Estimates of Shrimp Trawl Bycatch of Red Snapper (Lutjanus campechanus) in the Gulf of Mexico

Transcription

1 Fishery Stock Assessment Models 817 Alaska Sea Grant College Program AK-SG-98-01, 1998 Estimates of Shrimp Trawl Bycatch of Red Snapper (Lutjanus campechanus) in the Gulf of Mexico B.J. Gallaway LGL Ecological Research Associates, Inc., Bryan, Texas M. Longnecker Texas A&M University, Department of Statistics, College Station, Texas J.G. Cole LGL Ecological Research Associates, Inc., Bryan, Texas R.M. Meyer Meyers Chuck, Alaska Abstract Estimation of red snapper bycatch in the shrimp trawl fishery of the Gulf of Mexico has been a contentious issue. Estimates are generated by the National Marine Fisheries Service (NMFS) using a general linear model which establishes a relationship between resource trawl survey data and catch data from the fishery obtained by observers on shrimp fishing vessels. The more complete time series of resource trawl data is then used to predict commercial vessel CPUE which is multiplied by total fishing effort to determine bycatch. The estimates are characterized by exceptionally low R 2 values and highly skewed residuals (70% of the catch observations were zeros). We have attempted to improve the estimates by using fewer and larger time-space cells, pooling catch and effort data to reduce the number of zeros contained in the analysis, incorporating significant interactions, and using epochs to guard against nonstationarity. The R 2 values for the revised models are 2 to 3 times higher than the R 2 for the base case, and the distribution of the residuals is greatly improved. The revised estimates in recent years average on the order of 30 to 47% lower than the NMFS estimates. Nevertheless, bycatch levels are high (26 to 32

2 818 Gallaway et al. Shrimp Trawl Bycatch of Red Snapper million per year) and are increasing due to increasing abundance of juveniles. However, the age of structure of the bycatch may consist of a much larger fraction of age-0 fish and fewer age-1 fish than has been thought. Bycatch Reduction Devices (BRDs) do not effectively exclude age-0 red snapper. Thus, the existing stock recovery policy based on NMFS GLM bycatch estimates and using BRDs to reduce shrimp trawl mortality of juvenile red snapper may be ineffective. Introduction The fishery for red snapper (Lutjanus campechanus) in the Gulf of Mexico began over 150 years ago off Pensacola, Florida (Goodyear 1995) and by 1872 it had developed as a separate industry. The stocks offshore of Pensacola were greatly depleted between 1865 and 1883, causing the fishery to shift first to the Florida middle grounds (1883 to 1885), and then farther southward along the west Florida coast during 1885 to 1910 (Camber 1955). In 1892, the fishery also expanded to (1) newly discovered red snapper grounds in the Campeche Banks off Mexico, and (2) to the western Gulf of Mexico between the mouth of the Mississippi River to about Galveston, Texas. In the eastern U.S. gulf and Campeche, Mexico, subsequent geographic expansions of the fishery were driven by dwindling stocks in the areas previously fished. The U.S. fishing fleet was excluded from Mexican waters in the early 1980s and as a result the effort was redirected at the remaining stocks in U.S. waters. By this time, the U.S. stock was essentially restricted to the western Gulf part of the range from Mississippi/Alabama to Texas. This area has been fished commercially since 1892, and also lies in the heart of the Gulf of Mexico shrimping grounds. Many juvenile (age-0 and age-1) red snapper are taken as bycatch in the shrimp fishery (Nichols et al. 1987, 1990; Nichols 1990; Nichols and Pellegrin 1992; Nichols 1996). Although there is some debate (e.g., Rothschild et al. 1997), most consider the gulf red snapper stock to be, at present, severely overfished (Goodyear 1995, MRAG Americas, Inc. 1997). Management actions began in the mid-1980s and a stock rebuilding plan has been developed. Some of the management measures implemented between 1984 and 1996 include size and bag limits for the recreational fisheries, commercial and recreational quotas, prohibition of traps and longline gears in certain areas, and prohibition of commercial sale of red snapper from shrimp trawls. Also, Turtle Excluder Devices (TEDs) were mandated for use in the gulf shrimp fishery in These mechanical separation devices likely exclude large fish as well as turtles, and may have some effect on reducing take of juvenile red snapper (unpublished data). Collectively, these actions appear to have had positive effects on the stock as reflected by increases in both stock and recruitment and, possibly, the increased size of harvested fish (Schirripi and Legault 1997, Rothschild et al. 1997). A key component of the stock rebuilding plan, yet to be im-

3 Symposium on Fishery Stock Assessment Models 819 plemented, is to reduce mortality from shrimp trawl bycatch of age-0 and, in particular, age-1 red snapper through the use of bycatch reduction devices (BRDs). BRDs are more effective at excluding age-1-sized than age-0- sized red snapper (Nichols et al. 1995, NMFS 1996). Quantification of bycatch levels and the age-0 and age-1 fractions represented in this incidental catch are necessary for stock assessment and rebuilding evaluations. Estimates of bycatch are provided by the National Marine Fisheries Service (NMFS) as described in Goodyear (1995). In summary, the NMFS bycatch estimates are generated from a general linear model (GLM) applied to two datasets generated from shrimp trawl catches of red snapper. One dataset consists of catch-per-tow data which are provided from resource surveys conducted by NMFS, predominantly in summer and fall of each year (Nichols and Pellegrin 1989, Goodyear 1995). Features of this program since 1985 include a semi-synoptic sampling of the entire western Gulf of Mexico in summer and fall based on a random sampling design and the use of a standard shrimp trawl. Although the scope and design of the resource surveys have varied through time, continuous data are available for the fall season offshore Louisiana since For simplicity, we shall refer to these data as SEAMAP data, and include results from the Fall Groundfish and Summer SEAMAP programs. The second dataset comes from records of finfish catch and fishing effort compiled on an individual tow basis by observers placed on shrimp fishing vessels specifically to quantify the bycatch including red snapper. Observer data are collected year round, but observer programs are not conducted every year. Even when conducted, only a small fraction of the fleet is sampled. Observer data are available for 1972 to 1982, and 1992 to 1996 periods. We shall refer to these data as Observer (OBSR) data. The structure of the GLM model used by NMFS to estimate commercial catch per unit effort (CPUE) for a single net is: Log (CPUE + 1) ijklmn = mean + dataset i + year j + season k + area l + depth m + e ijklmn for an array of space (4 areas 2 depths) time (three 4-mo seasons or trimesters) cells over the 24-y period, 1972 to In effect, the GLM calibrates shrimp vessel catch rates and resource trawl surveys during the periods and areas that had observations in common, and then uses the resource trawl data to index shrimp trawl bycatch (Nichols et al. 1987, 1990; Nichols and Pellegrin 1992). The GLM-based estimators of the mean log (CPUE+1) are transformed to an unbiased estimate of the commercial CPUE. These estimates are then multiplied by 2 (the assumed average number of nets) times the effort estimated for that time-space cell in hours fished. The catch estimates are then summed to provide an overall bycatch estimate. This GLM approach has been selected because observer programs have not been conducted in each space-time cell and the stock assessments require annual estimates of bycatch mortality. Use of the

4 820 Gallaway et al. Shrimp Trawl Bycatch of Red Snapper SEAMAP data enables an estimate for each year as well as an update of the previous year s estimates. We believe the structure of the NMFS GLM model is problematic and that improvements can be made. The basic problems are that (1) on the order of 70% of the tows in each dataset have zero catch of red snapper, (2) less than 50% of the space-time cells have been sampled overall (29% of the OBSR cells and 46% of the SEAMAP cells), (3) interactions between the five factors in the GLM model are ignored, and (4) there is no consideration of possible effects of nonstationarity (effects of explanatory variables changing over time). Our approach toward addressing these problems is to (1) combine catch and effort over several tows to reduce the number of zeros, (2) use fewer but larger strata to reduce the number of empty cells, (3) consider significant interactions, and (4) conduct separate models for early and late epochs to protect against nonstationarity. Our hypothesis was that the above described changes would result in a better fit of the revised models (higher R 2 and lower residual error) as compared to models using the NMFS GLM structure, and a more normal distribution of residuals. The findings of the revised approach are presented and then discussed in terms of their ramifications with respect to management measures designed to recover red snapper stocks. The Data As noted above, the basic data used in the analysis are of three types: observer data, resource trawl survey data, and shrimp fishing effort data. The observer data are available for only two periods: 1972 to 1982 (historical) and 1992 to 1996 (modern). The historical observer data and what we have defined as the SEAMAP data were obtained from S. Nichols (NMFS, Pascagoula Laboratory) and the modern observer data were obtained from J. Nance (NMFS, Galveston Laboratory). Dr. Nance also provided the shrimp fishing effort data which is estimated by statistical reporting grid and depth zone within each grid. The spatial distributions of the OBSR tow data, historical and modern, are shown in Figs. 1 and 2, along with the distribution of the tows that contained red snapper. The historical ( ) sample sizes are small as compared to the sample sizes obtained in the 1992 to 1996 program, especially when considered on an annual basis (Figs. 1 and 2). The distribution of the SEAMAP data for 1972 to 1984 and tows containing red snapper during this period are shown in Fig. 3; the same data for 1985 to 1996 are shown in Fig. 4. The fall groundfish component of the SEAMAP dataset originated in 1972, and through 1984 sampling was mainly restricted to the so-called primary area off Louisiana (see Goodyear 1995). In 1985, the fall component of the NMFS resource surveys was expanded to encompass the entire geographic region from Pensacola, Florida to Brownsville, Texas. The summer component of the resource trawl surveys (SEAMAP) originated in 1982 and sampled the entire western Gulf

5 Symposium on Fishery Stock Assessment Models 821 Figure 1. Distribution of observer tows and tows containing red snapper, 1972 to 1982, in Gulf of Mexico statistical reporting grids The 10- fathom contour is provided as a reference depth. Longitude W provides the boundary between North and South Regions used in the GLM.

6 822 Gallaway et al. Shrimp Trawl Bycatch of Red Snapper Figure 2. Distribution of observer tows and tows containing red snapper, 1992 to 1996, in Gulf of Mexico statistical reporting grids The 10-fathom contour is provided as a reference depth. Longitude W provides the boundary between North and South Regions used in the GLM.

7 Symposium on Fishery Stock Assessment Models 823 Figure 3. Distribution of SEAMAP tows and tows containing red snapper, 1972 to 1984, in Gulf of Mexico statistical reporting grids The 10-fathom contour is provided as a reference depth. Longitude W provides the boundary between North and South Regions used in the GLM.

8 824 Gallaway et al. Shrimp Trawl Bycatch of Red Snapper Figure 4. Distribution of SEAMAP tows and tows containing red snapper, 1985 to 1996, in Gulf of Mexico statistical reporting grids The 10-fathom contour is provided as a reference depth. Longitude W provides the boundary between North and South Regions used in the GLM.

9 Symposium on Fishery Stock Assessment Models 825 from Pensacola, Florida to Brownsville, Texas. Summer and fall surveys thus became directly comparable in terms of spatial coverage in 1985, and shortly thereafter (1987) in terms of detailed sampling protocol. The data for years prior to 1985 are strongly dominated by fall samples taken in the primary area off Louisiana, but much better balance in seasonal and spatial sampling has been achieved since 1985 (compare Figs. 3 and 4). Shrimp fishing effort data are estimated by statistical grid and depth zone based on a census of shrimp landings obtained from seafood dealers by port agents. The Gulf of Mexico is stratified into 221 spatial cells for each month. Port agents assign the landings to these cells based on their knowledge of the shrimp fishery and interviews. Interviews are conducted with a subset of the landings, targeting larger vessels, to obtain CPUE data of the shrimp fleet in each of these cells every month. Where data are missing, the CPUE of a cell is imputed using a GLM model. These estimates are then used to calculate an estimate of total shrimping effort in each of these cells every month. The values are summed to provide an estimate of total shrimp fishing effort. Model Structure and Approach We use only two spatial strata in our models, both restricted to the western gulf. The North Region consists essentially of statistical grids and the South Region includes grids (Figs. 1-4). Juvenile red snapper are scarce in the gulf region east of grid 10, and shrimp fishing effort there is small compared to effort in the modeled area. The South Region corresponds roughly to the Texas Transitional Faunal Province of Pulley (1952) and the Dry Sub-Humid Climatological Zone of Parker (1960). The North Region has been a major source of red snapper in the Gulf of Mexico since Some major differences between the North and South Regions are shown in Table 1. Our models use only two periods January to August and September to December rather than three as used by NMFS. SEAMAP data for the January-April period were relatively sparse and winter shrimping effort is typically low compared to summer and fall. Further, age-1 juveniles predominate from January-August; age-0 fish are most abundant in the catches in September-December periods. We did not incorporate separate depth cells in our models, but rather used two cases. For Case I, the models were constructed using OBSR, SEAMAP, and shrimp fishing effort based on samples obtained for all depths greater than 5 fathoms for most of the North Region (as detailed below), and used data for all offshore depths for the remaining North Region and all of the South Region. For Case II, the models were based on data from depths greater than 10 fathoms. Examination of Figs. 1 and 2 shows that red snapper juveniles are infrequently encountered in commercial shrimp tows inside of 10 fathoms, and become even less frequent with proximity to the mainland. The historical ( ) SEAMAP data reflect a similar

10 826 Gallaway et al. Shrimp Trawl Bycatch of Red Snapper Table 1. Rationale for region determinations. North Region (Grids 10-18) South Region (Grids 19-21) Positive flow estuaries River influenced nearshore waters Hypoxia prevalent Currents predominantly to west in Zones Bathymetry gradient gentle Natural banks at shelf edge High density of petroleum platforms Neutral or negative flow estuaries More saline nearshore Hypoxia not prevalent Currents predominantly to north Bathymetry gradient steep Natural banks on mid-shelf Low density of petroleum platforms The South Region corresponds to the Texas Transitional Faunal Province of Pulley (1952) and the Dry Sub-humid Climatological Zone of Parker (1960). Boundary between Stat Areas 18 and 19 displaced west to W offshore to correspond with inshore boundary; i.e., dogleg removed (see Figs. 1-4). trend (Fig. 3), but red snapper have occurred with some regularity inside of 10 fathoms in statistical grids 17-21, especially in recent years (Fig. 4). Gallaway and Cole (1997) have shown that high-value habitat for juvenile red snapper (and brown shrimp) in the western Gulf of Mexico to be largely restricted to depths (>10 fathoms based upon results of habitat suitability modeling [USFWS 1980, 1982]). We constructed two temporal versions (epochs) for each model case. The first epoch covered the period 1972 to 1984; the second covered 1985 to In 1985, the spatial coverage of the fall component of the NMFS resource surveys was expanded from the so-called primary area off Louisiana (see Goodyear 1995) to encompass the entire geographic region from Pensacola, Florida to Brownsville, Texas. From this year, the summer and fall surveys became directly comparable in terms of spatial coverage and, shortly thereafter (1987), in terms of detailed sampling protocol. Also, the first red snapper management actions were taken in about that time frame (actually 1984), and it was not until about 1985 that good compliance with the mandatory dealer reporting of shrimp landings regulation was achieved (Pers. comm., M. Hightower, NMFS, Galveston). The latter should have resulted in more certain estimates of shrimping effort since effort is derived from the landings census. Each model case thus has 100 cells for each dataset: 52 cells for epoch I (2 regions 2 seasons 13 years) plus 48 cells for epoch II (2 regions 2 seasons 12 years). Of these, 53% of the OBSR cells and 78% of the SEAMAP cells were sampled overall. This compares to the NMFS model structure which has 576 cells for each dataset (4 regions 2 depths 3 seasons 24 years). Under this structure, only 29% of the OBSR cells and 46% of the SEAMAP cells were sampled overall.

11 Symposium on Fishery Stock Assessment Models 827 We also reduced the percent of zero values by using catch per trip per week as the basic observation unit rather than catch per tow. In the SEAMAP data, the percent of zeros declined from 70.6 to 6.9%; and in the OBSR data from 70.9 to 33.2%. Further, since the resource survey tows are of short duration (e.g., 15 min), combining tows over weekly intervals reduced the imbalance between SEAMAP and OBSR sample sizes. The data files used by NMFS to calculate their red snapper annual bycatch estimates were obtained from the NMFS Pascagoula Laboratory. This dataset included SEAMAP station and catch data for the years , and station and catch data from the early observer programs between 1972 and 1982 (Fig. 5). SAS programs were run to merge the station and catch data into a new dataset (Mod_Data) which included the new variables period (NMFS trimester), year, week (SAS date divided by 7), region (North or South), and dataset (SEAMAP or OBSR). The raw 1996 SEAMAP station, environment, and catch data, obtained from another department of the NMFS Pascagoula Laboratory were merged to form a new dataset (SEAMAP96) including the same variables from Mod_Data. The modern observer station and catch data were obtained from the NMFS Galveston Laboratory and were merged to create the modern Observer dataset, again including the same variables as the other sets (Fig. 5). Assignment of the region fields in each of the data records was based on the combination of depth and longitude of the sampling station. All stations located west of W were assigned to the south zone; stations between 87 W and 94 W and greater than 5 fathoms depth were assigned to the north region. Stations between 94 W and W were also assigned to the north region, regardless of depth. These three datasets were reduced to include only samples which contained a net operation code that indicated an undamaged tow, and these were then merged with a dummy dataset which included records for all year, period, and region cell combinations to be predicted by the SAS GLM (the catch per unit effort for each of the cell combinations was entered as a missing value) to form the model dataset. The catch and effort data in model dataset were then summarized by year, period, region, week, and cruise number to create the dataset BY- CRUISE. This catch and effort data was used to create a catch per unit of effort (CPUE) value for each year, period, region, week, and cruise number combination. The natural logarithm of the CPUE + 1 value was assigned to create the variable LNCPUE. Data for years prior to 1985 were extracted from BYCRUISE to create the CRUISE_EARLY dataset; post 1984 data were used to create the CRUISE_LATE dataset. In each dataset, trimester 2 values were reassigned to trimester 1. The SAS procedure GLM was then run on each of the two new datasets, modeling LNCPUE as a function of the class variables dataset, year, period, and region, and their interactions. For the CRUISE_LATE dataset, the year by region interactions were found to be significant, and

12 828 Gallaway et al. Shrimp Trawl Bycatch of Red Snapper Figure 5. Schematic of the steps in the estimation of red snapper bycatch showing the sources and types of data, the modifications made to the data, and the sequence of the analysis.

13 Symposium on Fishery Stock Assessment Models 829 were thus included in the analysis. No other significant interactions were found in either dataset. The modified GLM results produced a pair of datasets with predicted LNCPUE for each year, period, region, dataset combination. The observer dataset values were translated by using the following formula: e (OBSR Predicted LNCPUE GLM MSE) 1 to create an unbiased estimate of the predicted observer cell values. These cell values were multiplied by the NMFS estimate of fishing effort for each cell (from the NMFS Galveston Laboratory dataset) to create a bycatch estimate for each year, period, and region combination. These estimates were then summed and compared to the annual bycatch estimates produced by the NMFS Pascagoula Laboratory (Fig. 5). All of the bycatch in the January-August period was treated as if it was entirely age-1 fish. A few age-0 fish would be expected during June-August, but the fraction would be very low (less than 1 or 2% during June and July; on the order of 12% in August, Table 83 in Goodyear 1995). In fall, a substantial fraction of the catch (13 to 39%; Table 83, Goodyear 1995) can be age-1 fish although the catches are dominated by age-0 fish. The datasets provided to us included length-frequency data for juvenile red snapper in both the SEAMAP (1988 to 1995) and OBSR ( ) datasets. We plotted length frequency histograms for each of our two regions. These were used to allocate the total fall bycatch for each of the two regions into age- 0 and age-1 fractions. Results and Discussion The changes we instituted resulted in better performance of the models as compared to the NMFS estimates used as the base case (Table 2). Adjusted R 2 values for the historical period 1972 to 1984 and the more recent period (1985 to 1996) were 0.22 and 0.37, respectively, for Case I; and 0.22 and 0.38 for Case II. This compares to an R 2 of 0.13 for the NMFS base model. Likewise, skewness and kurtosis values of the modified GLM models were, considering sample size effects, greatly improved as compared to the base model, especially the model for the period 1985 to 1996 (Table 2). A part of the explanation for the better performance relates to improved balance between SEAMAP and OBSR data in recent years, particularly as compared to the historical period (compare Figs. 1-4). Using two models, one for each epoch, served to reduce the impact of the historical imbalances (which were severe in the early epoch) on the data for recent years. Pooling of effort and catch over several tows eliminated a large fraction of the zeros, and using larger spatial strata reduced the number of empty cells. The net result is better model performance, especially for the recent data.

14 830 Gallaway et al. Shrimp Trawl Bycatch of Red Snapper Table 2. Results of ANOVA and univariate residuals analyses. Basic ANOVA comparisons Root LCPUE Model structure n F P > F R 2 CV MSE Mean Case I Epoch Case I Epoch Case II Epoch Case II Epoch NMFS ( ) a 25, Results of residual analysis Shapiro-Wilk Model structure Skewness Kurtosis statistics P < W Case I Epoch Case I Epoch Case II Epoch Case II Epoch NMFS ( ) a b <0.01 c a Calculated by LGL using stated model structure and data provided by NMFS. b Kolmogorov D c P > D Bycatch Level The ramifications of our changes in GLM structure to the bycatch estimates as compared to the NMFS results are shown by Table 3 and the bottom panel of Fig. 6. Our estimates for the period 1972 to 1984 vary around the NMFS base-case estimates showing an 11% reduction overall for Case I, and a 10% reduction for Case II. Four of the 13 annual estimates were actually higher in our revised models than annual estimates yielded by the NMFS model structure. In contrast, a net reduction of about 30% for Case I and 47% for Case II is seen for the recent epoch (1985 to 1995), and only one estimate from the revised models is higher (2%) than a corresponding estimate from the base case (Table 3). Plots of the GLM estimates of annual CPUE and the corresponding levels of effort for each year illustrate the nature of the differences in

15 Symposium on Fishery Stock Assessment Models 831 Table 3. A comparison of shrimp trawl fishing bycatch estimates for juvenile red snapper in the western Gulf of Mexico, 1972 to GLM: Case I Epoch I Case II Epoch I Year NMFS This study Reduction (%) This study Reduction (%) ,100,000 48,267, ,181, ,200,000 14,370, ,318, ,800,000 21,655, ,579, ,200,000 8,338, ,627, ,300,000 36,092, ,343, ,800,000 27,219, ,203, ,700,000 17,007, ,564, ,000,000 16,593, ,668, ,300,000 19,416, ,804, ,300,000 57,943, ,250, ,900,000 26,039, ,439, ,300,000 12,853, ,250, ,600,000 9,215, ,126, Totals 353,500,00 315,013, ,359, GLM: Case I Epoch II Case II Epoch II Year NMFS This study Reduction (%) This study Reduction (%) ,100,000 11,712, ,011, ,400,000 7,347, ,500, ,400,000 14,450, ,235, ,500,000 10,346, ,386, ,400,000 13,101, ,528, ,200,000 42,656, ,959, ,100,000 28,266, ,953, ,700,000 26,012, ,551, ,900,000 24,354, ,216, ,400,000 33,405, ,292, ,000,000 43,184, ,895, Totals 364,100, ,838, ,531, NMFS estimates are from Goodyear (1995).

16 832 Gallaway et al. Shrimp Trawl Bycatch of Red Snapper Figure 6. Annual trends in shrimp fishing effort, CPUE of red snapper in shrimp trawls, and red snapper bycatch in the western Gulf of Mexico, 1972 to 1996.

17 Symposium on Fishery Stock Assessment Models 833 bycatch estimates among the three cases (NMFS; Case I and Case II of this study, Fig. 6). In the NMFS model, CPUE is low and shrimp fishing effort is high relative to our cases. This is because the NMFS model incorporates data for Florida where both juvenile abundance and shrimp fishing effort are low; and, more important, data from shallow nearshore areas in the western gulf where red snapper are either absent or low abundance, but shrimp fishing effort is high. Incorporating these data has a smoothing effect on the CPUE time series. In our cases, we excluded data for areas which, in our opinion, do not constitute suitable habitat for this species (Case I); and, in Case II, restricted the analysis to high-value habitats. As a result our values of CPUE are high relative to the NMFS case, but the corresponding effort multipliers are smaller (Fig. 6). With only one exception (1990), the annual CPUE values for Case II are higher than the values for Case I, reflecting the higher abundance of juvenile red snapper in areas greater than 10 fathoms in depth as compared to more shallow areas. The 1990 CPUE exception was largely attributable to a few large collections of red snapper which were obtained in the Fall SEAMAP sampling program just inside the 10 fathom contour in the North Region in the vicinity of the mouth of the Mississippi River where the bathymetry gradients are steep. The results of the bycatch analyses show that the trend of increase in red snapper bycatch since the mid-1980s is attributable to increased CPUE of juvenile red snapper (Fig. 6). Effort has remained relatively stable, and, in the most recent years, has even declined. Fishery-independent recruitment indices for red snapper have exhibited a similar trend of increase over the same time frame and the range occupied by recruits appears to be expanding (Goodyear 1997, Schirripa and Legault 1997). Age Composition of the Bycatch The stock assessment requires that the bycatch be partitioned into age-0 and age-1 fractions for each year class beginning with Age-0 year class fish are abundant mainly in fall (September to December) and are treated as age-1 fish from January through the following December. The length frequency data for the fall collections of juvenile red snapper showed marked differences by region (Fig. 7). Juvenile fish of a size to suggest age 1 were not apparent in the samples from the South Region except in the 1992 SEAMAP collections (1 out of 8 years). In contrast, the length distributions were clearly bimodal in 6 of 8 fall collections from the North Region (Fig. 7). The fall bycatch totals for the South Region in 1992 to 1994 were allocated to age fractions based on the OBSR length frequencies, and the mean of the 1992 to 1994 OBSR bycatch frequency data for this region was used to allocate the bycatch to age fractions in years in which there were no OBSR size data (i.e., 1982 to 1991, 1995). The same approach was used in the North Region. Based on the above, and using the more conservative Case I results, we estimate that the combined bycatch of the 1982 to 1992 year classes was on the order of 198 million fish, of which 65% were age-0 fish and 35%

18 834 Gallaway et al. Shrimp Trawl Bycatch of Red Snapper Figure 7. Length frequency of juvenile red snapper taken in the shrimp fishery observer program (OBSR) and National Marine Fisheries Service resource trawl surveys (SEAMAP) in fall in the North and South Regions of the Gulf of Mexico used in the analyses, 1988 to 1995.

19 Symposium on Fishery Stock Assessment Models 835 Figure 8. Estimated total bycatch and age composition of the bycatch for the 1982 to 1992 year classes based on comparisons of the results of Goodyear (1995) to results from this study.

20 836 Gallaway et al. Shrimp Trawl Bycatch of Red Snapper were age-1 fish. The corresponding estimates used in the Goodyear (1995) stock assessment are 311 million fish, of which only 44% were age 0, and 56% were age 1 (Fig. 8). The difference appears to stem from NMFS using length data combined from both datasets as well as over all regions to estimate the age composition of the fall collections, whereas we partition the catch data to age on a region-specific basis. The overall reduction indicated for total bycatch from our estimates as compared to the NMFS estimates (36%) may not seriously affect the conclusions of the stock assessment (Goodyear 1996). However, the marked differences in age composition of the bycatch are a matter of concern since BRDs do not effectively exclude age-0-sized red snapper. Their mandated use thus may not be an effective means of lowering shrimp trawl bycatch mortality. Independent Estimates Rothschild et al. (1997) have independently estimated red snapper bycatch in the Gulf of Mexico shrimp fishery. Their approach was to use only the observer data and they initially focused on the depths (5-20 fathoms), locations (grids 16-21), and time of year where the great majority of red snapper bycatch is taken (May-December). They used standard ratio estimation techniques to derive CPUE, and multiplied these by the corresponding effort to obtain bycatch on a per net basis. To complete the time series for years with no OBSR data ( ), the computed mean CPUE values for the available data were multiplied by the corresponding ratios of shrimp effort yielding estimates for all years. The correspondence between predicted and observed values in recent years was good, but there was less similarity between predicted and observed values for the historical period. They advised that the historical estimates should be viewed with reasonable caution. The results obtained for recent years by Rothschild et al. (1997) indicated a bycatch level between 7 and 8 million juvenile fish per trawl net, which, assuming an average of two nets, provides a lower bound estimate of 15 million. Based on their professional opinion, this lower bound was inflated by 50% to account for multiple nets, catches during other times and places, etc. They estimated that total bycatch was perhaps on the order of 20 to 25 million fish (Rothschild et al. 1997). They suggested the age-0 component might be on the order of 15 million fish or about 75% of the total. Rothschild et al. (1997) also commented on the scarcity of age-1 fish off the Texas coast during fall, providing independent confirmation of our findings regarding regional differences in age composition of the bycatch. Conclusions The results of our analysis and those of Rothschild et al. (1997) would suggest that bycatch of the 1982 to 1992 year classes has been overestimated in the Goodyear (1995) stock assessment by 36%. Nevertheless, the

21 Symposium on Fishery Stock Assessment Models 837 number of juvenile red snapper taken incidental to shrimp trawling is large (in our study the 1992 to 1996 average was 26 million for Case II and 32 million for Case I), and bycatch is increasing as juveniles are becoming more abundant (Case I value was 43 million in 1995). The more important finding of this study is that the age-0 fraction of the catch is likely much larger and the age-1 fraction much lower than has been previously assumed. Since BRDs do not effectively exclude age-0 fish, the anticipated level of gain from BRD use may not be realized. Recruitment, the distributional range of recruits, and stock level of red snapper are apparently all increasing in the Gulf of Mexico, as is size of harvested fish (Schirripa and Legault 1997, Rothschild et al. 1997). The commercial fishery presently lasts only days before the quota is filled, and the recreational quota in 1997 was, for the first time ever, reached before the year ended. All of this is occurring in the face of unabated and growing bycatch levels. Acknowledgments We especially thank Scott Nichols and Ken Savastano of the NMFS Pascagoula Laboratory for the SEAMAP and historical OBSR data, and Jim Nance of the NMFS Galveston Laboratory for the modern OBSR data and the shrimp fishing effort data files. We have been involved in three peer reviews of the bycatch estimation procedures that are being used in the Gulf of Mexico shrimp fishery. The panelists who conducted the reviews helped shape our approach. The initial panel consisted of C.E. Gates, J.P. Geaghan, D.W. Hayne, J.M. Hoenig, and G.E. Lewis; the second panel consisted of M.C. Christman, L.P. Fanning, D.B. Hayes, and M.S. Kaiser; the third panel included M.K. McAllister, A. Sinclair, K.T. Stokes, J. Sutinen, and T.E. Target. All provided valuable insights, but the approach we used and the opinions we express herein do not necessarily represent the views of any of the listed panelists. We especially thank Wayne Swingle of the Gulf of Mexico Fishery Management Council and John Witzig of the NMFS Office of Science and Technology who organized and distributed results of the peer reviews. The Texas Shrimp Association, Inc. and the Biological Resources Division of the United States Geological Survey provided funding for this study on a collaborative basis. References Camber, C.I A survey of the red snapper fishery of the Gulf of Mexico with special reference to the Campeche Banks. Technical Series of Florida State Board of Conservation 12:1-64. Gallaway, B.J., and J.G. Cole Cumulative ecological significance of oil and gas structures in the Gulf of Mexico: A Gulf of Mexico fisheries habitat suitability model Phase II model description. U.S. Geological Survey, Biological Resources Division, USGS/BRD/CR, and Minerals Management Service, Gulf of Mexico OCS Region, New Orleans, LA, OCS Study MMS pp.

22 838 Gallaway et al. Shrimp Trawl Bycatch of Red Snapper Goodyear, C.P Red snapper in U.S. waters of the Gulf of Mexico. NMFS Southeast Fisheries Science Center, Miami Laboratory, Coastal Resources Division, Miami, FL. MIA-95/ pp. Goodyear, C.P Red snapper bycatch mortality: Implications of possible estimate bias on parameters of the recovery plan. Report to the Gulf of Mexico Fishery Management Council. 15 pp. Goodyear, C.P An evaluation of the minimum reduction in the 1997 red snapper shrimp bycatch mortality rate consistent with the 2019 recovery target. Report to the Gulf of Mexico Fishery Management Council. 14 pp. MRAG Americas, Inc Consolidated report on the peer review of red snapper (Lutjanus campechanus) research and management in the Gulf of Mexico. NMFS, Office of Science and Technology, December Nichols, S The spatial and temporal distribution of the bycatch of red snapper by the shrimp fishery in the offshore waters of the U.S. Gulf of Mexico. NMFS, Mississippi Laboratories, Pascagoula, MS. 52 pp. Nichols, S An update on some issues relating to the distribution of red snapper bycatch. NMFS, Southeast Fisheries Center, Mississippi Laboratories, Pascagoula, MS. 9 pp. Nichols S., and G.J. Pellegrin Jr Trends in catch per unit effort for 157 taxa caught in the Gulf of Mexico fall groundfish survey 1972 to NMFS, Mississippi Laboratories, Pascagoula, MS. 10 pp. Nichols, S., and G.J. Pellegrin Jr Revision and update of estimates of shrimp fleet bycatch NMFS, Southeast Fisheries Center, Mississippi Laboratories, Pascagoula, MS. 17 pp. Nichols, S., A. Shah, G. Pellegrin Jr., and K. Mullin Estimates of annual shrimp fleet bycatch for thirteen finfish species in the offshore waters of the Gulf of Mexico. NMFS, Southeast Fisheries Center, Mississippi Laboratories, Pascagoula, MS. 29 pp. Nichols, S., A. Shah, G.J. Pellegrin Jr., and K. Mullin Updated estimates of shrimp fleet bycatch in the offshore waters of the U.S. Gulf of Mexico NMFS, Mississippi Laboratories, Pascagoula, MS. 22 pp. Nichols, S., J. Nance, C.P. Goodyear, A. Shah, and J. Watson Some considerations in determining bycatch reduction requirements. NMFS, Southeast Fisheries Science Center, Miami, FL. 18 pp. NMFS Summary report on the status of bycatch reduction device development. NMFS, Southeast Fisheries Science Center, Mississippi Laboratory, Pascagoula, MS. 34 pp. Parker, R.H Ecology and distributional pattern of marine invertebrates, northern Gulf of Mexico. In: F.P. Shepard, F.B. Pheleger, and T.H. van Andel (eds.), Recent sediments, northwest Gulf of Mexico. American Association of Petroleum Geologists, Tulsa, OK, pp Pulley, T.E A zoogeographic study based on the bivalves of the Gulf of Mexico. Ph.D. thesis. Harvard University, Cambridge, MA.

23 Symposium on Fishery Stock Assessment Models 839 Rothschild, B.J., A.F. Sharov, and A.Y. Bobyrev Red snapper stock assessment and management for the Gulf of Mexico. Report submitted to NOAA, NMFS Office of Science and Technology, by University of Massachusetts, Center for Marine Science and Technology, North Dartmouth, MA. 173 pp. Schirripa, M.J., and C.M. Legault Status of the red snapper in U.S. waters of the Gulf of Mexico. NMFS, Southeast Fisheries Science Center, Miami Laboratory, MIA-97/ pp. USFWS Habitat evaluation procedures (HEP). U.S. Fish and Wildlife Service, Division of Ecological Services, ESM 102. USFWS Habitat suitability index models: Appendix A. Guidelines for riverine and lacustrine applications of fish HSI models with habitat evaluation procedures. U.S. Fish and Wildlife Service, Division of Ecological Services, FWS/OBS- 82/10.A. 53 pp.

24

25 Fishery Stock Assessment Models 841 Alaska Sea Grant College Program AK-SG-98-01, 1998 A Summary of Assessment Information for Managing Alaska Groundfish Stocks Jane DiCosimo North Pacific Fishery Management Council, Anchorage, Alaska Abstract The North Pacific Fishery Management Council (NPFMC) is responsible for effectively managing the groundfish fisheries in the Bering Sea (BS), Aleutian Islands (AI), and Gulf of Alaska (GOA). These fisheries target walleye pollock (Theragra chalcogramma), Pacific cod (Gadus macrocephalus), sablefish (Anoplopoma fimbria), Atka mackerel (Pleurogrammus monopterygius) and numerous flatfish (Pleuronectes sp. and Hippoglossoides sp.) and rockfish (Sebastes sp. and Sebastolobus sp.) species. The stocks are routinely evaluated by National Marine Fisheries Service (NMFS) scientists, who are relied on by the NPFMC to recommend harvest levels that will maintain healthy stocks. Alaska Department of Fish and Game scientists evaluate the stocks of three species of demersal shelf rockfish, for which the NPFMC has delegated management authority to the State of Alaska. Introduction Stock assessments have evolved since 1978 in response to changes in target species, data collection, and assessment methodology. Currently, biomass for most groundfish stocks is estimated using a stock synthesis model described by Methot (1990). The lack of age data, however, has prevented the traditional application of the synthesis model to some stocks of flatfish, rockfish, GOA Atka mackerel, and squid. Instead, stock biomass is estimated using an area-swept index from trawl survey data. There also have been changes in methodology for estimating optimal harvest rates and overfishing rates. These harvest rates, when applied to estimated biomass for individual stocks, result in a preliminary recommendation for acceptable biological catch (ABC) and an overfishing level (OFL). For most stocks, ABC estimates are determined by calculating the fishing mortality which reduces the equilibrium level of spawning biomass

26 842 DiCosimo Assessment Information for Alaska Groundfish per recruit to 40% of its unfished level (F 40% ) (Clark 1993). The OFL for most stocks is currently based on an F 30% rate. In the absence of maturity and growth information, ABC is based on an F = 0.75M harvest strategy, and overfishing is based on F = M. Historical Assessment Methodologies The groundfish assessments are compiled into a stock assessment and fishery evaluation (SAFE) report, which is prepared and reviewed annually for each fishery management plan. These assessments undergo a thorough peer review process, first by the council s groundfish plan teams which review the stock assessments and then by the council s Scientific and Statistical Committee (SSC), both of which are composed of biologists and economists from state and federal agencies, and academia. Conservation-based scientific assessment advice, along with strict adherence to scientific advice by managers, has resulted in relatively healthy stocks of groundfish in the North Pacific. Of 37 North Pacific groundfish stocks examined, three-year trends in relative abundance, based on stock assessment and survey catch-per-unit-effort (CPUE) trends, indicate 8 stocks are increasing, 8 stocks are stable, 13 stocks are decreasing, and 8 stocks are of unknown status (Table 1). Total groundfish harvests in 1996 relative to their respective ABCs are shown in (Table 2). Harvests for all species except GOA Pacific cod and BS/AI and GOA Pacific ocean perch (POP) (Sebastes alutus) were held at or below the respective ABC. Overruns occurred due to difficulty of in-season management of Pacific cod and POP. The council s reliance on the plan teams and SSC is well-documented. During (Table 3), the council exceeded the SSC ABC recommendation only twice in 344 ABC determinations. In 1980, the council set the AI sablefish ABC at 4,500 t, higher than the SSC recommendation of 3,700 t, but lower than the plan team recommendation of 9,600 t. In 1992, the council set the GOA pollock ABC midway between the SSC and plan team recommendations. The council set ABC lower than the SSC recommendation in five instances. At no time has the council set total allowable catch (TAC), or quota, higher than ABC. In fact, the council set TAC lower than ABC for 54% of the determinations and set TAC equal to ABC in 44% of determinations. ABCs were not specified by the council in 2% of cases; these occurred in the GOA between 1987 and Based on current criteria in the Magnuson-Stevens Act, under which the council is authorized to manage North Pacific fish stocks, NMFS has determined that of 106 groundfish stocks under the council s jurisdiction, 0 are overfished or approaching overfished condition, 64 are not overfished, and 42 are of unknown status (NMFS 1997). Nationally, 86 stocks are overfished, 10 are approaching overfished, 183 are not overfished, and 448 are unknown. Stock status may change under revised guidelines that should be published in late 1998.

27 Symposium on Fishery Stock Assessment Models 843 Table 1. Relative abundance (exploitable biomass in t) of Bering Sea (BS), Aleutian Islands (AI), and Gulf of Alaska (GOA) groundfish, Biomass 1996 Mean Reference relative 3-year Species Area biomass biomass years to mean trend Pollock BS 6,200,000 8,540, Below Down AI 230, , Below Down Bogoslof 680,000 1,210, Below Stable GOA 574,000 1,495, Below Down Pacific cod BSAI 1,106,000 1,364, Below Up GOA 314, , Above Up Deepwater flatfish GOA 101, , Below Down Yellowfin sole BSAI 2,862,000 1,979, Above Stable Greenland turbot BSAI 127, , Below Down Arrowtooth flounder BSAI 556, , Above Stable GOA 1,640, , Above Up Rock sole BSAI 2,183,000 1,187, Above Stable Rex sole GOA 72,300 85, Below Down Flathead sole BSAI 616, , Above Stable GOA 206, , Stable NK Other flatfish BSAI 589, , Near Stable Shallow water flatfish GOA 316, , Stable Down Sablefish BS 24,100 41, Below Stable AI 24,100 54, Below Down GOA 271, , Below Down Pacific ocean perch BS 72,500 97, Below Down AI 324, , Above Up GOA 774, , Above Up Sharpchin/northern AI 96,800 96, Near NK Northern rockfish GOA 85,000 73, Above Up Shortraker/rougheye AI 45,600 45, Near NK GOA 65,000 68, Near Down Pelagic shelf rockfish GOA 78,000 56, Above Up Other slope rockfish GOA 131, , Near Up Other red rockfish BS 29,700 NK NK NK NK Other rockfish BS 7,100 5, NK NK AI 13,600 13, NK NK Thornyheads GOA 47,000 65, Below Down Atka mackerel AI 576, , Below Down GOA NK NK NK NK NK Squid BSAI NK NK NK NK NK Other species BSAI 621, , Above Stable NK = not known.

28 844 DiCosimo Assessment Information for Alaska Groundfish Table 2. Exploitable biomass and harvest specifications (t) of Bering Sea (BS), Aleutian Islands (AI), and Gulf of Alaska (GOA) groundfish, Species Area Biomass OFL ABC TAC Pollock BS 6,120,000 1,980,000 1,130,000 1,130,000 AI 100,000 38,000 28,000 28,000 Bogoslof 558,000 43,800 32,100 32,100 GOA 1,097, ,270 79,980 79,980 Pacific cod BSAI 1,590, , , ,000 GOA 562, ,000 81,500 69,115 Deepwater flatfish GOA 101,430 9,440 7,710 7,710 Yellowfin sole BSAI 2,530, , , ,000 Greenland turbot BSAI 118,000 22,600 12,350 9,000 Arrowtooth flounder BSAI 587, , ,000 20,760 GOA 1,639, , ,840 35,000 Rock sole BSAI 2,390, , ,000 97,185 Rex sole GOA 72,330 11,920 9,150 9,150 Flathead sole BSAI 632, , ,000 43,500 GOA 206,340 34,010 26,110 9,040 Shallow water flatfish GOA 315,590 59,540 43,150 18,630 Other flatfish BSAI 616, ,000 97,500 50,750 Sablefish BS 17,900 2,750 1,308 1,100 AI 18,600 2,860 1,367 1,200 GOA 206,060 35,950 14,520 14,520 Other slope rockfish GOA 103,710 7,560 5,260 2,170 Northern rockfish GOA 83,370 9,420 5,000 5,000 Pelagic shelf rockfish GOA 55,640 8,400 5,140 5,140 Demersal shelf rockfish GOA 60,510 1, Pacific ocean perch BS 72,500 5,400 2,800 2,800 AI 324,000 25,300 12, ,800 GOA 242,300 19,760 12,990 9,190 Sharpchin/northern AI 96,800 5,810 4,360 4,360 Shortraker/rougheye AI 45,600 1, GOA 65,380 2,740 1,590 1,590 Other red rockfish BS 29,700 1,400 1,050 1,050 Other rockfish BS 7, AI 13, Thornyheads GOA 46,110 2,400 1,700 1,700 Atka mackerel AI 450,000 81,600 66,700 66,700 GOA NK 6,200 1,000 1,000 Squid BSAI NK 2,620 1,970 1,970 Other species BSAI 688, ,000 25,800 25,800 GOA NK NK NK 13,470 TOTAL (all species) BSAI 17,004,800 3,998,839 2,464,130 2,000,000 TOTAL (all species) GOA 4,797, , , ,815 TOTAL (all species) BOTH 21,802,560 4,783,699 2,957,180 2,282,815 OFL = overfishing level; TAC = total allowable catch; ABC = allowable biological catch; NK = not known.