Interactions Between Commercial Fishing and Walleye Pollock • Wilson, Hollowed, Shima, Walline, and Stienessen 59 Interactions Between Commercial Fishing and Walleye Pollock Christopher D. Wilson, Anne B. Hollowed, Michiyo Shima, Paul Walline, and Sarah Stienessen Reprinted from the Alaska Fishery Research Bulletin Vol. 10 No. 1, Summer 2003 The Alaska Fishery Research Bulletin can be found on the World Wide Web at URL: http://www.state.ak.us/adfg/geninfo/pubs/afrb/afrbhome.htm 60 Articles IAnltaersakcat iFonissh Beerytw Reeens eCaormchm Berucilalel tFinis 1hi0n(g1 )a:n6d1 –W7a7ll.e 2ye0 0P3o.llock • Wilson, Hollowed, Shima, Walline, and Stienessen 61 Copyright © 2003 by the Alaska Department of Fish and Game Interactions Between Commercial Fishing and Walleye Pollock Christopher D. Wilson, Anne B. Hollowed, Michiyo Shima, Paul Walline, and Sarah Stienessen ABSTRACT: Results from the first two years of a multiyear fishery interaction study near Kodiak Island in the Gulf of Alaska are presented. Findings from acoustic surveys, which were conducted in August 2000 and 2001, provide important information that begins to address the question of whether the abundance and spatial patterns of various species, including walleye pollock Theragra chalcogramma are impacted by commercial fishing activities over short spatio-temporal scales. The biomass and distribution of walleye pollock were stable over periods of days to weeks although during the second year an unusual, extremely dense, small-scale walleye pollock aggregation was detected during one of several survey passes. Several morphological descriptors of the walleye pollock echosign layers were evaluated to better understand whether differences at the scale of the fish aggregations occurred in response to fishing. Variography was also used to quantify walleye pollock spatial patterns. Results from the second year, when the commercial fishery took place within the study area, do not suggest a significant link between fishing activities and changes in estimates of juvenile and adult walleye pollock geographical distribution, biomass, or vertical distribu- tion. It will be important, however, to evaluate whether these trends persist during subsequent years. INTRODUCTION 1989; Merrick et al. 1987). The home range of a for- A multiyear field experiment was initiated in August aging Steller sea lion could be considered a localized 2000 near Kodiak Island in the Gulf of Alaska. The aim area. A reduction in prey availability may result from of this research was to characterize the effects of com- a reduction in prey abundance and/or a disruption in mercial fishing activity on the distribution and abun- the spatial patterns of the Steller sea lion prey. The dance of walleye pollock Theragra chalcogramma spatio-temporal extent of the perturbation to the prey over short temporal scales of days to weeks. The work field could determine the impact on the foraging suc- forms part of a larger research effort designed to de- cess of the Steller sea lion predator. For example, fish- termine whether commercial fishing activities impact ing removals may cause a decline in the abundance of the prey availability of walleye pollock and other for- a prey species within a localized area, but recovery to age fish species (e.g., capelin Mallotus villosus) to en- pre-fishery levels may be so quick that impacts to dangered Steller sea lions Eumetopias jubatus (Fadely predator foraging success would be negligible. Alter- et al. 2003). natively, disturbances from fishing operations may The impetus for this work was the need to under- elicit longer-term behavioral responses by prey species stand mechanisms that contributed to the precipitous that might affect spatial patterns and impact Steller sea decline in the western stock of Steller sea lions which lion foraging behaviors. Disturbed fish might move began in the 1970s (Loughlin 1998). One of several deeper in the water column to form smaller, denser ag- explanations that have been offered to account for this gregations, which may adversely impact the foraging decline is that large-scale commercial fisheries, such behavior of Steller sea lions. Unfortunately, no data as those for walleye pollock and Atka mackerel exist to answer two important questions regarding in- Pleurogrammus monopterygius, compete with Steller teractions among commercial fishing, Steller sea lions, sea lion populations by reducing the availability of po- and their prey. Firstly, do commercial fishing activities tential prey in localized areas (Loughlin and Merrick affect the distribution and abundance of Steller sea lion Authors: CHRISTOPHER D. WILSON, ANNE B. HOLLOWED, PAUL WALLINE, and SARAH STIENESSEN work with the Alaska Fisheries Science Center, NMFS, NOAA, 7600 Sand Point Way NE, Seattle WA 98115. E-mail: [email protected]. MICHIYO SHIMA is with Imperial College Consultants, 47 Prince’s Gate, South Kensington, London SW7 2QA UK. Acknowledgments: We wish to thank Jay Stinson and Tim Regan for their insight and helpful comments during the initial stages of this research. Comments from David Somerton, Martin Dorn, Bill Karp, and two anonymous reviewers improved an earlier version of this paper. This research would not have been possible without the contributions of the officers and crew of the research vessel Miller Freeman. Reference to product names does not imply endorsement by the National Marine Fisheries Service, NOAA. 61 62 Articles prey species significantly? And secondly, if the fish- mercial walleye pollock fishery was also scheduled to ery induces perturbations in prey spatial patterns and/ open in the area during August. The study was de- or abundance, how do these perturbations impact signed to extend over several years because natural Steller sea lion foraging success? shifts in ocean conditions and/or variations in the age The primary goal of this study was to investigate composition of the walleye pollock stock might influ- whether measurable changes in spatial patterns (i.e., ence responses to fishing activities. vertical distribution, fish school characteristics) and abundance occurred in walleye pollock at scales rel- Field Methods evant to Steller sea lion foraging (Merrick and Loughlin 1997). This paper reports results from the Multiple surveys of the control and treatment troughs first two years of the field study to examine the spatio- were conducted during daylight hours (about 15 hours/ temporal characteristics of walleye pollock before and day) over several weeks in August 2000 and August during a commercial fishing season. Although results 2001 using acoustic survey methods routinely em- for capelin are briefly discussed, complete results for ployed by Alaska Fisheries Science Center scientists this species will be reported elsewhere. Future direc- (Karp and Walters 1994; Traynor 1997). The surveys tions for this type of fishery-interaction research are consisted of a series of uniformly-spaced (3 nmi) par- also discussed. allel transects (Figure 1). A complete sampling of all transects within a trough was considered a survey pass. The work during the first year was completed in METHODS the absence of the August fishery and during the sec- ond year was conducted before and during the fishery. The work during the first year served two purposes. Study Area and Season First, the area had not been previously surveyed with The east side of Kodiak Island in the Gulf of Alaska acoustic methods during summer, so it was important was chosen as the study area for the fishery interaction to evaluate the feasibility of this approach in the study work for several reasons (Figure 1). Two adjacent sub- location in August. Second, it was important to char- marine troughs with similar topographical features acterize the natural variability in the temporal and spa- characterized the area. Barnabas Trough served as a tial patterns of the walleye pollock distribution and treatment site where commercial fishing was allowed abundance over the 2–4 week field season. Two sur- and Chiniak Trough served as a control site where fish- vey passes were conducted within each trough in Au- ing was prohibited. The proximity of the two troughs gust 2000. During the second year of the study, two minimized vessel travel time and enabled a more syn- survey passes were conducted within each trough be- optic sampling schedule. Additionally, a commercial fore the fishery commenced in Barnabas Trough. trawl fishery for walleye pollock occurs within the area These were followed by one pass in Chiniak Trough so that implementation of an experimental fishery per- and two passes in Barnabas Trough during the fishery turbation was relatively easy. Although not a require- in Barnabas Trough to investigate whether fishery-in- ment for the experiment, the area along the east side duced changes occurred in the fish distribution. The of Kodiak Island is also characterized by at least six fourth, partial pass in Barnabas Trough was only con- haulout sites for Steller seal lions where 508 animals ducted over the area where walleye pollock had been were counted in June–July 2002 (Sease and encountered during earlier passes. Thus, no commer- Gudmundson 2002; Figure 1). Thus, other closely as- cial fishing was allowed in either Barnabas or Chiniak sociated research efforts have been initiated in the area, Troughs during survey passes 1–2 in either year, and which focus directly on the behavior of the Steller sea commercial fishing operations occurred in Barnabas lions in relation to their prey distributions (Fadely et Trough during passes 3–4 in 2001. A similar number al. 2003; Shima et al. 2003). of days elapsed between repeated surveys in each Surveys for the fishery interaction experiment oc- trough during both years (Table 1). curred in August when recently weaned Steller sea lion The acoustic data for this study were collected juveniles (1-year-olds) were considered vulnerable to with a calibrated Simrad1 EK 500 echosounder oper- nutritional stress due to their high caloric needs per ating at 38 kHz. The nominal pulse length was 1 ms, unit body weight and inexperience at capturing prey beam width was 6.9°, and ping rate was 1 s-1. Echo- (Winship et al. 2002; Loughlin et al. In press). A com- integration data from the sounder were initially logged 1 Reference to product names does not imply endorsement by the National Marine Fisheries Service, NOAA. Interactions Between Commercial Fishing and Walleye Pollock • Wilson, Hollowed, Shima, Walline, and Stienessen 63 Figure 1. Fishery interaction study area off the east side of Kodiak Island showing survey transects used for all passes during August 2000 and 2001, and survey trawl locations (open circles) for August 2001. Similar numbers of trawl hauls were conducted during August 2000. Polygons (gray lines) in Barnabas Trough represent areas where commercial trawl hauls were made in August 2001. Stars represent locations of six Steller sea lion haulout sites. 64 Articles Table 1. Walleye pollock biomass estimates (thousands of t) for Barnabas and Chiniak Troughs from the fishery interaction study off the east side of Kodiak Island. Error bounds are shown in parentheses (see text for explanation). Whether a survey pass was conducted prior to or during the August commercial fishery is indicated. Chiniak Trough Barnabas Trough Survey Pass Date Adult Juvenile Date Adult Juvenile Study year 2000 1 (pre-fishery) 8–11 Aug 6.7 (5.9, 7.5) 5.9 (4.3, 7.5) 11–14 Aug 13.1 (10.7, 15.5) 0 2 (pre-fishery) 14–17 Aug 6.2 (5.4, 7.0) 8.0 (5.8, 10.2) 17–19 Aug 10.8 (8.2, 13.4) 0 Study year 2001 1 (pre-fishery) 9–11Aug 3.5 (2.9, 4.0) 17.2 (14.5, 20.0) 11–14 Aug 12.7 (9.7, 15.8) 7.8 (4.9, 10.6) 2 (pre-fishery) 14–16 Aug 2.9 (2.3, 3.5) 19.7 (16.2, 23. 2) 16–19 Aug 4.8 (4.1, 5.5) 9.3 (8.2, 10.4) 3 (fishery) 23–26 Aug 3.7 (3.2, 4.1) 18.7 (14.8, 22.6) 26–29 Aug 7.6 (5.9, 9.2) 10.7 (9.1, 12.2) 4 (fishery) 29–30 Aug 4.6 (3.7, 5.5) 10.9 (8.0, 13.8) with a horizontal resolution of about 5–6 m (dependent uted to walleye pollock was calculated for each geo- on vessel speed which averaged 5–6 m s-1) and a ver- graphic area based on the average of all s values from A tical resolution of 0.1–0.5 m. These data were pro- within that area. The s estimate for each geographic A cessed with the Echoview software (SonarData 2002) area was then scaled to length-specific fish numbers for the echo trace classification analysis. The Simrad and biomass using walleye pollock length distribu- BI500 software (Knudsen 1990) was used to log the tions, a length-weight relationship derived from trawl data and to classify the echosign into different groups catches, and a standard target strength to walleye pol- based on taxonomic or size-group considerations (see lock length relationship (Traynor 1996). Estimates of below and Results). The classified data were binned into length-specific biomass were then summed across fish cells with a vertical resolution of 5 m and horizontal lengths and geographic strata to provide total estimates resolution of 185 m (0.1 nmi) for subsequent analyses. for each trough. Because the acoustic data were only Biological samples were collected with trawls dur- collected from a nominal depth of 14 m to within about ing all surveys to identify the species and size compo- 0.5 m of the bottom echo, the resulting biomass esti- sitions of selected echosign and to collect other mates may not represent the total biomass within the information needed to estimate abundance and distri- troughs. bution patterns (Wilson and Guttormsen 1997). A large midwater Aleutian wing trawl and smaller midwater Data Analysis Marinovich trawl were used to target midwater echosign, and a poly Nor’eastern bottom trawl (poly To evaluate whether significant large-scale differences Nor’eastern) was used to target near-bottom echosign. existed between the walleye pollock geographical dis- The codends of the Aleutian wing trawl and poly tributions between the pre-fishing and fishing periods, Nor’eastern were fitted with 32 mm (1 1/4 in) mesh a statistical test was used which is based on a modified codend liners and the Marinovich trawl with a 3.2 mm Cramer-von Mises (CvM) statistic (Syrjala 1996). This (1/8 in) mesh liner except in August 2001, when a 9.5 test is dependent on the spatial scale selected. This mm (3/8 in) mesh liner was used in the Aleutian wing procedure calculates a test statistic as the sum of the trawl. The smaller liner was used the second year to squares of the differences between the cumulative dis- improve the retention of smaller capelin. Walleye pol- tribution functions of walleye pollock density (i.e., s ) A lock were sampled to determine sex, fork length (to from the two samples. Observations were normalized nearest cm), body weight (to nearest 2 g), age, matu- to remove the effect of differing population sizes be- rity, and ovary weight of selected females. tween passes. Significance of the test statistic was de- Walleye pollock biomass estimates were derived termined with a randomization test. To test for by partitioning the echogram and catch data into geo- differences between the pre-fishing and fishing peri- graphic areas within each trough so that the areas were ods in the treatment and control troughs, the acoustic characterized by similar walleye pollock echo signa- data were initially block-averaged into 3 nmi sections tures and fish length distributions. An estimate of the (i.e., distance equal to transect spacing) along transects acoustic backscattering or nautical area scattering co- for each survey pass. Because differences between efficient (s ; defined in MacLennan et al. 2002) attrib- passes were slight within the pre-fishing or fishing pe- A Interactions Between Commercial Fishing and Walleye Pollock • Wilson, Hollowed, Shima, Walline, and Stienessen 65 riods for either trough, means at each 3 nmi location aggregation were evaluated before the final values were calculated using the block-averaged passes were chosen. An S threshold of -70 dB in conjunction v within each period when multiple passes were con- with other criteria values (i.e., minimum school length ducted within the period (i.e., both Chiniak and (40 m), minimum height (5 m), minimum connected Barnabas Troughs for pre-fishery, Barnabas Trough for length (5 m), minimum connected height (2 m), maxi- fishery period). For the fishery period in Barnabas mum vertical linking distance (5 m), maximum hori- Trough, observations at similar locations during pass zontal linking distance (20 m)) provided the best 3 and the partial pass 4 were averaged and inserted into definition of the walleye pollock aggregations when the pass 3 data set. These means of the block-averaged compared to the original echograms. These final val- data were tested for differences using the modified ues were used in the analysis of all survey passes. As CvM test. emphasized by Reid (2000) and Freon et al. (1996), the To determine whether differences occurred be- criterion values will likely contain substantial and un- tween estimates of walleye pollock abundance for each known biases in defining aggregations, but if kept con- period, relative estimation errors were generated using stant (as in the present study), they should provide a model-based one-dimensional geostatistical proce- useful information about the variability of the aggre- dure (Petitgas 1993a; Williamson and Traynor 1996). gation structure. Caution is needed, however, in mak- The estimation variance obtained from geostatistical ing inferences about the exact dimensions of fish analysis is an indicator of the precision of the biom- aggregations with these types of data (Reid 2000). A ass estimate (Rivoirard et al. 2000). Error bounds or few areas were excluded from the present analysis, intervals were constructed by adding or subtracting where substantial backscattering from unidentified twice the relative estimation error from the mean. In organisms overlapped with that from walleye pollock. this study, mean estimates of biomass were considered Several of the aggregation-size and -shape descrip- significantly different when intervals constructed in tors generated by the Echoview software (SonarData this way did not overlap. 2002) were used to evaluate whether differences ex- To evaluate whether the walleye pollock changed isted in the walleye pollock aggregations between the position in the water column between the pre-fishing pre-fishery and fishery periods during the second year. and fishing periods, estimates were calculated of the Descriptors used in this analysis, which are defined in mean fish depth and the mean fish depth above the Nero and Magnuson (1989), included aggregation bottom. Ninety-five percent confidence intervals were height, length, area, perimeter, and mean volume back- generated using bootstrapping methods (Mooney and scattering coefficient (S ). Fractal dimensions of the v Duval 1993), and estimates were compared with a z- aggregations, which relate school perimeter to school test (Zar 1984). area, were also generated (Nero and Magnuson 1989; Two analytical procedures were used to describe Barange 1994). An increase in the fractal value, which the structure of the walleye pollock distribution in Au- indicates a more complex aggregation shape, may be gust 2001 at different spatial scales, and to evaluate indicative of a redistribution process of the fish aggre- whether the scale of patchiness changed between the gation (e.g., potential indicator of disruption of wall- pre-fishery and fishery periods. These included an eye pollock layer into smaller groups). Where various echo trace classification (Reid 2000) and variographic descriptors were highly correlated (i.e., length, area, analysis (Petitgas 2001) of the walleye pollock perimeter), a representative morphological variable is echosign, which are described below. presented (i.e., length). Statistical significance among Walleye pollock echo traces were classified using the descriptor estimates was based on analysis of vari- the Echoview software (SonarData 2002), which in- ance results and the Student’s t statistic with cluded school or patch recognition algorithms, and also Bonferroni adjustment (Zar 1984). A log transform of generated estimates of various descriptive parameters some of the variables was needed to stabilize the vari- of the aggregations. Most of the walleye pollock ance. Statistical test results were considered significant echosign was distributed in pelagic and demersal lay- at P < 0.05. Ninety-five percent confidence intervals ers (Reid 2000; Swartzman et al. 1994) rather than were based on traditional sample-based methods. aggregations more typically chararcterized as schools Variography was used to examine the spatial struc- (Partridge et al. 1980). For the purposes of this study, ture of walleye pollock distributions. Experimental these finite walleye pollock layers are referred to as semi-variograms were constructed for the adult and aggregations or patches. Several mean volume back- juvenile walleye pollock for each pass and trough us- scattering strength (S ) thresholds (-70 to -60 dB) and ing the 0.1 nmi echo integration data (i.e., s ). A 2-pa- v A a range of values for other criteria used to define an rameter (range and nugget) spherical model was fitted 66 Articles to each semi-variogram using a weighted least squares RESULTS algorithm (Johnston et al. 2001). The range corre- sponds to the distance at which the semi-variogram The survey for the first year of the study was con- reaches its asymptote. The sill defines the asymptotic ducted between 8–20 August 2000. Two survey passes height of the variogram (i.e., the maximum variability were completed within each trough, along with 35 in the data). The two components of the sill are the hauls conducted with the Aleutian wing trawl, 5 hauls nugget and partial sill. The nugget is the fitted semi- with the poly Nor’eastern, and 5 hauls with the variance at a lag of zero. It is composed of measure- Marinovich trawl. The echo integration-trawl survey ment error and variation at scales smaller than the lag for the second year of the study was conducted be- size (about 1 km in the present study). The partial sill tween 9–31 August 2001. Multiple survey passes were is the difference between the sill and the nugget and completed within each trough prior to the fishery along is the part of the semi-variance due to autocorrelation. with 41 hauls conducted with the Aleutian wing trawl Model estimates for the range, nugget, and partial sill and 16 hauls with the poly Nor’eastern. are reported to characterize walleye pollock spatial Acoustic backscattering attributed to walleye pol- structure between the pre-fishery and fishery periods. lock and capelin was easily recognized from other Isotropy was assumed because data were not available backscattering in Barnabas and Chiniak Troughs each to evaluate differences in the spatial structure as a year (Figure 2). Most of the backscattering was as- function of direction at scales smaller than the inter- signed to 4 types of fish echosign: 1) adult walleye transect distance (3 nmi). pollock, 2) juvenile (mostly age-1) walleye pollock, 3) 25 50 ) 75 m ( h 100 Juvenile walleye pollock t p e 125 D 150 Adult walleye pollock Adult walleye pollock 175 200 3636.0 3637.0 3638.0 3639.0 25 50 ) m 75 ( h pt 100 Capelin e D 125 150 175 2956.0 2957.0 2958.0 2959.0 Distance (nautical miles) Figure 2. Example echograms illustrating echosign layers attributed to walleye pollock (A) and capelin (B) during August 2001. Distance refers to cumulative distance traveled by the vessel during the survey. Interactions Between Commercial Fishing and Walleye Pollock • Wilson, Hollowed, Shima, Walline, and Stienessen 67 a mixture of capelin/age-0 walleye pollock in August Geographical Distribution 2000 or capelin in August 2001, and 4) other fishes. The size composition of adult walleye pollock was The geographical distribution of walleye pollock and similar between years and troughs whereas the juve- capelin within each trough exhibited similarities be- nile walleye pollock were largely absent in Barnabas tween years although some notable differences existed Trough during the first year (Figure 3). Juveniles were for juvenile walleye pollock. Adult walleye pollock present in both troughs during the second year, al- were distributed throughout Chiniak Trough, and in though the size composition indicated that some age- Barnabas Trough they tended to concentrate more to- 2 fish (about 30 cm modal fork length) were present wards the northern half of the trough during both years in Chiniak Trough but not Barnabas Trough. (Figure 4). Juvenile walleye pollock were broadly dis- tributed in Chiniak Trough but virtually absent in Barnabas Trough during the first year; they occurred August 2001 Fishery in both troughs during the second year, with distribu- Catch data have been compiled for 27 of 28 vessels tions similar to the adults (Figure 5). The mixture of that were fishing in Barnabas Trough during the ex- age-0 walleye pollock and capelin was broadly distrib- periment in August 2001 (Figure 1). These data ac- uted in both troughs during the August 2000 survey. count for about 99% of the total catch removed from The following year, age-0 walleye pollock were only this trough based on the National Marine Fisheries detected at the east end of one transect in Chiniak Service logbook data and shoreside database (NMFS, Trough during the second survey pass. Capelin were Alaska Region, P.O. Box 21668, Juneau AK 99802- often present over the shallower edges of Chiniak 1668). The 27 vessels spent 1,074 hours fishing to Trough, but were concentrated in the deeper waters complete 167 hauls during 22–31 August (i.e., during within the southern half of Barnabas Trough during the Chiniak Trough survey pass 3 (23–26 August) and second year. These basic geographical patterns for Barnabas Trough passes 3 (26–29 August) and 4 (29– each group occurred during each pass each year. Thus, 30 August)). Vessel deliveries during this period indi- although the distributional patterns between the two cated 2,850 t of walleye pollock were removed from troughs exhibited some intra- and interannual differ- Barnabas Trough. Based on historical fishing trends, ences, they were similar enough to justify their use as this did not appear to be an unusual level of effort. treatment and control sites. During the second year no significant difference was detected in the geographical distributions between the pre-fishery and fishery periods for juvenile wall- eye pollock in Barnabas Trough (CvM test, P = 0.453), although the difference was marginally significant for juveniles in Chiniak Trough (P = 0.049). Because no fishing was allowed in the control site, this difference suggested that the temporal variability in the juvenile distribution patterns differed between the troughs. No difference was detected between the periods for adults in Chiniak Trough (P = 0.362). Although a significant difference was detected between periods for adults in Barnabas Trough (P = 0.017), this result was due to the presence of a small, but extremely dense aggregation of adults that was only observed during one of the two pre-fishery passes (pass 1) along the east end of transects 6 and 7 (Figure 4). This dense adult aggre- gation also complicated biomass estimates (see below). However, when the single extreme observation along transect 6 was replaced with the pass 2 (pre-fishery) observation from the same location, no significant dif- ference was detected between the pre-fishery and fish- ery periods for adults (P = 0.108). This illustrates that Figure 3. Walleye pollock size composition estimates for the statistical significance of the CvM test is sensitive Chiniak and Barnabas Troughs during August 2000 and to single large values. Thus, the statistical significance August 2001. 68 Articles may not indicate that a fundamental difference exists Trough as Chiniak Trough (Table 1). Juvenile walleye in the underlying population represented in the pre- pollock were scarce in Barnabas Trough but present in fishing and fishing periods. quantities similar to adults in Chiniak Trough. Differ- ences in estimates between passes ranged between 7% Abundance and 36%. The results suggest that the biomass of adult walleye pollock was relatively stable over a period of Biomass estimates for August 2000 indicated that 1–2 weeks in both troughs and offers support for us- adults were about twice as abundant in Barnabas ing the two troughs as treatment and control sites. Figure 4. Acoustic backscatter attributed to adult walleye pollock along transects during a typical pass from the pre-fishery and fishery periods in August 2001. See text for explanation. Vertical (z -axis) scale is 0 to 12,000 m2 nmi-2.