ℎ , where n is the ∑�𝑛𝑦ℎ𝑁𝑦ℎ� number of fish in a length bin (l) for an individual year (y) and h𝑃a𝑦u𝑙l =samp∑le𝑁 (𝑦hℎ) and N is the total number of fish in a haul (h) for year (y) for each fleet. Change in fishery multinomial sample sizes for the length data Initial sample sizes for the two fisheries for each year were determined as the minimum of 100 + (number of hauls sampled/mean number of hauls sampled/100) or the number of hauls sampled. This scheme was intended to reduce the influence of within sample and across haul autocorrelation in very large, single year, sample sizes on model fit. Change in recruitment estimation In this year’s assessment we explore four models. Model 1 is the 2011 reference model where recruitment was modeled as two separate Beverton-Holt stock recruitment (BH) curves with steepness of 0.9 and sigma R of 0.6, but with a difference in productivity (R ) between the early 0 recruitment (1965-1970) and later recruitment. Recruitment deviations were not estimated for the 1965-1970 recruitments and they were assumed to follow the BH curve with no error. Model 2 and Model 3 follow the models presented at the September 2012 Plan Team and October 2012 SSC meetings (Appendix 5.1). All recruitment is modeled as a single BH curve with either no autocorrelation (Model 2) or 0.6 autocorrelation (Model 3), steepness of 0.79 and sigma R of 0.6. Recruitment deviations are estimated separately for the pre-1975 and the later recruitment deviations. Recruitment is modeled back to 1945 to allow the model to ramp up to the estimated abundance levels needed to support the large pre-1975 fishery, but for which we have no length or age composition data. Model 3 follows suggestions by Dr. Grant Thompson to start the model in 1977 and ignore the pre-1977 catches where no length or age data were available. In this model recruitment was estimated as a BH curve with steepness of 0.79, sigma r of 0.6, and no autocorrelation. Recruitment deviations from 1977-1989 were estimated separately from the post-1989 recruitment where better length composition data were available. Changes in Selectivity for all fisheries and surveys There was focused effort to explore appropriate selectivity curves for the 2012 assessment. The main difference between the 2011 Reference model selectivity and the 2012 candidate model selectivities is in how the male and female selectivity curves were allowed to differ. A new method for fitting curves that differ between male and females was implemented in the latest version of SS3 (V 2.24). Summary of Results There was a major revision of the Greenland turbot stock assessment model and data for this year. The changes in the weight at age and selectivities had the net effect of reducing the current biomass estimate while increasing the reference points for this species. In addition to changes to the assessment model and data, there was a input error in 2009-2011 projection models that resulted in underestimates of the initial female spawning biomass (B ), and therefore all 100% biomass reference points. From the 2012 Authors’ preferred reference model (Model 2) this year’s estimate for B of 119,217 t is more than double last year’s estimate of 53,900 t, but 100% similar to the 2008 estimate of 109,328 t. The 2012 status of the stock is B , much lower than 21% last year’s projected status for 2012 of B and the 2008 estimate of B . The change in status 89% 52% was mostly due to fixing the input error and improvements in the shapes of the selectivity curves chosen in 2012. Due to these changes the stock is now in Tier 3b and therefore the ABC and OFL recommendations were further reduced by the descending portion in the control rule. The 2013 recommended ABC is only 26% of the projected 2013 ABC from last year’s model. However, the projected 2013 estimated total biomass in this year’s model is higher than projected from the 2011 Reference model. This is due to strong 2008 and an especially large 2009 year classes observed in both the survey and fisheries size composition data. These two year classes are expected to be larger than any other recruitment event since the 1970’s and will begin to have an increasing influence on spawning stock biomass starting in 2014. Model 2 estimated that the BSAI Greenland turbot fishery is not overfishing the stock, that the stock is not currently overfished, and that the stock is not approaching an overfished condition. It should be noted however, that Model 3 in this assessment estimates that the BSAI Greenland turbot stock is in an Overfished condition. The only difference between Model 3 and Model 2 is the inclusion of autocorrelation in the recruitment deviations. Model 3 is the best fitting model and the only reason this model was not selected by the stock assessment authors is due to the fact that inclusion of autocorrelation in SS3 has not yet been thoroughly vetted. As estimated or As estimated or specified last year for: recommended this year for: Quantity 2012 2013 2013 2014 M (natural mortality rate) 0.112 0.112 0.112 0.112 Tier 3a 3a 3b 3b Projected total (age 1+) 76,850 73,910 80,989 94,752 Female s pa wning biomass (t) Projected 47,687 41,441 23,485 26,537 B 53,900 53,900 119,217 119,217 100% B 21,560 21,560 47,686 47,686 40% B 18,870 18,870 41,726 41,726 35% F 0.453 0.453 0.14 0.16 OFL maxF 0.367 0.367 0.12 0.13 ABC F 0.367 0.367 0.12 0.13 ABC OFL (t) 11,658 9,697 2,539 3,266 maxABC (t) 9,660 8,029 2,064 2,655 ABC (t) 9,660 8,029 2,064 2,655 EBS 7,226 6,006 1,612 2,074 Aleutian Islands 2,434 2,023 452 581 As determined last year As determined this year Status 2010 2011 2011 2012 Overfishing No n/a No n/a Overfished n/a No n/a No Approaching overfished n/a No n/a No Responses to SSC and Plan Team Comments on Assessments in General Retrospective analysis From the December 2011 SSC minutes: The SSC is pleased to see that many assessment authors have examined retrospective bias in the assessment and encourages the authors and Plan Teams to determine guidelines for how to best evaluate and present retrospective patterns associated with estimates of biomass and recruitment. We recommend that all assessment authors (Tier 3 and higher) bring retrospective analyses forward in next year’s assessments. From the September 2012 Plan Team minutes: The Teams recommend that authors conduct a retrospective analysis back 10 years (thus, back to 2002 for the 2012 assessments), and show the patterns for spawning biomass (both the time series of estimates and the time series of proportional changes relative to the 2012 run). This is consistent with a December 2011 NPFMC SSC request for stock assessment authors to conduct a retrospective analysis. The base model used for the retrospective analysis should be the author’s recommended model, even if it differs from the accepted model from previous year. In response to these requests, we conducted a within-model retrospective analysis back 10 years using the recommended model (Model 2). Responses to SSC and Plan Team Comments Specific to this Assessment SSC Comments The SSC commends the assessment authors for their efforts to improve this assessment model and address SSC and Plan team concerns. The SSC looks forward to additional improvements in next year’s assessment. Authors - Thank you. Plan Team Comments For the November meeting, the Team recommends that the author present two or possibly three models: 1) a reference model, which is the original 2011 model with updated and corrected data; 2) an alternative model similar to the author’s preferred model from the preliminary assessment with a few modifications (see below for details); and 3) a third model of the author’s choosing, included at the author’s discretion. 1) Early recruitments. Noting the potential influence of catches from earlier years (i.e., 1960s) on reference points, the Plan Team recommends further evaluation of that influence by starting the model at different points in time with single large catches, rather than a time series of catches, and including this change in Model 2 for November at the author’s discretion and if the analysis can be completed in time. If this evaluation cannot be conducted in time for the November 2012 meeting, the Team recommends that it be completed for the September 2013 meeting. Authors - Three model configurations were explored beyond the 2011 Reference Model. The first two start in 1945 with all catch and the third starts the model in 1977 without previous catch. 2) Selectivity patterns The Plan Team recommends that only the logistic selectivity curve be used for the ABL longline survey in Model 2 for November. Authors - This was done. 3) Models with fitted catchability For November, the Plan Team recommends that the Model 2 estimate shelf survey catchability with as diffuse a prior as possible. The Team also recommends further exploration of alternative catchability assumptions for the September 2013 meeting. Authors - Model 2 has a lognormal prior on shelf catchability of ln(q) = -0.69385 and ln(St.Dev) = 0.4. Models 3 and 4 both have more restricted priors on shelf catchability with ln(q) = -0.69385 and ln(St.Dev) = 0.1. 4) Alternative values for Sigma R. For November, the Team recommends fixing Sigma R at a value of 0.6 in Model 2, while allowing a small amount of autocorrelation. Authors - All candidate models had sigma R = 0.6. Model 3 allowed for a small amount of autocorrelation (rho = 0.6) in the recruitment deviations. Introduction This year the BSAI Greenland turbot stock assessment will be lead by Dr. Steven Barbeaux. Although the stock will continue to be modeled using the same software as previous assessments (Stock Synthesis 3) there are a number of changes within the model. This is an attempt to better capture the complex population dynamics of this species due its unique life history and distribution across two geopolitical boundaries (the US-Russian EEZ and the Northern extent of the AFSC surveys). We will present the 2011 model configuration (Model 1) fit to the most recent data as well as three alternative candidate models (Model 2, Model 3, and Model 4) with special emphasis on the author’s preferred model (Model 2). Life History Greenland turbot (Reinhardtius hippoglossoides) is a Pleuronectidae (right eyed) flatfish that has a circumpolar distribution inhabiting the North Atlantic, Arctic and North Pacific Oceans. The American Fisheries Society uses “Greenland halibut” as the common name for Reinhardtius hippoglossoides instead of Greenland turbot. To avoid confusion with the Pacific halibut, Hippoglossus stenolepis, common name of Greenland turbot which is also the “official” market name in the US and Canada (AFS 1991) is retained. In the Pacific Ocean, Greenland turbot have been found from the Sea of Japan to the waters off Baja California. Specimens have been found across the Arctic in both the Beaufort (Chiperzak et al. 1995) and Chuchki seas. This species primarily inhabits the deeper slope and shelf waters (between 100 m to 2000 m; Fig. 5.1) in bottom temperatures ranging from -2°C to 5°C. The area of highest density of Greenland turbot in the Pacific Ocean is in the northern Bering Sea, straddling the border between US and Russian exclusive economic zones. Juveniles are believed to spend the first 3 or 4 years of their lives on the continental shelf and then move to the continental slope (Alton et al. 1988; Sohn 2009; Fig. 5.2). Adult Greenland turbot distribution in the Bering Sea appears to be dependent on size and maturity as larger more mature fish migrate to deeper warmer waters. In the annual summer shelf trawl surveys conducted by the Alaska Fisheries Science Center (AFSC) the distribution by size shows a clear preference by the smaller fish for shallower (< 100m) and colder shelf waters (< 0°C). The larger specimens were in higher concentrations in deeper (> 100 m), warmer waters (> 0°C) (Fig. 5.3 and Fig. 5.4). Juveniles are absent in the Aleutian Islands regions, suggesting that the population in the Aleutians originates from the EBS or elsewhere. In this assessment, Greenland turbot found in the two regions are assumed to represent a single management stock. NMFS initiated a tagging study in 1997 to supplement earlier international programs. Results from conventional and archival tag return data suggest that individuals can range distances of several thousands of kilometers and spend summer periods in deep water in some years and in other years spend time on the shallower EBS shelf region. Greenland turbot are sexually dimorphic with females achieving a larger maximum size and having a faster growth rate. For this assessment, data from the AFSC slope and shelf surveys were pooled to obtain growth curves for both male and female Greenland Turbot (Fig. 5.5). This sexual dimorphic growth is consistent with trends observed in the North Atlantic. Collections in the North Atlantic suggest that males may have higher mortality than females. Evidence from the Bering Sea shelf and slope surveys suggest males reach a maximum size much smaller than females, but that mortality may not be higher than in females. Prior to 1985 Greenland turbot and arrowtooth flounder were managed together. Since then, the Council has recognized the need for separate management quotas given large differences in the market value between these species. Furthermore, the abundance trends for these two species are clearly distinct (e.g., Wilderbuer and Sample 1992). Fishery Catches of Greenland turbot and arrowtooth flounder were not reported separately during the 1960s. During that period, combined catches of the two species ranged from 10,000 to 58,000 t annually and averaged 33,700 t. Beginning in the 1970s the fishery for Greenland turbot intensified with catches of this species reaching a peak from 1972 to 1976 of between 63,000 t and 78,000 t annually (Fig. 5.6). Catches declined after implementation of the MFCMA in 1977, but were still relatively high in 1980-83 with an annual range of 48,000 to 57,000 t (Table 5.1). Since 1983, however, trawl harvests declined steadily to a low of 7,100 t in 1988 before increasing slightly to 8,822 t in 1989 and 9,619 t in 1990. This overall decline is due mainly to catch restrictions placed on the fishery because of apparent low levels of recruitment. From 1990- 1995 Council set the ABC’s (and TACs) to 7,000 t as an added conservation measure citing concerns about recruitment. Since 1996 the ABC levels have varied but averaged 6,540 t (with catch for that period averaging 4,468 t). The majority of the catch over time has been concentrated in deeper waters (> 150 m) along the shelf edge ringing the eastern Bering Sea (Fig. 5. 7 and Fig. 5. 8), but Greenland turbot has been consistently caught in the shallow water on the shelf as bycatch in the trawl fisheries (Table 5.2 and Table 5.3). Catch of Greenland turbot is generally dispersed along the shelf and shelf edge in the northern most portion of the management area. Since 2008 however at a 400km2 resolution the cells with the highest amount of catch have been in the Eastern Aleutian Islands (Fig 5.9), suggesting high densities of Greenland turbot in these areas. These areas of high Greenland turbot catch in the Aleutians are coincident with the appearance of the Kamchatka and arrowtooth flounder fishery. This fishery has the highest catch of Greenland turbot outside of the directed fishery. For 2008 and in the preliminary catch data for 2012, Greenland turbot catch in the Arrowtooth/Kamchatka fishery has exceeded the directed catch. In 2008 through 2012, trawl-caught Greenland turbot exceeded the level of catch by longline vessels (Table 5.3). The shift in the proportion of catch by sector was due in part to changes arising from Amendment 80 passed in 2007. Amendment 80 to the BSAI Fishery Management Plan (FMP) was designed to improve retention and utilization of fishery resources. The longline fleet generally targets pre-spawning aggregations of Greenland turbot; the fishery opens May 1 but usually occurs June-Aug in the EBS to avoid killer whale predation. Catch information prior to 1990 included only the tonnage of Greenland turbot retained Bering Sea fishing vessels or processed onshore (as reported by PacFIN). Discard levels of Greenland turbot have typically been highest in the sablefish fisheries (at about one half of all sources of Greenland turbot discards during 1992-2003) while Pacific cod fisheries and the “flatfish” fisheries also have contributed substantially to the discard levels (Table 5.2). About 9.2% of all Greenland turbot caught in groundfish fisheries were discarded (on average) during 2004-2012. The overall discard rate of Greenland turbot has dropped substantially in recent years from a high of 82% discarded in 1992 down to only 2% in 2011 and so far in 2012. By gear-type and region, trawl catch was most significant in the Aleutian Islands in 2009 through 2012 (Table 5.4), whereas in the EBS there was high trawl catch in 2008, but then a switch to higher longline catches in 2009 through 2012 (Table 5.3). By target fishery, the gain in trawl- fishery has occurred primarily in the Greenland turbot target fishery in 2009 and arrowtooth flounder/Kamchatka fisheries in 2008 - 2012 (Table 5.3). Data Fisheries data in this assessment were split into the Longline (including all fixed gear) and Trawl fisheries. Both the Trawl and Longline data include observations and catch from targeted catch and bycatch. There are also data from three surveys, the Shelf and Slope surveys are bottom trawl surveys conducted by the RACE Division of the Alaska Fisheries Science Center and the Auke Bay Laboratory (ABL) Longline survey has been conducted by the ABL out of Juneau, Alaska. The type of data and relevant years from each can be found in Table 5.5 and Figure 5.10. Fishery data Catch The catch data were used as presented above for both the longline and trawl fisheries. The early catches included Greenland turbot and arrowtooth flounder together. To separate them, the ratio of the two species for the years 1960-64 were assumed to be the same as the mean ratio caught by USSR vessels from 1965-69. Size and age composition Extensive length frequency compositions have been collected by the NMFS observer program from the period 1980 to 2012. The length composition data from the trawl and longline fishery are presented in the Appendix 5.2 (along with the expected values from the assessment model) and absolute sample sizes for the period of the domestic fishery by sex and fishery from 1989- 2012 are given in Table 5.6 Catch totals from research and other sources Annual research catches (t, 1977 - 2012) from NMFS longline and trawl surveys are estimated as follows: Year 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 NMFS BT surveys 62.5 48.3 103.0 123.6 15.0 0.6 175.1 26.1 0.5 18.5 0.6 0.7 11.4 0.9 1.4 8.5 1.4 Longline surveys 3 3 6 11 9 7 8 7 11 6 16 10 10 22 23 23 Year 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 NMFS BT surveys 1.5 4.6 1.4 1.0 6.6 1.1 6.6 1.1 12.8 0.7 3.0 0.6 4.8 0.4 6.6 1.0 4.9 Longline surveys 1.1 3.5 n/a n/a 0.36 n/a n/a An updated database for 2010 sport and research catches indicates the following for Greenland turbot: Source t 2010 Aleutian Island Bottom Trawl Survey 0.530 2010 Bering Sea Acoustic Survey 0.000 2010 Bering Sea Bottom Trawl Survey 0.816 2010 Bering Sea Slope Survey 5.210 2010 Northern Bering Sea Bottom Trawl Survey 0.004 Blue King Crab Pot 0.056 IPHC (halibut commission) 2.989 NMFS LL survey 0.364 Recent analyses examining the bycatch of Greenland turbot in directed halibut fisheries indicate an average of just over 109 t from 2001-2010 with about 49 t average since 2006 (NMFS Regional Office). EBS slope and shelf bottom trawl survey The older juveniles and adults on the slope had been surveyed every third year from 1979-1991 (also in 1981) as part of a U.S.-Japan cooperative agreement. From 1979-1985, the slope surveys were conducted by Japanese shore-based (Hokuten) trawlers chartered by the Japan Fisheries Agency. In 1988, the NOAA ship Miller Freeman was used to survey the resources on the EBS slope region. In this same year, chartered Japanese vessels performed side-by-side experiments with the Miller Freeman for calibration purposes. However, the Miller Freeman sampled a smaller area and fewer stations in 1988 than the previous years. The Miller Freeman sampled 133 stations over a depth interval of 200-800 m while during earlier slope surveys the Japanese vessels usually sampled 200-300 stations over a depth interval of 200-1000 m. In 2002, the AFSC re-established the bottom trawl survey of the upper continental slope of the eastern Bering Sea and a second survey was conducted in 2004. Planned biennial slope surveys lapsed (the 2006 survey was canceled) but resumed in the summer of 2008, 2010, and 2012 (Table 5.7). Although the size composition data for surveys prior to 2002 were used in this assessment the abundance estimates were not. This was decided after discussions with Dr. Jerry Hoff, the current Slope survey Chief Scientist in which Dr. Hoff stated that the older Slope survey data were not comparable to the most recent surveys, and may have not been conducted consistently enough in the early years to be considered a time series. The surveys differed in vessel power, in gear used, and in the ability of the surveyors to determine whether the gear was in contact with the bottom. The trawl slope-surveys are likely to represent under-estimates of the BSAI-wide biomass of Greenland turbot since fish are found consistently in other regions. A similar issue likely affects the distribution of Greenland turbot on the shelf region, particularly given the extent of the cold pool and warm conditions in recent years (Ianelli et al. 2011). The Shelf and recent Slope survey biomass estimates are therefore treated as a relative abundance index and a separate catchability parameter were fit for each. The estimated biomass of Greenland turbot in this region has fluctuated over the years. When US-Japanese slope surveys were conducted in 1979, 1981, 1982 and 1985, the combined survey biomass estimates from the shelf and slope indicate a decline in EBS abundance. After 1985, the combined shelf plus slope biomass estimates (comparable since similar depths were sampled) averaged 55,000 t, with a 2004 level of 57,500 t. The average shelf-survey biomass estimate during the last 19 years (1993-2012) was 24,600 t. The number of hauls and the levels of Greenland turbot sampling in the shelf surveys were presented in Table 5.8. In 2011 and 2010 the abundance estimates from the shelf surveys indicate a significant increase of Greenland turbot recruitment but also the proportion of tows with Greenland turbot present has increased (Fig. 5.11). These observations suggest that the extent of the spatial distribution has remained relatively constant prior to 2010 (with a slight increase) and that the most recent surveys have both higher densities and broader spatial distribution. Although the 2012 EBS slope biomass estimate of 17,984 t was down from 2010 estimate of 19,873 t, the population numbers in 2012 of 11,839,700 fish was more than double the 2010 estimate of 5,839,126 fish. The 2012 Slope survey abundance estimate was the highest population estimate since the Slope survey was reinstated in 2002. Most of the change in population estimates is due to the changes in Greenland turbot abundance found in the two shallowest strata between 200 and 600 m depth strata (Table 5.9 and Table 5.10). In the 200- 400 m strata the population was more than 8 times that of the 2010 survey estimate and the 400- 600 m strata was more than double the 2010 estimate. These high numbers, but low abundance is a reflection of the large number of smaller fish moving into the slope region from the shelf due to the large 2007 through 2009 year classes as evidenced by the large number of fish between 30 cm and 50 cm observed in this survey (Fig. 5.12). Survey size composition A time series of estimated size composition of the population was available for both surveys. The slope surveys typically sample more turbot than the shelf trawl surveys; consequently, the number of fish measured in the slope surveys is greater. The shelf survey appears to be useful