sy i» •Ap~ P63 Pistt U.S. Department of Commerce Volume 108 Number 3 July 2010 U.S. Department of Commerce Gary Locke Secretary of Commerce National Oceanic and Atmospheric Administration Jane Lubchenco, Ph.D. Administrator of NOAA National Marine Fisheries Service Eric C. Schwaab Assistant Administrator for Fisheries ★ ^rES Of Scientific Editor Richard D. Brodeur, Ph.D. Associate Editor Julie Scheurer National Marine Fisheries Service Northwest Fisheries Science Center 2030 S. Marine Science Dr. Newport, Oregon 97365-5296 Managing Editor Sharyn Matriotti National Marine Fisheries Service Scientific Publications Office 7600 Sand Point Way NE Seattle, Washington 98115-0070 The Fishery Bulletin (ISSN 0090-0656) is published quarterly by the Scientific Publications Office, National Marine Fisheries Service, NOAA, 7600 Sand Point Way NE, BIN C15700, Seattle, WA 98115-0070. Periodicals postage is paid at Seattle, WA. POSTMASTER: Send address changes for subscriptions to Fish- ery Bulletin, Superintendent of Docu- ments, Attn.: Chief, Mail List Branch, Mail Stop SSOM, Washington, DC 20402- 9373. Although the contents of this publica- tion have not been copyrighted and may be reprinted entirely, reference to source is appreciated. The Secretary of Commerce has deter- mined that the publication of this peri- odical is necessary according to law for the transaction of public business of this Department. Use of funds for printing of this periodical has been approved by the Director of the Office of Management and Budget. For sale by the Superintendent of Documents, U.S. Government Printing Office, Washington, DC 20402. Subscrip- tion price per year: $36.00 domestic and $50.40 foreign. Cost per single issue: $21.00 domestic and $29.40 foreign. See back for order form. Editorial Committee John Carlson Kevin Craig Jeff Leis Rich McBride Rick Methot Adam Moles Frank Parrish Dave Somerton Ed Trippel Mary Yoklavich National Marine Fisheries Service, Panama City, Florida Florida State University, Tallahassee, Florida Australian Museum, Sydney, New South Wales, Australia National Marine Fisheries Service, Woods Hole, Massachusetts National Marine Fisheries Service, Seattle, Washington National Marine Fisheries Service, Auke Bay, Alaska National Marine Fisheries Service, Honolulu, Hawaii National Marine Fisheries Service, Seattle, Washington Department of Fisheries and Oceans, St. Andrews, New Brunswick, Canada National Marine Fisheries Service, Santa Cruz, California Fishery Bulletin web site: www.fisherybulletin.noaa.gov The Fishery Bulletin carries original research reports and technical notes on investigations in fishery science, engineering, and economics. It began as the Bulletin of the United States Fish Commission in 1881; it became the Bulletin of the Bureau of Fisheries in 1904 and the Fishery Bulletin of the Fish and Wildlife Service in 1941. Separates were issued as documents through volume 46; the last document was No. 1103. Beginning with volume 47 in 1931 and continuing through volume 62 in 1963, each separate appeared as a numbered bulletin. A new system began in 1963 with volume 63 in which papers are bound together in a single issue of the bulletin. Beginning with volume 70, number 1, January 1972, the Fishery Bulletin became a periodical, issued quarterly. In this form, it is available by subscription from the Superintendent of Documents, U.S. Government Printing Office, Washington, DC 20402. It is also available free in limited numbers to libraries, research institutions, State and Federal agencies, and in exchange for other scientific publications. U.S. Department of Commerce Seattle, Washington Volume 108 Number 3 July 2010 Fishery Bulletin Contents Articles 251-267 Hobbs, Roderick C, and Janice M. Waite Abundance of harbor porpoise ( Phocoeno phocoena) in three Alaskan regions, corrected for observer errors due to perception bias and species misidentification, and corrected for animals submerged from view 268-281 Poisson, Francois, Jean-Claude Gaertner, Marc Taquet, Jean-Pierre Durbec, and Keith Bigelow Effects of lunar cycle and fishing operations on longline-caught pelagic fish: fishing performance, capture time, and survival of fish The National Marine Fisheries Service (NMFS) does not approve, recommend, or endorse any proprie- tary product or proprietary material mentioned in this publication. No reference shall be made to NMFS, or to this publication furnished by NMFS, in any advertising or sales promotion which would indicate or imply that NMFS approves, recom- mends, or endorses any proprietary product or proprietary material mentioned herein, or which has as its purpose an intent to cause directly or indirectly the advertised product to be used or purchased because of this NMFS publication. The NMFS Scientific Publications Office is not responsible for the con- tents of the articles or for the stan- dard of English used in them. 282-304 McAllister, Murdoch K., Richard D. Stanley, and Paul Starr Using experiments and expert judgment to model catchability of Pacific rockfish in trawl surveys, with application to bocaccio (Sebastes paucispmis) off British Columbia 305-322 Patrick, Wesley S., Paul Spencer, Jason Link, Jason Cope, John Field, Donald Kobayashi, Peter Lawson, Todd Gedamke, Enric Cortes, Olav Ormseth, Keith Bigelow, and William Overholtz Using productivity and susceptibility indices to assess the vulnerability of United States fish stocks to overfishing 323-345 Munroe, Thomas A., and Steve W. Ross Distribution and life history of two diminutive flatfishes, Citharichthys gymnorhinus and C. cornutus (Pleuronectiformes: Paralichthyidae), in the western North Atlantic II Fishery Bulletin 108(3) 346-351 Passerotti, Michelle S., John K. Carlson, Andrew N. Piercy, and Steven E. Campana Age validation of great hammerhead shark (Sphyrna mokarran), determined by bomb radiocarbon analysis 352-362 Williams, Kresimir, Christopher N. Rooper, and Rick Towler Use of stereo camera systems for assessment of rockfish abundance in untrawlable areas and for recording pollock behavior during midwater trawls 363 Best paper awards for 2009 364 Guidelines for authors Subscription form (inside back cover) 251 Abundance of harbor porpoise ( Phocoena phocoena) in three Alaskan regions, corrected for observer errors due to perception bias and species misidentification, and corrected for animals submerged from view Email address for contact author: Rod.Hobbs@noaa.gov National Marine Mammal Laboratory Alaska Fisheries Science Center National Marine Fisheries Service 7600 Sand Point Way N.E Seattle, Washington 98115 Abstract — Estimating the abundance of cetaceans from aerial survey data requires careful attention to survey design and analysis. Once an aerial observer perceives a marine mammal or group of marine mammals, he or she has only a few seconds to iden- tify and enumerate the individuals sighted, as well as to determine the distance to the sighting and record this information. In line-transect survey analyses, it is assumed that the observer has correctly identified and enumerated the group or indi- vidual. We describe methods used to test this assumption and how survey data should be adjusted to account for observer errors. Harbor porpoises (Phocoena phocoena ) were censused during aerial surveys in the summer of 1997 in Southeast Alaska (9844 km survey effort), in the summer oft 1998 in the Gulf of Alaska (10,127 km), and in the summer of 1999 in the Bering Sea (7849 km). Sight- ings of harbor porpoise during a beluga whale ( Phocoena phocoena) survey in 1998 (1355 km) provided data on harbor porpoise abundance in Cook Inlet for the Gulf of Alaska stock. Sightings by primary observ- ers at side windows were compared to an independent observer at a belly window to estimate the probability of misidentification, underestimation of group size, and the probability that porpoise on the surface at the track- line were missed (perception bias, g(0)). There were 129, 96, and 201 sightings of harbor porpoises in the three stock areas, respectively. Both g(0) and effective strip width (the realized width of the survey track) depended on survey year, and g(0) also depended on the visibility reported by observers. Harbor porpoise abundance in 1997-99 was estimated at 11,146 animals for the Southeast Alaska stock, 31,046 animals for the Gulf of Alaska stock, and 48,515 animals for the Bering Sea stock. Manuscript submitted 2 March 2009. Manuscript accepted 4 March 2010. Fish. Bull. 108:251-267 (2010). The views and opinions expressed or implied in this article are those of the author (or authors) and do not necessarily reflect the position of the National Marine Fisheries Service, NOAA. Roderick C. Hobbs (contact author) Janice M. Waite Accurate estimation of abundance of cetaceans from survey data requires careful attention to both survey design and analysis (Buckland et al., 2001). Aerial surveys of cetaceans depend on rapid discovery, recognition, and recording of sightings of individuals and groups of animals by observers, in addition to accounting for animals that were missed because the observers did not notice them (perception bias) or because the animals were below the surface (availability bias) (Buckland et al., 2001). Although an experienced trained observer is efficient at recogni- tion of a species and recording data, it is necessary to include methods that can measure error rates of observers and account for them in the estima- tion of abundance. We present here the results of a series of aerial surveys designed to estimate the abundance of harbor porpoise (Phocoena phocoena ) in Alaskan waters. When, during an aerial line-tran- sect survey, an object or group of objects is encountered, an aerial observer has only a few seconds to complete several tasks: 1) perceive the objects, 2) identify the objects, 3) enumerate the objects, and 4) deter- mine the distance of the objects from the trackline. Items 1 and 4 are the major concern for the estimation of perception bias and for line-transect survey analysis, and it is generally assumed that the observer completes items 2 and 3 correctly or indicates uncertainty correctly (e.g., species code “unidentified porpoise” indicates uncertainty between Dali’s porpoise [ Phocoenoides dalli] and harbor por- poise). We develop methods to test the assumptions of correct species iden- tification and enumeration and apply them to the analysis of line-transect survey data and the estimation of abundance. From 1991 to 1993, the National Oceanic and Atmospheric Adminis- tration (NOAA) conducted aerial sur- veys in three regions of the Alaskan coast: 1) Cook Inlet and Bristol Bay in 1991; 2) in the waters around Ko- diak Island and south of the Alaska Peninsula in 1992; and 3) in the offshore waters of Southeast Alaska from Dixon Entrance to Prince Wil- liam Sound in 1993. The inside wa- ters of Southeast Alaska were sur- veyed in each of these years by NOAA crews aboard the NOAA RV John N. Cobb. The abundance estimates for these regions were combined to produce an abundance estimate for the Alaska stock of harbor porpoise (Dahlheim et al., 2000). Since then, the Alaska stock has been split in- to three stocks: Southeast Alaska (SEA), Gulf of Alaska (GOA), and the Bering Sea (BS) stocks (Fig. 1). The 1991-93 abundance estimate was subdivided to correspond with the new stock boundaries (Hill and DeMaster, 1998). To maintain up-to- date stock assessments, abundance 252 Fishery Bulletin 108(3) Uixon ~ Entrance 60'N- 55°N' 140 W 1 30 W Alaska Prince William Canada Bering Sea Bristol Stock Bay ^ Southeast Alaska stock Gulf of Alaska stock Gulf of Alaska 260 520 1,040 tm Kilometers Figure 1 The three harbor porpoise ( Phocoena phocoena ) stock regions in Alaska (Southeast Alaska, Gulf of Alaska, and Bering Sea). The gray shaded areas represent the areas surveyed in 1997- 99, subdivided into areas based on geographical features and depth zones. Dark gray offshore areas were surveyed at one third the effort level per square km than lighter gray nearshore areas. Black lines represent boundaries between stocks. estimates are required to be based on data not more than 8 years old (Wade and Angliss, 1997). To meet this requirement, abundance surveys of the harbor porpoise stocks in Alaskan waters were conducted from 1997 through 1999. An important consideration when conducting multi- year surveys is that animals may move from one sur- vey area to another among years and therefore may be counted more than once. Little is known about the year-to-year changes in the distribution of harbor por- poises in Alaska. For two studies of harbor porpoise (Phocoena phocoena ) on either side of the Atlantic, in the Danish Belt seas (Teilmann et al.1) and in the Gulf of Maine (Read and Westgate, 1997), it was concluded 1 Teilmann, J., R. Dietz, F. Larsen, G. Desportes, and B. Geertsen. 2003. Seasonal migrations and population structure of harbour porpoises (Phocoena phocoena) in the North Sea and inner Danish waters based on satellite telemetry. Abstract in proceedings of annual meeting of the European Cetacean Society, Tenerife, Spain. that porpoises follow similar movement patterns from year to year and typical ranges of up to 200 km. This finding indicates that a net movement in response to interannual variation in habitat could occur over a range of 100 km. Although this could result in a bias in the estimation of abundance for each stock, depending on the year of the survey, it is not likely that this is a significant occurrence. Each stock region comprises 800 to 1200 km of shoreline so at most approximately 10% to 15% of the region is potentially subject to a net shift in distribution from or into the adjacent stock. Also, the stock boundaries have been chosen to correspond with areas of low harbor porpoise density and therefore there are few animals available to make a shift. These two arguments suggest that if a net shift does occur, it af- fects at most a small percentage of the population. This study had three objectives: 1 to present the results of an aerial survey of three harbor porpoise stocks in Alaskan waters during the summers of 1997, 1998, and 1999; Hobbs and Waite: Abundance of Phocoena phocoena in three Alaska regions 253 2 to produce a correction factor for perception bias that was specific to harbor porpoise in Alaskan waters and to the surveys presented here and to develop methods to test observer performance, by using data collected during these surveys; and 3 to generate abundance estimates for the three stocks of harbor porpoise in Alaska during 1997-99. Materials and methods Survey design Aerial surveys were conducted during June and July beginning with the SEA stock in 1997, proceeding west- ward through the GOA stock in 1998, and on to the BS stock in 1999. Each study region was divided into areas; 70 areas in Southeast Alaska, 39 areas in the Gulf of Alaska, and 4 areas in the Bering Sea (primar- ily in Bristol Bay) based on geographical features for inside waters, such as straits and inlets, and two depth zones for offshore waters (Fig. 1). Southeast Alaska was divided into more areas because of its complicated system of waterways, whereas Bristol Bay has rela- tively homogenous features and therefore was divided into fewer survey areas. Survey effort was stratified by area in Southeast Alaska based on harbor porpoise encounter rates calculated from sightings made in previ- ous surveys (Dahlheim et ah, 1993, 19942). The survey transect design each year varied depending on the body of water. In general, the transects in offshore waters were stratified by depth and distance from shore, after an alternating two short and one long sawtooth transect pattern, so that survey effort in the nearshore strata was about three times that in the offshore strata. The 1991-93 surveys were designed with fixed distances of 28 km offshore for the short and 74 km offshore for the long sawtooth tracklines. Our surveys were designed to include the area surveyed in 1991-1993 but also to cover the continental shelf if it extended beyond the original survey. Each set of sawtooths had two crite- ria and the further offshore of the criteria determined the length of the line. Specifically, in 1997, the short transects in the sawtooth transect pattern extended to a distance of 31 km offshore or the 183-m (100 fm) depth contour, and the long transects extended 74 km or to the 1829-m (1000 fm) depth contour, whichever was farthest from shore. In the 1998 GOA survey, the shelf fell much more gradually in places and funding limited the total survey time. Therefore, the nearshore strata transects were reduced to a distance of 28 km or to the 91-m (50 fm) depth contour, whichever was 2 Dahlheim, M., A. York, J. Waite, and R. Towel], 1993. Abun- dance and distribution of harbor porpoise (Phocoena phocoena) in Southeast Alaska and Western Gulf of Alaska, 1992. 1992 Annual report to the Marine Mammal Protection Act (MMPA) Assessment Program, 52 p. Office of Protected Resources. NMFS, NOAA, 1335 East-West Highway, Silver Spring, MD. farthest from shore, whereas the long transects followed the same criteria as in 1997. Because the entire Bering Sea survey region is shallower than 183 m (100 fm), it was covered equally with short transects out to 30 km along the shore and with parallel north— south lines through the center approximately 18 km apart. Smaller bays and inlets were treated separately and stratified by the width of the mouth of the bay or inlet. A subset was chosen to approximate the survey effort by area for the other survey regions, and selection was made on the basis of convenience (i.e., bays and inlets close to the end of survey effort lines were chosen). Line-transect surveys were flown at an altitude of 152.5 m and a speed of 185 km/h in a DeHavilland Twin Otter aircraft. Survey areas were chosen each day to complete coverage of contiguous areas during weather with winds below 15 knots and at a ceiling above 1000 ft (305 m). Survey lines were broken off and other tracklines with better conditions were sought if the Beaufort sea state exceeded 3 or if visibility dropped to poor for a significant period (at the discretion of the team leader). A primary observer (also referred to as a “side observer”) was stationed at the left and right bubble windows of the plane; these positions allowed them to see water directly below the plane. To collect additional sightings and data to estimate perception bias for this study, an independent observer was sta- tioned at a belly window located in the floor at the back of the plane (this observer is also referred to as the “belly observer”). This window provided a circular field of view 100 m (30°) to either side of the trackline and 200 m along the trackline. Five observers rotated in 40-minute shifts through five positions: the right and left bubble windows (primary observers), the belly window (independent observer), a computer station, and a rest position. A headset system was used by the primary observers and computer operator to communi- cate openly, and the independent observer was isolated and used a string attached to the arm or ankle of the computer operator to indicate a sighting and a notepad to relay information. A simple short hand was developed so that the belly observers would not need to take their eyes off of the trackline. Survey data were recorded directly to a laptop com- puter in the airplane using a Turbo PASCAL (vers. 5.0, Borland Software Corp., Austin, TX) language-based software customized for the survey. The software in- cluded a proprietary routine (Survey, vers. 3.2, Cascadia Research, Olympia, WA) which read the text output of a global positioning system (GPS) unit connected directly to the serial port of the computer. The date, time, and position of the aircraft were automatically entered into the survey data every minute or whenever other data were entered by the recorder. At the start of each tran- sect, waypoint numbers, observer positions, and envi- ronmental conditions were recorded. Environmental con- ditions included percent cloud cover, Beaufort sea state, visibility (a subjective rating of sighting conditions by each observer at the following levels (excellent, good, fair, poor, and unacceptable), and glare (none, minor, 254 Fishery Bulletin 108(3) bad, or reflective) experienced by each observer. Visibil- ity was defined as the observer’s subjective assessment of the conditions for the likelihood of seeing a harbor porpoise and the observer’s assessment of the effect of glare, sea state, as well as less quantifiable factors such as turbidity, sun angle, unusual weather conditions, and fatigue on the observer’s ability to sight a harbor por- poise. The observers reported these environmental data as changes in such data were noticed along a transect. For each sighting, the observer notified the computer operator when the beam line of the plane crossed the animal’s location. The primary observers used inclinom- eters to obtain the vertical angle below the horizontal to convert the perpendicular distance of the animal from the trackline (Lerczak and Hobbs, 1998). To determine the distance of a sighting from the trackline indicated by a center line on the belly window, the window was subdivided with a grease pencil into six 10°-bins (out to 30° to either side of the trackline for an averaged eye height), labeled 1-6 from port to starboard. When alerted to a sighting by the primary or independent observers, the computer operator immediately entered the sighting by using a hot key assigned to an observer (which recorded the observer’s initials and which cap- tured the time and position from the GPS unit). The hot key also opened a window for entering species name, vertical angle or angle bin, group size, and any notable animal behavior. Matching sightings from side and belly windows Sighting data (time, perpendicular distance, species, and group size) collected on the same transects were compared between side and belly observers. For compari- son purposes, left- and right-side sighting angles were converted to corresponding belly observer bin number. Sightings were considered matches (same group seen by both observers) if they 1) occurred within 5 seconds of each other; 2) were not greater than one 10° bin dif- ference; and 3) met other conditions such as a species of similar size or of hierarchical relation (e.g., harbor porpoise matched to unidentified small cetacean) and similar group size. Matched sightings were used 1) to estimate an empirical average angle for each belly window bin, based on the angles measured from the side windows; 2) to identify circumstances resulting in unreliable species identifications (see Error's in spe- cies identification in Appendix I); 3) to estimate bias in group-size estimates by the belly observer; 4) to estimate perception bias and g(0) (here g(0) accounts only for the consequences of perception bias; correction for availability bias is treated separately as described below); and 5) to eliminate duplicate sightings from the distance analysis. Correction for bias in group-size estimates determined by belly observers Initial inspection of the data when both the side and belly observers reported a sighting indicated that the group size estimate of the belly observer was occasion- ally less than that provided by the side observer — a result of the restricted visual field and limited observa- tion time for the belly observer. For each of these pairs, the count by the side observer was divided by the count by the belly observers. These ratios were then grouped by belly observer group size and averaged to estimate a correction for each group size reported by the belly observer. The standard error for each correction factor was estimated by the usual formula. The correction was applied to all group sizes from belly sightings included in the average estimate of group size. Distance smearing Angle rounding occurred in both the side observer data and the belly observer data. In the case of the side observer data, peaks in frequency occurred on multiples of 5°. The rounding of angles often occurred after the sighting was out of the field of view and the observer estimated the angle from a remembered location. The accuracy for these remembered locations may not have been any better than 5°, and therefore created a ten- dency for observers to use a close 5° increment number rather than one of the marks in between. Belly observer data were assigned to a bin and were thus automati- cally rounded. To remove these effects, side sightings were dithered uniformly over 13 m (2.5° on either side of the reported angle) and belly sightings were dithered uniformly over 26 m (5° on either side of the reported angle, the center of sighting bins). The dithering dis- tance was chosen empirically as the minimum distance necessary to remove the rounding effect. The dithering was repeated several times and the cumulative distri- bution of the sightings by distance to the trackline was examined. An instance of the dithering which gave a visually smooth distribution was retained and used as the data set for further analysis to estimate the sight- ing distribution. Estimation of perception bias and g( 0) All three years of data were combined to estimate per- ception bias from comparisons of the primary and the independent observer sightings. Logistic regression with a generalized linear model (the GLM function in S-PLUS, Lucent Technologies, Murray Hill, NJ) and an offset algorithm for comparison of paired sightings (Buckland et al., 1993) was used to estimate the perception bias of the side and belly observers on the trackline. Review of the ratio of matched to unmatched sighting for the belly and side observers by 25-m bins indicated that the two inner bins were consistent with each other (0-25 m and 25-50 m), whereas the outer bin (50-75 m) differed. Con- sequently, perception bias was estimated by using only sightings within 50 m of the trackline (approximately 20° at the standard survey altitude or bins 2 through 5 in the belly window). Sightings beyond this cutoff dis- tance were excluded. Possible covariates in the logistic regression were visibility, sea state, cloud cover, glare, Hobbs and Waite: Abundance of Phocoena phocoena in three Alaska regions 255 group size, observer, and survey year. Covariates were initially tested individually to identify functional forms or groupings that could reduce the number of parameters necessary to represent them. The discrete covariates (visibility, sea state, glare, group size, observer, and survey year) and the continuous covariate cloud cover (grouped into five categories: 0-20%, 20—40%, 40-60%, 60-80%, 80-100%), were examined individually as cat- egorical factors. The coefficients from the categorical analysis were then charted against their hierarchical ranks. Where the coefficients appeared to fit a simple functional form of the hierarchical ranks (line, square root, natural logarithm, exponential) or could be grouped to reduce the number of parameters, the analysis was repeated with this alternative. The parameters for the function or grouping were estimated by using the regres- sion described above and compared to the result of the categorical factor by using Akaike’s information criterion (AIC). The function or grouping was used in the subsequent analysis if it improved the AIC. From this preliminary analysis, visibility and sea state were found to have a nearly linear effect and were treated as linear functions by using the hierarchical number as the value and by setting the best condition to one; group size was also considered to be linear. Cloud cover, glare, observer, and survey year were considered as cat- egorical data. Significant covariates were then combined in the GLM model and removed in a stepwise manner until the AIC had been minimized. The perception bias for each observer position and each transect segment was estimated from the final model and combined to estimate g{0) for each transect segment (see Appendix II for details). The program DISTANCE, vers. 3.5 (Thomas et al., 1998) allowed only a global g(0) and thus did not ac- commodate g(0) to be estimated for each transect seg- ment from environmental and observer covariates. It was possible to circumvent this limitation by adjusting the length of each trackline to allow an estimate of density in the vicinity of each trackline because g(0) and length are multiplied together to estimate density. The estimates ofg(0) for each transect segment were averaged for all three years weighted by the transect lengths to estimate an average glO). The length of each transect segment was multiplied by its estimated g(0) divided by the average g(0) to generate an adjusted transect segment length which accounted for the g( 0). The adjusted transect segment lengths were then used in DISTANCE in place of the actual lengths and the av- erage g(0) calculated above was used as the global g(0) in DISTANCE. The standard error for the global g( 0) was estimated as the weighted average of the standard errors of the g( 0) estimates for the individual transect segments. Estimation of abundance The line-transect analysis program DISTANCE (vers. 3.5) was used to estimate the observed density of harbor porpoise in each surveyed region. Two sighting prob- ability curves were estimated so that for transect seg- ments with usable belly observer effort data, sightings from the side and belly observers could be combined and duplicates removed or, when no belly observer data were available, sightings from the side observers only could be used. To identify significant effects of possible covariates for estimated strip width (presence or absence of a belly observer, survey year, individual observers, visibility levels, glare types, percent cloud cover, and sea state), each factor was considered separately as a covariate and the one with the lowest value for the AIC was retained. This process was repeated with the remaining possible covariates in an additive manner until further addition of covariates did not lower the AIC. Distances were pooled into 50-m bins to allow application of the esti- mate of perception bias. Densities were estimated for the individual areas with usable survey data. Unsurveyed areas such as the small bays and inlets were assigned the average densities from the surveyed areas of that stratum. These densities were then averaged, weighted by the area of each survey region, to estimate an aver- age observed density and abundance for each stock. Variances were calculated as in Buckland et al. (2001). The correction factor for availability bias is the inverse of the estimate of availability from Laake et al. (1997) (2.96 = (l/0.338), CV=0.18). This factor was applied as a multiplier to the observed abundance estimates to produce the abundance estimate for each stock. Incorporation of other survey data The vast area comprising Alaska waters made it impos- sible to survey all areas where harbor porpoises occur. Harbor porpoise sighting data were available from a concurrent NMFS beluga whale line-transect survey in Cook Inlet. For this survey, an Aero Commander aircraft with bubble windows was used; however, the windows were smaller than those of the Twin Otter aircraft, and the observers could not see directly below the plane. Survey methods were similar, except that the beluga whale survey was conducted at an altitude of 244 m and the primary focus was beluga whales. The search effort, therefore, was not concentrated as close to the trackline as it would have been if the survey had been designed to survey harbor porpoise. NMFS National Marine Mammal Laboratory has conducted these beluga whale surveys each year since 1993. We estimated abundance for harbor porpoise in Cook Inlet using the 1998 survey data, a strip width estimated from all beluga surveys (1993 to 1999), and the correction for availability bias from Laake et al. (1997). Perception bias could not be estimated for this survey. This abundance estimate was added to the abundance estimate from the GOA survey to produce a combined estimate for that stock. Minimum abundance estimate A minimum abundance estimate, Wmin, defined in Wade and Angliss (1997) as the lower 20th percentile of the lognormal error distribution, is used in management 256 Fishery Bulletin 108(3) 58* N- -58°N 56*N- -56° N ■60'N 144°W 140”W 1 136°W 132*W 144 °W 140“W i 136°W 1 132’W Gulf of Alaska 0 62 5125 250 375 500 l Kilometers Canada Figure 2 Completed survey transects and sightings (circles) of harbor porpoise ( Phocoena phocoena ) during the 1997 aerial survey in the Southeast Alaska stock region. decisions by NMFS. This quantity was calculated for each stock from the completed abundance estimates as N ■ =Ne y min -0.842 Vlnd+CV ( AO2 ) where N = estimated abundance; and CV(N) = the estimated coefficient of variation of N. Results The 1997 line-transect aerial survey was conducted from 27 May to 7 June and 10-28 July 1997 in the inside waters of southeastern Alaska, Yakutat Bay, Icy Bay, and in offshore waters from Dixon Entrance to Cape Suckling (Fig. 2). Necessary repairs on the survey plane resulted in an unplanned month-long break in the survey, and adverse weather prevented a second survey of offshore waters. A total of 9844 km were surveyed. The 1998 survey was conducted from 27 May to 28 July 1998 in Prince William Sound, the western Gulf of Alaska (from Cape Suckling to the west side of Kodiak Island), and Shelikof Strait (Fig. 3). Gaps in the survey effort occurred on account of inclement weather, primar- ily off the Kenai Peninsula and the southern side of the Alaska Peninsula west of Kodiak Island. A total of 9486 km were surveyed. The 1999 survey was conducted 11 June to 4 July in Bristol Bay and associated bays. In addition, an area south of the Alaska Peninsula west of Chignik Bay was surveyed that was not completed in 1998 (Figs. 3 and 4). A total of 8490 km were surveyed. The 1999 data for the Gulf of Alaska was included with the 1998 data to estimate abundance of harbor porpoise for the Gulf of Alaska. Sightings of harbor porpoise for each region (Figs. 2-4) were more common in nearshore areas, but oc- curred throughout the depth range surveyed during all three surveys. High densities of harbor porpoise were found in Yakutat Bay and near Wrangell (Fig. 2), be- tween Prince William Sound and Cape Suckling, on the southeast side of Kodiak Island, southwest of Chignik Bay (Fig. 3), and in a few small bays on the northern side of Bristol Bay (Fig. 4). Corrections for species misidentification and for under-counting group sizes by belly observers A cursory examination of the discrepancies in species identification between side and belly observers from the 1997 and 1998 seasons indicated that discrepan- cies occurred primarily when inexperienced observers were at the belly window and had fewer than 10 days of Hobbs and Waite: Abundance of Phocoena phocoena in three Alaska regions 257 164°W 160°W 156°W 152°W 148°W 144W Figure 3 Completed survey transects and sightings (circles) of harbor porpoise (Phocoena phocoena) during the 1998 aerial survey and the 1998 beluga whale aerial survey (triangles) in the Gulf of Alaska stock region. Also shown are transects and sightings of harbor porpoise made during the 1999 aerial survey south of the Alaska Peninsula and west of Chignik Bay. survey experience. Based on this ad hoc assessment of discrepancies, an experienced observer was defined as one who had 10 or more days of observation experience on the survey. Species discrepancies involving harbor porpoise were those that led to a misidentification of harbor porpoise as either a Dali’s porpoise or harbor seal ( Phoca vitulina). Within the data from the first two years of the survey, there were 68 species identifications of harbor porpoise, Dali’s porpoise, and harbor seal from paired observ- ers. Of these 68 identifications, 52 were determined by paired experienced observers in the side and belly (one discrepancy), 12 were determined by an experienced side observer paired with an inexperienced belly ob- server (4 discrepancies), and 4 were determined by an inexperienced side observer paired with an experienced belly observer (no discrepancies). No correlation be- tween discrepancies and environmental conditions was found. To verify the observation that the inexperienced observers in the belly position had a higher than aver- age misidentification rate and determine if the rate was unacceptable, four possible models were compared and AIC was used to identify the most parsimonious model. The models were 1) side observers and experienced and inexperienced belly observers were all different (four parameters); 2) experienced and inexperienced side ob- servers and experienced belly observers were equivalent and inexperienced belly observers were different (two parameters); 3) experienced and inexperienced observ- ers were different but side and belly were equivalent (two parameters); and 4) all observers were equivalent (one parameter). For model 1, probabilities of a correct identification were >0.99 for both experienced and in- experienced side observers, 0.98 for experienced belly observers, and 0.67 for inexperienced belly observers with an AIC of 13.0. For model 2, side observers with experienced belly observers had a probability of 0.98, and inexperienced belly observers had a probability of 0.67 with an AIC of 9.1. Model 3 resulted in a prob- ability of 0.99 for experienced observers and >0.76 for inexperienced observers with an AIC of 11.6. Model 4, a probability for all observers, was 0.96 with an AIC of 17.6. The most parsimonious model (lowest AIC) was model 2, indicating that an inexperienced observer in the belly position had a low reliability for species iden- tification. Consequently, observation effort and sightings by inexperienced observers in the belly during their first 10 survey days were treated as practice and were not included in the subsequent analysis. Although it would be possible to estimate a ^(0) that accounted 258 Fishery Bulletin 108(3) 164°W 160°W 156°W Completed survey transects and sightings (circles) of harbor porpoise ( Phocoena phocoena) during the 1999 aerial survey in the Bering Sea stock region. Transect and sightings on the south side of the Alaska Peninsula are shown in Figure 3 as part of the Gulf of Alaska survey. for the inexperienced observers in the belly, we did not have the estimate of the density for the other species necessary to complete this calculation. Group size was typically underestimated by the belly observer in comparison to the side observers. In sight- ings by both the belly and side observers, separate corrections were calculated for group sizes reported as one individual and groups reported as two individuals by the belly observer. Of 30 groups reported as one harbor porpoise by the belly observer, 25 were reported as a group size of one by the side observer and 5 were reported as a group size of 2, yielding a multiplicative correction of 1.167 (CV=0.059) groups of size one seen only by the belly observer. Likewise, of 12 groups re- ported as two harbor porpoise by the belly observer, 11 were reported as a group size of 2 by the side observer and one was reported as a group size of 3, yielding a multiplicative correction of 1.042 (CV=0.080) for groups of 2 observed from the belly of the aircraft . The group- size estimate from the side observer was used when a group was reported by both the side and belly observers. The correction was only applied to group size when the group was seen only by the belly observer. A correc- tion for total animals was estimated as the sum of the group sizes with the belly-window-derived groups sizes corrected, divided by the sum of the group sizes with the belly-window-derived group sizes uncorrected. As a result, a multiplier of 1.018 (CV=0.006) was applied to the abundance estimates. It was necessary to apply a general correction rather than correct individual group sizes to avoid problems with the g(0) estimate and DIS- TANCE analysis arising from non-integer group sizes. Estimation of perception bias and g(0) Comparisons between sightings by the belly observer and the side observers indicated that each missed a small but significant fraction of the near-surface animals on the trackline. A total of 129 potential matches between experienced belly observers and independent side observ- ers within 50 m of the trackline were examined for perception bias and g(0) estimation. Although several of the potential covariates were significant by them- selves, only visibility as a continuous variable remained in the stepwise elimination. A significant difference in estimated strip width between survey year 1997 and the years 1998 and 1999 was identified in the distance analysis and, therefore, year was included as a covariate as well. Although year was not a significant coefficient, there was a significant turnover in personnel from the Hobbs and Waite: Abundance of Phocoena phocoena in three Alaska regions 259 Figure 5 Relative probability of detection of harbor porpoise ( Phocoena phocoena) by distance from the survey trackline (km) determined with a half-normal model for the 1997 sighting data of harbor porpoise in the Southeast Alaska stock region. The histogram shows the distribution of sightings in 0.05-km groups. first to the second and third years, and 1997 was the first year of a belly observer for the survey leaders. The belly observer had a significantly higher probability of sighting an available group than the side observers for any particular sighting, and the belly observers routinely reported better visibility than did the side observers. This difference in perception bias was accounted for by the difference in reported visibility. The logistic regres- sion coefficients were as follows: constant=1.187 ±0.542, =2.19); for year=0 for 1997 and 1 for 1998, 1999, coef- ficients.296 ±0.290, (£=1.02); for visibility=l (excellent), 2 (good), 3 (fair), 4 (poor), 5 (unacceptable) as a continu- ous variable, coefficient=— 0.502 ±0.217, (£=-2.31). The model for probability of sighting of an available group for a single observer is then 1. 187+0. 296 y ear-0. 502visibility P(sighting\ year, visibility) = - - el,187+0.296ytor_o.so2l,fel-6ffi» • Heterogeneity in probability of sighting a harbor por- poise resulted in a decrease of a difference of roughly 0.11 for each reduction in visibility and an increase by roughly 0.06 from 1997 to 1998-99 (Table 1). The average observed g( 0) values (perception bias only) for the SEA stock, the GOA stock, and BS stock of harbor porpoise were 0.641 ±0.069, 0.729 ±0.048, and 0.748 ±0.046, respectively, and yielded average perception bias correction factors of 1.560 (CVS. 108), 1.372 (CVS. 066), and 1.337 (CVS. 062), respectively (Table 2). Estimated strip width for observations Variation in effective strip width (ESW) occurred for the configuration of observers and visibility as reported by the observers. Few sightings occurred beyond 400 m and therefore this distance was chosen as the trunca- tion point for distance from the trackline and sightings beyond this distance were not included in the analysis. Effort was separated into effort with and without a belly observer. ESW without a belly observer was the ESW on one side of the plane covered by a side observer. ESW with a belly observer represented the effort of one side observer and half of the belly observer because the belly observer’s field of view was divided by the trackline (note that duplicate sightings were removed such that where sightings were reported by both the side and belly observer, only the side sighting record was used). The ESW for 1997 was significantly different from that of 1998 and 1999; therefore they were treated separately (Table 1). In the 1997 data, significant variation in ESW was related to visibility level and in the 1998 and 1999 data, to the presence or absence of the belly observer. The best fit for the detection function was a half-normal curve for the 1997 data set and a half-normal curve with a one term cosine correction for the 1998—99 data set (Figs. 5 and 6, Table 2). The ESW of the survey team in 1997 decreased by roughly 20% per step change in visibility. When this decrease in ESW was combined with approximately a 12% decrease in g(0) with each step in visibility, the product (Table 1) indicated an 260 Fishery Bulletin 108(3) Table 1 Estimated g(0) (probability of detecting an animal at the surface on the trackline (perception bias) and effective strip width (ESW) in km for harbor porpoise in Alaska determined from surveys from 1997 through 1999. Data were obtained from indi- vidual observers and teams of observers, all of whom reported the same visibility code. The product of these two values, gtO) and ESW, is a measure of the relative effectiveness of the observer team under different conditions. Single observer=single observer at either the right, left, or belly window of aircraft. Team of observers=a team of observers at the right, left, and belly window of the aircraft. Single observer Team of observers 1997 1998-99 1997 1998-99 Visibility Value SE Value SE Value SE Value SE 1 (excellent) g<0) 0.66 0.08 0.73 0.06 0.89 0.01 0.93 0.01 ESW 0.252 0.022 0.139 0.008 0.252 0.022 0.118 0.007 g(0)ESW 0.166 0.009 0.101 0.004 0.224 0.002 0.11 0.001 2 (good) g(0) 0.55 0.06 0.62 0.04 0.79 0.02 0.85 0.01 ESW 0.2 0.017 0.139 0.008 0.2 0.017 0.118 0.007 g(0)ESW 0.11 0.004 0.086 0.002 0.158 0.003 0.1 0.001 3 (fair) g(0) 0.42 0.07 0.49 0.06 0.66 0.04 0.74 0.02 ESW 0.153 0.013 0.139 0.008 0.153 0.013 0.118 0.007 g(0)ESW 0.064 0.002 0.068 0.002 0.101 0.003 0.087 0.001 4 (poor) £(0) 0.31 0.1 0.37 0.09 0.52 0.11 0.61 0.08 ESW 0.116 0.01 0.139 0.008 0.116 0.01 0.118 0.007 g(0)ESW 0.036 0.001 0.051 0.002 0.06 0.003 0.072 0.004 5 (unacceptable) £(0) 0.21 0.11 0.26 0.12 0.38 0.19 0.46 0.16 approximately 30% decrease in effective effort with each step in visibility and that survey effort during poor conditions had less than one quarter of the effective- ness of effort during the best conditions (Fig. 7). In the 1998-99 surveys (Fig. 8), the ESW was narrower overall and slightly broader when the belly observer was not present. Although this seems counterintuitive, it is the result of a peak that occurred near the trackline when the belly sightings were included and which made the distribution away from the trackline relatively lower and resulted in the narrower ESW (Table 1, Fig. 8). When ESW andg(O) were multiplied together, the added value of the belly observer was 10% under the best conditions and nearly 50% under poor conditions. Density and abundance of harbor porpoise Abundance estimates of harbor porpoise increased from east to west as did estimates of average density by stock (0.10, 0.19, and 0.44 porpoise/km-, respectively). Aver- age observed harbor porpoise densities (uncorrected for availability or perception biases) for the SEA, the GOA, and the BS stocks were 0.033 groups/km2 (CV=17.2%), 0.062 (CV=11.9%), and 0.153 (CV=13.2%), respectively. Approximately 5% of the study areas, consisting primar- ily of inlets and channels, were unsurveyed. Density estimates for these unsurveyed areas were extrapolated from similar surveyed areas in the same general region (Table 3). The correction factor of 2.96 (CV=0.180) (Laake et al., 1997) was applied to each abundance estimate to account for availability bias. The full corrections for vis- ibility bias (correction for perception bias x correction for availability bias) were 4.62 (SEA, CV=21%), 4.06 (GOA, CV=19%), and 3.96 (BS, CV=19%). For the Cook Inlet survey, the effective strip width (0.280 km, CV=0.281) was based on 44 sightings from the 1993 tol999 surveys. Truncation of the sighting strip by discarding sightings less than 0.1 km from the trackline or greater than 0.6 km from the trackline on each side of the plane was necessary to obtain a good fit of the detection function. The best fit for the detection function for the Cook Inlet data, based on AIC, was a hazard-rate curve with a cosine correction (Fig. 9). The 1998 beluga whale survey in Cook Inlet resulted in eight harbor porpoise sightings along 1355 km of track- line. No data were available to estimate the perception bias for this survey and its format was sufficiently dif- ferent from the harbor porpoise survey with the result that it was uncertain whether the perception bias cor- rection would be approximately correct. Consequently, only the correction for availability (2.96, CV= 0.180) was applied. This results in a rather conservative estimate with a known negative bias which we feel is preferable to one with an unknown bias. The abundance estimate for the SEA stock of harbor porpoise was 11,146 animals (CV=24.2%; Nmin=9116, Hobbs and Waite: Abundance of Phocoena phocoena in three Alaska regions 261 Table 2 Survey parameters and abundance estimates for harbor porpoise ( Phocoena phocoena) stocks off Southeast Alaska, the Gulf of Alaska, and the Bering Sea in 1997, 1998, and 1999, respectively. Cook Inlet survey results are taken from a survey for beluga whales which followed a protocol similar to the harbor porpoise surveys. No perception bias correction was available for the 1998 Cook Inlet survey; consequently observed abundance is used in place of total abundance for this area. The Cook Inlet abundance was included in the Gulf of Alaska stock total abundance. Extrapolated areas were small inlets and unsurveyed areas where the density of harbor porpoise was assumed to be the same as similar surveyed areas for the purpose of estimating total abundance. Coefficient of variation (CV) of a statistic is the standard error of the statistic divided by the statistic; confidence intervals are calculated by using a log-normal distribution and Nmin = Afe-°842'/ina+cv40% survival. Hook timers indicated that only 8.4% of the swordfish were alive after 8 hours of capture, and two shark species (blue shark [Prionace glauca ] and oceanic whitetip shark [ Carcharhinus longimanus ]) showed a greater resilience to capture: 29.3% and 23.5% were alive after 8 hours, respectively. Our results have impli- cations for current fishing practices and we comment on the possibilities of modifying fishing strategies in order to reduce operational costs, bycatch, loss of target fish at sea, and detri- mental impacts on the environment. Manuscript submitted 19 August 2009. Manuscript accepted 22 March 2010. Fish. Bull. 108:268-281 (2010). The views and opinions expressed or implied in this article are those of the author (or authors) and do not necessarily reflect the position of the National Marine Fisheries Service, NOAA. Empirical studies have shown that surface longlines set at night are more productive for capturing sword- fish ( Xiphias gladius) than lon- glines set during the day (Kume and Joseph, 1969). Studies with ultra- sonic telemetry have proved useful for understanding habitat distribu- tion and for providing some insights into short-term horizontal and verti- cal movements of swordfish (Carey and Robinson, 1981; Carey, 1990). Recently, more advanced tagging technology devices, such as archival tags (Takahashi et al., 2003) and pop- up satellite archival tags (Canese et al., 2008; Neilson et al., 2009), have provided comprehensive information on the diel behavior of swordfish and their movement patterns, both of which indicate the ability of sword- fish to navigate to long distance feed- ing areas in the Pacific and Atlantic oceans and Mediterranean Sea. Ver- tical movement data obtained from these various studies have shown diel diving patterns to be very consistent, as have been observed in other pelagic fishes (Musyl et al., 2003). During the daytime, swordfish can spend most of their time at depths between 250 and 650 m (Canese et al., 2008) and as deep as 900 m (Takahashi et al., 2003). At crepuscular hours, they swim up and down through a consid- erable range of depths following the movements of organisms in the deep sound-scattering layers but restrict their movements to forage in the uniform surface mixed-layer during the night. They also feed during the day and can demonstrate basking behavior. Carey and Robinson (1981) suggested that swordfish swimming activity in the water column is closely associated with prey locations and concluded that lunar illumination was a determining factor influencing verti- cal migrations. Swordfish are “strike and return” feeders (Nakamura, 1983) targeting fast-moving prey. Sword- fish can achieve a maximum speed of Poisson et al.: Effects of lunar cycle and fishing operations on longline catches 269 130 km/h (Lee et al., 2009) and Fritsches et al. (2005) demonstrated that their large eyes and pupils are ide- ally adapted to detect rapid movements in dim light. Swordfish rely on visual cues at small scales (meters) and ambient light conditions are likely to be a major factor influencing feeding behavior. There is strong anecdotal information from captains who target large pelagic fish that the lunar phase af- fects fishing success for swordfish and other pelagic fish. The effect of the lunar cycle on swordfish catch- ability has been investigated in longline (Bigelow et al., 1999; Neves Dos Santos and Garcia, 2005; Damalas et al., 2007) and driftnet fisheries (Di Natale and Manga- no, 1995), but the results obtained are not directly comparable because they were conducted in differ- ent areas and fisheries. Generalized additive models (GAMs) have been used to examine the relative in- fluence of environmental conditions and operational factors on pelagic longline catch rates (Bigelow et al., 1999; Walsh and Kleiber, 2001; Tserpes et al., 2008). In the case of the Reunion Island longline fishery, Guyomard et al. (2004) demonstrated significant effects (among others) of geostrophic currents generated by sea level anomalies (SLA) and lunar day on swordfish catch and catch per unit of effort (CPUE, number of fish per 1000 hooks). I n 1991, the first swordfish longliner began operating from Reunion Island, a French overseas territory in the southwestern Indian Ocean (21°5'S lat., 53°28'E long.), Two main factors promoted the development of this fishery: 1) the success of the Asian fleet that was based at Reunion Island and 2) a new tax regulation, offering exemption for certain investments in French overseas territory, which encouraged French fishing companies to come to Reunion Island. The annual average catch in the Reunion Island longline fishery during the years 1996-2000 was 2670 metric tons (t) and the composi- tion by weight was 1730 t (65%) of swordfish (Xiphias gladius), 320 t (12%) of albacore ( Thunnus alalunga), 270 t (10%) of yellowfin tuna (T. albacares), 130 t (5%) of bigeye tuna (T. obesus ), 90 t (3%) of billfish (Indo- Pacific black marlin [ Makaira indica ], Indo-Pacific blue marlin [M. mazara], shortbill spearfish [ Tetrapturus angustirostri s], and Indo-Pacific sailfish [Istiophorus platypterus]), 60 t (2%) of sharks, and 70 t (3%) of other species. The main goal of the current study was to investi- gate the performance of the domestic longline fishery at Reunion Island with regard to several variables to determine whether lunar periodicity affected the catch of pelagic species. A three-year logbook data series (1998-2000) was used. In addition, data were aug- mented by deploying time-depth recorders (TDRs) and hook-time recorders on a number of commercial longline sets. Information on capture time from hook timers allowed us to investigate the survival of large pelagic fish on longlines. Hourly catch rates were correlated with lunar illumination but also with the tidal phase which could be considered as an indicator of induced local currents. Finally, to test the assumption that feeding behavior of large pelagic fish was predicated on vision (sword- fish can differentiate between blue-green wavelengths [Fritsches et al., 2005]), we investigated whether green chemical lightsticks influenced catch rates and species composition. With the combined results for catch rates and species composition, we comment on the possibili- ties of modifying fishing operations and strategy 1) to reduce operational costs during fishing trips, 2) to re- duce loss of target fish at sea, and 3) to reduce bycatch mortality in accordance with the United Nations guide- lines for responsible fisheries (FAO, 2003). Materials and methods Operational characteristics of the domestic longline fleet The Reunion Island swordfish longline fishery is a sur- face fishery. An adapted “American style” gear is used, where the monofilament mainline (3.5-4 mm in diam- eter) longline is set at dusk and fished during the night. The mainline was deployed from a hydraulically powered spool over the stern and fishing depth was controlled by adjusting branchline (~20 m) and floatline lengths (~10-90m), distance between buoys (~50 m), number of hooks between floats (6-10 hooks), and vessel speed (Fig. 1). In strong currents, lead weights can also be added to branchlines to control mainline depth. The fishery uses J hooks (size 8/0 or 9/0) baited with squid (Illex spp.) and green chemical lightsticks were attached ~1 m above the hook on every third branchline. The domestic longline fleet can be categorized by two vessel sizes (classes): a small-medium boat class (vessels <16 m), and a large boat class (vessels >16 m). The fishery underwent dramatic changes between 1992 and 2000 expanding from 5 to 23 active small- medium boats and from zero to 9 large boats, the latter with a peak of 14 in 1998. No longliners were authorized to fish within a zone of 30 nautical miles (~55 km) from Reunion Island in order to avoid con- flicts with the artisanal tuna fishery. Thus, these vessels explored new fishing areas within and beyond the French Exclusive Economic Zone (EEZ) (Fig. 2). The number of days at sea varied according to ves- sel size, vessel capacity, and weather conditions. The smallest boats stayed at sea 2-3 days, medium-size vessels generally stayed 6-8 days; whereas some of the largest vessels stayed at sea for up to 30 days. The large boats gradually expanded to fishing grounds to the west and southwest of Reunion Island within the EEZ in association with deep seamounts and sea sur- face temperature (SST) fronts (Fig. 2; area C) and to the Mozambique Channel (area B) and to Seychelles Island waters in 1997 and 1998 (area A). A decline of large boats from the fleet brought notable changes; therefore catches in 1999 and in 2000 were taken mainly around Reunion Island between 20s and 259S lat. and 50°-60°E long. Only a few vessels ventured far from Reunion Island in 2000. Table 1 summarizes 270 Fishery Bulletin 108(3) the fishery and environmental indices and subsequent statistical analyses. Logbook (database 1) During the “Programme Palangre Reunion” (PPR), a rigorous data-collecting strategy was implemented. Logbooks were distributed to all domestic vessels, along with species identification guides, including guides for identifying sharks and sea turtles. Because these logbooks were completed on a voluntary basis, logbook data were cross checked against landing receipts to estimate the monthly logbook coverage. Logbook data made up the major source of information used to ana- lyze the effect of operational and gear-setting practices on catches. Analyses focused on data collected between 1998 and 2000 and which encompassed several lunar cycles. Log- book submission was high by the domestic fleet and the fleet operated between 20° and 23°S lat. and 53° and 57° long. (Fig. 2, area D). Database 1 consisted of 2009 longline sets where seven species were easily identified: albacore, bigeye and yellowfin tuna, dolphinfish, sword- fish, blue ( Prionace glauca ) and oceanic whitetip sharks ( Carcharhinus longimanus), and three broader shark species groups (mako sharks [Isurus spp.], hammer- head sharks [Sphyr?ia spp.], and other sharks grouped together as “mixed sharks”). Experimental fishing (databases 2 and 3) Fish behavior on a daily scale (fine scale) was investi- gated on portions of the longlines equipped with hook timers to estimate fish capture time and with time- depth recorders (TDRs) attached in the middle position between two consecutive floats, which is theoretically the deepest point reached by the mainline. On 160 sets, 284 TDRs were used to estimate hook depth (TDR depth+branchline length). The number of hook timers deployed per set varied from 61 to 408 according to the boat size and weather conditions. For thirty-three trips, 28,974 hook timers were used during 160 sets. These experimental data constituted database 2. The time between the setting of the hook and capture time was represented as a capture index and stratified into four classes (0-4 h, 4-7 h, 7-10 h and >10 h). Database 3 Poisson et al. : Effects of lunar cycle and fishing operations on longline catches 271 40 ° E 45 ° E 50° E 5” S 10 ° S 15 0 S 20 ” S 25” S 30 1 S 5°S 15 ” S 20 ” S 25 ° S 30” S 40 ° E 45 ” E 55 ” E Figure 2 Location of Reunion Island in the southwestern Indian Ocean. The areas shown with diagonal lines are the three fishing grounds of the fleet between 1998 and 2000, the Seychelle waters (area A), the Mozambique Channel (area B), the west and southwest of Reunion Island within the EEZ in association with deep sea- mounts and sea surface temperature (SST) fronts (area C). Logbook data used to analyse the effect of operational and gear setting practices on catches where located in area D (20°-23°S lat. and 53°-57°E long.) represented other longline field trials where CPUE and catch compositions were com- pared by using gear with various densities of chemical lightsticks. Environment (database 4) Database 4 consisted of two environmental databases that were linked to experimen- tal databases to investigate cyclical lunar influence on CPUE. Lunar days were coded in chronological order with values between 1 and 30. A lunar phase index was allo- cated to each fishing day according to the four phases of the moon (new moon, first quarter, full moon, and last quarter). Full and new moons refer to the day of each full or new moon ±2 days. Lunar illumina- tion was calculated for each fish caught, according to its hooking hour, based on the angle of elevation of the moon (on a 24-h cycle). These data were obtained by using an astronomical software package called LunarPhase (http://www.nightskyobserver. com/LunarPhase/index.htm, accessed Janu- ary 2000) and were stratified into three classes: 1) angle <45° (low illumination); 2) high illumination (angle >45°-90°); and 3) dark (new moon). An important underlying assumption inherent in this approach is that the cloud coverage is not considered when calculating the index. An index for tidal phase was assigned to each hooking time and location by the French Service Hydrographique et Oceanographique de la Marine (SHOM). The tidal index consisted of four nominal phases (ebb, high, flood, and low) and is thus a good approximation of the theoretical sea level height changes, although it does not account for current velocity. The fishery operates throughout the year and no temporal discontinuities in fishing effort were evident because effort was quite homoge- neous for indices of lunar illumination and tide phase. Statistical methods Lunar effect on fishing performance was investigated at two different scales. At a larger scale, we conducted a between-class analysis (Doledec and Chessel, 1990, 1994; Gaertner et al., 2005; Bigot et al., 2008) in order to approximate the influence of lunar days (phase) on the CPUE of all the species recorded in logbooks (database 1), where “lunar days” was a categorical factor. For that purpose, we sought axes for the between-group analysis that would best discriminate the centers of gravity of each lunar day and allow us to investigate associations between lunar phases and the variability of CPUE for each species. In addition, a permutation test, which extended the test of Romesburg (1985) to all kinds of variables (Manly, 1991), was carried out to test the significance of the between-group variability. We used nonparametric LOESS regression to further investigate variability in CPUE for species that were most affected by the lunar day. On a finer scale, we carried out a multiple correspondence analysis (MCA) — equivalent to normalized principal component analysis (PCA) to determine the most favourable catch factors on a daily scale (Tenenhaus and Young, 1985; Mazouni et al., 1996). MCA allowed us to visualize the associations between each of the five major species caught during experimental sets and three lunar-related factors stud- ied on a daily scale (tide phase, capture time, lunar illumination). A between-group centered principal component anal- ysis (CPCA, R-mode) and a between-group factorial correspondence analysis (FCA) were conducted to ex- amine the influence of lightstick density on CPUE and catch composition of experimental sets (database 3). The ADE-4 software (Thioulouse et al., 1997, http//pbil. 272 Fishery Bulletin 108(3) Table 1 Variables, indices, and types of analyses performed during our study to elucidate the direct and indirect effects of lunar cycle and other operational factors affecting fishing performance, fish behavior, capture time, and fish survival for the domestic Reunion Island longiine fishery. Database Scale Variable Index Type of analyses 1 Logbook Set (day) Number of fish (7 species, 3 broad groups) number of hooks CPUE (number of fish per 1000 hooks) Between classes analysis, Permutation test, Nonparametric loess regression 2 Experiments with hook timers (HT) Hook (hour) Setting time Capture time Elapsed time after setting as capture time index (0-4 h, 4-7 h, 7-10 h, and >10 h) Multiple correspondence analysis 3 Experiments with lightsticks Set (day) Number of fish (7 species, 3 broad groups) Number of hooks Lightstick density Lightstick density and CPUE Catch composition (percentage of fish per species Centered principal component analysis (CPCA) Factorial correspondence analysis (FCA) 4 Environmental Set ( day) Lunar day Lunar day (1 to 30) and lunar phase index (new moon, first quarter, full moon, last quarter) Hook (hour) Angle of elevation of the moon tide characteristics Lunar illumination index: low illumination intensity (angle <45°), high illumination intensity angle (45°-90°) and dark (new moon) Tide phase index (ebb, high, flood and low) univ-lyonl.fr/ADE-4, accessed January 2000) was used to perform calculations and graphical displays for be- tween-group analyses. Results Effect of the moon at a large scale Time series covered by logbook data The CPCA indi- cated that lunar day represented only 13.3% of the total variability, although the influence of the lunar day on CPUE was highly significant (permutation test, PcO.Ol). The first two axes explained 71% and 17% of the between-lunar-day variability and illustrated the effect of lunar days on CPUE (Fig. 3). Albacore CPUE reached maximum values during the full moon (between the 13th and 19th day of the lunar cycle), whereas CPUE was less important during the new moon (between the 25th and 6th day of the lunar cycle). Swordfish was the unique component of the second axis (Fig. 3) and the projection of lunar days on this axis indicated that the highest CPUE occurred mainly during two short periods of time within the lunar cycle, between the 7th and 9th day and between the 23rd and 26th day. None of the other four species analyzed ap- peared to be affected by the lunar day (Fig. 3). Case study of albacore and swordfish For albacore, LOESS regression confirmed that the highest CPUE was obtained during the full moon (Fig. 4A). In contrast, there was no apparent influence of tidal fluctuation (Fig. 4C). The situation for swordfish was more complex. Variation of CPUE according to lunar cycle provided less contrast than that for albacore. LOESS regression confirmed that the highest CPUE was obtained during the first and last quarter of the moon as characterized by the lowest tidal ranges. Nevertheless, the relationship between swordfish CPUE and lunar intensity was not consistent because the CPUE progressively decreased Poisson et al. : Effects of lunar cycle and fishing operations on longline catches 273 during the last days of the last quarter and reached the lowest values around the full moon phase (Fig. 4B). Moreover, CPUE decreased as tidal amplitude increased. Large confidence intervals around the mean CPUE per- taining to the new moon indicated a higher variability in fishing performance during this phase. The lowest CPUE was recorded between the 14th and 17th day of the lunar cycle during the full moon (which is charac- terized by the largest tidal ranges and maximum lunar illumination). Effect of lunar luminescence, tidal phase, and tidal velocity of local induced currents on catches Sample sizes of five species were considered adequate to conduct detailed analyses of lunar-related variables on a fine scale, although 15 species and 2 broader species groups were identified in the catches (Table 2). Swordfish were caught rapidly after the gear was set (Fig. 5) and during days characterized by weak lunar illumination because most of the individuals were confined in an area corresponding simultaneously to the negative part of both axes (Fig. 6). The occurrence of swordfish on the negative side of the first axis and mainly on the negative axis of the second axis indicates that swordfish were caught during low tides and to a lesser extent during flood tides characterized by a low current. These results reinforce earlier conclusions obtained at a large scale, although the results from the experimental data were weaker than the data from logbooks. As with swordfish, bigeye tuna are likely to be caught rapidly after the gear is set during periods of weak lu- nar illumination (Figs. 5 and 6). Bigeye tuna and blue shark exhibited an opposite distribution on the second axis, which indicated that blue sharks were caught late during the soak time of the longline set and when lumi- nescence was minimal. The limited number of albacore and yellowfin caught in the experimental data may have restricted the statistical analysis. No particular favora- ble or unfavorable conditions were identified for albacore on a daily scale (Fig. 6), in contrast to the influence of lunar intensity observed on a monthly scale. Effect of chemical lightsticks A marginally significant influence of lightstick density on CPUE and catch composition was observed (permuta- tion tests, P<0.05); but the majority of the CPUE vari- ability was not correlated with lightstick density (93.6% for PCA and 91.3% for FCA, Table 3). Capture depths, times, and fish survival TDRs indicated that the deepest depth of the mainline was 110 m; most of the longlines were deployed between 30 and 110 meters (Fig. 7); consequently most hooks fished between 50 and 130 meters. Hook timers indicated that the number of commercial species caught (swordfish and tunas) declined with soak time (i.e., with the time the gear was in the water [Fig. 5A] ) and indicated an Axis 2 A between-group projection of (A) the species, and of (B) the lunar days of the month (represented as numbers between 1 to 30), on the first plane of the between- group centered principal component analysis (CPCA) conducted on catch per unit of effort of albacore (Alb, Thunnus alalunga)\ swordfish (Swo; Xiphias gladius ); bigeye tuna (Bet, Thunnus obesus ); yellowfin tuna (Yft, Thunnus albacares); sharks (Shk), and dolphinfish (Dol, Coryphaena hippurus). Scale box in the upper right corner of each plot indicates the limits for the first and second axes. opposite trend for five bycatch species (Fig. 5B). Blue marlin and dolphinfish were mostly caught after the longline had soaked for 8 hours or longer. In particular, pelagic stingrays, and sailfish to a lesser extent, were caught mostly during gear retrieval. In contrast, more than 60% of the sharks were caught within the initial seven hours of fishing. Fish survival at longline retrieval varied widely among species (Table 4). Survival was high for species such as dolphinfish and black marlin which strike the hooks mainly during retrieval. Over 40% of the main elasmobranch species (blue shark, oceanic whitetip. 274 Fishery Bulletin 108(3) and pelagic stingray), as well as sailfish, were alive upon retrieval. Among commercially important spe- cies, 49% of the bigeye tuna were alive upon retrieval, whereas only 3 of 79 albacore were alive. Swordfish survival was also low (20%) and no relationship was evident between fish size and mortality. Bigeye tuna and two shark species exhibited the highest survival when the longline was at a settled depth. Six species (represented by sample sizes >30 individuals) exhibited long survival times of up to 14 hours after capture (Table 4). Discussion Lunar illumination and sensory abilities of large pelagic fish At a broad scale, our results indicate that the phases of the moon, which presumably affect ambient light levels, have a significant but limited influence on the night catch rates of albacore and swordfish in the Reunion Island-based swordfish longline fishery, whereas such a lunar influence was not found for other species. We found that the highest swordfish CPUE occurred during the first and last quarters of the lunar cycle — a finding that is consistent with the results obtained for the same fishery in the Indian Ocean (Guyomard et ah, 2004). Other studies conducted in various geo- graphical areas with distinctive oceanic features, have indicated that lunar influ- ences on swordfish CPUE are not consist- ent. The highest swordfish CPUE occurred during the full moon phase in the Hawaii- based swordfish fishery (Bigelow et al., 1999) and in the central Atlantic sword- fish fishery (Draganik and Cholyst, 1988). In the U.S. longline fishery in the west- ern Atlantic, more hooks were deployed in the 2-week period around full moon, but Podesta et ah (1993) could not dem- onstrate a significant correlation between CPUE and lunar illumination, whereas Hazin et ah (2002) and Damalas et ah (2007) showed positive effects of other lu- nar phases on swordfish CPUE. No sig- nificant relationship between lunar phase and swordfish abundance could be dem- onstrated in artisanal swordfish fisheries off the Cuban coast (Moreno et ah, 1991). In the case of a swordfish gillnet fishery operating in the Mediterranean Sea, Di Natale and Mangano (1995) showed that the lowest catch rate occurred during the full moon. Lastly, it is apparent that swordfish size and maturity may confound catch rates during different lunar phases (Draganik and Cholyst, 1988; Neves Dos Santos and Garcia, 2005). One possible explanation for the ab- sence of a consistent pattern in associa- tion with lunar phase could be attributed to prey availability. In the Atlantic Ocean, swordfish appear to exhibit feeding plas- ticity (i.e., are opportunistic feeders and exhibit various search strategies for prey) based on forage abundance and prey size; larger swordfish tend to eat larger prey than smaller swordfish (Chancollon et ah, 2006). In the Pacific Ocean, Young et ah (2006) showed no significant relation- Poisson et al.: Effects of lunar cycle and fishing operations on longline catches 275 Table 2 Number and percentage of catch per species during experimental sets conducted from commercial vessels Island-based longline fishery between June 1998 and November 2000. in the domestic Reunion Common name Species Number of individuals Percentage Swordfish Xiphias gladius 389 47.8 Blue shark Prionace glauca 92 11.3 Bigeye tuna Thunnus obesus 86 10.6 Albacore Thunnus alalunga 79 9.7 Yellowfin tuna Thunnus albacares 66 8.1 Common dolphinfish Coryphaena hippurus 48 5.9 Oceanic whitetip shark Carcharhinus longimanus 17 2.1 Pelagic stingray Pteroplatytrygon violacea 12 1.5 Indo-Pacific sailfish Istiophorus platypterus 7 0.9 Hammerhead sharks Sphyrna spp. 4 0.5 Indo-Pacific black marlin Makaira indica 3 0.4 Escolar Lepidocybium flavobrunneum 3 0.4 Indo-Pacific blue marlin Makaira mazara 2 0.2 Wahoo Acanthocybium solandri 2 0.2 Shortfin mako shark Isurus oxyrinchus 2 0.2 Barracuda Sphyraena spp. 1 0.1 Leatherback turtle Dermoclielys coriacea 1 0.1 Total hook timers triggered 2115 Hook timers triggered with catch 814 Hook timers triggered without catch 1301 Table 3 Decomposition of inertia between and within groups according to the lightstick den- sity factor for centered principal component analysis (CPCA), factorial correspon- dence analysis (FCA), and associated permutation tests. Between lightstick density CPCA Between lightstick density AFC Total inertia 3.234 102 1.597 Between groups 2.046 101 (6.32%) 1.391 10-1 (8.7%) Within groups 3.030 102 (93.68%) 1.458(91.3%) Permutation test P<0.05 P<0.05 ship between prey biomass and lunar phase but found that prey size significantly increased with swordfish size and that this in- crease coincided with a dietary shift from fish to cephalopods (Palko et al., 1981). This finding indicates that swordfish have the ability to forage at considerable depth and temperature, which are afforded by a suite of physi- ological adaptations to enable opportunistic feeding within the deep sound-scattering layer (DSL) (Josse et al., 1999; Musyl et al., 2003), Gilly et al. (2006) depicted lunar influence on the vertical migration of squid, which, in turn, would have a direct effect on the distribution and vulnerability of swordfish. We found that higher swordfish CPUE correlated with small tides. Moreover, at a finer scale, we found that higher swordfish CPUE occurred with lower tidal fluctuations which coincided with the possible genera- tion of low-velocity oceanic currents. These results were consistent with the results obtained when applying GAMs on similar scales (Guyomard et al., 2004), where the meridional component (V) of geostrophic currents derived from sea level anomaly (SLA) data was the most significant environmental factor within one of the models tested. It was likely that the low positive V val- ues (i.e., low velocity currents to the north) were more beneficial for catch rates, whereas the trend became clearly negative for higher velocity values. We assumed that the influence of the tide and associ- ated local current velocities would be complex and affect the dispersion of organisms within the DSL and other associated organisms of the mixed layer. In addition, current could affect turbidity by advecting organisms and particulate matter and alter the shape and depth obtained by the longline (Bigelow et al., 2006). To gain insights into swordfish behavior during complex pat- terns of current velocities, acoustic telemetry could be used to provide short-term horizontal movement data that could help in elucidating the effect of the tide on the foraging behavior of large pelagic fish. Analysis 276 Fishery Bulletin 108(3) 10 h; “moon luminescence index” 3a = low, 3b=high, and 3c = dark tracted by offal and spent bait discarded during the haul but also by the distress signals of captured fish (Myrberg et al., 1969) Our results indicate that swordfish and bigeye tuna exhibit active predation when lunar illumination is weak. These results are consistent with the findings of Fristches et al. (2005), who showed that retinas of swordfish and bigeye tuna provide visual acuity and sensitivity to blue-green light and thus these apex pred- ators are efficient visual hunters in dim light. We found that 60% to 80% of swordfish were caught after the gear soaked 4-6 hours and these results indicated a possible positive effect of bait freshness on the attractiveness of bait to swordfish and bigeye tuna. The effect of the chemical lightsticks on catch rate also decreased with time and as their glowing intensity waned. The role of the lightsticks is still not clear, al- though it is thought that they either attract predators to the bait or they attract small fish and squid (Hazin et al., 2005), or both, and as such, were considered as an important refinement of the longline gear. A signifi- cant but low association of lightstick density on CPUE was confirmed by our analyses. Bigelow et al. (1999) found that increasing the proportion of lightsticks from 278 Fishery Bulletin 108(3) Table 4 Number of fish caught per species (No.), number of fish alive at time of hauling (no.), range of size per species (lower jaw fork length for billfish and swordfish, and fork length for other species in cm), maximum survival time per species, percentage of individuals alive at time of hauling, and percentage alive eight hours after hooking. Common name Species No. no. Length (cm) Maximum survival time (hours) Alive at time of hauling (%) Alive after 8 hours (%) Swordfish Xiphias gladius 389 76 93-242 14 19.5 8.4 Bigeye tuna Thunnus obesus 86 42 65-160 14 48.8 26.7 Albacore Thunnus alalunga 79 3 105-113 8 3.8 1.2 Yellowfin tuna Thunnus albacares 66 23 99-150 14 34.8 13.6 Common dolphinfish Coryphaena hippurus 48 32 83-120 14 66.7 8.3 Oceanic whitetip shark Carcharhinus longimanus 17 7 120-151 14 41.1 23.5 Blue shark Prionace glauca 92 45 150-240 14 48.9 29.3 Indo-Pacific sailfish Istiophorus platypterus 7 3 ?-163 4 42.8 — Indo-Pacific black marlin Makaira indica 3 3 238-240 2 100 — Indo-Pacific blue marlin Makaira mazara 2 0 — — 0 — Pelagic stingray Pteroplatytrygon violacea 12 5 — 2 41.2 — 0% to 40% on the hooks increased CPUE, but that the effect was not increased beyond this threshold. In contrast, blue shark had an increase in CPUE when the proportion of lightsticks was increased. Beyond an interest in understanding the influence of lunar perio- dicity and operational factors on the behavior of large pelagic fishes, these potential mechanisms are also of interest to fishermen, and our results also have impli- cations for current and future fishing practices. Frequency (%) 0 5 10 15 20 25 Figure 7 Pooled data from the time-depth recorders located on the mainline in the middle position between two consecutive floats during experimental sets (n = 284) in the domestic Reunion Island longline fishery. Adjustments of fishing strategy and future research needs This study may help fishermen to modify fishing opera- tions and select a fishing strategy to increase economic benefits and reduce the impact of bycatch mortality. Cap- tains would be able to switch fishing practices to target one species or another according to lunar phase. More- over, shifting from expensive squid to cheaper mackerel bait to catch albacore could reduce the operational costs during trips in October and November when albacore CPUE is seasonally highest. We suggest that shortening the soak time dur- ing the fishing operation should be beneficial be- cause the major portion of the catch occurred in the first few hours of the operation (presumably bait decreases in quality with time and there is an increase in bait loss) (Lokkeborg and Pina, 1997). Hook-timers and direct observations indi- cated that swordfish could escape for several hours after capture. Many triggered hook-timers were retrieved without a fish (1301 cases) and they may have been triggered for a variety of reasons, such as being triggered during deployment or re- trieval operations, being activated by squids, or may have been triggered by the escape of fishes or turtles. Although we cannot assume that high es- cape rates have occurred, we had the opportunity to film a swordfish escaping during the hauling of the gear. It succeeded in unhooking itself at the surface but sank immediately from exhaustion. Therefore, even if the fate of the escaped animal was unknown, we assumed that a large propor- tion of swordfish might die from ingestion of the J hooks. In addition, long soaking periods in warm temperatures increase the degradation of flesh. Poisson et al.: Effects of lunar cycle and fishing operations on longline catches 279 reducing the market price of the fish. Shortening the soaking time could reduce the chance of depredation by large marine mammals (e.g., false killer whale [ Pseu - dorca crassidens ], shortfin pilot whale [ Globicephala macrorhyncus ]) and sharks on longline-caught fish. In the United States, limits on the length of a pelagic longline set have been proposed as a management meas- ure to reduce bycatch (Kerstetter, 2008). It has been demonstrated that such a restriction would reduce the interaction rate of longlines with marine mammals in the Mid-Atlantic Bight by approximately 26%. 1 The survival rates for blue shark and oceanic whitetip shark were estimated to be 49% and 41%, respectively. The survival rate of blue shark at haulback after a soak during the night was lower than that during day longline sets: 100% (Boggs, 1992), 80-90% (Campana et al., 2005), 69% (Diaz and Serafy, 2005), and 87% (Fran- cis et al., 2001). Differences in survival rates among studies may result from hook types, leader material (monofilament or wire), and handling procedures, al- though survival rates between day and night longlining should be further investigated. Nevertheless, reducing the soaking period would increase the number of sharks released alive. The release of live bycaught sharks (Moyes et al., 2006) and billfish (Kerstetter and Graves, 2006, 2008) is by far the best management measure to reduce longline fishing mortality of these species. From a cost benefit perspective, fishermen believe that chemical lightsticks improve fishing performance but they limit the number deployed because of the price. However, our data strongly indicate that the use of lightsticks did not increase swordfish catch by very much. Lightsticks are suspected to attract sea turtles to the vicinity of longlines (Wang et al., 2007) and thus may increase their incidental catch; however, light- sticks have a limited lifespan and are not reusable and thus are an environmental concern. Thousands of spent lightsticks are discarded at sea and constitute a potential toxicant to marine flora and fauna. In the case of Reunion Island, local fishermen were keen to retain used chemical lightsticks onboard, store them during the fishing trip, and offload them when returning to port after a significant awareness campaign about the negative environmental impact of lightsticks. In light of our results and for ecological concerns (Ivar do Sul et al., 2009; Pinho et al., 2009), the use of chemical lightsticks should be reconsidered. Recently, the work- ing group of the General Fisheries Commission for the Mediterranean (GFCM) has proposed a ban on chemical lightsticks and any light source in the pelagic longline fishery in the Mediterranean Sea.2 1 APLTRT (Atlantic Pelagic Longline Take Reduction Team). 2006. Atlantic pelagic longline take reduction plan, 97 p. Submitted to the National Marine Fisheries Service Southeast Regional Office, St. Petersburg, Florida. 2 Report of the transversal working group on bycatch/incidental catches; Rome, Italy, 15-16 September 2008, 17 p. General Fisheries Commmission for the Mediterranean (GFCM), (http://www.gfcm.org/gfcm). Complementary three-dimensional acoustic telem- etry experiments are needed to better understand the movements of swordfish and to test the hypothesis that tidal and oceanic currents may influence their foraging behavior and fishing operations associated with their foraging behavior. We also recommend additional stud- ies to understand the interaction of lunar luminescence and swordfish size because of concerns for the sustain- ability of swordfish stocks and for the protection of certain age classes (Poisson and Fauvel, 2009). Acknowledgments Funding for the PPR programme was supported by the European Union (FEDER), the Conseils Regional, and General de La Reunion. We express our gratitude to the fishing industry of Reunion Island for their outstand- ing support. We are very grateful to J. F. Reynaud, C. Marjolet, D. Guyomard, and M. Vanpouille for their work completed within the framework of the project. We also acknowledge R. Galzin (University of Perpignan, France) for his support. We thank M. Musyl for provid- ing helpful advice. We thank the editor and the three anonymous reviewers who improved the manuscript with insightful suggestions and P. Lopez for his input on the improvement of the illustrations. Literature cited Bigelow, K. A., C. H. Boggs, and X. He. 1999. Environmental effects on swordfish and blue shark catch rates in the US North Pacific longline fishery. Fish. Oceanogr. 8:178-198. Bigelow, K., M. K. Musyl, F. Poisson, and P Kleiber. 2006. Pelagic longline gear depth and shoaling. Fish. Res. 77:173-183. Bigot, L., A. Gremare, J. M. Amouroux, P. Frouin, O. Maire, and J.-C. Gaertner. 2008. Assessment of the ecological quality status of soft- bottoms in Reunion Island (tropical Southwest Indian Ocean) using AZTI Marine Biotic Indices. Mar. Pollut. Bull. 56:704-722. Boggs, C. H. 1992. Depth, capture time and hooked longevityof longline-caught pelagic fish — timing bites of fish with chips. Fish. Bull. 90:642-658. Bonfil, R. 1994. Overview of world elasmobranch fisheries, 119 p. FAO Fish. Tech. Paper 341. FAO., Rome. Brill, R. W„ D. B. Holts, R. K. C. Chang, S. Sullivan, H. Dewar, and F. G. Carey. 1993. Vertical and horizontal movements of striped marlin ( Tetrapturus audax) near the Hawaiian-Islands, determined by ultrasonic telemetry, with simultane- ous measurement of oceanic currents. Mar. Biol. 117: 567-574. Campana, S. E., L. Marks, W. Joyce, and N. E. Kohler. 2005. Catch, by-catch, and indices of population status of blue shark (Prionace glauca) in the Canadian Atlantic. Col. Vol. Sci. Pap. ICCAT 58(3):891-934. 280 Fishery Bulletin 108(3) Canese, S., F. Garibaldi, L. Orsi Relini, and S. Greco. 2008. Swordfish tagging with pop-up satellite tags in the Mediterranean Sea. Col.Vol. Sci. Pap. ICCAT 62(4):1052-1057. Carey, F. G. 1990. Further acoustic telemetry observations of swordfish. In Proceedings of the second international billfish symposium, part 2 (R. H. Stroud, ed.); Kailua- Kona, Hawaii 1-5 August 1988, p. 103-122. Natl. Coalition for Mar. Conserv., Savannah, GA. Carey, F. G., and B. H. Robinson. 1981. Daily patterns in the activities of swordfish, Xiphias gladius, observed by acoustic telemetry. Fish. Bull. 79:277-292. Chancollon, O., C. Pusineri, and V. Ridoux. 2006. Food and feeding ecology of Northeast Atlantic swordfish ( Xiphias gladius) off the Bay of Biscay. ICES J. Mar. Sci. 63:1075-1085. Damalas, D., P. Megalofonou, and M. Apostolopoulou. 2007. Environmental, spatial, temporal and operational effects on swordfish ( Xiphias gladius) catch rates of eastern Mediterranean Sea longline fisheries. Fish. Res. 84:233-246. Di Natale, A., and A. Mangano. 1995. Moon phases influence on CPUE: A first analysis of swordfish driftnet catch data from the Italian fleet between 1990 and 1991. Col. Vol. Sci. Pap. ICCAT 44(l):264-267. Diaz, G. A., and J. E. Serafy. 2005. Longline-caught blue shark ( Prionace glauca ): factors affecting the numbers available for live release. Fish. Bull. 103:720-724. Doledec, S., and D. Chessel. 1990. Rythmes saisonniers et composantes stationnelles en milieu aquatique. II. Prise en compte et elimina- tion d'effets dans un tableau faunistique. Acta Oecol. 10:207-232. [In French.] 1994. Co-inertia analysis: an alternative method for studying species-environment relationships. Freshw. Biol. 31:277-294. Draganik, B., and J. Cholyst. 1988. Temperature and moonlight as stimulators for feeding activity by swordfish. Col. Vol. Sci. Pap. ICCAT 27(1):305-314. FAO (Food and Agriculture Organization of the United Nations). 2003. The ecosystem approach to fisheries. Technical Guidelines for Responsible Fisheries 4:1-112. Francis, M. P., L. H. Griggs and S. J. Baird. 2001. Pelagic shark bycatch in the New Zealand tuna longline fishery. Mar. Freshw. Res. 52:165-178. Fritsches, K. A., R. W. Brill, and E. J. Warrant. 2005. Warm eyes provide superior vision in sword- fishes. Curr. Biol. 15:55-58. Gaertner, J. -C., J. A. Bertrand, L. G. de Sola, J. P. Durbec, E. Ferrandis, and A. Souplet. 2005. Large spatial scale variation of demersal fish assemblage structure on the continental shelf of the NW Mediterranean Sea. Mar. Ecol. Prog. Ser. 297:245-257. Gilly, W. F., U. Markaida, C. H. Baxter, B. A. Block, A. Boustany, L. Zeidberg, K. Reisenbichler, B. Robison, G. Bazzino, and C. Salinas. 2006. Vertical and horizontal migrations by the jumbo squid Dosidicus gigas revealed by electronic tagging. Mar. Ecol. Prog. Ser. 324:1-17. Gilman, E., S. Clarke, N. Brothers, J. Alfaro-Shigueto, J. Man- delman, J. Mangel, S. Petersen, S. Piovano, N. Thomson, P. Dalzell, M. Donoso, M. Goren and T. Werner. 2008. Shark interactions in pelagic longline fisheries. Mar. Pollut. 32:1-18. Guyomard, D., M. Desruisseaux, F. Poisson, M. Taquet, and M. Petit. 2004. GAM analysis of operational and environmental factors affecting swordfish (Xiphias gladius) catch and CPUE of the Reunion Island longline fishery, in the south western Indian Ocean. Report IOTC-2004-WPB-08 of the fourth session of the IOTC working party on billfish. Mauritius; 27 September-1 October 2004, 38 p. Indian Ocean Tuna Commission, Victoria, Seychelles. Hazin, H. G., F. H. V. Hazin, P. Travassos, and K. Erzini. 2005. Effect of light-sticks and electralunie attrac- tors on surface-longline catches of swordfish (Xiphias gladius, Linnaeus, 1959) in the southwest equatorial Atlantic. Fish. Res. 72:271-277. Hazin, H. G., F. H. V. Hazin, P. Travassos, S. Hamilton, and F. P. Ribeiro. 2002. Influence of the phases of the moon on the relative abundance of swordfish (Xiphias gladius, Linnaeus, 1758) caught in the equatorial Atlantic Ocean. Col. Vol. Sci. Pap. ICCAT 54(5):1586-1589. Ivar do Sul, J. A., O. Rodrigues, I. R. Santos, G. Fillmann, and A. Matthiensen. 2009. Skin irritation and histopathologic alterations in rats exposed to lightstick contents, UV radiation and seawater. Ecotoxicol. Environ. Safety 72:2020-2024. Josse, E., A. Bertrand, and L. Dagorn. 1999. An acoustic approach to study tuna aggregated around fish aggregating devices in French Polyne- sia: methods and validation. Aquat. Living Resour. 12:303-313. Kerstetter, D. W. 2008. Measuring the length of a pelagic longline set: Applications for management. N. Am. J. Fish. Manag.28:378-385. Kerstetter, D. W., and J. E. Graves. 2006. Survival of white marlin ( Tetrapturus albidus) released from commercial pelagic longline gear in the western North Atlantic. Fish. Bull. 104:434-444. 2008. Postrelease survival of sailfish caught by com- mercial pelagic longline gear in the southern Gulf of Mexico. N. Am. J. Fish. Manag. 28:1578-1586. Kume, S., and J. Joseph. 1969. Size composition and sexual maturity of bill- fish caught by the Japanese longline fishery in the Pacific Ocean east of 130°W. Far Seas Fish. Res. Lab. 2:115-162. Lee, H. -J., Y. -J. Jong, L. M. Chang, and W. -L. Wu. 2009. Propulsion strategy analysis of high-speed swordfish. Trans. Jpn. Soc. Aeronaut. Space Sci. 52:11-20. Lokkeborg, S., and T. Pina. 1997. Effects of setting time, setting direction and soak time on longline catch rates. Fish. Res. 32:213-222. Manly, B. F. J. 1991. Randomization and Monte Carlo methods in biol- ogy, 281 p. Chapman and Hall, London. Mazouni, N., J. -C. Gaertner, J. -M. Deslous-Paoli, S. Landrein, and M. Geringer d’Oedenberg 1996. Nutrient and oxygen exchanges at the water-sedi- ment interface in a shellfish farming lagoon (Thau, France). J. Exp. Mar. Biol. Ecol. 205:91-113. Poisson et al.: Effects of lunar cycle and fishing operations on longline catches 281 Moreno, S., J. Pol, and L. Munoz 1991. Influence of the moon on the abundance of swordfish. Col. Vol. Sci. Pap. ICCAT 35(21:508-510. Moyes, C. D., N. Fragoso, M. K. Musyl, and R. W. Brill. 2006. Predicting postrelease survival in large pelagic fish. Trans. Am. Fish. Soc. 135:1389-1397. Musyl, M. K., R. W. Brill, C. H. Boggs, D. S. Curran, T. K. Kazama, and M. P. Seki. 2003. Vertical movements of bigeye tuna ( Thunnus obesus) associated with islands, buoys, and seamounts near the main Hawaiian Islands from archival tagging data. Fish. Oceanogr. 12:152-169. Myrberg, A. A., A. Banner, and J. D. Richard. 1969. Shark attraction using a video-acoustic system. Mar. Biol. 2:264-276. Nakamura, I. 1983. Systematics of the billfishes (Xiphiidae and Istiophoridae). Publ. Seto Mar. Biol. Lab. 28:255- 396. Neilson, J. D., S. Smith, F. Royer, S. D. Paul, J. M. Porter, and M. Lutcavage. 2009. Investigations of horizontal movements of Atlantic swordfish using pop-up satellite archival tags. In Tag- ging and tracking of marine animals with electronic devices (J. L. Nielsen, H. Arrizabalaga, N. Fragoso, A. Hobday, M. Lutcavage, and J. Sibert, eds.) p. 145- 159. Springer, New York. Neves Dos Santos, M., and A. Garcia. 2005. The influence of the moon phase on the CPUEs for the Portuguese swordfish (Xiphias gladius L., 1758) fishery. Col. Vol. Sci. Pap. ICCAT 58(41:1466-1469. Ortega-Garcia, S., G. Ponce-Diaz, R. O’Hara, and J. Merila. 2008. The relative importance of lunar phase and environ- mental conditions on striped marlin ( Tetrapturus audax) catches in sport fishing. Fish. Res. 93:190-194. Palko, B. J., G. L. Beardsley, and W. J. Richards. 1981. Synopsis of the biology of the swordfish, Xiphias gla- dius Linnaeus. NOAA Tech. Rep. NMFS Circ. 441, 21 p. Pinho, G. L. L., P. M. Ihara, and G. Fillmann. 2009. Does light-stick content pose a threat to marine organisms? Environ. Toxicol. Pharmacol. 27:155- 157. Podesta, G. P, J. A. Browder, and J. J. Hoey. 1993. Exploring the association between swordfish catch rates and thermal fronts on united States longline grounds in the western north Atlantic. Cont. Shelf Res. 13:253-277. Poisson, F., and C. Fauvel. 2009. Reproductive dynamics of swordfish ( Xiphias gladius 1 in the southwestern Indian Ocean (Reunion Island). Part 2: fecundity and spawning pattern. Aquat. Living Resour. 22:59-68. Pusineri, C., O. Chancollon, J. Ringelstein, and V. Ridoux. 2008. Feeding niche segregation among the Northeast Atlantic community of oceanic top predators. Mar. Ecol. Prog. Ser. 361:21-34. Romesburg, H. 2008. Scaling laws of marine predator search behaviour. Nature 451:1098-U1095. Takahashi, M., H. Okamura, K. Yokawa, and M. Okazaki. 2003. Swimming behaviour and migration of a sword- fish recorded by an archival tag. Mar. Freshw. Res. 54:527-534. Tenenhaus, M., and F. W. Young. 1985. An analysis and synthesis of multiple corre- spondence-analysis, optimal-scaling, dual scaling, homogeneity analysis and other methods for quanti- fying categorical multivariate data. Psychometrika 50:91-119. Thioulouse, J., D. Chessel, S. Doledec, and J. M. Olivier. 1997. ADE-4: A multivariate analysis and graphical display software. Stat. Comput. 7:75-83. Tserpes, G., P. Peristeraki, and V. D. Valavanis. 2008. Distribution of swordfish in the eastern Mediter- ranean, in relation to environmental factors and the species biology. Hydrobiologia 612:241-250. Walsh, W. A., and P. Kleiber. 2001. Generalized additive model and regression tree analyses of blue shark ( Prionace glauca ) catch rates by the Hawaii-based commercial longline fishery. Fish. Res. 53:115-131. Wang, J. H., L. C. Boles, B. Higgins and K. J. Lohmann. 2007. Behavioral responses of sea turtles to light- sticks used in longline fisheries. Anim. Conservat. 10:176-182. Young, J., M. Lansdell, S. Riddoch, and A. Revill. 2006. Feeding ecology of broadbill swordfish, Xiphias gladius, off eastern Australia in relation to physi- cal and environmental variables. Bull. Mar. Sci. 79:793-809. 282 Using experiments and expert judgment to model catchability of Pacific rockfishes in trawl surveys, with application to bocaccio ( Sebastes paucispinis ) off British Columbia Email address for contact author: m.mcallister@fisheries.ubc.ca 1 Fisheries Centre, Aquatic Ecosystems Research Laboratory (AERL) 2202 Main Mall The University of British Columbia Vancouver, British Columbia V6T 1Z4, Canada 2 Marine Ecosystem and Aquaculture Division Science Branch, Fisheries and Oceans Canada Pacific Biological Station Nanaimo, British Columbia V9T 6N7, Canada 3 Canadian Groundfish Research and Conservation Society 1406 Rose Ann Drive Nanaimo, British Columbia V9T 4K8, Canada Abstract — The time series of abun- dance indices for many groundfish populations, as determined from trawl surveys, are often imprecise and short, causing stock assessment esti- mates of abundance to be imprecise. To improve precision, prior probability distributions (priors) have been devel- oped for parameters in stock assess- ment models by using meta-analysis, expert judgment on catchability, and empirically based modeling. This arti- cle presents a synthetic approach for formulating priors for rockfish trawl survey catchability ( qgross ). A multi- variate prior for qgross for different surveys is formulated by using 1) a correction factor for bias in estimating fish density between trawlable and untrawlable areas, 2) expert judgment on trawl net catchability, 3) observa- tions from trawl survey experiments, and 4) data on the fraction of popu- lation biomass in each of the areas surveyed. The method is illustrated by using bocaccio ( Sebastes paucipinis) in British Columbia. Results indicate that expert judgment can be updated markedly by observing the catch-rate ratio from different trawl gears in the same areas. The marginal priors for qaross are consistent with empiri- cal estimates obtained by fitting a stock assessment model to the survey data under a noninformative prior for q cross- Despite high prior uncertainty (prior coefficients of variation >0.8) and high prior correlation between q gross’ the prior for qgross still enhances the precision of key stock assessment quantities. Manuscript submitted 8 May 2009. Manuscript accepted 5 April 2010. Fish. Bull. 108:282-304(2010). The views and opinions expressed or implied in this article are those of the author (or authors) and do not necessarily reflect the position of the National Marine Fisheries Service, NOAA. Murdoch K. McAllister (contact author)1 Richard D. Stanley2 Paul Starr3 Rockfishes (Sebastes spp.) are a group of groundfish species on the west coast of North America, many of which are commonly exploited; over 25 stocks are assessed and individually managed in the United States and Canada (DFO, 2008; NPFMC, 2008; PFMC, 2008). Stock assessment models for rock- fish are typically fitted to a variety of data, such as estimates of popula- tion biomass determined from survey trawl-swept areas. These swept-area biomass estimates are usually treated as relative indices of abundance because of the unknown relationship between the availability of the target population to the survey net. Factors affecting this relationship include the proportion of fish present within the path of the net that on average enter the net, the proportion of the popula- tion that is potentially available to be captured by the survey gear and the relative density of rockfish in traw- lable and untrawlable areas. Treated as a relative abundance index, a single scalar parameter is typically estimated, called “bulk catchability” or “q gross’ scale the model-predicted population biomass to the swept-area biomass values (Millar and Methot, 2002). Owing to the trawl survey data being available for only a portion of the history of a stock’s exploitation and because of moderate to large amounts of variation in interannual error, stock assessment estimates of qgross are often imprecise and may not provide reliable estimates of popu- lation biomass (Millar and Methot, 2002). In order to reduce the large uncer- tainty common to estimates of qgross and population abundance for rock- fishes and many other assessed fish- es, stock assessment scientists have quantified Bayesian prior probabil- ity density functions (pdfs) for qgross- Among these quantifications, there have been efforts to quantify expert judgment (e.g., Punt et al., 1993; McAllister and Ianelli, 1997; Boyer et ah, 2001) on factors affecting survey catchability. Others have performed hierarchical analyses of stock assess- ments for different rockfish species to quantify the mean and variance in q gross across different populations for surveys, using the same gear (Millar and Methot, 2002). Although these McAllister et al.: Using experiments and expert judgment to model catchability of Pacific rockfishes 283 analyses have improved the conceptual understand- ing of the processes contributing to trawl catchability during surveys and have shown how expert judgment and other information can be used to help form priors, no unifying framework exists that integrates the judg- ment from multiple experts with other available data on survey catchability. In this article, we present a synthetic approach for integrating inputs on technical parameters elicited from experts and survey data not used in a stock assessment to form a prior for qgross that can be used to improve the value of biomass estimates from trawl-swept areas for use in stock assessments. This approach can be applied to swept-area abundance estimates across a number of areas and over multiple types of trawl gear (e.g., shrimp and groundfish trawl nets). Each expert is assigned equal prior weight which is then updated from ratios of relative catch rates between different survey gear types. We use estimates of the fraction of the total population biomass that lies within the boundaries of each sur- vey area and that accounts for the fraction of the area within the boundary of each survey that is trawlable. A factor is applied to prevent the expert inputs from being overly certain. Finally, because this method is applied to different surveys of the same stock, these parameters are not independent (in some surveys the same gears were used) and there is spatial covariance in estimates of the fraction of biomass in the different survey areas (the prior pdf formed accounts for the correlations in the qgross parameters between surveys). We illustrate the method with an application to bocaccio ( Sebastes paucispinis) off British Columbia (B.C.) that relied upon technical information obtained from interviews with a dozen trawl captains. We show the sensitivity of the results to assumptions about potential differences in rockfish density between trawlable and untrawlable sub- strates, the amount of uncertainty in expert inputs, and how results from different experts should be integrated. The impact on the overall stock assessment results are illustrated by comparing results obtained with and without informative priors for qaross- Bocaccio in British Columbia were chosen to illus- trate the new method to formulate a prior for qgross because this species presents an instance in which the time series of abundance indexes available for stock assessment are mostly too short or imprecise to en- able estimation of parameters of population dynamics and abundance trends. An informative prior for survey q is essential to achieve these ends. Bocaccio range from the Alaska Peninsula to Baja California (Love et al., 2002). In British Columbia, adult bocaccio exhibit a widespread distribution mainly on the outer coast (Fig. 1). Most catches are taken close to the bottom over depths of 60-200 m near the break-in-slope of the continental shelf, as well as at the edges of troughs in Queen Charlotte Sound (QCS) and Hecate Strait (HS). Adult bocaccio can be semipelagic and are found over a variety of bottom types, although harvesters suggest they favour proximity to high relief and rocky bottom. In British Columbia, bocaccio are caught by trawl and hook-and-line gear along with many other groundfish species, including Pacific ocean perch (S. cilutus), yel- lowtail rockfish (S. flavidus), canary rockfish (S. pin- niger ), and lingcod ( Ophiodon elongatus). Indices from seven trawl surveys (Fig. 1) were used in our study. Four of the surveys, 1) the west coast of Vancouver Island groundfish (WCVI Gfish), 2) Queen Charlotte Sound groundfish (QCS Gfish), 3) Hecate Strait groundfish (HS Gfish), and 4) west coast of Haida Gwaii groundfish (WCHG Gfish) represent a set of nonoverlapping bottom trawl surveys that were started between 2003 and 2006 to collectively survey most of the B.C. coastal shelf between 50 and 500 m of bottom depth. The focus of these surveys was to provide relative indices of all groundfish species af- fected by the groundfish bottom trawl fishery in B.C. waters. For all four surveys, the Atlantic Western II groundfish bottom trawl was used and the surveys were conducted by the Canada Department of Fisher- ies and Oceans (DFO) staff on either the government research trawler (surveys 1 and 2) or chartered trawler (surveys 3 and 4). Two of the surveys, the WCVI shrimp (survey 5) and QCS shrimp (survey 6) are conducted by DFO staff on board the same DFO research trawler (Boutillier et al., 1998). These surveys use a shrimp trawl and were de- signed to provide relative indices of shrimp abundance on two specific shrimp fishing grounds. For the seventh survey, the U.S. triennial survey a Nor’Eastern ground- fish bottom trawl was used. This survey was designed to monitor groundfish abundance in U.S. waters, but in some years covered a small portion of southern B.C. waters. This survey stopped covering Canadian waters after 2001. Methods General model structure for trawl survey catchability ( qgr0S5 ) See Table 1 for descriptions of all symbols used in this paper and Figure 2 for a schematic outline of the inputs, sub-models and outputs of the q prior model. We define catchability ( qgross ) as the ratio of biomass of rockfish in a particular survey area to the population biomass of a given rockfish population that is on average vulnerable to trawl survey gear on account of gear selectivity (i.e., the fully vulnerable population biomass). qgross is typi- cally considered to be the long-term average value and is applied as a scalar to the fully vulnerable popula- tion biomass (B ) modeled in a stock assessment model to predict the index of biomass obtained from a given trawl survey. The predicted swept-area biomass (/) is obtained from the product of By and qaross- iy= Q gross X By. For rockfish, it has been generally acknowledged that there are three main factors that may cause the value 284 Fishery Bulletin 108(3) 136°W 134“W 132"W 130°W 128°W 126"W 124°W 122“W I36°W 134°W 132°W 13Q°W ! 128°W i 126°W i 124°W 122°W i !_ B Surveys ' Haida Gwaii Groundfish j?H^| Hecate Strait Groundfish Queen Charlotte Sound Groundfish Queen Charlotte Sound Shrimp i=E==i West Coast Vancouver Island Groundfish HHm West Coast Vancouver Island Shrimp lll|||]iii U S. Triennial 3 British Columbia Vancouver Island 200 100 0 200 Kilometers Can. USA^ -54°N -53°N -52°N ■51 °N ■50“N ■49’N -48°N Figure 1 (A) Locations where bocaccio ( Sebastes paucispinis ) were caught in commercial and research trawls, 2004-08. The 200-m depth contour is shown by a thin gray line. (B) Locations of trawl surveys in outside waters of British Columbia. for q„ross to deviate from unity and may be conceived to act multiplicatively (Millar and Methot, 2002): Qgross ~~ Qnet x Qtrawlable X ^-available ’ where qnet = the fraction of exploitable biomass that is within the path (i.e., between the trawl doors) of a given type of survey net and that is on average captured by the net; McAllister et al.: Using experiments and expert judgment to model catchability of Pacific rockfishes 285 Table 1 Definition of symbols for the indices, model parameters, and model variables used to determine trawl catchability for bocaccio ( Sebastes paucispinis) off British Columbia. Symbol Description Indices pop entire fish population of interest that is potentially susceptible to capture by the survey gear of interest n trawl net type, e.g., in study l=the groundfish AWII used in the Department of Fisheries and Ocean’s groundfish trawl surveys, in study 2=the shrimp trawl, in study 3=the Nor’ Eastern used in the U.S. triennial groundfish trawl survey c interviewed captain s survey Model parameters qsross the parameter that scales the model-predicted population biomass to a trawl survey swept area biomass value <4> where fTs = the fraction of sea bottom in a surveyed area that is trawlable in survey area s; and a = the ratio of target species density (t/km2) in untrawlable to density in trawlable habitat. Although a may vary with survey area, we typically do not have data that would allow us to estimate these separate factors and a therefore was assumed not to depend on s. In contrast, it is common to have data on the fraction of trawlable area within each survey area and therefore one can thus compute factor g for each survey area. Should some experiment provide data on a, then the prior for a, P(a), could be statistically updated with the following equation: P(a\data)xP(a)P(data\a), (5) where P(a\data) = the posterior for a, given the data; and P(data\a) = the probability of the data, given a. Note that no such data were available for our case study application. fTs can be treated as a beta random variable with binomial data on the number of trawlable sites for each area, or it can be fixed, if the number of trials in each area is very large. q availability-^* fraction of total exploitable species abundance in each surveyed region We developed a protocol to approximate the percentage of the coastwide target species exploitable biomass that is available to each of the surveys. This protocol computes and treats as a random variable the fraction of the total coastwide swept-area biomass in each survey area and accounts for potential differences between survey areas in the fraction of trawlable ground in each survey area. For this protocol, tow-by-tow data from the most exten- sive trawl survey (i.e., the “reference survey”) were used. The protocol is illustrated with bocaccio as an example and includes the following assumptions: 1 The relative distribution of stock biomass among areas has been constant over a reference set of years and two or more groundfish trawl swept-area bio- mass estimates are available for all survey areas. For bocaccio this reference set covers the years from 2003 to 2007. 2 The proportion of untrawlable area to trawlable area within a surveyed region varies among regions and can be approximated from the observed frequencies of trawlable and untrawlable sites in the cumula- tive set of randomly allocated survey locations in each survey area. In fact, the locations found to be untrawlable are removed from the set of locations to be considered for research trawling in future years. However, in recent calculations we have found that the impact of this sequential removal of untrawlable locations on the estimated fraction of trawlable area is very small because the fraction of untrawlable areas is overall relatively small (less than about 20%). 3 The ratio of target species density in trawlable and untrawlable areas (a) is the same across areas and time. 4 The habitat for the target species is assumed to be the seabed area between a fixed depth range, e.g., 100-300 m in bottom depth for bocaccio. Building on the above assumptions, we assume that the proportion of the coastwide target stock biomass available to each survey is the ratio of the swept-ar- ea biomass (adjusted for untrawlable area) estimated from trawl surveys during the reference-year period to coastwide stratified swept-area biomass of the target species (Table 2). For bocaccio, the coastwide swept- area biomass was the sum of the swept-area biomasses computed in each survey area from the DFO groundfish surveys plus a swept-area biomass estimate from the unsurveyed area not covered by the DFO groundfish surveys (Fig. IB). This coastwide swept-area biomass includes regions over 100-300 m in bottom depth on the outer coast or in Hecate Strait (see, for example, in Fig. IB the gap off the southwest coast of Haida Gwaii between the west coast of Haida Gwaii (WCHG) and QCS surveys). For the latter survey, a global estimate of species density was made across all trawl areas for the reference year period. For surveys other than the reference survey (e.g., for bocaccio in the West coast of Vancouver Island (WCVI) shrimp survey, Queen Charlotte Sound (QCS) shrimp survey, and U.S. triennial surveys), the biomass of the target species present in the areas covered by the sur- veys was computed from densities observed in the ref- McAllister et al.: Using experiments and expert judgment to model catchability of Pacific rockfishes 289 Table 2 Estimates of bocaccio ( Sebastes paucispinis ) biomass with the swept-area method and based on Department of Fisheries and Oceans (DFO) groundfish surveys in each region in the years 2003-07. Percentage of coast-wide biomass refers to the fraction of coast-wide swept-area biomass (in regions 1-4, 8) that is on average found inside each region and is used in computing survey catchability. “SE ln(6io)” is the standard error for the mean of the natural logarithm of swept-area estimates of stock biomass in each region, with the mean determined from swept-area estimates in different years. “%Trawlable” is the estimate of the percentage of the region that was found to be trawlable based on random sampling of locations in each region for trawling and large sample sizes (>300 sites in each region). GFish=groundfish; WCVI=west coast of Vancouver Island; QCS = Queen Char- lotte Sound; HS = Hecate Strait. WCHG=west coast of Haida Gwaii. The total is obtained only from the DFO groundfish survey because the other surveys are contained within it. Survey number and region Biomass (kg)1 Percentage of coast-wide biomass Years of data SE ln(bio) % Trawable 1 WCVI Gfish 375,207 50 2004, 2006 0.0540 71.6 2 QCS Gfish 247,966 33 2003-05, 2007 0.2950 76.5 3 HS Gfish 35,340 5 2005, 2007 0.2670 78.3 4 WCHG Gfish 11,255 2 2006, 2007 0.0507 87.2 5 WCVI shrimp 20,787 3 2004, 2006 0.0262 100.0 6 QCS shrimp 27,767 4 2003-05 2.1010 100.0 7 U.S. triennial Gfish 180,599 24 2004, 2006 1.8590 82.0 8 Unsurveyed 76,664 10 2003-07 0.2020 0.0 Total 746,432 1 Computed from the product of the average density from the trawlable area and the total area of each survey area. erence survey data set. Note that the regions covered by both shrimp surveys and the U.S. triennial survey lay within areas covered by the reference groundfish survey. For example, the biomass available in the WCVI shrimp trawl survey was based on the density observed in the tows conducted during the WCVI groundfish sur- vey, within the area covered by the shrimp survey. The fraction of total stock biomass, Ss in a given area s, can be obtained by g It ,s(fT’S+a(1~fT,s)) (g) s 11 areas ^ Itj [frj + a{l- f r,j )) 7=1 where nareas = number of regions for the reference survey plus 1 to account for the coastal habitat for the species that is outside of the sur- veyed area. For the each region in the survey q prior model, ITs can be considered a lognormal random variable; IT s ~ lognormal (hi ( 7™^), S.Es2). (7) The cross-year median for the lognormal density func- tion, I™? , in Equation 7 can be computed from the empirical mean swept-area biomass estimate (I^WSA ) in each area s and the standard error (SE): I™* = * exp< -SE2 / 2), ( 8a) where tSw.A. 1T,s nyr,s XrSw.A y7\s,y y= 1 (8b) se‘: = X(l"(7?V)-ln('S'4));and (8c) y= i (8d) where n - number of years in which a swept area biomass estimate is available for a given survey in the reference year set. The reference swept-area estimate for unsurveyed regions was obtained from stratified estimates of species density from trawled areas and the estimated habitat area outside of surveyed areas. Because the average catch per tow is based on tows over trawlable bottom, the swept-area estimate was adjusted to account for the estimate of the fraction of trawlable area in the survey area and the average relative difference in bocaccio density between trawlable and untrawlable bocaccio habitat (see Eq. 3). 290 Fishery Bulletin 108(3) Approach to acquiring information from trawl experts Our approach relies on a large number of experts being interviewed to seek their judgments on the credibil- ity of hypothesized values for factors affecting trawl- net catchability ( qnet ). A sufficiently large number of experts is required to characterize the range of dif- ferences in opinion among experts (e.g., Martin et ah, 2005; Uusitalo et al., 2005). This can be achieved by continuing to sample until the distribution of inputs stabilizes, e.g., the means and standard deviations in inputs change by less than 10% for each new expert interviewed. In the interviews each trawl captain was asked to specify the most likely, minimum plausible, and maximum plausible average values for a set of key factors conjectured to determine qnet and these values were then used to formulate a triangular dis- tribution for each factor for each survey net specific to each captain. The component factors of qnet formulated below represent “average” effects. Thus the minimum and maximum input values for each key factor do not reflect a predicted response for one case (i.e., from the population of all tows), but the minimum and maximum values for the average value across all tows combined. The pdfs formulated thus represent density functions of the mean value for a given factor, not the population of values from all conceivable tows. The probabilistic modeling approach that was applied to synthesize the captains’ inputs was similar to that taken by Uusitalo et al. (2005) and Martin et al. (2005) to formulate priors based on interviews with several different experts. For each net type, the resulting qnet was modeled as a mixture of the distributions result- ing from the specifications from each of the interviewed captains. Answers to our questions allowed us to simplify the process of catching a bocaccio to six steps based on four questions: Step 1 Resolve the relative distribution in the water column (ax). Question 1) What is your best estimate (and minimum and maximum) of the percentage of target species that would be near-bottom (within 3-4 m) as the vessel passed overhead ? Rockfish, particularly bocaccio, are presumed to oc- cupy the water column from surface to bottom, but their density increases with depth. The factor, av for the relative distribution of the target species in the water column (zone A, Fig. 3) defines the proportion of fish below headrope height, as the vessel passes over the fish. For this step, three assumptions are made: 1 fish below the headrope, as the vessel passes over them, continue to stay below the height of the head- rope until they arrive at the mouth of the net; 2 fish outside the doors (horizontally), continue to stay outside the doors; and 3 aT is the same for all nets (this is reasonable because most of the nets have headline heights of around 3 m and only the U.S. triennial net is higher). Step 2 Resolve the proportion of off-bottom target species that “dive” into the “kill zone” (i.e., the area immediately in front of the opening of the trawl net), (i.e., zone E, Fig. 3) (a2). Question 2) What percentage of those fish initially off-bottom would dive into the kill zone? The factor a2 is the proportion of fish in zone B that would dive into the kill zone from those initially above the head rope of a given type of net (zone B, Fig. 3). For factor a2, the following assumptions are made: 1 all fish below the headrope, stay below the headrope until at the mouth of the net; 2 fish dive in response to vessel noise and warps; and 3 dive rate is equal for all net -warp-vessel combinations. Step 3 Resolve the proportion of fish which lie in the “dead” zone, i.e., the zone between the doors but external to the trawl warps. The answers to questions 1 and 2 provided the per- centage of fish that were initially in the path of the trawl doors that would lie in the capture zone as the doors approached (between the doors and below the headrope) (zones C-D, Fig. 3). The disposition of the fish horizontally would then be partially determined by whether they lay directly in the path of the net between or outside the wingtips but still within the door path. Fish in zone C were assumed to stay there as the net approached (zone E). Fish in zone D would have to be herded inwards to area C by the sweeps and bridles (Fig. 3). Discussions with some captains indicated that for fish that lie within 6 m of doors inside the door path there is zero catchability. As the trawl warps approach the doors near the bottom, they spread out towards the doors, possibly scaring near-bottom fish out of the kill zone. Therefore, as the doors approach the fish, the fish are assumed to be distributed across the path of the doors in one of three sectors, in proportion to the linear dimensions of that sector (Fig. 3), namely: 1 in the path of the net (i.e., between wingtips, zone C, Fig. 3, Tables 3 and 4); 2 in the path of the sweep and bridles but more than 6 m (for survey nets used for bocaccio) inside of the door path (herding zone, Dl, Fig. 3B) (factor a31n)\ and 3 in the path of sweep and bridles but within 6 m of the doors (dead zone, i.e., horizontal area in which all fish are expected to escape capture) (D2, Fig. 3B). Step 4 Resolve arithmetic correction for the relative proportions of fish remaining in front of the net (between wingtips) or in front of sweeps and bridles (inside of the dead zone). After allowing fish in the dead zone to escape, we estimated the proportions of remaining fish that ei- ther lie in front of the net (zone C, Fig. 3) (factor a4 „) or the “herdable” section of the sweeps and bridles McAllister et al.: Using experiments and expert |udgment to model catchability of Pacific rockfishes 291 Table 3 Relative distribution of fish in different sectors and parts of the kill zone of the gear as the gear approaches a stationary fish. The factors, proportion of the linear distance between the trawl doors that is within wingtips (a3J) and proportion in the dead zone (a32) are used in the trawl survey catchability model. Distances are in meters. See Figure 3 for a schematic diagram of a trawl net. Net type Nominal door spread Nominal wing spread Nominal distance between doors, outside of wings Dead zone in herding area Effective herding zone Proportion remaining in herding zone (a3,2) Proportion within wingtips (a3,F Proportion removed by dead zone AWII trawl 63.3 14.4 48.9 6.0 36.9 0.583 0.227 0.190 (groundfish) Nor’Eastern trawl 58.9 13.4 45.5 6.0 33.5 0.569 0.228 0.204 (U.S. triennial) Shrimp trawl 26.5 10.6 15.9 6.0 3.9 0.147 0.400 0.453 Table 4 Relative proportions (P) of remaining fish in areas C and D1 (from columns 6 and 7 in Table 3, Fig. 3). Both of these factors (a4 and a5, respectively) are used in the trawl-survey catchability model. Net P between wingtips (a4) P in herding zone (a5) AWII trawl (groundfish) 0.281 0.719 Nor’ Eastern trawl 0.286 0.714 (U.S. triennial survey) Shrimp trawl 0.731 0.269 (zone Dl, Fig. 3B, Tables 3, 4) (factor a5n) (see Eq. 10 for cap- tains who explicitly accounted for the dead zone and Eq. 11 for those who did not). Step 5 Determine the propor- tion of fish that will be herded from the path of the sweeps and bridles (zone Dl) into the path of net (inside the wingtips) (zone C). Question 3) What percentage of the fish in front of the bridles and sweeps would be herded into the path of the net ? The factor a6 concerns the remaining fish in sweeps and bridles path and for this step the following assump- tions are made: 1 fish initially in front of the net, stay in front of the net; and 2 factor a6 is the same for all nets. Step 6 Determine the proportions of fish that are cap- tured of those that end up in front of the net (a7nc). Question 4) What percentage of the fish that make it to area E will be captured and retained by the net ? Finally, of the fish that have ended up in front of the net (zone E in front of footrope, Fig. 3), what percentage will be captured and retained in the net? Step 1 Draw a value for the ratio of fish density in untrawlable areas to fish density in trawlable areas, a, from the density function for it (see Eq. 3 and below for specifications). Step 2 To generate a value for the fraction of total population biomass vulnerable to survey gear in each area, Ss, draw a value for swept-area biomass in each of the eight coastal areas, using the lognormal density function and the empirical swept-area value as the median and the variance in the natural logarithm of the estimate (Eqs. 6-8, Table 2). Step 3 For each captain, draw a value for the proportion of fish below the headrope (a4) using the parameters of the triangular distribution provided by each captain. Steps in the algorithm to compute a prior probability density function for estimates of catchability WinBUGS 1.4 (Lunn et al., 2000) was applied to synthe- size the inputs from the trawl captains and other techni- cal settings and to produce output density functions for the qgross values for each of the surveys. The steps of the algorithm applied are provided below. Step 4 For each captain, draw a value for the propor- tion of fish above the headrope that stay above the head- rope as the net approaches (a2), using the parameters of the triangular distribution provided for each captain. For each captain, compute the proportion of fish en- tering the path of the net and doors from those in the water column that are in the path of the net and the doors (a1 2), such that 292 Fishery Bulletin 108(3) cq 2 =l-(l-a1 )*a2. (9) Step 5 For each captain, draw a value for the propor- tion of fish that will successfully be herded from the path of sweeps and bridles to the path of the net (one captain) or herded from the path of the doors to the path of the net (the other captains) (a6n). Step 6 For each captain, compute the fraction of fish between the doors that end up in front of the net, given the proportion of doorspread that is between the wing- tips (a3 1 n) (step 3 of previous section) for the following sections of the gear — the dead zone (a3 2 n), between the dead zone and the wingtips (of the area not in the dead zone) (a4n), and between the wingtips (of the area not in the dead zone) (a5 n) — and the fraction of fish herded into the path of the net (a6 n). For the captains that conditioned herding of fish into the front of the net on those fish that swim in the zone between the doors and the wingtips, the following for- mula applies: a3.6,n,l = ^ 1 _ a3,l ,n X a6 ,n + a3,l ,n ■ ) For the captain that conditioned herding of fish into the front of the net on those fish that swim in the area that does not include the dead zone, the following formula applies: a 3.6,n,2 =^-a3,2,n^X^a4,;i X a6 ,n +a5,n^ Step 7 For each net type (n) and captain (c), draw a value for the proportion of fish that are captured of those that end up in front of the net (a7 n c). To do this, use the parameters of the triangular distribution provided for each captain for each net type. Step 8 Compute qnet for each net type (n) and captain (c): Qnet,n,c = °1.2,c X °3.6 ,n,c X °7 ,n,c* d^) Step 9 Compute the qaross for each survey (s) for each captain (c): Q gross, s,c = Qnettn,c X Un,c XSs/(ggX a8 n ), (13) where Un c - the uncertainty random variable for each net type and captain; Ss = the random variable for the fraction of exploitable stock biomass in the region s; gs = the random variable accounting for traw- lable area in region s; and a8n - the fixed correction factor applied where the wingtip distance had been applied to compute the swept-area biomass. U n c, is applied to each qnet n c to ensure that the density functions are not overly precise (i.e., it applies a multi- plicative uncertainty factor). Such factors have been applied in other situations where it is presumed that the distributions offered by experts are far too certain (e.g., Boyer et al., 2001). In our application, an uncertainty factor was drawn from a lognormal density function with a coefficient of varia- tion (CV) of 0.5 and a median of 1 for each captain and net type. See the discussion for further justifications for including this factor and for the choice of the value for the CV. Step 10 Give each captain’s qgross equal prior weight in the final qaross distribution such that the chance of including a given captain’s input has equal prior prob- ability. We applied a C-dimensional Dirichlet density function where C is the number of captains. This was applied as the multivariate prior pdf for the relative weight given to each captain’s qgross distribution for a given survey. All C input parameters for this density function were set to 0.5, which gives a relatively uninformative prior for the weight placed on each captain. In each Monte Carlo iteration, one of the C captain’s qgross val- ues was randomly chosen for the qaross random variable for each of the seven research surveys. Thus, without any Bayesian updating with new data, each captain’s inputs are given equal weight in the output probability distribution qgross for each regional survey. Step 1 1 Use observations of the ratio of average catch rates from the different survey gears, e.g., shrimp trawl and groundfish trawl, from comparative gear experi- ments in specific locations (intended or unintended) to update the qnet c density functions for these survey nets (see Eqs. 14-15 below). The ratios of observed average catch rates for the different survey nets will give more weight to captain inputs that are more consistent with the observed ratios for these two net types. Step 12 Apply WinBUGS (or other Bayesian integra- tion software) to produce two or more sets of Markov chain results for the qnet parameters for each net type and qgross parameters for each survey; apply diagnostics to remove the burn-in and summarize the posterior results. Step 1 3 Evaluate the posterior correlations between the Qgross parameters for the different surveys and identify a suitable multivariate density function to summarize the results. Because the qgross distributions for the different sur- vey regions in our application were computed with identical input values for qnei across survey regions, the qgross variables tended to be highly correlated across survey regions. There is the potential for multimodal- ity in the marginal density functions for qgross for the different survey areas and thus a mixture distribution may be appropriate. Should the results for each survey be unimodal, then a multivariate lognormal density function should be a good candidate. McAllister et al: Using experiments and expert judgment to model catchability of Pacific rockfishes 293 Implementation of the method for bocaccio In the q prior model for British Columbia bocaccio, we treated the factor a as a random variable, having a triangular prior distribution with a minimum of 1, maximum of 10, and mode at 3. fT is provided by survey area (Table 2) and treated as known because the number of sites sampled per survey area was high in all areas (300-1000 depending on the survey area). Shelf regions without trawl surveys (unshaded regions in Fig. IB) were excluded from the original set of groundfish sur- veys because the fraction of trawlable seabed was known to be very low. Thus, fT in areas where there are no surveys is presumed to be 0% The average and standard error (SE) in the average of the natural logarithm of the available estimates of an- nual swept-area biomass for each survey region (Table 2) were computed and applied in the q prior model to generate from a lognormal density function samples of potential stock biomass in each region and the potential fraction of total stock biomass in each region. Some of these standard errors were very large and created large uncertainty in the fraction of stock biomass for each of the regions. Note that in instances in which the SE is less than 0.15, we recommend that this value be set to 0.15, because, in general, the minimum CV in a swept-area biomass, accounting for all sources of error variability (i.e., SE divided by the mean) for relative stock size, should be no less than 0.15 (the CV of a lognormal distribution is Vexp(a2)-1, where a is the SD in the natural logarithm of the random variable). The empirical values for SE may be low because of small sample sizes (e.g., n- 2 years) and chance. We believe that because of the highly clumped spatial distribution of bocaccio, longer time series would yield higher values for SE than were obtained when the empirical values happened to be less than 0.15. The U.S. triennial survey and the two shrimp surveys are contained within the WCVI and QCS groundfish surveys (Fig. IB). These larger surveys that contain the smaller, more localized ones are called here “containing surveys.” For the smaller or “contained” surveys, the random variable (RV) for ITs was limited to the product of the fraction of area occupied by the contained sur- vey and the RV for ITs for the containing survey. This computation presumes that bocaccio density in the con- tained survey is no larger than that in the containing survey and limits the biomass for the contained survey to no more than that expected if the density was the same between the contained and containing survey. In our application, we consulted with 12 commercial trawl captains — each with at least 10 years of experi- ence in trawling for rockfish. All captains had experi- ence (11-22 years) with types of trawls used in the DFO groundfish and U.S. triennial surveys, i.e., both ground- fish and shrimp trawl nets, and with total groundfish landings ranging from 6800 to 275,000 t. Captains 1-4 were interviewed in groups of two and the remaining captains were interviewed separately. An attempt was made in each interview to provide the same explanation for the requested information, although the interview was conducted in an informal conversational manner. The format undoubtedly varied in subtle ways over the course of the 12 interviews. During our interviews with trawl captains, we characterized “typical survey fishing” as occurring on average at 150 m depth from June to July from 1 h after sunrise to 1 h before sun- set. This fixed interval of time was necessary because trawl captains preferred to answer the questions while considering specific fishing conditions (i.e., time, depth, season, etc.) For bocaccio, only one of the captains presumed that the a6 proportion reflected the proportion of fish between the dead zone and the path of the net that are herded into the path of the net. The rest of the captains presumed that this proportion reflected the fraction of fish between the doors and the path of the net that are herded into the path of the net. The door- spread of the U.S. triennial survey Nor’ Eastern trawl net was not measured. We assumed it had the same ratio of wingtip to doorspread as that of the Atlantic western (WII) trawl net used in the DFO groundfish survey. For the qnet interview questions, each captain was asked for catch estimates for each of the three nets. The nets are towed at different speeds, have different vertical openings and, perhaps most importantly, the mouth opening of the shrimp trawl is not configured, so that the headrope overhangs the footrope (known as a “cape”). The net parameters are as follows: 1 DFO Atlantic Western trawl: towed at ~3 knots, and having a 3.7-m vertical opening; 2 U.S. Nor’ Eastern trawl: towed at ~3 knots and having a 7.1-m vertical opening; and 2 DFO shrimp trawl: towed at ~2 knots and having a 2.7-m vertical opening. Most groundfish trawls have a shorter headrope than footrope so that the headrope precedes the foot- rope through the water providing a “cape” or “hood.” As a fish encounters the footrope, it cannot escape by swimming directly up. On the shrimp trawl, however, the headrope and footrope are virtually in line. Pre- sumably, when the bocaccio detect the proximity of the mouth opening of the shrimp trawl, the net front is already effectively a 2.7-m vertical “wall” of footrope, disturbed sediment, and headrope. It is reasonable to assume that some bocaccio would escape vertically. When a bocaccio encounters the groundfish footrope, however, it is surrounded on four sides (wings, cape, and the bottom). We assumed that the relatively large bocaccio did not escape through the net and that the probability of retention was 1. The value for this fac- tor, a7n, depends on the net (a71 for the AWII trawl, a7 2 for the triennial Nor’Eastern, a7 3 for the DFO shrimp trawl). Our approach derives catchability based on doorspread, and therefore the U.S. triennial and the WCVI shrimp trawl estimates first had to be altered by the ratio of wingspread to doorspread (aSn). 294 Fishery Bulletin 108(3) Table 5 Biomass estimates of bocaccio ( Sebastes paucispinis) based on survey tows using the shrimp trawl survey net and the Atlantic western (WII) groundfish trawl survey net, Estimates were based on survey positions shown in Figures 4 and 5. Results are shown for Queen Charlotte Sound (2003-07 for shrimp trawl and 2004-05 for AWII groundfish) and west coast of Vancouver Island (2004, 2006). The ratio of catch rates between these two survey gears was used to screen the plausibility of values for trawl-net catchability for these two nets that was determined from the interviews of the trawl captains. Shrimp trawl AWII groundfish trawl Region Number of tows Biomass (kg) Number of tows Biomass (kg) Queen Charlotte Sound 212 4993 52 39,746 West coast of Vancouver Island 141 5258 66 20,787 The shrimp and groundfish survey gears were ap- plied in the same years in the survey area of the WC- VI shrimp survey and the QCS shrimp survey. The observed mean ratio of bocaccio density between the trawl and shrimp survey nets for QCS for the years 2003, 2004, 2005, and 2007 was 8.76, with a SE in the natural logarithms of the estimates of 0.59 (Table 5, Fig. 4). The observed mean ratio for density estimates for the WCVI shrimp survey region between the ground- fish and shrimp nets for the years 2004 and 2006 was 3.95, with a SE of 0.116 (Fig. 5). This latter SE was increased to 0.3 for the statistical estimation, because it was judged unlikely that the precision could be so high and there were only two years of survey data to provide this estimate. In each Monte Carlo iteration, the natural logarithms of the computed qnet values cho- sen for the shrimp and groundfish surveys were taken and the logarithm of the qnet for the shrimp survey was subtracted from the logarithm of the qnet for the groundfish survey. This difference was used as the expected log ratio for these survey catch rates for these two types of trawl nets. A lognormal density function was then applied to compute the probability of the observed ratio, given the model predicted ratio of qnet for these two nets: Figure 4 Mapped zones and trawl tow positions (symbols) of the Queen Charlotte Sound groundfish and shrimp trawl surveys where the two surveys overlapped. These two surveys provided an unintended reference source of the catch-rate ratio data for these two survey gears that were then used to update the groundfish- to-shrimp trawl-net catchability ratios for each of the experts. Polygons sur- rounding the overlapped survey areas were connected by hand to delimit the outer boundaries. Open and closed circles indicate the absence and presence, respectively, of bocaccio ( Sebastes paucispinis) in the groundfish surveys. Open and closed triangles indicate the absence and presence, respectively, of bocaccio in the shrimp surveys. Ki-j =log<9n=;,s)- log< s) and i*j, r-°Ki-j log normal(lrs t asi_2). (15) where the subscripts i and j denote the DFO groundfish survey and shrimp survey nets; respectively, and r_obsi_j = the observed ratio of density values from groundfish and shrimp nets for survey area s. As indicated above, there are two observed ratios for bocaccio; one for the WCVI and one for QCS. For WCVI ,osi_j was 0.3, and for QCS, os ■ • was 0.59. The WinBUGS results for bocac- cio were numerically stable after McAllister et al.: Using experiments and expert judgment to model catchability of Pacific rockfishes 295 rapid burn-in and rapid mixing. The Gelman-Rubin (Brooks and Gelman, 1998) statistic was ap- plied to assess the burn-in period, which was judged to be about 500 iterations. A total of 40,000 itera- tions with two chains were judged to be sufficient to provide precise approximations of the target den- sity function. Using the results after the burn-in, we found that the ratio of Monte Carlo error (analogous to standard error in the sampled posterior mean) to the posterior standard deviations (SDs) for all outputted variables was far less than the minimum standard of 5% (Best and Thom- as, 2000). Results 127"30'W 12?°W 126‘30'W 126"W 125"3Q'W Figure 5 Mapped zones and trawl tow positions (symbols) of the west coast Vancouver Island groundfish and shrimp trawl surveys where the two surveys overlapped. These two surveys provided the catch-rate ratio data for these two survey gears that were then used to update the groundfish-to-shrimp trawl-net catchability ratios for each of the experts. Polygons surrounding the overlapped survey areas were connected by hand and delimit the outer boundaries. Open and closed circles indicate the absence and presence, respectively, of bocaccio ( Sebastes paucispinis) in the groundfish surveys. Open and closed triangles indicate the absence and presence, respectively, of bocaccio in the shrimp surveys. We first considered the individual distributions computed from each captain’s inputs for the catchabil- ity of each of the three net types (qnet). For each of the three net types, a wide range of plausible values for qnet were obtained from the 12 interviewed captains and there was considerable variability between the captains and some of the distributions were nonoverlap- ping (Fig. 6). The CVs in the qnet distributions by captain for each net varied from about 0.1 to 0.6, reflecting con- siderable variability in individual levels of uncertainty in the qnet inputs. The qgross values obtained for each captain and for each net type with no updating and no uncertainty fac- tor showed considerably wider distributions and more overlap in all cases between the captains than the qnet distributions for each captain (Figs. 6 and 7). The qgross distributions for the different surveys showed varying amounts of overlap between the captains with the WCVI shrimp survey showing the least amount of overlap and the U.S. triennial survey showing the most overlap be- cause of very low precision in qgross among captains. The low precision was primarily due to the high uncertainty in the fraction of stock biomass in the U.S. triennial survey area (Table 2). The CVs in the qgross distribu- tions by captain ranged from about 0.3 to 0.7 for the DFO groundfish surveys and the WCVI shrimp survey (Fig. 7). However, the QCS shrimp survey showed high CVs of about 1.5-1. 8 because of the added uncertainty in accounting for the fraction of the stock in each sur- vey area and the ratio of bocaccio density in untraw- lable and trawlable areas. The qaross distributions for the shrimp trawl surveys (e.g., for WCVI) were centered considerably lower than those provided for the ground- fish surveys partly because of low values for qnet and because a small fraction of the stock falling in these areas. Also for the shrimp trawl survey areas, the frac- tion trawlable was 100%, whereas the groundfish trawl survey areas this was closer to 70-80% (Table 2). We next consider different approaches to combining the qnet distributions from the different experts into a single quet distribution for each net type. When equal weighting was applied to the inputs from the different captains without Bayesian updating and without the uncertainty factor, the qnet distributions for each net were multimodal (Fig. 8). Under these same conditions, the combined distributions for qaross for each net showed varying amounts of departure from unimodality; the qgross distribution for the WCVI shrimp survey showed the most pronounced bimodality (Fig. 9). When the un- certainty factor was applied without Bayesian updating, the qnet distributions showed less pronounced multimo- dality (Fig. 8); multimodality was no longer seen in any of the qaross distributions and the distributions became slightly wider (Fig. 9, Table 6). We compared the ratios in values of qnet for the groundfish survey to qnet for the shrimp survey (pro- vided by the captains) with the observed ratios in values of qnet for the groundfish survey to qnet for the shrimp survey in the WCVI and QCS surveys 296 Fishery Bulletin 108(3) (Fig. 10). s/blOAlthough the observed ratios for WCVI and QCS were about 4 and 9, the ratios obtained from the captains’ inputs ranged from about 1 to 32 (Table 7). The CVs in the expertly determined ratios ranged from about 0.1 to 0.5, conveying a range of degrees of uncertainty between captains (Table 7). Where the cap- tains’ ratios deviated most from the observed ratios and had the smallest CVs, the posterior probability for the captain was effectively zero or very low. This situation occurred for six of the captains (Table 7). In contrast, ratios that deviated considerably from the observed ones but had high uncertainty, e.g., for a captain whose mean ratio was 0.7 and CV was 0.4, still retained some posterior weight whereas captains with mean ratios of 1.3 and 1.5 but much smaller CVs had effectively zero posterior weight. The posteriors for the six captains that retained most of the weight ranged from about 0.06 to 0.38. These posteriors showed, if anything, a negative correlation (about -0.10 to -0.30) with the three mea- sures of trawling experience (years of experience, and tons of groundfish and bocaccio landed) (Table 8). There- fore, although an expert’s amount of experience may be a reliable indicator of his or her technical proficiency (i.e., the estimates of total groundfish landings and to- tal bocaccio catch are strongly positively correlated with years of experience), experience does not indicate the re- liability of the information that he or she may provide. When Bayesian updating was applied without the un- certainty factor, most of the weight shifted from three modes for qnet to predominantly two modes for each net type (Fig. 8). For example, the third mode over the high- est values in the shrimp qnet posterior was eliminated. The central tendencies for the shrimp qnet and qgross val- ues decreased by about 40%, whereas that for the DFO AWII nets increased by no more than about 5% with the Bayesian update. Under these same conditions, preci- sion in the qaross distribution for each survey increased only for the shrimp surveys (Table 6). Bimodality was no longer present in the qgross distributions (Fig. 9). In contrast to the instance with no uncertainty factor and no Bayesian updating, when Bayesian updating and the uncertainty factor were applied, the qnet distributions by captain all overlapped for each of the three nets (Fig. 10). The mean value for qnet for the shrimp trawl was lower and more uncertain than for the two ground- fish nets (Fig. 8). None of the qnet and qgross composite distributions showed bimodality and results were not quite as precise as with the analogous case without the un- certainty factor (Figs. 8 and 9, Table 6). The CVs in the groundfish survey qgross val- ues ranged from about 0.77 to 0.83; and 95% probability intervals (Pis) ranged be- tween 22- and 25-fold between the bounds (Table 9). The CVs for the shrimp survey qgross values were considerably higher at about 1.5 and 2.7 (95% Pis of about 84- and 3100-fold). This high uncertainty was largely due to large differences between the inputs provided by captains but also due to higher uncertainty in the fraction of stock in these surveys (Table 2). The U.S. trien- nial survey qgross had a high CV (1.7) and an 800-fold 95% probability interval (PI). This high uncertainty is also due mainly to the high uncertainty in the fraction of the stock in this survey (Table 2). One key factor is the ratio of fish density in untrawlable areas to that in trawlable areas, a. When this factor was set to 1 and the Bayesian update and uncertainty fac- tor were applied, the central tendencies of the posterior distributions for all of the surveys approximately doubled, indicat- ing that the effect of this parameter is to decrease qaross for all surveys and to lead to increased population biomass estimates (Table 6). Doubling the mode and maxi- mum value for a from 3 and 10 to 6 and 20 caused the mean for qgross to decrease Figure 6 Probability density functions for trawl survey net catchability (qnet) for each captain for the (A) Atlantic western (AWII), (B) shrimp, and (C) Nor’Eastern trawl nets, based on inputs provided for each captain without any Bayesian updating and without the uncertainty factor applied. McAllister et al.: Using experiments and expert judgment to model catchability of Pacific rockfishes 297 to between 63% and 81% of the reference case values. The CVs for qgross for the different surveys decreased slightly when a was set to 1 and increased slightly when its input distribution mode and maximum were doubled (Table 6). The qgross values for the different surveys showed varying amounts of positive correlation, which resulted from the use of the same or very similar nets in all of these surveys and the captains prescribing highly correlated prior inputs as in the case of the AWII and Nor’Eastern nets (Table 10, Fig. 8 shows a high degree of similarity in the qnet outputs for these two nets). The Qgross values f°r the DFO groundfish survey nets showed the highest correlations with values up to about 0.96. The qaross for the U.S. triennial survey showed the low- est correlations with the other nets because of the high amount of uncertainty in the fraction of the population in this survey (correlations between 0.14 and 0.35). The QCS shrimp q gg also showed low correlations with the other surveys also because of the high uncertainty in the fraction of the population in this survey area. In our application, density functions for qgross were unimodal and in all instances positively skewed. Thus, a multivariate lognormal density function was formu- lated to summarize the joint prior density function for Qa,oss f°r fhe six survey time series used in the stock assessment. This multivariate density function was for- mulated using the posterior median and covariance out- puts from the WinBUGS (Tables 10 and 11). The prior results for qaross for six of the surveys were compared with posterior results for qgross from a stock assessment of Bristish Columbia bocaccio for which a noninforma- tive prior for qaross was used (Table 11). All posterior medians for qaross obtained from the stock assessment with noninformative priors for p,r,.oss (Table 11) were in- side of the 95% Pis for the informative qgross prior (Table 11). However, in most instances the posterior medians were larger than the prior medians, indicating that the stock assessment data tend to produce stock biomass values lower than those indicated by the qgross density function obtained in this study. Some key stock assessment quantities are also shown that were obtained with a noninformative prior and the informative qgross prior (Table 12). The posterior mean values for stock biomass at maximum sustainable yield, stock biomass in 2008, and replacement yield changed slightly with the use of the informative qgross prior. In contrast, the posterior CVs for the current stock bio- mass and replacement yield decreased substantially with the use of the informative prior for qgross. 298 Fishery Bulletin 108(3) Discussion and conclusions We provide an approach to formulating a Bayesian prior for trawl survey catchability for rockfish that can integrate subjective judgment from several experts on factors affecting net catchability with data obtained from field experiments and trawl research surveys. The approach is useful for situations where a stock assess- ment model is to be fitted to one or more survey indices and although directed at rockfish surveys, it could be extended to other groundfish species. For this approach, data are used from a series of related trawl surveys covering most of the British Columbia continental shelf, all based on the same trawl gear, to provide estimates of the fraction of the population in each survey area. Bias correction factors are formulated and can be updated with experimental data that allow us to evaluate hypoth- esized average differences in target fish species density between trawlable and untrawlable areas and the frac- tion of area in the survey that is treated as untrawlable. The current approach presumes that experts all have experience with all of the types of nets that are used in the surveys or with nets that are very similar. The approach updates each captain’s inputs on the basis of the con- sistency with observations from experi- ments where the ratios of catch rates between different types of survey nets were evaluated. This updating process reduced the degree of uncertainty in the prior by considerably modifying the posterior distributions, particularly for the most poorly understood gear — the shrimp trawl. A few different procedures were ap- plied to counteract the adverse effects on stock assessment results that may result from the tendency of individual experts to provide distributions that are too certain. One procedure was to apply an uncertainty factor (Boyer et al., 2001); this is discussed further below. The second was to apply a mix- ture distribution approach to incorpo- rate judgment from different experts. However, this procedure may cause the resulting posterior distributions to be multimodal — a feature that may arise when the narrow distributions offered by the various experts fall into differ- ent modes. Such a result is less likely when there is a large number of con- tributing experts. Attempts to directly estimate catch- ability with trawl nets ( qnet ) have met with limited success, particularly for rockfish. The principal difficulty lies in measuring the abundance of fish that is positioned in front of the net. Krieger and Sigler (1996) attempted to estimate catchability of Pacific ocean perch (S. alutu s) for a bottom trawl on the basis of observations from a submersible vessel. They reported es- timates for qnet of 0.97-1.27 for trawl catchability based on wingtip spread. Using the AWII to calculate an ap- proximate ratio of doorspread to wing- spread of about 4.4 (AWII, Table 3), they determined a doorspread catch- ability of 0.22-0.29 for Pacific ocean McAllister et al.: Using experiments and expert judgment to model catchability of Pacific rockfishes 299 Figure 9 Probability density functions for trawl survey catchability (qgross) for the (A) west coast Vancouver Island (WCVI) groundfish, (B) WCVI shrimp, and (C) U.S. triennial groundfish surveys, based on inputs provided for each captain with and without any Bayesian updating, and with and without the uncertainty factor applied. perch. Korotkov1 used an underwater camera-mounted sled towed in front of the trawl to provide a ground-truth of actual fish density. He estimated a doorspread catchability of 0. 1-0.4 for unspecified species of groundfish. Our estimates of qnet for bocaccio for three different trawl net types ranging from about 0.1 to 0.3 are similar to his estimates. Scientists at the Alaska Fisheries Science Center (NMFS) have spent many years attempting to esti- mate catchability of the trawl used in their west coast groundfish surveys. Their most successful work was with flatfish for which they observed maxi- mum door-spread catchability for large arrowtooth flounder ( Atheresthes sto- mias) of 0.47 (Somerton et al., 2007). The catchability for a flatfish such as arrowtooth flounder could be expected to be higher than that for rockfishes because bottom trawl nets are gen- erally designed to capture flatfishes, which tend to stay very close to the bottom as opposed to many rockfishes, such as bocaccio, which tend to dis- tribute themselves higher in the water column. Previous efforts at indirectly form- ing a prior for qgross involved seeking expert judgment and, in some instanc- es, adding auxiliary data to the com- ponents of qgross and then integrat- ing these components within a Monte Carlo framework to formulate a pdf for qgross (McAllister and Ianelli, 1997; Boyer et al., 2001). Different approach- es were used to elicit information from experts. In some instances, the experts were interviewed separately to gain in- formation on key factors determining Qgross Punt et al. 1993, McAllister and Ianelli, 1997; Mosqueira, 2005). Another approach was to put several experts in the same room so that they could form a consensus on these factors (Boyer et al., 2001). An im- portant limitation to these approaches has been that the posterior distributions often tend to be very narrow as a result that too few experts were consulted or that divergent opinions were forced into a consensus. That experts often hold divergent views while each being certain about his or her knowledge has long been recognized as a problem when forming priors based on expert input. In the last decade, a number of analysts have suggested that it is important to retain this diver- 1 Korotkov, V.K. 1984. Fish behaviour in a catching zone and influence of bottom trawl rig elements on selectivity. Int. Council. Explor. Sea, Council Meeting 1984. B:15. sity as an output of the analysis and that it is unwise to eliminate the diversity by averaging across experts (Burgman et al., 1993; Chrome et al., 1996; Uusitalo et al., 2005). Some researchers have advocated assigning weights to experts according to their level of expertise (Burgman et al., 1993); others have assigned equal weighting to expert input, providing that all of these experts initially qualify to provide expert judgment (Martin et al., 2005; Uusitalo et al., 2005). It may not be desirable to assign different weights when the num- ber of available experts is relatively few, because it is possible that a single expert may end up with all of the weight, thus defeating the purpose of the exercise. We recommend applying an uncertainty factor to the for qnet variable obtained from each expert’s inputs to 300 Fishery Bulletin 108(3) Table 6 Posterior means and coefficients of variation (CV) in the natural logarithm for bulk catchability (qgrosa) under four different runs of the trawl survey catchability model with and without the Bayesian update and with and without the uncertainty factor applied. The values in the first column give the prior mean and CV for qgross that could be used in a stock assessment. WCVI =west coast of Vancouver Island; QCS = Queen Charlotte Sound; HS = Hecate Strait. WCHG=west coast of Haida Gwaii. Bayesian Bayesian update, update, uncertainty uncertainty factor, same factor, Bayesian Bayesian No Bayesian No Bayesian density in very high update, update, update, update, trawable and densities in uncertainty no uncertainty uncertainty no uncertainty untrawable untrawable Survey number factor factor factor factor areas areas Mean CV Mean CV Mean CV Mean CV Mean CV Mean CV and region WCVI groundfish 0.070 0.77 0.067 0.67 0.068 0.76 0.065 0.66 0.145 0.68 0.046 0.85 QCS groundfish 0.046 0.80 0.044 0.71 0.044 0.79 0.042 0.70 0.094 0.71 0.030 0.89 HS groundfish 0.0067 0.83 0.0064 0.74 0.0064 0.82 0.0061 0.73 0.0137 0.75 0.0043 0.91 WCHG groundfish 0.0021 0.79 0.0020 0.69 0.0021 0.78 0.0020 0.68 0.0044 0.70 0.0014 0.87 WCVI shrimp 0.0030 1.46 0.0022 0.81 0.0072 1.63 0.0052 0.94 0.0062 1.36 0.0019 1.51 QCS shrimp 0.00036 2.74 0.00026 1.89 0.00086 2.92 0.00063 2.05 0.00054 2.52 0.00029 3.01 U.S. triennial groundfish 0.047 1.73 0.045 1.62 0.047 1.70 0.045 1.61 0.069 1.50 0.037 1.87 Table 7 The posterior mean ratio and posterior coefficient of variation (CV) for the ratios for trawl-net catchability ( qnet ) for the groundfish and shrimp trawl nets obtained from inputs provided by each of the captains. The posterior probability assigned to each captain’s inputs is given in the last column. Captain Mean ratio Mean CV Posterior probability assigned to each captain’s inputs 1 2.30 0.52 0.110 2 9.18 0.45 0.165 3 0.96 0.15 0.000 4 1.30 0.10 0.000 5 0.70 0.38 0.007 6 3.02 0.19 0.277 7 0.95 0.10 0.000 8 0.77 0.25 0.000 9 1.91 0.25 0.060 10 1.47 0.15 0.000 11 31.86 0.11 0.000 12 5.26 0.26 0.380 counteract the problem of experts be- ing overly certain (Chrome et al., 1996; Martin et al., 2005; Uusitalo et al., 2005). The use of highly precise prior distributions for q has at least two ad- verse consequences for fish stock as- sessment. First, when a highly precise prior for q is applied (e.g., CV<0.4), the estimates of quantities of interest (e.g., virgin biomass, B0 ) can be highly pre- cise and exclude values consistent with stock assessment data (Boyer et al., 2001). Second, simulation evaluations have shown that when highly precise priors for q (e.g., with prior CV<0.5) that are centered over values as little as 50% higher or lower than the ac- tual value, it takes many more years for stock assessment data to update precise priors than less precise priors centered over the same incorrect val- ues (McAllister and Kirkwood, 1998). Application of a multiplicative uncer- tainty factor with a median of 1 and a CV of no less than about 0.5 maintains McAllister et at: Using experiments and expert judgment to model catchability of Pacific rockfishes 301 Table 8 The correlation between measures of expertise and the posterior placed on each trawl captain’s input. These correlations are presented to evaluate the presumption that the reliability of a trawl captain’s judgment on trawl-net catchability increases with experience (i.e., the posterior probability on the captain should be positively correlated with years of experience). See Results section for further details. Years of Total Total Posterior probability experience landings (t) bocaccio catch (t) obtained on a captain Years of experience 1 Total landings (t) 0.66 1 Total bocaccio catch (t) 0.70 0.85 1 Posterior probability obtained on a captain -0.09 —0.29 -0.24 1 the central tendency of the experts’ dis- tributions and gives a prior CV in qnet for each expert that is no less than 0.5. Simulation evaluation has shown that prior CVs for q>0.5 enable stock assess- ment data to override a biased prior for q within relatively few (e.g., 5-10) years (McAllister and Kirkwood, 1998). The choice of a CV of 0.5 is somewhat arbi- trary but in our view necessary. A hierarchical meta-analysis of stock assessment data from different popu- lations of the same species group was applied to quantify the cross-stock cen- tral tendency and variability in q„ross for rockfish in the U.S. triennial sur- vey (Millar and Methot, 2002). This approach has the advantage of avoid- ing expert judgment altogether and is tractable providing there is uniformity in the survey gear used in the differ- ent surveys. However, differences in be- havioral responses to trawl gear among species could limit the validity of the assumption of exchangeability, which must be made in hierarchical model- ing, and thus limit the applicability of the results as a prior distribution to an unsampled population. The log- transformed mode in the posterior pdf of “bulk” catchability equated to about 1.27 between the wingtips. The ratio of doorspread to wingspread ratio for this survey is not available but is probably similar to the approximately 4.4:1 ratio of the AWII configuration used in the DFO groundfish surveys and translates to a doorspread catchability estimate of about 0.29. The priors provided in this study have higher CVs (0.8— 2.7) than previous priors on survey q difference is partly due to the high uncertainty in the obtained from different experts (ranging from about fraction of the population falling within each survey 0. 4-0.7, e.g., Punt et al. 1993; Boyer et al., 2001). This area which did not apply to the other studies because A C Figure 10 Probability density functions for trawl survey net catchability (qnet) for each captain for the (A) Atlantic western (AWII), (B) shrimp, and (C) Nor’Eastern trawl nets, based on inputs provided for each captain with Bayesian updating and with the uncertainty factor applied. 302 Fishery Bulletin 108(3) Table 9 Output statistics for qgross for the run with the Bayesian update and with the uncertainty factor applied. Posterior means (mean), standard deviations ( SD), and coefficients of variation (CV ) are given. The last three columns show the 2.5th, 50th, and 97.5th per- centiles. WCV=west coast of Vancouver Island; QCS = Queen Charlotte Sound; HS=Hecate Strait. WCHG =west coast of Haida Gwaii. Mean SD CV 2.5th 50th 97.5th 1 WCVI groundfish 0.070 0.054 0.77 0.010 0.055 0.212 2 QCS groundfish 0.046 0.037 0.80 0.006 0.036 0.143 3 HS groundfish 0.0067 0.0055 0.83 0.0008 0.0051 0.0215 4 WCHG groundfish 0.0021 0.0017 0.79 0.0003 0.0017 0.0067 5 WCVI shrimp 0.0030 0.0044 1.46 0.0002 0.0016 0.0143 6 QCS shrimp 0.00036 0.00098 2.74 0.0000008 0.0000730 0.00263 7 U.S. triennial groundfish 0.047 0.08225 1.73 0.0004 0.0161 0.287 Table 10 Posterior correlation matrix for the natural logarithm of the qgross values for the seven surveys off British Columbia, during bocaccio ( Sebastes paucispinis ) were captured. The index number in the first column and first row indicates the survey for which the correlations apply. See Table 9 for a key to the survey indices. 1 2 3 4 5 6 7 1 1.000 2 0.912 1.000 3 0.928 0.879 1.000 4 0.963 0.912 0.929 1.000 5 0.601 0.568 0.581 0.604 1.000 6 0.312 0.311 0.305 0.315 0.526 1.000 7 0.346 0.323 0.331 0.343 0.259 0.140 1.000 Table It Comparison of prior and posterior medians and standard deviations in the natural log of qgross SDllng) before and after a stock assessment with noninformative priors for qgross- See Tables 6 and 9 for the prior 95% probability intervals for qgross- NA means not available. WCVI=west coast of Vancouver Island; QCS = Queen Charlotte Sound; HS = Hecate Strait. WCHG=west coast of Haida Gwaii. Prior for qgross Posterior for qgross After fitting assessment model to data Uncertainty factor, Bayesian update with the noninformative survey q prior median q SD ilnq) median q SD (Inq) 1 WCVI groundfish 0.0534 0.78 0.1040 0.50 2 QCS groundfish 0.0343 0.80 0.0566 0.49 3 HS groundfish 0.00492 0.82 0.0030 0.52 4 WCHG groundfish 0.00160 0.79 NA NA 5 WCVI shrimp 0.00161 1.14 0.0091 0.35 6 QCS shrimp 0.000065 2.08 0.0023 0.46 7 U.S. triennial groundfish 0.0144 1.75 0.0924 0.35 McAllister et al : Using experiments and expert judgment to model catchability of Pacific rockfishes 303 Table 12 Posterior means and coefficients of variation (CVs) of key stock assessment quantities for bocaccio ( Sebastes paucispinis ) obtained with a noninformative and informative prior for qgross (see Tables 6 and 9 for the inputs for the informative prior). See Stanley et al. (2009) for the stock assessment method applied. B (t) refers to the population biomass that provides the maximum sustain- able yield in tons. B2008(t) refers to the estimated population biomass in the year 2008 in tons. Bmsy( t) CV B 2008 d) CV Replacement yield (t) CV Non-informative qgross prior 24,146 0.68 4697 2.27 310 1.24 Informative qgross prior 27,021 0.66 3022 0.83 236 0.65 it was felt that these other surveys covered most of the population’s range. Also, in contrast to the present study, in these other studies it was effectively assumed that inputs were obtained from only one expert and did not formally account for cross-expert uncertainty. All other studies so far have developed priors for q that have zero prior correlation (i.e., independence was assumed). In contrast, we developed a mixed-model structure for survey q that produces strong nonzero correlation in qaross values between different surveys — a necessary consequence because the information sources used to produce the qgross values for different surveys are not independent. The prior correlation between qgross values for different surveys was very high in some in- stances (up to 0.96) because different surveys were us- ing the same gear. This high correlation resulted in the same inputs for a given gear type feeding into the formulation of the qnet factor across different surveys. It is important to include this correlation in the prior for qL,ross in a stock assessment because it accounts for the dependencies between the qgross values for differ- ent surveys. Use of the marginal prior variances and assuming independence, i.e., applying zero correlation, would overstate the amount of prior information avail- able about qgross. In contrast to the norm, of which experts tend to be overly certain, all captains in this study expressed con- cern about their estimates. They commented that there had been few opportunities in their careers to compare actual catches with acoustic signals for bocaccio. Three captains said that they could not provide an estimate for at least one question. All captains expressed that they would have been more comfortable estimating these val- ues for other schooling rockfish, particularly yellowtail rockfish and widow rockfish (S. entomelas) because of the greater opportunity to correlate acoustic observations with observed catches. Furthermore, they commented that for bocaccio, as well as other species, catchability would be influenced by factors such as location and bottom type, time of day, state of the tide, and whether the fish were present in large schools or were solitary. The Bayesian computations in our approach that vet the expert-specified inputs against survey-observed val- ues for the same quantities had the result of excluding the inputs from about half of the captains. This situa- tion is undesirable from the point of view of an attempt to include different viewpoints. However, it provides an empirical basis for screening the inputs provided by different experts. More conventional measures of experience (e.g., years of experience, total groundfish landings, and total bocaccio catch) showed either no correlation or a negative correlation with the amount of posterior weight placed on the captains. This finding indicates that practitioners should avoid applying ap- parently sensible criteria to formulate weights to inputs from different experts. Comparison of empirical data with the expert advice within the context of a model ap- pears to provide a reasonably objective way of screening such advice and should be considered instead. One of the most poorly understood parameters is the ratio of rockfish density in untrawlable to that in trawlable areas. In this analysis, a subjective prior was applied which ranged between 1 and 10, with a mode at 3. This application had the effect of reducing the central tendency of the qgross by half for all surveys which would give larger estimates of population biomass. Doubling the width and mode of the input distribution for a fur- ther decreased the mean value for qgross, although by no more than about 37%. We have suggested a simple Bayesian approach to updating this prior, using esti- mates of a from experiments. Possible approaches to estimating a could include experiments designed to estimate relative density in trawlable and untrawlable locations with gillnets, hook-and-line sampling gear, or submersible vessels (Kreiger and Sigler, 1993). The qaross prior developed for British Columbia bocac- cio was applied in a recent stock assessment of this population. The availability of these priors was crucial because most of the survey index series were quite short and all had low precision. Although the prior CV was very high, i.e., no less than about 0.8, this mildly informative prior still helped to bound the range of plausible hypotheses about current stock size and re- placement yield. The posterior medians for qgross were also within the prior 95% Pis when a noninformative prior for it was applied, indicating that the uncertainty obtained in the priors was reasonable and that the priors were consistent with values indicated by the fit of the assessment model to the data. Thus, in this application, the method provided useful inputs for a stock assessment by bounding the range of values for estimated parameters and reducing uncertainty in key management quantities. The higher precision obtainable in stock assessment results when a noninformative prior 304 Fishery Bulletin 108(3) for q is replaced with an informative prior will reduce uncertainty concerning the status of the population and allow fisheries managers to apply harvest-control measures with less uncertain consequences. With our application, less pessimistic and less imprecise assess- ment results could lower the risk of implementing stock rebuilding policies that would cause unnecessary hard- ship on the fishing industry, and larger harvests could be taken with greater confidence that they would be sustainable. Acknowledgments A. Punt, T. Carruthers, and two anonymous reviewers are thanked for their comments on earlier drafts of this article. The authors thank the twelve commercial fishing captains for the time and effort that they provided to quantify factors impacting trawl catchability. M. Wilkins of the U.S. NOAA-NMFS provided data from the U.S. triennial survey. Literature cited Best, N., and A. Thomas. 2000. Bayesian graphical models and software for GLMs. In Generalized linear models: a Bayesian per- spective (D. K. Dey, S. K. Ghosh, and B. K. Mallick, eds.), p. 387-406. Marcel Dekker, New York. Boyer, D., C. Kirchner, M. K. McAllister, A. Staby, and B. Staalesen. 2001. The orange roughy fishery of Namibia: Lessons to be learned about managing a developing fishery. In Pro- ceedings of the symposium 2000: a decade of Namibian fisheries science; October 2000, Swakopmund, Namibia (A. I. L. Payne, ed.), 205-222 p. Special issue S. Afr. J. Mar. Sci. 23. Boutillier, J. A., J. A. Bond, H. Nguyen, and R. M. Harbo. 1998. Shrimp survey off the west coast of Vancou- ver Island and resulting management actions-May 1998. Can. Man. Rep. Fish. Aquat. Sci 2461:110 p. Brooks, S. P., and A. Gelman. 1998. General methods for monitoring convergence of iter- ative simulations. J. Comp. Graph. Stat. 7:434-455. Burgman, M. A., S. Ferson, and H. R. Akcakaya. 1993. Risk assessment in conservation biology, 314 p. Chapman and Hall. London. Chrome, F. H. J., M. R. Thomas, and L. A. Moore. 1996. A novel Bayesian approach to assessing impacts of rain forest logging. Ecol. Appl. 6:1104-1123. DFO (Fisheries and Oceans Canada). 2008. Pacific region amended integrated fisheries man- agement plan: groundfish (March 8, 2008 to February 20, 2009), 157 p. [Available at http://www.pac. dfo-mpo. gc.ca/index-eng.htm] Krieger, K. J. 1993. Distribution and abundance of rockfish determined from a submersible and by bottom trawling. Fish. Bull. 91:87-96. Krieger, K. J., and M. F. Sigler. 1996. Catchability coefficient for rockfish estimated from trawl and submersible surveys. Fish. Bull. 9:282-288. Love, M. S., M. Yoklavich, and L. Thorsteinsson. 2002. The rockfishes of the Northeast Pacific, 414 p. Univ. California Press, Berkeley and Los Angeles. Limn, D. J., A. Thomas, N. Best, and D. Spiegelhalter. 2000. WinBUGS — a Bayesian modelling framework: concepts, structure, and extensibility. Stat. Comput. 10:325-337. Martin, T. G., P. M. Kuhnert, K. Mengersen, and H. P. Possingham. 2005. The power of expert opinion in ecological models using Bayesian methods: impact of grazing on birds. Ecol. Appl. 15:266-280. McAllister, M. K., and J. N. Ianelli. 1997. Bayesian stock assessment using catch-age data and the sampling/ importance resampling algorithm. Can. J. Fish. Aquat. Sci. 54:284-300. McAllister, M. K., and G. P. Kirkwood. 1998. Using Bayesian decision analysis to help achieve a precautionary approach to managing newly developing fisheries. Can. J. Fish. Aquat. Sci. 55:2642-2661. Millar, R. B., and R. D. Methot. 2002. Age-structured meta-analysis of U.S. West Coast rockfish (Scorpaenidae) populations and hierarchical modelling of trawl survey catchabilities. Can. J. Fish. Aquat. Sci. 59:383-392. Mosqueira, I. 2005. Bayesian and geostatistical methods for the esti- mation of spawning stock biomass from pelagic egg surveys. Ph D., diss. 150 p. Univ. London, U. K. NPFMC (North Pacific Fishery Management Council). 2008. Stock assessment and fishery evaluation reports (groundfish, scallop, and crab). [Available from North Pacific Fishery Management Council, 605 West 4th Ave, Suite 306, Anchorage, AK 99501.] PFMC (Pacific Fishery Management Council). 2008. Status of the Pacific Coast groundfish fishery through 2008. Stock assessment and fishery evaluation: stock assessments, STAR panel reports, and rebuild- ing Analyses. [Available from North Pacific Fishery Management Council, 7700 NE Ambassador Place, Suite 101 Portland, OR 97220-1384] Punt, A. E., E. K. Pikitch, M. K. McAllister, and R. Hilborn, 1993. Stock assessment and decision analysis for the western stock of hoki ( Macruronus novaezelandiae) for 1993. New Zealand fisheries assessment research document, Wellington, New Zealand, 93/13. [Avail- able from New Zealand Ministry of Fisheries, ASB Bank House, 101-103 The Terrace, P.O. Box 1020, Wel- lington, NZ.] Somerton, D. A., P. T. Munro, and K. L.Weinberg. 2007. Whole-gear efficiency of a benthic survey trawl for flatfish. Fish. Bull. 105:278-291. Stanley, R. D., M. K. McAllister, P. Starr, and N. Olsen. 2009. Stock assessment for bocaccio ( Sebastes paucispi- nis ) in British Columbia waters. Can. Sci. Adv. Sec. Res. Doc. 2009/055, 214 p. [Available from http://www. meds-sdmm.dfo-mpo.gc.ca/csas/applications/Publica- tions/publicationlndex_e.asp] . Uusitalo, L., S. Kuikka, and A. Romakkaniemi. 2005. Estimation of Atlantic salmon smolt carrying capacity of rivers using expert knowledge. ICES J. Mar. Sci. 62:708-722. 305 Abstract — Assessing the vulner- ability of stocks to fishing practices in U.S. federal waters was recently highlighted by the National Marine Fisheries Service (NMFS), National Oceanic and Atmospheric Adminis- tration, as an important factor to consider when 1) identifying stocks that should be managed and protected under a fishery management plan; 2) grouping data-poor stocks into rel- evant management complexes; and 3) developing precautionary harvest control rules. To assist the regional fishery management councils in deter- mining vulnerability, NMFS elected to use a modified version of a pro- ductivity and susceptibility analy- sis (PSA) because it can be based on qualitative data, has a history of use in other fisheries, and is recom- mended by several organizations as a reasonable approach for evaluating risk. A number of productivity and susceptibility attributes for a stock are used in a PSA and from these attributes, index scores and mea- sures of uncertainty are computed and graphically displayed. To dem- onstrate the utility of the resulting vulnerability evaluation, we evalu- ated six U.S. fisheries targeting 162 stocks that exhibited varying degrees of productivity and susceptibility, and for which data quality varied. Overall, the PSA was capable of differentiat- ing the vulnerability of stocks along the gradient of susceptibility and productivity indices, although fixed thresholds separating low-, moderate-, and highly vulnerable species were not observed. The PSA can be used as a flexible tool that can incorporate regional-specific information on fish- ery and management activity. Manuscript submitted 12 August 2009. Manuscript accepted 22 April 2010. Fish. Bull. 108:305-322 (2010). The views and opinions expressed or implied in this article are those of the author (or authors) and do not necessarily reflect the position of the National Marine Fisheries Service, NOAA. Using productivity and susceptibility indices to assess the vulnerability of United States fish stocks to overfishing Wesley S. Patrick (contact author)1 Paul Spencer2 Jason Link3 Jason Cope4 John Field5 Donald Kobayashi6 Peter Lawson7 Todd Gedamke8 Ernie Cortes9 OJaw Ormseth Keith Bigelow6 William Overholtz3 Email address for contact author: Wesley.Patrick@noaa.gov 6 Pacific Islands Fisheries Science Center National Marine Fisheries Service National Oceanographic and Atmospheric Administration 2570 Dole Street Honolulu, Hawaii 96822 1 Office of Sustainable Fisheries National Marine Fisheries Service National Oceanographic and Atmospheric Administration 1315 East-West Highway Silver Spring, Maryland 20910 2 Alaska Fisheries Science Center National Marine Fisheries Service National Oceanographic and Atmospheric Administration 7600 Sand Point Way Seattle, Washington 981 1 5 3 Northeast Fisheries Science Center National Marine Fisheries Service National Oceanographic and Atmospheric Administration 166 Water Street Woods Hole, Masssachusetts 02543 4 Northwest Fisheries Science Center National Marine Fisheries Service National Oceanographic and Atmospheric Administration 2725 Montlake Boulevard East Seattle, Washington 98112 5 Southwest Fisheries Science Center National Marine Fisheries Service National Oceanographic and Atmospheric Administration 110 Shaffer Road Santa Cruz, California 95060 7 Northwest Fisheries Science Center National Marine Fisheries Service National Oceanographic and Atmospheric Administration 2030 South Marine Science Drive Newport, Oregon 97365 8 Southeast Fisheries Science Center National Marine Fisheries Service National Oceanographic and Atmospheric Administration 75 Virginia Beach Drive Miami, Florida 33149 9 Southeast Fisheries Science Center National Marine Fisheries Service National Oceanographic and Atmospheric Administration 3500 Delwood Beach Road Panama City, Florida 32408 The need to ascertain the status of fish stocks is a common issue for fisheries management agencies the world over. Stock assessments are usually man- dated by various national or interna- tional laws and frequently include an evaluation of a stock’s current biomass and fishing mortality rate compared to some reference level, often maximum sustainable yield (MSY). Because of the data requirements for evaluating the status of stocks, however, a large proportion of the world’s fishery man- agers and scientists lack the ability to adequately assess the status of their stocks (Mora et al. 2009). In the past, many of these data-poor stocks have been managed by using a “harvest control rule” that was based on the overfishing limit for, and biomass of, the stock. However, with little knowl- edge of a stock’s status it is difficult to appropriately apply precautionary management (Restrepo and Powers, 306 Fishery Bulletin 108(3) 1999; Katsukawa, 2004). Today, however, many man- agers and scientists are turning to risk assessments to try to better manage stocks for which there are directed measures of stock status (e.g., Lane and Stephenson, 1998; Peterman, 2004; Fletcher et al., 2005; Astles et al„ 2006). Risk assessments for data-poor stocks usually follow some type of semiquantitative method. In previous ex- amples of semiquantitative risk assessments, scientists have evaluated fishery impacts on bycatch and targeted species (Francis, 1992; Lane and Stephenson, 1998; Stobutzki et al., 2001a,), extinction risk (Musick, 1999; Roberts and Hawkins, 1999; Cheung et al., 2005; Mace et al., 2008), and impacts on ecosystem viability (Jen- nings et al., 1999; Fletcher et al., 2005; Astles et al., 2006). These approaches allow for the inclusion of less quantitative information and a wide range of factors and can complement both stock and ecosystem assessments. In the United States, scientists of the National Ma- rine Fisheries Service (NMFS), National Oceanic and Atmospheric Administration, recently developed a risk assessment to assist managers and scientists in evalu- ating the vulnerability of stocks to overfishing ( Patrick et al., 2009). Vulnerability is a measurement of a stock’s productivity and its susceptibility to a fishery. Pro- ductivity refers to the capacity of the stock to recover rapidly when depleted, whereas susceptibility is the potential for the stock to be impacted by the fishery. In general, vulnerability is an important factor to consider when organizing stock complexes, developing buffers between target and limit fishing mortality reference points, and determining which stocks should be man- aged under a fishery management plan. This article de- scribes the method developed by scientists at NMFS for determining vulnerability, explores the various caveats and nuances in its underlying calculations, and presents an overview of its application to six U.S. fisheries. Materials and methods Determining vulnerability of stocks Several risk assessment methods were reviewed to deter- mine which approach would be flexible and broadly applicable across fisheries and regions. A modified ver- sion of a productivity and susceptibility analysis (PSA) was selected as the best approach for examining the vulnerability of stocks, owing to its history of use in other fisheries (Milton, 2001; Stobutzki et al., 2001a, 2001b; Braccini et al., 2006; Griffiths et al., 2006; Zhou and Griffiths, 2008) and owing to recommendations by several organizations and working groups as a reason- able approach for determining risk (Hobday et al.1,2; Rosenberg et al.3; Smith et al., 2007). 1 Hobday, A. J., A. Smith, and I. Stobutzki. 2004. Ecological risk assessment for Australian Commonwealth fisheries, 172 p. Report R01/0934 for the Australian Fisheries Manage- ment Authority, Canberra, Australia. The PSA was originally developed to classify differ- ences in bycatch sustainability in the Australian prawn fishery (Milton, 2001; Stobutzki et al., 2001b) by evalu- ating the productivity (p) of bycatch stocks and their susceptibility (s) to the fishery. The values for p and s were determined by providing a score ranging from 1 to 3 for a standardized set of attributes related to each index (i.e., 7 productivity and 6 susceptibility attri- butes). When data were lacking, scores could be based on similar taxa or given the most vulnerable score as a precautionary approach. The scores were then aver- aged for each index and displayed graphically on an x-y scatter plot (Fig. 1). The two-dimensional nature of the PSA leads directly to the calculation of an overall vul- nerability score ( v ) of a species, defined as the Euclidean distance of productivity and susceptibility scores: v = ^[(P-Z0)2+(S-y0)2], (1) where x0 and y0 are the ( x , y ) origin coordinates, respec- tively. Stocks that received a low productivity score and a high susceptibility score are considered to be the most vulnerable to overfishing, whereas stocks with a high productivity score and low susceptibility score are con- sidered to be the least vulnerable. Since 2001, the PSA has been modified by others to evaluate habitat, community, and management compo- nents of a fishery (Hobday et al.2; Rosenburg et al.3). In general, these modifications have included expanding the number of attributes for scoring, exploring additive and multiplicative models for combining scores, and ex- amining a variety of alternative treatments for missing data. In the next section we review our application of a PSA to provide a uniform framework for evaluating the wide variety of fish stocks managed within the United States. Identifying productivity and susceptibility attributes With the expansion of the PSA to evaluate other man- agement factors (e.g., habitat impacts, ecosystem consid- erations, management efficacy), the number of attributes that could be considered in a PSA has increased con- siderably— in some instances to approximately sev- enty-five (Hobday et al.2; Rosenberg et al.3). Although ~75 attributes have been recommended, Hobday et al.2 noted that the use of more than six attributes per index 2 Hobday, A. J., A. Smith, H. Webb, R. Daley, S. Wayte, C. Bulman, J. Dowdney, A. Williams, M. Sporcic, J. Dambacher, M. Fuller, T. Walker. 2007. Ecological risk assessment for the effects of fishing: methodology, 174 p. Report R04/1072 for the Australian Fisheries Management Authority, Can- berra, Australia. 3 Rosenberg, A., D. Agnew, E. Babcock, A. Cooper, C. Mogensen, R. O’Boyle, J. Powers, G. Stefansson, and J. Swasey. 2007. Setting annual catch limits for U.S. fisher- ies: An expert working group report, 36 p. MRAG Americas, Washington, D.C. Patrick et al.: Use of productivity and susceptibility indices to assess vulnerability of fish stocks to overfishing 307 (e.g., productivity, susceptibility, habitat) does little to improve the accuracy of an assessment. Development of our PSA began with an initial examination and reduc- tion of these 75 attributes to 35 after removing those perceived as redundant or not directly related to our definition of vulnerability. The remaining attributes were evaluated in two phases. In phase 1, our team provided individual scores (i.e., “yes,” “no,” or “maybe”) to determine whether each attribute was 1) appropriate for calculating productivity or susceptibility of a stock; 2) useful at different scales (i.e., for stocks of various sizes and spatial distributions); and 3) capable of being calculated for most fisheries (i.e., for data availability). Attributes receiving a majority of “yes” scores for all three questions were retained. In phase 2, attributes receiving mixed scores, as well as new attributes not previously identified, were evaluated in a group dis- cussion. Through this process, 18 (9 productivity, 9 susceptibility) of the 35 attributes were selected and four new attributes were added, including 1) recruit- ment pattern; 2) management strategy; 3) fishing rate in relation to natural mortality; and 4) desirability or value of the fishery. Overall, 22 attributes were selected for the analysis (10 productivity, 12 susceptibility). The large set of attributes to be scored, compared to previous versions of the PSA, is largely a result of the susceptibil- ity index, including both catchability and management attributes (see Susceptibility attributes section below). We also recognized that the PSA would mainly be used to evaluate extremely data-poor stocks; thus, a larger set of attributes would be useful to ensure that an adequate number of attributes were scored. Productivity attributes Many of the productivity attributes are based on Musick’s (1999) qualitative extinction risk assessment and the PSA of Stobutzki et al. (2001b). However, the scoring thresholds have been modified in many cases to better suit the distribution of life history characteris- tics observed in U.S. fish stocks (Table 1). Information on maximum length, maximum age, age-at-maturity, natural mortality, and von Bertalanffy growth coeffi- cient were available for more than 140 stocks considered to be representative of U.S. fisheries (see Patrick et al., 2009). For these attributes, a range of scoring categories was evaluated by using analysis of variance (ANOVA) and post hoc tests to identify attribute scoring thresh- olds that produced significantly different bins of data. To ensure consistency in these attributes, the optimal scoring thresholds from the ANOVA were also compared to published relationships among maximum age and natural mortality (Alverson and Carney, 1975; Hoenig, 1983), von Bertalanffy growth coefficient (Froese and Binohlan, 2000), and age at maturity (Froese and Bino- hlan, 2000). Overall, we found this approach produced sensible categories compared to the approach of inde- pendently dividing each attribute into equal bins or using a quantile method. We defined the following 10 productivity attributes. 3.0 2.5 s 2.0 1.5 ★ jJ% 1*T a *** -m v A .. ,v' ^Ir - %&&& , 4k. A 4 Jr ■ AS 1 e# ST □ * 3.0 2.5 (High) 2.0 Productivity score 1.0 (Low) Alaska shark complex ?BSAI skate complex CA nearshore groundfish CA current pelagics NE groundfish HI swordfish longline HI tuna longline SA GOM longline 3.0 2.5 1 2.0 1.5 1.0 B o ,,r~: a : V • A A A l- ub A m, -i. vy,'~U' at a. A- Jk - Hg; . V 3.0 2.5 (High) 2.0 Productivity score 1.5 1.0 ( Low ) Data quality good ,x Data quality moderate Figure 1 A Data quality poor (A) Overall distribution of productivity and susceptibility x-y plot for the 166 stocks evaluated in this study, dif- ferentiated by fishery. BSAI = Bering Sea and Aleutian Islands. SA GOM= South Atlantic and Gulf of Mexico. (B) Associated data quality of each datum point of the 166 stocks evaluated in this study (see Appendix 1 for a list of the species in these fisheries). Intrinsic growth rate (r) This is the intrinsic rate of population growth or maximum population growth that would occur in the absence of fishing at the lowest population size (Gedamke et al., 2007). Density-depen- 308 Fishery Bulletin 108(3) Patrick et al.: Use of productivity and susceptibility indices to assess vulnerability of fish stocks to overfishing 309 dent compensation is at a maximum in these depleted conditions and therefore r is a direct measure of stock productivity. The scoring thresholds were taken from Musick (1999), who stated that r should take precedence over other productivity attributes because it combines many of the other attributes defined below. Maximum age (tmax) Maximum age is related to natu- ral mortality rate (M), where M is inversely related to maximum age (Hoenig, 1983). The scoring thresholds were based on the ANOVA applied to the observed fish stocks considered to be representative of U.S. fisheries (see Patrick et al., 2009). The tmax for a majority of these fish ranges between 10 and 30 years. Maximum size (LmaJ Maximum size is also correlated with productivity, and large fish tend to have lower levels of productivity (Roberts and Hawkins, 1999), although this relationship varies phylogenetically and is strongest within higher taxonomic levels (e.g., genus, family). The scoring thresholds were based on the ANOVA applied to the observed fish stocks considered to be representative of U.S. fisheries (see Patrick et al., 2009). The Lmax for a majority of these fish ranges between 60 and 150 cm total length (TL). Growth coefficient (k) The von Bertalanffy growth coefficient measures how rapidly a fish reaches its maxi- mum size. Long-lived, low-productivity stocks tend to have low values of k (Froese and Binohlan, 2000). The scoring thresholds of 0.15 and 0.25 were based on the ANOVA applied to the observed fish stocks considered to be representative of U.S. fisheries (see Patrick et al., 2009). This observed range of k is roughly consistent with the values obtained from Froese and Binohlan’s (2000) empirical relationship k = 3/ tmax of 0.1 and 0.3, based upon tmax values of 10 and 30. Natural mortality (M) Natural mortality rate directly reflects population productivity because stocks with high rates of natural mortality will require high levels of production to maintain population levels. For several methods of estimating M, one must rely on the negative relationship between M and tmax, including Hoenig’s (1983) regression based upon empirical data, the quan- tile method that depends upon exponential mortality rates (Hoenig, 1983), and Alverson and Carney’s (1975) relationship between mortality, growth, and tmax. The scoring thresholds from the ANOVA applied to the fish stocks considered to be representative of U.S. fisheries were 0.2 and 0.4, roughly consistent with those produced from Hoenig’s (1983) empirical regression of 0.14 and 0.4, based on tmax values of 10 and 30. Fecundity Fecundity (i.e. , the number of eggs produced by a female for a given spawning event or period) varies with size and age of the spawner; therefore we followed Musick’s (1999) recommendation that fecundity should be measured at the age of first maturity. As Musick (1999) noted, low values of fecundity imply low popula- tion productivity, but high values of fecundity do not necessarily imply high population productivity; thus, this attribute may be more useful at the lower fecun- dity values. The scoring thresholds were taken from Musick (1999) and were fecundities values of 1,000 and 100,000. Breeding strategy The breeding strategy of a stock provides an indication of the level of mortality that may be expected for the offspring in the first stages of life. To estimate offspring mortality, we used Winemiller’s (1989) index of parental investment. The index ranges from 0 to 14 and is scored according to 1) the place- ment of larvae or zygotes (i.e., in a nest or in the water column; score ranges from 0 to 2); 2) the length of time of parental protection of zygotes or larvae (score ranges from 0 to 4); and 3) the length of gestation period or nutritional contribution (score ranges from 0 to 8). To translate Winemiller’s index into our ranking system, we examined King and McFarlane’s (2003) parental investment scores for 42 North Pacific stocks. These 42 stocks covered a wide range of life histories and habitats, including 10 surface pelagic, three mid-water pelagic, three deep-water pelagic, 18 near-shore ben- thic, and nine offshore benthic stocks. Thirty-one per- cent of the stocks had a Winemiller score of zero, and 40 percent had a Winemiller score of 4 or higher; therefore 0 and 4 were used as the scoring thresholds. Recruitment pattern Stocks with sporadic and infre- quent recruitment success often are long lived and thus might be expected to have lower levels of productivity (Musick, 1999). This attribute is intended as a coarse index to distinguish stocks with sporadic recruitment patterns and high frequency of year-class failures from those with relatively steady recruitment. Thus, the pro- portion of years in which recruitment was above aver- age (e.g., the percentage of successful year classes over a 10-year period) was used for this attribute. Because this attribute was viewed as a coarse index, we chose 10% and 75%- as the scoring thresholds, so that scores of 1 and 3 allowed us to identify relatively extreme dif- ferences in recruitment patterns. Age-at-maturity (fmof) Age at maturity tends to be strongly related to both maximum age (tmax) and natu- ral mortality (M), where long-lived, lower-productivity stocks will have higher ages at maturity than short-lived stocks (Beverton, 1992). The scoring thresholds from the ANOVA applied to the fish stocks considered to be representative of U.S. fisheries were ages 2 and 4. These values are lower than those obtained from Froese and Binohlan’s (2000) empirical relationship between tmat and tmax, which were ages 3 and 9 based upon values of tm ax of 10 and 30. However, Froese and Binohlan (2000) used data from many fish stocks around the world, which may not be representative of U.S. stocks. For the PSA, thresholds that were obtained from the ANOVA were applied to stocks considered representative of U.S. fisheries. 310 Fishery Bulletin 108(3) Mean trophic ievel The position of a stock within the larger fish community can be used to infer stock pro- ductivity; lower-trophic-level stocks are generally more productive than higher-trophic-level stocks. The trophic level of a stock can be computed as a function of the tro- phic levels of the organisms in its diet. For this attribute, stocks with trophic levels higher than 3.5 were catego- rized as low-productivity stocks and stocks with trophic levels less than 2.5 were categorized as high-productiv- ity stocks, and moderate-productivity stocks would fall between these bounds. These scoring thresholds roughly categorize piscivores to higher trophic levels, omnivores to intermediate trophic levels, and planktivores to lower trophic levels (Pauly et al., 1998) and carry the assump- tion that the food web analysis did not consider microbial loops as an individual trophic level. Susceptibility attributes Previous applications have been focused on the catch- ability and mortality of stocks, and other attributes, such as management effectiveness and effects of fish- ing gear on habitat quality, have been addressed in subsequent analyses (Hobday et al.2). Our susceptibil- ity index includes all these attributes in an effort to make the results of our analysis more transparent and understandable. We defined 12 susceptibility attributes; the first seven relate to catchability and the other five measure management factors. Like the susceptibility attributes of Hobday et al.2, catchability attributes provide information on the likeli- hood of a stock’s capture by a particular fishery, given the stock’s range, habitat preferences, behavioral re- sponses, and morphological characteristics that may affect its susceptibility to the fishing gear deployed in that fishery. For management attributes, one must con- sider how the fishery is managed: for example, fisheries with conservative management measures in place that effectively control the amount of catch are less likely to overfish. For some of these attributes, the criteria are somewhat general in order to accommodate the wide range of fisheries and management systems. Areal overlap This attribute pertains to the extent of geographic overlap between the known distribution of a stock and the distribution of the fishery. Greater over- lap implies greater susceptibility, because some degree of geographical overlap is necessary for a fishery to impact a stock. The simplest approach to determining areal overlap is to evaluate, either qualitatively or quantitatively, the proportion of the spatial distribu- tion of a given stock that overlaps that of the fishery, based on known geographical distributions of both. Geographic concentration Geographic concentration is the extent to which the stock is concentrated into small areas. We included this attribute because a stock with a relatively even distribution across its range may be less susceptible than a highly aggregated stock. For some species, a useful measure of this attribute is the proportion of an area of interest occupied by a specified percentage of the stock (Swain and Sinclair, 1994), which can be computed if survey data exist (see Patrick et al., 2009). For many stocks, this measure gives a general index of areal coverage that relates well to geographic concentration. However, some stocks can be concentrated in a small number of locations throughout a survey area (i.e., a “patchy” stock that is distributed over the survey area). Thus, some refinements to the index may be necessary to characterize geographic concentration in these cases. Vertical overlap Like geographic overlap, this attribute concerns the position of the stock within the water column (e.g., demersal or pelagic) in relation to the fishing gear. Information on the depth at which gear is deployed (e.g., depth range of hooks for a pelagic longline fishery) and the depth preference of the species (e.g., obtained from archival tagging or other sources) can be used to estimate the degree of vertical overlap between fishing gear and a stock. Seasonal migrations Seasonal migrations either to or from the fishery area (i.e., spawning or feeding migra- tions) could affect the overlap between the stock and the fishery. This attribute also pertains to cases where the location of the fishery changes seasonally, and therefore may be relevant for stocks captured as bycatch. Schooling, aggregation, and other behaviors This attri- bute encompasses behavioral responses of both individ- ual fish and the stock in response to fishing. Individual responses may include, for example, herding or gear- avoidance behavior that would affect catchability. An example of a population-level response is a reduction in the area of stock distribution with reduction in popula- tion size, potentially leading to increases in catchability (MacCall, 1990). Morphological characteristics affecting capture This attribute pertains to the ability of the fishing gear to capture fish according to their morphological character- istics (e.g., body shape, spiny versus soft rayed fins). On a population level, this attribute refers to gear selectivity as it varies with fish size and age. Scoring this attribute, one should take into consideration what portion of the population size or age composition is accessible to the fishing gear or gears in question. Particular attention should be paid to the size or age at maturity in relation to capture. Desirability or value of the fishery For this attribute, one assumes that highly valued fish stocks are more susceptible to overfishing or becoming overfished by recreational or commercial fishermen because of increased fishing effort. To identify the value of the fish, we used the price per pound or annual landings value for commercial stocks (using the higher of the two values; see Table 2) or the retention rates for rec- reational fisheries. Patrick et al.: Use of productivity and susceptibility indices to assess vulnerability of fish stocks to overfishing 311 Management strategy The susceptibility of a stock to overfishing may largely depend on the effectiveness of fishery management procedures used to control catch (Roughgarden and Smith, 1996; Sethi et al., 2005; Dankel et al., 2008). Stocks managed by using catch limits that allow for fishery closure before the catch limit is exceeded (i.e., in-season or proactive accountability measures) are considered to have a low susceptibility to overfishing. Stocks managed by using catch limits and reactive accountability measures (e.g., catch levels determined after the fishing season) are considered to be moderately susceptible to overfishing or to becoming overfished. Lastly, stocks that have neither catch limits nor accountability measures are considered to be highly susceptible to overfishing. Fishing mortality rate (in relation to M) This attribute is applicable to stocks for which estimates of both fish- ing and natural mortality rates ( F and M) are available. Because sustainable fisheries management typically involves conserving the reproductive potential of a stock, it is recommended that the average F on mature fish be used where possible, as opposed to the fully selected or “peak” F. We base our thresholds on the conservative rule of thumb that the M should be an upper limit of F (Thompson, 1993), and thus F/M should not exceed 1. For this attribute, we define intermediate F/M values as those between 0.5 and 1.0; values above 1.0 and below 0.5 are defined as high and low susceptibility, respectively. Biomass of spawners Analogous to fishing mortality rate, a comparison of the current stock biomass (BCUR rent/ expected unfished levels (B0) offers information on the extent to which fishing has potentially depleted the stock and the stock’s realized susceptibility to over- fishing. If B0 is not available, one could compare BCUR rent against the maximum observed biomass from a time series of population size estimates (e.g., from a research survey). If a time series is used, it should be of adequate length, and it should be recognized that the maximum observed survey estimates may not correspond to the true maximum biomass and that substantial observation errors in estimates may be present. Additionally, stocks may decline in abundance because of environmental fac- tors unrelated to their susceptibility to the fishery, and therefore this situation should be considered by scientists when evaluating depletion estimates. Notwithstanding these issues, which can be addressed with the data quality score described below, some measure of cur- rent stock abundance was viewed as a useful attribute. Survival after capture and release Fish survival after capture and release varies by species, region, depth, gear type, and even market conditions, and thus can affect the susceptibility of the stock (Davis, 2002). Consider- ations of barotraumatic effects, discarding methods, and gear invasiveness (e.g., gears with hooks or nets would likely be more invasive than traps) are particularly relevant. Fishery impact on habitat A fishery may have an indi- rect effect on a species through adverse impacts on habi- tat (Benaka, 1999; Barnes and Thomas, 2005). Within the United States, a definition of the level of impact is the focus of environmental impact statements and essential fish habitat evaluations (see Rosenberg et al., 2000). To align with NMFS evaluations of impact, the scoring thresholds for this attribute were categorized as minimal, temporary, or mitigated. Defining attribute scores and weights Depending on the specific stock being evaluated, not all of the productivity and susceptibility attributes listed in Tables 1 and 2 will be equally useful in determining the vulnerability of a stock. In previous versions of the PSA, an attribute weighting scheme was used in which higher weights were applied to the more important attributes (Stobutzki et al., 2001b; Hobday et al.1; Rosenberg et al.3). We used a default weight of 2 for the productivity and susceptibility attributes, where attribute weights can be adjusted within a scale from 0 to 4 to customize the application to each fishery. In determining the proper weighting of each attribute, users should consider the relevance of the attribute for describing productivity or susceptibility rather than the availability of data for that attribute (e.g., data-poor attributes should not auto- matically receive low weightings). In some rare cases, it is also anticipated that some attributes will receive a weighting of zero, which cause them to be removed from the analysis, because the attribute has no rela- tion to the fishery and its stocks. Some attributes (e.g., management strategy, fishing mortality rate, biomass of spawners, etc.) may also be removed from the analysis to avoid double-counting if they are considered in a more overarching risk analysis, for which the results of the PSA are only one component. Like Milton (2001) and Stobutzki et al. (2001b), we defined the criteria for a score of 1, 2, or 3 to a produc- tivity or susceptibility attribute (see Table 1). However, our approach provides users the flexibility to apply in- termediate scores (e.g., 1.5 or 2.5) when the attribute value spans two categories. Owing to the subjective nature of semiquantitative analyses, scores should be applied in a consistent manner to reduce scoring bias (Lichtensten and Newman, 1967; Janis, 1983; Von Win- terfeldt and Edwards, 1986; Bell et al., 1988), such as by employing the Delphi method (see Okoli and Paw- lowski, 2004 and Landeta, 2006). Data-quality index As a precautionary measure for risk assessment scor- ing, the highest-level risk score can be used when data are missing to account for uncertainty and to avoid identifying a high-risk stock as low risk (Hardwood, 2000; Milton, 2001; Stobutzki et al., 2001b; Astles et al., 2006). Although precautionary, that approach also confounds the issues of data quality with risk assess- ment. For example, a data-poor stock may receive Fishery Bulletin 108(3) 312 Table 2 Susceptibility attributes and rankings used to determine the vulnerability of a stock becoming overfished Susceptibility attribute Definition Areal overlap The extent of geographic overlap between the known distribution of a stock and the distribution of the fishery. Geographic concentration The extent to which the stock is concentrated into small areas. Vertical overlap The position of the stock within the water column (i.e., whether is demersal or pelagic) in relation to the fishing gear. Seasonal migrations Seasonal migrations (i.e. spawning or feeding migrations) either to or from the fishery area could affect the overlap between the stock and the fishery. Schooling, aggregation, and other behavioral responses Behavioral responses of both individual fish and the stock in response to fishing. Morphological characteristics affecting capture The ability of the fishing gear to capture fish based on their morphological characteristics (e.g., body shape, spiny versus soft rayed fins, etc.). Desirability or value of the fishery The assumption that highly valued fish stocks are more susceptible to overfishing or to becoming overfished by recreational or commercial fishermen owing to increased effort. Management strategy The susceptibility of a stock to overfishing may largely depend on the effectiveness of fishery management procedures used to control catch. Fishing rate relative to M As a conservative rule of thumb, it is recommended that M should be the upper limit of F so as to conserve the reproductive potential of a stock. Biomass of spawners (SSB) or other proxies The extent to which fishing has depleted the biomass of a stock in relation to expected unfished levels offers information on realized susceptibility. Survival after capture and release Fish survival after capture and release varies by species, region, and gear type or even market conditions, and thus can affect the susceptibility of the stock. Impact of fisheries on essential fish habitat or habitat in general for nontargeted fish A fishery may have an indirect effect on a species by adverse impacts on habitat. a high-risk evaluation either from an abundance of (Table 3). The data-quality score is computed for the missing data or from the risk assessment of the avail- productivity and susceptibility scores as a weighted able data, with the result that the risk scores may be average and implies the overall quality of the data or inflated (Hobday et al.1). In contrast, we considered belief in the score rather than the actual type of data missing data within the larger context of data quality, used in the analysis. Like Hobday et al.2, we divided and report the overall quality of data as a separate the data-quality scores into three groupings (poor >3.5; value. moderate 2. 0-3. 5; and good <2.0) for display purposes. A data-quality index was developed to represent the This information, along with more detailed descrip- information quality of individual vulnerability scores tions of data quality (e.g., mean score, range), is a based on five tiers, ranging from best data (or high be- quick and useful means of providing decision-makers lief in the score) to no data (or little belief in the score) with details on the uncertainty of the vulnerability Patrick et al.: Use of productivity and susceptibility indices to assess vulnerability of fish stocks to overfishing 313 Ranking Low (1) Moderate (2) High (3) <25% of stock present in the area fished. Between 25% and 50% of the stock present in the area fished. >50% of stock present in the area fished. Stock is distributed in >50% of its total range Stock is distributed in 25% to 50% of its total range Stock is distributed in <25% of its total range. <25% of stock present in the depths fished. Between 25% and 50% of the stock present in the depths fished. >50% of stock present in the depths fished Seasonal migrations decrease overlap with the fishery. Seasonal migrations do not substantially affect the overlap with the fishery. Seasonal migrations increase overlap with the fishery. Behavioral responses of fish de- crease the catchability of the gear. Behavioral responses of fish do not substantially affect the catchability of the gear. Behavioral responses of fish increase the catchability of the gear (i.e., hyperstability of catch per unit of effort with schooling behavior). Species shows low susceptibility to gear selectivity. Species shows moderate susceptiblity to gear selectivity. Species shows high susceptiblity to gear selectivity. Stock is not highly valued or desired by the fishery (<$ 1/lb; <$500K/yr landed; <33% retention). Stock is moderately valued or desired by the fishery ($ l-$2. 25/lb; $500K-$ 10, OOOK/yr landed; 33—66% retention). Stock is highly valued or desired by the fishery (>$2. 25/lb; >$10, OOOK/yr landed; >66% retention). Targeted stocks have catch limits and proactive accountability mea- sures; nontarget stocks are closely monitored. Targeted stocks have catch limits and reactive accountability measures. Targeted stocks do not have catch limits or accountability measures; nontargeted stocks are not closely monitored. <0.5 0. 5-1.0 >1 B is >40% of B0 (or maximum observed from time series of bio- mass estimates). B is between 25% and 40% of B0 (or maximum observed from time series of biomass estimates). B is <25% of B0 (or maximum observed from time series of biomass estimates). Probability of survival >67% 33% < probability of survival <67% Probability of survival <33% Adverse effects absent, minimal or temporary. Adverse effects more than minimal or temporary but are mitigated. Adverse effects more than minimal or temporary and are not mitigated. scores. Such uncertainty in the data would help with the interpretation of the overall vulnerability score and also help in targeting areas of further research and data needs. Example case studies To demonstrate the utility of our PSA scoring process, we evaluated six U.S. fisheries including the Northeast groundfish multispecies, highly migratory Atlantic shark complexes, California nearshore groundfish fin- fish assemblage, California Current coastal pelagic species, skates of the Bering Sea and Aleutian Islands (BSAI) management area (a bycatch fishery of the BSAI groundfish fishery), and the Hawaii-based pelagic long- line fishery (both the tuna and swordfish sectors). In total, 162 stocks were evaluated (Appendix 1). These fisheries were chosen because they were expected to display varying degrees of productivity, susceptibility, and data quality. For descriptions of these fisheries and details on how our PSA scoring procedure was applied to each fishery, see Patrick et al. (2009). 314 Fishery Bulletin 108(3) Table 3 The five tiers of data quality used when evaluating the productivity and susceptibility of an individual stock. Data quality tier Description Example 1 Best data. Information is based on collected data for the stock and area of interest that is established and substantial Data-rich stock assessment; published literature for which multiple methods are used, etc. 2 Adequate data Information is based on limited coverage and corroboration, or for some other reason is deemed not as reliable as tier-1 data Limited temporal or spatial data, relatively old information, etc. 3 Limited data. Estimates with high variation and limited confidence and may be based on studies of similar taxa or life history strategies Similar genus or family, etc. 4 Very limited data. Information based on expert opinion or on general literature reviews from a wide range of species, or outside of region General data not referenced 5 No data. When there are no data on which to make even an expert opinion, the person using the PSA should give this attribute a “data quality” score of 5 and not provide a “productivity” or “susceptibility” score so as not to bias those index scores. When plotted, the susceptibility or productivity index score will be based on one less attribute, and will be highlighted as such by its related quality score. 3.0 2.5 = 2.0 / A* V \ <## • ^ O o v- \ / / ' N / / /\ / y >- * JA /, 1.0 - 3.0 (High) 2.5 2.0 1.5 1.0 Productivity Score (Low) i I Blue .shark <\y1 Longfin ltiako shark t ) Conation thresher shark Bitteye thresher shark A Oceanic whitetip shark O Silky shark Figure 2 Comparison of vulnerabilities among common shark species in the highly migratory Atlantic shark com- plexes (gray), Hawaii-based pelagic longline — tuna sector (white), and Hawaii-based pelagic longline — swordfish sector (black). Results and discussion Range of vulnerability scores The managed stocks evaluated in this report represent both targeted (n=71; 44%) and nontargeted species ( /r = 9 1 ; 56%) that were included in fishery management plans to prevent overfishing and rebuild overfished stocks. The stocks generally displayed vulnerability scores greater than 1.0 (Fig. 1). Species evaluated within the Atlantic highly migratory shark complexes were found to be the most vulnerable, averaging vulnerability scores of 2.17, and California Current coastal pelagic species were the least vulnerable, averaging 1.29. Although different groups of species will exhibit dif- ferent ranges of productivity and susceptibility scores, it is interesting to note that in some cases even the same species may exhibit different productivity scores. For example, the productivity scores for the blue ( Prio - nace glauca), bigeye thresher ( Alopias superciliosus), longfin mako (Isurus paucus), oceanic whitetip (Car- charhinus longimanus ), silky (C. falciformis), and com- mon thresher (A. vulpinus) sharks differed between the highly migratory Atlantic shark complexes and the Hawaii-based pelagic longline fishery example applica- tions (Fig. 2). These differences are likely related to intraspecific variations in life history patterns (Cope, 2006) and to the use of different weightings in the vul- nerability analysis (see Patrick et ah, 2009). In contrast, the species in the Hawaii-based pelagic longline fishery (both the tuna and swordfish sectors) showed an expanded range of productivity and suscep- Patrick et al Use of productivity and susceptibility indices to assess vulnerability of fish stocks to overfishing 315 Table 4 Summary of the productivity and susceptibility scoring frequencies and correlations to the overall index or category score. Cor- relations were based on stock attributes scores (1—3) (see Tables 1 and 2) that were compared to a modified categorical score for the stock, the latter of which did not include the related attribute score. Frequency Pearson correlation Category No. scored scored coefficient P-value Productivity r Maximum age Maximum size von Bertalanffy growth coefficient ( k ) Estimated natural mortality (M) Measured fecundity Breeding strategy Recruitment pattern Age at maturity Mean trophic level Susceptibility Catchability Areal overlap Geographic concentration Vertical overlap Seasonal migrations Schooling, aggregation, and other behavioral responses Morphology affecting capture Desirability or value of the fishery Management Management strategy Fishing rate in relation to M Biomass of spawners (SSB) or other proxies Survival after capture and release Fishery impact to essential fish habitat (EFH) or habitat in general for nontargeted fish 128 96% 0.596 <0.001 126 95% 0.674 <0.001 128 96% 0.592 <0.001 129 97% 0.656 <0.001 127 95% 0.785 <0.001 126 95% 0.509 <0.001 133 100% 0.568 <0.001 84 63% -0.211 0.054 125 94% 0.802 <0.001 132 99% 0.439 <0.001 123 92% 0.333 <0.001 133 100% 0.345 <0.001 133 100% 0.772 <0.001 49 37% 0.058 0.692 87 65% 0.340 0.001 132 99% 0.319 <0.001 133 100% 0.504 <0.001 133 100% 0.154 0.077 79 59% 0.510 <0.001 78 59% 0.389 <0.001 126 95% 0.201 0.024 133 100% 0.286 0.001 tibility scores. The swordfish sector overall exhibited a slightly reduced susceptibility when compared to the tuna sector, probably due to the higher level of tar- geting in the tuna sector of the fishery (Fig. 1). The restricted range in some of the example applications may reflect the species chosen for these examples, and a more expanded range may be observed if the PSA were applied to all species in a fishery management plan (FMP). For example, BSAI skate complexes are managed as bycatch within the BSAI Groundfish FMP, which includes a range of life-history types, including rockfish and flatfish, and the productivity and suscepti- bility scores for these species would likely contrast with those obtained for skates. A restricted range of scores from a PSA may moti- vate some to modify the attribute scoring thresholds to produce greater contrast. But because the overall goal of the present PSA is to estimate vulnerability in relation to an overall standard appropriate for the range of managed species, a lack of contrast in vulner- ability scores may simply reflect a limited breadth of species diversity. It may be advantageous in some cases to modify the attribute scoring thresholds to increase the contrast within a given region or FMP (see Field et al., in press), while recognizing that the vulnerability scores for that particular fishery no longer represent the risk of overfishing based on the original scoring criteria described here. Data availability and data quality From our example applications, data availability was relatively high for the majority of the attributes evalu- ated, averaging 88% and ranging from 37% to 100% in scoring frequency (Table 4). However, the quality of these data was considered moderate (i.e., medium data quality scores of 2—3), except for the Northeast multi- species groundfish fishery (Fig. 1). The high degree of data quality for those targeted stocks reflects the relatively long time series of fishery and survey data. In general, a relationship between susceptibility and data quality is intuitive (i.e., valuable stocks are likely 316 Fishery Bulletin 108(3) the most susceptible owing to targeting, and priority is therefore given to the collection of data for valuable target fisheries). The degree of consistency within the productivity and susceptibility scores was determined from correlations of a particular attribute to its overall productivity or susceptibility score (after removal of the attribute being evaluated). In this analysis, susceptibility attributes related to management were separated from other sus- ceptibility attributes. All but two of the attributes had relatively high correlation coefficients, with an overall average correlation of 0.43 and ranging from -0.21 to 0.80 (Table 4). The correlation coefficients for recruit- ment pattern (-0.21) and seasonal migration (0.06) were unusually low and could reflect the narrow range of ob- served recruitment patterns or seasonal migrations, as is evident from each attribute being scored 90% of the time as a moderate risk. Although these attributes were not informative for the majority of the stocks we exam- ined, we anticipate that these attributes may prove to be more useful for other fisheries. As previously noted, in these cases the attribute weight can be adjusted to reflect its utility. □ Allantic shark complex ★ Nf siroundfish A CA nearshore groundfish © Ml swordfish longline ( A current pelagics HI tuna longlinc Figure 3 A subset of the stocks from the example applications (n=50) for which the status (stock is either overfished ^ current^ msy^ or being overfished \BCURRENT60 mm SL. The smallest mature females are 41.5 mm SL, and by 55.1 mm SL virtually all are mature. Juveniles are found with adults on the outer shelf. Only 214 C. gymnorhinus were located in 42 different field collections (35-201 m, with 90% between 61 and 120 m) off the east coast of the United States, Bahamas, and eastern Caribbean Sea. Adults are found as far north as the shelf off Cape Hatteras, NC (35°N, 75°W). This diminutive species (to 52.4 mm SL) is among the smallest flatfishes but males (n = 131; 20.3—52.4 mm SL) attain a slightly larger maxi- mum size than that of females (« = 58; 26.2- 48.0 mm SL). Males begin to mature between 29 and 35 mm SL and reach 100% maturity by 35-40 mm SL. Some females are mature at 29 mm SL, and all females >35.1 mm SL are mature. Overlooked specimens in museum collections and literature enabled us to correct long-standing inaccuracies in northern distribu- tional limits that appear in contem- porary literature and electronic data bases for these species. Associated locality-data for these specimens allow for proper evaluation of distri- butional information for these species in relation to hypotheses regarding shifts in species ranges due to climate change effects. Manuscript submitted 27 May 2009. Manuscript accepted 26 April 2010. Fish. Bull. 108:323-345 (2010). The views and opinions expressed or implied in this article are those of the author (or authors) and do not necessarily reflect the position of the National Marine Fisheries Service, NOAA. Thomas A. Munroe (contact author)1 Steve W. Ross2 Email address for contact author: munroet@si.edu 1 National Systematics Laboratory/NMFS/NEFSC Smithsonian Institution Post Office Box 37012 NHB, WC 57, MRC-153 Washington, DC 20013-7012 2 University of North Carolina at Wilmington Center for Marine Science 5600 Marvin Moss Lane Wilmington, North Carolina 28409 Citharichthys cornutus (Gunther, 1880), the horned whiff, and C. gymno- rhinus Gutherz and Blackman, 1970, the anglefin whiff, are small-size, poorly known flatfishes inhabiting substrata located on the middle and outer continental shelves — primarily in subtropical and tropical waters of the western Atlantic Ocean (Gutherz, 1967; Gutherz and Blackman, 1970; Topp and Hoff, 1972; Figueiredo and Menezes, 2000; 2003). Larvae of both species have been collected off Nova Scotia, Canada, but these were consid- ered strays from more southern locali- ties (Scott and Scott, 1988). These two species are morphologically similar and, although broadly sympatric in the western North Atlantic Ocean, are seldom taken together in the same collections. Their relatively small size (maximum to 55 mm standard length [SL] in C. gymnorhinus and 91 mm SL in C. cornutus), lack of commercial importance, infrequency of capture, and the probable low abundance of both species have resulted in their largely being ignored. Published, detailed life history in- formation for C. cornutus is sparse. Consequently, our knowledge of the biology, ecology, and geographic distri- bution of this species is limited to rel- atively few observations on maximum size (Parr, 1931; Longley and Hildeb- rand, 1941; Gutherz, 1967), size at maturity (Parr, 1931 [in part]; Long- ley and Hildebrand, 1941), reports of hermaphrodites (Gutherz, 1969), and general descriptions of geographic and bathymetric distributions (Parr, 1931 [in part]; Longley and Hildeb- rand, 1941; Gutherz, 1967; Topp and Hoff, 1972). Citharichthys cornutus is found from temperate regions of the North Atlantic Ocean off New Jersey (Goode, 1880; reported as Ci- tharichthys unicornis, now considered a junior synonym of C. cornutus, see Norman, 1934; Fowler, 1952; Steves et al., 1999) and Hudson Submarine Canyon (Fahay, 2007) to subtropical waters off southern Brazil (Gunther, 1880) and Uruguay (Figueiredo and Menezes, 2000; 2003). From the lim- ited data, it appears that C. cornu- tus reaches sizes to about 89 mm SL (Gutherz, 1967), and although this species is known from depths rang- ing from 20 to 408 m (Gutherz, 1967; Topp and Hoff, 1972), it has been cap- tured most frequently between 130 and 370 m (Gutherz, 1967). Most published sources commonly cited for distributional information on C. cornutus are inaccurate. Data in Gutherz (1967) are the basis for the geographic range most frequently 324 Fishery Bulletin 108(3) reported for adult C. cornutus (Robins and Ray, 1986; Munroe, 2003; McEachran and Fechhelm, 2005). Gutherz (1967) described the range as the continen- tal shelf from off Georgia throughout warm-temperate and tropical regions of the western Atlantic to Brazil. Several earlier reports (Goode, 1880; Goode and Bean, 1895; Fowler, 1952), however, had already documented occurrences of C. cornutus (as C. unicornis) from more northern localities including those as far north as the outer continental shelf off New Jersey (see also Fahay, 2007). These earlier reports of C. unicornis ( =C . cornu- tus) from more northern locales were overlooked in most recent studies where information has been compiled for this species (with exception of Fahay, 2007). More information is available on the life history and distribution of C. gymnorhinus, which reportedly reach- es a maximum size of about 55 mm SL (Gutherz and Blackman, 1970; Topp and Hoff, 1972). Citharichthys gymnorhinus is one of the smallest species of the genus and is also among the smallest of flatfishes (Munroe, 2005). It is found on the mid- to outer continental shelf at depths of 35-201 m, but has been collected most fre- quently between 30 and 90 m (Gutherz and Blackman, 1970; Topp and Hoff, 1972; Walsh et al., 2006). Cithar- ichthys gymnorhinus inhabits subtropical and tropical regions of the western North Atlantic Ocean (Gutherz and Blackman, 1970; Topp and Hoff, 1972) from North Carolina (Quattrini and Ross, 2006) to Guyana (Topp and Hoff, 1972). Although not rare, C. gymnorhinus has been captured less frequently than C. cornutus, and seldom has it been taken in abundance, especially on the continental shelf off the eastern United States. No summaries of biological information exist for C. gymnorhinus found off the east coast of the United States, and information from this region is restricted to limited geographic and bathymetric data based on few specimens — the data appearing in tables and appendices of various reports (see below). Most biological and ecological data for this species, including observations on habitat, depth of occurrence, size, size at maturity, and geographical distribution, are based on 47 specimens collected from the Florida Keys to Guyana, and the majority of these specimens were taken on the west Florida shelf during cruises of the RV Hourglass (Topp and Hoff, 1972). Re- cent summaries of information on adult C. gymnorhinus are based almost entirely on data originally presented in Gutherz and Blackman (1970) and Topp and Hoff (1972) and indicate a northernmost geographical limit for adults, either as the Bahamas (Robins and Ray, 1986; McEachran and Fechhelm, 2005; Lyczkowski- Shultz and Bond, 2006), Florida (Fahay, 2007), the Florida Keys (Robins and Ray, 1986; McEachran and Fechhelm, 2005), or perhaps the continental shelf as far north as off North Carolina (Munroe, 2003). As- sessment of distributional information in these recent reviews indicated that the northernmost limits in the geographic range reported for C. gymnorhinus were inaccurate. Earlier published records of adult C. gym- norhinus from waters north of the Florida Keys and Bahamas, including those from off Georgia (Tucker, 1982) and South Carolina (Wenner et al., 1979a), were overlooked in these recent summaries. Recently, Walsh et al. (2006) again collected juveniles and adults of this species off Georgia, and Quattrini and Ross (2006) reported catching adults on the continental shelf off North Carolina, thereby documenting the northernmost latitude known for adult C. gymnorhinus. Although the captures in Quattrini and Ross (2006) are the first pub- lished records for C. gymnorhinus as far north as North Carolina, our examination of museum lots uncovered a specimen taken off North Carolina during the first part of the 20th century (see below). The objectives of this study are to update and aug- ment biological and distributional information for these two diminutive flatfish species. Data were gleaned from three sources: 1) specimens in fish collections, including specimens for which some information may have already appeared in published and gray literature, and includ- ing some specimens that we re-identified; 2) specimens in fish collections not previously reported; and 3) from recently collected specimens of both species captured off the southeastern United States. Additional informa- tion from other specimens reported in the literature, although not examined by us but where identifications were deemed reliable, is also included in the data sum- maries. The cumulative contributions of information from the above sources allowed us to note obscure dis- tributional records for both species and also to compile more accurate summaries of life history, and ecological and distributional information for these flatfishes. In summarizing such data, we were able to correct long- standing inaccuracies in the reported distributions of these species and to evaluate this new distributional information in relation to contemporary hypotheses regarding shifts in ranges of continental shelf fishes due to effects of climate change. Materials and methods This study was initiated with the collection of both C. gymnorhinus and C. cornutus from the continental shelf off North Carolina (Quattrini and Ross, 2006; Ross, unpubl. data). Recognizing that captures of both species off North Carolina represent significant contributions to our knowledge of the distribution and ecology of these species, we initiated a complete review of available information on these fishes. Pertinent literature was examined to identify and validate published records of both C. gymnorhinus and C. cornutus (including records for type specimens of C. unicornis Goode, a junior sub- jective synonym of C. cornutus', see Norman, 1934) from localities off the eastern United States and adjacent areas in the Bahamas and northeastern Caribbean Sea. Major fish collections likely to have holdings of these species from this region were also surveyed and speci- mens were examined (Appendices 1 and 2), or data were taken from internet databases where identifications were deemed reliable. Details for institutional fish collections Munroe and Ross: Distribution and life history of Citharichthys gymnorhinus and C. cornutus 325 Figure 1 Adult Citharichthys cornutus from the western North Atlantic Ocean. (A) Male; North Carolina State Museum (NCSM) 47664; 65.1 mm standard length (SL); featuring secondary sexual dimorphisms in cephalic spination and wide interorbital space. ( B ) Female (NCSM 47664; 62.1 mm SL) with no cephalic spination and narrower interorbital space. Note also the elongate, pigmented ovary. designated by acronyms in this study can be found at http://www.asih.org/codons.pdf. Additionally, specimens of both species were identified from materials collected during recent NMFS-Northeast Fishery Science Center (NEFSC) groundfish surveys, and data associated with these specimens were also included in this study. All benthic specimens examined or noted through lit- erature and museum searches were collected primarily by various types of bottom trawl, and a few specimens were also taken in benthic dredges. Descriptive geo- graphic locations of collections taken on the continental shelf along the southeastern United States designated in Appendices 1 and 2 and elsewhere were based on their latitudinal positions in relation to terrestrial state boundaries, which may not necessarily coincide with the state boundaries on the continental shelf. Fishes examined were identified, enumerated, and measured to the nearest mm SL, unless otherwise noted. Species were identified according to characters outlined in Gutherz and Blackman (1970). Citharichthys cornutus (Fig. 1, A and B) is distinguished from C. gymnorhinus (Fig. 2, A and B) in having scales on the snout (absent in C. gymnorhinus), 6 ocular-side pelvic fin rays (vs. 5), 40 or more lateral-line scales (vs. <40 scales in lateral line), and a dark spot in the axil of the ocular-side pectoral fin (vs. no dark spot in axil of pectoral fin). Male C. cornutus do not have large black spots in the middle of their dorsal and anal fins that are characteristic of male C. gymnorhinus (compare Figs. 1A and 2A), and male C. cornutus also have a much larger interorbital space compared with that of male C. gymnorhinus. Sex and maturity of individuals were determined (where possible) by examining external sexually di- morphic characters and by macroscopic examination of gonads with light transmitted through the abdominal 326 Fishery Bulletin 108(3) Figure 2 Adult Citharichthys gymnorhinus from the western North Atlantic Ocean. (A) Male; North Carolina State Museum (NCSM) 47657; 49.5 mm standard length (SL); featuring secondary sexual dimorphisms in cephalic spina- tion, wider interorbital space and black blotches on dorsal and anal fins. (B) Female (NCSM 47662; 39.3 mm SL) lacking cephalic spination or black spots on dorsal and anal fins, and with narrower interorbital space. Note also the elongate, pigmented ovary. region of the body. Adult C. gymnorhinus (Gutherz and Blackman, 1970; Topp and Hoff, 1972; this study) and C. cornutus (Parr, 1931; Gutherz and Blackman, 1970; this study) feature distinct sexual dimorphisms that facilitate macroscopic determination of sex of matur- ing and mature individuals of both sexes. Female C. cornutus lack dimorphic sexual features characteristic of male C. cornutus (compare Fig. 1, A and B), includ- ing the absence of rostral and cephalic spines, in hav- ing a much narrower interorbital space (usually 42 mm SL. In contrast, female C. gymnorhinus (Fig. 2B) lack cephalic spination, black pigmented blotches on dorsal and anal fins, and the dusky blind-side pigmentation characteristic of males. Also, with the onset of maturity, the ovaries undergo a conspicuous extensive posterior elongation (easily ob- served with light transmitted through the body), which, Munroe and Ross: Distribution and life history of Citharichthys gymnorhinus and C. cornutus 327 as also occurs in C. cornutus , may extend for more than one-half of the standard length of the individual. Im- mature females of both species are identified by their shorter, more triangularly shaped ovaries that have either not yet begun to elongate, or that are only in the earliest stages of posterior elongation. Results and discussion Biological and ecological information from 594 C. cor- nutus (Appendix 1) and 214 C. gymnorhinus (Appendix 2) were included in this study. Of the 594 C. cornutus, size data were available for 566 individuals. Sizes were not available or were not taken for 28 of the C. cornutus ; however, geographic and bathymetric data associated with these individuals were included in appropriate sum- maries. Size information was taken from 196 of the 214 C. gymnorhinus included in the study. One individual (a mature male) had a regenerated caudal region, and therefore it could not be measured accurately, and size data from this specimen were excluded. Depth of capture information was known for 578 of the 594 C. cornutus and for all 214 of the C. gymnorhinus included in the study. Sex and maturity information was compiled from 430 of 566 C. cornutus that were measured. Sex could not be determined macroscopically for C. cornutus smaller than 20.0 mm. For 135 C. cornutus , although size in- formation was available, no information was provided to determine the sex of these fishes. One specimen was damaged during collection and its sex could not be determined. Sex could not be determined for another individual of 68 mm because the gonad appeared to be undeveloped (macroscopic appearance was neither that of a typical testes or ovary) and although this individ- ual is well within the size range for adults (see below), it does not feature any external sexually dimorphic characters typical of adult males, which also precluded macroscopic determination of its sex. Sex and maturity were determined for 190 of 214 C. gymnorhinus. We did not examine 25 C. gymnorhinus, including seven for which size data were available (8-38 mm); thus no information on their sex or maturity was available. Citharichthys cornutus (Gunther 1880) Taxonomic note Gunther (1880) described Rhomboidichthys cornutus from specimens taken on the continental shelf in the western South Atlantic off Brazil. Later that year, Goode (1880) provided a description of Citharichthys unicornis based on three specimens (USNM 26003: 3 syntypes) collected on the outer continental shelf off southern New England (actually at a latitude off New Jersey; see Fowler, 1952). Norman (1934) first considered these two nominal species to be conspecific and this decision has been followed by subsequent authors, but the status of these nominal species is in need of further study. During the interim between Goode’s (1880) description of C. unicornis and Norman’s (1934) placement of this species in the synonymy of C. cornutus, several stud- ies were published in which this species was listed or in which ecological, distributional, and systematic information was reported under the name C. unicornis. These earliest reports of C. cornutus from the western North Atlantic (Jordan and Gilbert, 1883; Gunther, 1887; Jordan and Goss, 1889; Goode and Bean, 1895; Jordan and Evermann, 1898; Evermann and Marsh, 1902; Parr, 1931) were overlooked in nearly all contem- porary compilations of information on the species, in part because C. unicornis Goode was not recognized as a junior synonym for C. cornutus (Gunther). The series of specimens reported in Goode and Bean (1895) comprises at least two species, including C. uni- cornis (-cornutus) and C. gymnorhinus (see below). Parr’s study (1931) may also have contained a mixture of both species, but because he did not list any specific collection data for the 68 specimens from the USNM and MCZ identified as C. unicornis in his study, the possibility that two species were intermingled cannot be proven definitively. The majority of specimens of C. unicornis available to Parr from the USNM fish collec- tion were those collected earlier during cruises of the RV Albatross and reported in Goode and Bean (1895). Several of these lots comprise mostly specimens of C. gymnorhinus ; therefore, if these were the same speci- mens examined by Parr, it seems likely that his data set was compromised because it would have contained a mixture of at least two species. Geographic distribution Fifty-six different field collections (Appendix 1) con- taining juvenile and adult specimens of C. cornutus encompassed the geographic range from the outer con- tinental shelf off New Jersey to the continental shelves off the northern coasts of the Bahamas, Cuba, Puerto Rico, and the Lesser Antilles (Fig. 3). All but five collec- tions occurred south of 35°N latitude. Among collections examined were four from off New Jersey (we could not find specimens from the second location off NJ reported by Goode, 1880), seven from off North Carolina, 22 off South Carolina, three off Georgia, nine off Florida, five from the Bahamas, one off Cuba, two off Puerto Rico, and three others from the eastern Caribbean Sea. One early record (USNM 111520) purportedly of this species from off North Carolina was misidentified as Citharich- thys unicornis (=C. cornutus) by Hildebrand (1941). This 56-mm total length (TL) male, taken on 13 September 1914, off Cape Lookout by the RV Fish Hawk at a depth of about 92 m (50 fm) is actually C. gymnorhinus, and the significance of this specimen is discussed under the account for that species. Our knowledge concerning the geographic occurrence of C. cornutus in the western North Atlantic Ocean is now improved through the inclusion of previously published distributional data, by highlighting data for specimens previously listed only in published tables 328 Fishery Bulletin 108(3) Figure 3 Capture locations of specimens examined, or, of specimens cited in literature for which there are reliable identifications, for Citharichthys cornutus (?;=594) and C. gymnorhinus (n= 214). Points may indicate more than one fish and more than a single capture. Capture records for other specimens of C. cornutus from the Dry Tortugas Islands Region based on Longley and Hildebrand (1941) are not plotted because precise locality data were not provided for these specimens. haps Yucatan, Mexico (Castro-Aguirre et al., 1999), the Bahamas, the Greater Antilles (Puerto Rico and Cuba), and Lesser Antilles (off Virgin Islands, St. Kitts Island), on the outer continental shelf throughout the Caribbean Sea, and off the Atlantic coast of South America to Uruguay. Larvae are known from Canadian Atlantic waters; several individuals have been collected or appendices, and by inclusion of unpublished infor- mation associated with museum specimens and other uncatalogued specimens (Appendix 1). Based on these sources, the range of adult C. cornutus extends from the outer continental shelf off New Jersey (40°05.593'N, 70°42.75'W) and New York southward along the outer shelf of the eastern United States to Texas, and per- Munroe and Ross: Distribution and life history of Citharichthys gymnorhmus and C. cornutus 329 at 41°33'N, 54°55'W (MCZ 77935) and 41°05'N, 66°31'W (Scott and Scott, 1988; Fahay, 2007). The geographical distribution reported here is largely consistent with that known for the species as reported in Fahay (2007), but it is different from range informa- tion contained in literature accounts published during the last half of the 20th century. Goode (1880) first reported C. cornutus (as C. unicornis) from the western North Atlantic from two stations located on the outer continental shelf south of Rhode Island (40°02'54"N, 70°23'40"W and 40°02'36"N, 70°22'58"W). This capture site, actually off New Jersey (Fowler, 1952), is located on the shelf between the head of Alvin Submarine Canyon (40°00'N, 70°30'W) and a feature called “The Mud Patch” located just to the northwest of the canyon. The same distributional information for the species (deep waters of the Gulf Stream off Rhode Island) was reported by Jordan and Gilbert (1883) and Jordan and Goss (1889) from the data presented in the original description of the species by Goode (1880). Gunther (1887) also listed this locality for C. unicornis on the basis of Goode’s specimens. Fowler (1952) again listed C. cornutus among fishes of New Jersey on the basis of “offshore records” that most likely refer to specimens contained in Goode’s (1880) original description of C. unicornis, because no other specimens were listed to indicate otherwise. In their treatise on deep-sea fishes, Goode and Bean (1895) again repeated the same dis- tributional information for the syntypes of C. unicornis captured on the outer shelf off northern New Jersey, but they also provided data for additional specimens taken off South Carolina (33°18'N, 77°07'W) and in the Gulf of Mexico (28'36-38'N, 85°52-53'W). These new specimens increased the knowledge about the geo- graphic distribution of this species along the south- east continental shelf of the United States. Although Goode and Bean also reported C. unicornis from the Straits of Florida off the Florida Keys (24°25'45"N, 81'46'W), re-examination of these specimens from RV Albatross station 2318 (USNM 45610; USNM 45677; USNM 143120) and RV Albatross station 2316 (USNM 129946) revealed they are not C. cornutus, but rather are C. gymnorhinus (see data summaries below for that species). Scott and Scott (1988) considered their record of a larval C. cornutus, taken at 41°05'N, 66°31'W in 1982, to represent the northernmost point of the range of the species. They also considered that larvae of this spe- cies reported from the vicinity of the Canadian Atlan- tic are most likely strays from more southern locales. They missed earlier references documenting adults from the outer shelf region at 40°N and cited Georgia as the northern limit for the species following informa- tion in Tucker (1982). Other larval C. cornutus known from this general area (MCZ fish collection), including Georges Bank (Fahay, 2007), indicate the possibility that larvae caught in this region could also be produced by more localized spawning of C. cornutus. FishBase (Froese and Pauly, 2010; www.fishbase.org) also shows the western North Atlantic distribution of this species as Canada to Georgia based on the larva reported in Scott and Scott (1988). Additional records of this species from off the south- eastern coast of the United States and nearby areas include those in an unpublished dissertation by Staiger (1970), who detailed distribution of this species in the Straits of Florida region based on 143 specimens. He reported that C. cornutus occurs along the continental margin of the Straits from off the Dry Tortugas to Miami, also in the central Straits near the Cay Sal Bank, and along the insular margin of the Straits from the Santaren Channel to the Little Bahama Bank. Quattrini and Ross (2006) caught this species on the continental shelf off North Carolina but provided no further comments on the distribution of the species (because their specimens are included in the present study). Eight specimens of C. cornutus have also been taken at two locations off Puerto Rico (Evermann and Marsh, 1902; this study) and off the northern coast of Cuba near Provincia de Matanzas (Vergara Rodriguez, 1974). Cervigon (1996) recorded the species from off Venezuela but indicated that it is poorly documented from this area. Other studies published during the last century con- tained more general, and often vague, information or commentary on the distribution of this species. For example, Jordan and Evermann (1898) listed it as oc- curring in deep waters of the Gulf Stream. Parr (1931) did not report any specific capture information for the approximately 68 specimens (largest series examined to that date) in his study. Instead, he mentioned that although very little was known about the distribution of C. cornutus, it was generally regarded as occurring in deep waters of the Gulf Stream. Norman (1934) consid- ered C. cornutus to have a disjunct distribution with a northern range in deep waters of the Gulf Stream and a southern distribution off the coast of Brazil. Longley and Hildebrand (1941) reported the species as occurring in the Gulf Stream from the Dry Tortugas to at least off the southeast coast of New England (presumably from earlier literature records off New Jersey). Topp and Hoff (1972) listed this as a deepwater species in the Gulf of Mexico. Lyczkowski- Shultz and Bond (2006) described the range of C. cornutus as the Atlantic and Gulf coasts of the United States, without specifying any geographic limits. In a summary of distributional in- formation for C. cornutus, Gutherz (1967) reported that its geographic range included the outer continental shelf along the Atlantic and Gulf coasts of the United States from Georgia to Texas, the Bahamas, the Greater An- tilles, off Yucatan, Mexico, throughout the Caribbean, and off the Atlantic coast of South America to Brazil. Although Gutherz (1967) adequately summarized cap- ture locations for the species in the middle portions (extreme southeastern United States, Gulf of Mexico, and Caribbean Sea) of its geographic range, he did not mention previous captures of the species from north of Georgia. Beginning with Gutherz (1967), captures of adult C. cornutus from off New Jersey (Goode, 1880; and cited 330 Fishery Bulletin 108(3) in other studies) and South Carolina (Goode and Bean, 1895; Wenner et ah, 1979a; 1979c; 1979d; 1980) were overlooked in nearly all subsequent contemporary litera- ture, and misinformation regarding the northern limits of distribution for C. cornutus was perpetuated. For example, Hoese and Moore (1977) reported the distribu- tion of C. cornutus as the northwestern Gulf of Mexico, and from Georgia throughout the Caribbean to Brazil, later (Hoese and Moore, 1998) adding the Bahamas to this distribution. Tucker (1982: Table 1) also listed the geographic range for adults as Georgia to Brazil (al- though he illustrated in a map that larvae were known from areas north of Georgia at about Cape Fear, NC). Robins and Ray (1986) and Boschung (1992) repeated the same distributional information for the northern point of the geographic range (Georgia) for C. cornutus as that appearing in Gutherz (1967). Cervigon (1996) listed the geographic distribution as the eastern United States, Bahamas, and northern Gulf of Mexico to Bra- zil. Castro-Aguirre et al. (1999) tentatively listed the species from the Veracruz, Mexico, region on the basis of a study by Lozano-Vilano et al. (1993), and reported the geographic distribution of the species from Georgia, Florida, and Gulf of Mexico to Brazil, including the Bahamas and Antilles. Figueiredo and Menezes (2000; 2003) also listed the northern limit of the geographic range of C. cornutus as Georgia, as did Saavedra-Diaz et al. (2000). Munroe (2003) reported the distribution as the continental shelf off the Atlantic and Gulf coasts of the United States from North Carolina to Texas, which distribution is essentially the same as that given in McEachran and Fechhelm (2005). Historical captures of C. cornutus from off New Jersey were also overlooked (J. A. Moore and K. E. Hartel, personal commun.1) by Moore et al. (2003) and Hartel et al. (2008) in check- lists of the deepwater (>200 m) resident fishes from the Mid-Atlantic Bight area south of New England. Among contemporary literature, only Fahay (2007) has provided a more accurate assessment of the north- ernmost occurrences of C. cornutus. Although he notes that the geographic range usually reported in the lit- erature is from Georgia to Brazil, including the Gulf of Mexico and Caribbean Sea, he also has observed that adults are fairly common as far north as Cape Hatteras and Hudson Canyon and that larvae are collected as far north as Georges Bank. A recent capture of one specimen of C. cornutus from off New York is reported by Steves et al. (1999: their Table 3) and a photograph provided by M. Fahay (personal commun.2) indicates that at least five other specimens were taken recently on the continental shelf off New Jersey. 1 Moore, Jon A., and Karsten E. Hartel. 2008. Honors Col- lege, Florida Atlantic University, Jupiter, FL 33458 and Museum of Comparative Zoology, Harvard Univ., 26 Oxford Street, Cambridge, MA 02138. 2 Fahay, Michael. 2006. (Retired.) James J. Howard Marine Sciences Laboratory at Sandy Hook, Northeast Fisheries Science Center, National Marine Fisheries Service, National Oceanic and Atmospheric Administration, 74 Magruder Road, Highlands, NJ 07732. Bathymetric distribution Overall, the 578 specimens for which information on depth of capture was available were captured at depths ranging from 68 to 287 m (Appendix 1). The majority of specimens (496 of 578 = 85%) were taken between 101-200 m (Fig. 4A). Only 31 (5.3%) C. cornutus in this study were taken shallower than 100 m, and only 61 (10.5%) specimens were collected deeper than 200 m (Fig. 4A). Specimens examined in this study were collected within the bathymetric range generally reported for this species; however, our specimens did not represent the depth extremes reported for this species. This spe- cies is usually reported from outer shelf depths where these fishes are associated with soft bottoms, including sand-mud substrata (Staiger, 1970; McEachran and Fechhelm, 2005; Fahay, 2007). For example, Goode and Bean (1895) reported that their specimens were taken between about 83 and 285 m; however, their study included both C. cornutus and C. gymnorhinus. The depth range for specimens of C. cornutus in Goode and Bean is actually 174-285 m. Off the Dry Tortugas, Longley and Hildebrand (1941) reported the distribution of C. cornutus as benthic habitats in the Gulf Stream at depths of about 81 to somewhat less than 185 m, and that abundance was highest near 120 m. Staiger (1970) reported a depth range for the species in the Florida Straits region from 83 to 260 m. Gutherz (1967) listed the depth range for C. cornutus throughout its geographic range as 27-366 m and that captures gener- ally exceeded 137 m; this range was repeated in several publications (Tucker, 1982; McEachran and Fechhelm, 2005; and Fahay, 2007). Topp and Hoff (1972) reported that off West Florida this species is known from the outer shelf or shelf edge; those in the Gulf of Mexico can be found at depths exceeding 350 m (5 out of 38 of their specimens were collected at these depths), and in the Caribbean it is usually found deeper than 137 m. Boschung (1992) listed the depth range for C. cornutus off Alabama as 24-172 m. Robins and Ray (1986) indi- cated a depth range of 30-400 m, usually deeper than 140 m for the species. Cervigon (1996) reported that this species inhabits depths between 30 and 400 m, but generally less than 300 m. Hoese and Moore (1977; 1998) considered this to be a deepwater species ranging from about 28-368 m, generally >138 m. Saavedra- Diaz et al. (2000) reported a depth range of 24-400 m. Munroe (2003) and Lyczkowski- Shultz and Bond (2006) reported this species at depths of 20-370 m, but gener- ally deeper than 130 m. Off southern Brazil, Figueiredo and Menezes (2000) noted that C. cornutus is captured in fisheries conducted at depths of 20-192 m, and a maximum depth record of 365 m has been documented for this species in this region. Size Our specimens ranged in size from 17.2 to 81.3 mm SL (Fig. 5A). Overall, approximately 48% (272 of 566) of Munroe and Ross: Distribution and life history of Cithanchthys gymnorhinus and C. cornutus 331 A Cithanchthys cornutus B Cithanchthys gymnorhinus Depth Figure 4 Mean depth of capture (at 10-m intervals) for two species of Citharichthys captured off the east coast of the United States and nearby regions. (A) Citharichthys cornutus (n = 578). (B) Citharichthys gymnorhinus (n = 214). the specimens measured 60 mm or larger; but only 27 (4.7%) exceeded 70 mm. Males (259 of 430, ca. 60% of total fish for which sex was determined) ranged in size from 20.2 to 79.1 mm (Fig. 6A). Females (171 of 430 = ca. 40% of total fish for which sex was determined) attained similar sizes (28.0-81.3 mm) to those recorded for males (Fig. 6B). Despite attaining nearly the same maximum size, males overall were usually larger than females and slightly more than twice as many males (169 or 39.3% of 430 fish for which sex was determined) reached 60 mm or larger than did females (48 or 11% of total fish for which sex was determined). Size distributions in our samples are similar to those recorded for the species in other studies, but are slightly smaller than the maximum size (about 91 mm SL) re- ported for the species (Longley and Hildebrand, 1941; Gutherz, 1967; Cervigon, 1996). From a size range of 32-68 mm SL (n- 68), Parr (1931) concluded that C. cornutus is a small species reaching only about 70 mm SL. But, as noted above, his results were likely based on a mixture of both C. cornutus and C. gymnorhinus. Norman (1934) reported sizes of 47-87 mm SL for six males, and 58 mm SL for one female. For C. cornutus collected off the Dry Tortugas, the largest size obtained by Longley and Hildebrand (1941) was slightly larger than 90 mm (TL?). Gutherz (1967) illustrated a male measuring 89 mm SL, but later (Gutherz, 1969), re- ported a maximum length for C. cornutus of only about 75 mm SL. Staiger (1970:64) examined 85 males and 35 females from the Straits of Florida that ranged in size between 26 and 70 mm SL. Topp and Hoff (1972) measured 38 specimens between 47.5 and 74.1 mm SL. Cervigon (1996) listed a maximum size of 91 mm SL for the species. Saavedra-Diaz et al. (2000) reported sizes for seven specimens taken off Colombia ranging from 46.7 to 61.8 mm SL. In other studies (Robins and Ray, 1986; Munroe, 2003; McEachran and Fechhelm, 2005), a maximum size of 100 mm TL was recorded (apparent- ly rounded upward on the basis of other literature). Size at maturity Sex was determined for 430 of the 566 (ca. 76%) indi- viduals measured. Among males (Fig. 6A), 32 of 259 (about 12%), ranging from 20.0 to 71.4 mm, are imma- ture; whereas, 227 (87%), measuring 36.5-79.1 mm, are mature. All nine males < 35.0 mm are immature, and 3 of 4 males between 35 and 40 mm are also immature. 332 Fishery Bulletin 108(3) A Citharichthys cornutus 20.0 25.0 30.0 35.0 40.0 45.0 50.0 55.0 60.0 65.0 70.0 75.0 80.0 81.3 Standard length (mm) B Citharichthys gymnorhinus 75 n 6.0-10.0 10.1- 15.1- 20.1- 25.1- 30.1- 35.1- 40.1- 45.1- 50.1- 15.0 20.0 25.0 30.0 35.0 40.0 45.0 50.0 55.0 Standard length (mm) Figure 5 Size distribution (standard length) for two species of Citharichthys captured off the east coast of the United States and nearby regions. (A) Citharichthys cornutus (n=566). (B) Citharichthys gymnorhinus (n = 196). Between 40 and 45 mm, 7 of 10 males are mature, but in the next size category (45-50 mm) only 4 of 9 males are mature. Between 50 and 55 mm, 90% of the males are mature, and between 55 and 65 mm, 95-98% of the males (42 of 44 and 126 of 129, respectively) are mature. Nearly all males (n = 38 of 40) >65.1 mm are mature. Among the 171 females examined (Fig. 6B), 18 (28.0-55.1 mm) are immature, whereas mature fe- males (rc = 153) range in size between 41.5 and 81.3 mm. Among the smallest females, 14 of 15 (93%) rang- ing between 28.0 and 45.0 mm are immature. Between 45.1 and 50.0 mm, only 1 of 3 females are immature, and among females >50 mm, only three (51.1, 51.8, and 55.1 mm) are immature. Of mature females, 2 of 3 fe- males between 45.1 and 50.0 mm are mature, whereas between 50.1 and 55.0 mm about 93% (28 of 30) of the females are mature, and all but one female > 55.1 mm are mature. The largest collection of C. cornutus from off the east- ern seaboard of the United States (NCSM 47664) con- tains 314 fish ranging in size from 20.0 to 68.0 mm. The sample comprises 177 males (20-68 mm), 136 females (28-64 mm), and 1 individual (68 mm) of unknown sex. The sex ratio for this sample is 1.3 :1.0 males to females. Among males, 164 are mature (48-68 mm), and 13 oth- ers (20-49 mm) are immature. Most of the 136 females (n=121, 47-64 mm) are mature and only 15 (28-46 mm) are immature. Co-occurrence of juveniles and adult males and females in the same collection indicates that both sexes and both life stages occupy similar depths and probably occur in the same microhabitat(s). Before this study, little information was available regarding size at maturity for C. cornutus. Earlier lit- erature on C. cornutus emphasized the sexual dimor- phisms exhibited by this species and the relative sizes when these dimorphic features become evident. Parr (1931:17) provided the most detailed account on size, morphological features, and striking sexual dimorphism exhibited in this species (as C. unicornis) based on at least 37 males and 31 females ranging from 32 to 68 mm SL (note again that his specimens may have been a mixture of C. cornutus and C. gymnorhinus). Norman Munroe and Ross: Distribution and life history of Citharichthys gymnorhmus and C. cornutus 333 A Males 1 1 3 4 4 10 13 10 44 129 24 14 2 15.5- 20.1- 25.1- 30.1- 35.1- 40.1- 45.1- 50.1- 55.1- 60.1- 65.1- 70.1- 75.1- 20.0 25.0 30.0 35.0 40.0 45.0 50.0 55.0 60.0 65.0 70.0 75.0 79.1 Standard length (mm) Standard length (mm) Figure 6 Size (standard length) and maturity information for 430 Citharichthys cornutus captured off the east coast of the United States and nearby regions. (A) Males (n = 259). (B) Females (rc = 171). Number above each column is the number of indi- viduals examined in that size range. (1934:153) examined seven specimens of C. cornutus (in- cluding types of Rhomboidichthys cornutus and several USNM specimens), noted sexual dimorphisms of this species, and also commented that a more complete de- scription of these dimorphic differences was available in Parr (1931). Longley and Hildebrand (1941:43) provided descriptive information on sexual dimorphism, as did Gutherz (1969), who also reported the capture of two adult hermaphroditic specimens from off Nicaragua. Longley and Hildebrand (1941) further commented that the sexes were easily distinguished at sizes of 55 mm (TL?) and greater because the posterior elongation of the ovaries was easily observed through the body wall of females. We also found that all females but one taken off the east coast of the United States that are >55 mm are sexually mature. Abundance Overall, the majority of specimens examined were taken on the outer continental shelf of the South Atlantic Bight. Ten different collections of C. cornutus from this region contained solitary individuals (Appendix 1), and another nine collections contained two or three specimens. The largest collections of C. cornutus from the east coast of the United States are all from North Carolina and southwards, including one collection off North Carolina with 12 specimens, seven off South Carolina (two with 334 Fishery Bulletin 108(3) 12, one each with 15, 24, 26, and 30 individuals), and the single largest collection (314 individuals) of C. cornutus from this region (NCSM 47664) from off South Carolina (at 33°25.389'N, 77°02.234'W). Other notable collections are those of 16 specimens from the Dry Tortugas region, and another collection containing 24 specimens (USNM 282789) made off the Bahamas. Citharichthys cornutus is found in relatively high den- sities in the Dry Tortugas region (Longley and Hildeb- rand, 1941). Although it is unclear how many specimens they encountered, Longley and Hildebrand (1941:43) noted that C. cornutus was “rather common” in this area, with sometimes 50 or more individuals taken to- gether. Other significant collections of C. cornutus from the eastern Gulf of Mexico containing 43, 47, 63, 100, 103, 172, and 321 specimens are curated in the AMNH fish collection (see Additional material section). Citharichthys gymnorhinus Geographic distribution We identified 214 juvenile and adult C. gymnorhinus in a total of 42 different field collections from the continen- tal shelf off North Carolina to the Dominican Republic, Virgin Islands, and Puerto Rico (Fig. 3; Appendix 2). Of collections containing C. gymnorhinus, five are from the shelf off North Carolina, including the northern- most capture of the species made off Cape Hatteras (35°05.51'N). Nine were made on the continental shelf off South Carolina, and at least two collections were made off Georgia. Most (n= 17) collections of C. gymnorhinus are from the continental shelf off Florida, including the Straits of Florida and the Dry Tortugas regions. Citharichthys gymnorhinus not only was less frequently captured at insular regions in the eastern Caribbean Sea than off the southeastern United States, but it is also known from relatively few specimens per collection at these Caribbean locations (e.g., five collections from Virgin Islands yielded only 21 specimens). Only two collections (n = 3 specimens) from off the Bahamas were found, and we note single captures of this species from off the northern coasts of Puerto Rico and the Domini- can Republic. The northernmost records for juvenile and adult C. gymnorhinus are those from the continental shelf just south of Cape Hatteras, NC (NMFS Survey specimens; NCSM records; Quattrini and Ross, 2006). Larval C. gymnorhinus have been collected from sites much far- ther north than those reported for adults. For example, Scott and Scott (1988) reported that a 13-mm larva of C. gymnorhinus , collected in 1982 at 41°21'N, 66°14'W, represented the first record for this species from Cana- dian waters and also was the northernmost record of occurrence for the species. Other captures of larval C. gymnorhinus from this general region include at least three lots (one taken at 41°35'28"N, 66°24'75"W; the others at 33°24'07"N and 32°58'37"N) curated in the Atlantic Reference Center (ARC) fish collection and one lot curated in the MCZ collection (MCZ 77935) taken at 41°33'N, 54°55'W. Scott and Scott (1988) considered that larval C. gymnorhinus occurring in Canadian wa- ters were strays from more southern locations. Museum records at ARC and MCZ document captures of larval C. gymriorhinus at or just beyond 41'N latitude over the course of several years, indicating that larval C. gym- norhinus at this latitude may be found more frequently than previously recognized. The record of C. gymnorhi- nus from off New Jersey (Able, 1992) is also based on larvae. Absence of adults in areas north of North Caro- lina may indicate that suitable habitat or appropriate environmental conditions are not available for juvenile settlement or for adult survival in these areas. Earlier occurrences of larval C. gymnorhinus in the South Atlantic Bight were detailed by Tucker (1982), who reported the northern range for larval C. gymnorhi- nus at about Cape Fear, NC. More recently, Powell et al. (2000) and Grothues et al. (2002) also listed this species among the larval fish assemblages off Cape Hatteras, NC, and Fahay (2007) reported collecting larvae in the area north of Cape Hatteras from January to November, with peak occurrence from August through September. Powell et al. (2000) considered that larval C. gymno- rhinus in Onslow Bay, NC, were most likely produced either by adults spawning in outer-shelf waters (55-185 m) nearby or were larvae produced by fishes spawning south of their study area. Off Georgia, C. gymnorhinus is only a minor component of the larval fish assemblage in these waters (Marancik et al., 2005). Before the work of Quattrini and Ross (2006), stud- ies of fishes off North Carolina did not list adult or juvenile C. gymnorhinus among species occurring there. Although specimens cited in Quattrini and Ross (2006) represent the first published record of adult C. gymrio- rhinus taken on the continental shelf off North Caro- lina, these are not the first specimens known from the area. Examination of catalogued museum lots and un- catalogued specimens from NMFS-NEFSC groundfish surveys (USNM uncat., see Appendix 2) reveals that other specimens had been captured off North Carolina before the Quattrini and Ross (2006) study. The earli- est collection of C. gymnorhinus from off North Caro- lina that we can document is that of a 45.4-mm male (USNM 111520) taken 13 September 1914, off Cape Lookout by the RV Fish Hawk, at a depth of about 92 m. Collection of this specimen pre-dates recognition and formal description of the species by Gutherz and Black- man (1970). Originally, Hildebrand (1941) misidentified this specimen as Citharichthys unicornis ( -C . cornutus-, see Norman, 1934); however, meristic features and color pattern, including dark pigment blotches on its dorsal and anal fins, reveal that it is an adult (45.4 mm SL) male C. gymnorhinus. One of the earliest reports of this species off South Carolina is that of Wenner et al. (1979a), who recorded a single specimen of C. gymnorhinus taken at 86 m during the 1973 fall trawl survey. Other specimens taken off South Carolina during the 1970s are curat- ed in several fish collections (AMNH, GMBL, UF; see Appendix 2). In his article describing larval develop- Munroe and Ross: Distribution and life history of Citharichthys gymnorhinus and C. cornutus 335 ment of C. gymnorhinus, Tucker (1982) described the northern limit for adult C. gymnorhinus as off Georgia (however, his Table 1 noted the geographic range as Florida to Guyana). More recently, Walsh et al. (2006) have collected five small specimens of this species be- tween 35 and 48 m on the inner continental shelf off Georgia. Gutherz and Blackman (1970) documented oc- currence, based on 34 specimens, of C. gymnorhinus off the Florida Keys, the Antilles, off northern and western Bahamas, northern Hispaniola, northern Puerto Rico, Tobago, and the Caribbean Sea off Colombia, Panama, and Nicaragua. An additional record of C. gymnorhinus (as an undescribed Citharichthys species) was included by Starck (1968) from off Alligator Reef, Florida. Topp and Hoff (1972) recorded the species from several addi- tional locations, including the continental shelf off west Florida, and also at sites off northern Cuba, the Virgin Islands, Venezuela, and Guyana. Boschung (1992) listed the geographic range as the northern Gulf of Mexico, and Bahamas to the western Caribbean and northern South America. Recent literature and an on-line database synthe- sizing information on C. gymnorhinus have perpetu- ated misinformation concerning the northern limits of distribution for this species. Robins and Ray (1986), Cervigon (1996), Saavedra-Diaz et al. (2000), Munroe (2003), McEachran and Fechhelm (2005), Lyczkowski- Shultz and Bond (2006), Fahay (2007), and Froese and Pauly (2010) indicate a northernmost geographical limit for adult C. gymnorhinus as the Bahamas or the Florida Keys, which is essentially the same distribution reported in Gutherz and Blackman (1970) and Topp and Hoff (1972). Earlier records of adult C. gymnorhi- nus north of the Florida Keys and the Bahamas pub- lished after studies by Gutherz and Blackman (1970) and Topp and Hoff (1972) were overlooked owing to a lack of thorough investigation. Perhaps these oversights resulted because citations of C. gymnorhinus from this area are infrequent and scattered among species lists that are included only in tables or appendices of re- gional studies (e.g., Wenner et al., 1979a; 1979b; 1979c; 1980) or because some records are not vouchered by specimens and are difficult to verify. For other studies (i.e., Tucker, 1982), conflicting information reported within the same work regarding the distribution of this species is confusing. Bathymetric distribution Citharichthys gymnorhinus examined in the present study were collected over a depth range from 35 to 201 m (Appendix 2; Fig. 4B). Approximately 90% (193 of 214) of these fish were taken at depths averaging 61-120 m. Only 16 individuals were captured shallower than 60 m, with the shallowest depths recorded (Walsh et al., 2006) for five individuals taken between 35 and 48 m off Georgia (specimens not examined by us). Capture locations for 48 C. gymnorhinus collected at depths averaging 100 m or more occurred in a variety of areas including those off North Carolina and South Carolina, in the Straits of Florida, and at the Bahamas. Only five individuals (2.3% of the total) examined in the present study were collected at depths averaging deeper than 120 m, and except for single specimens of C. gym- norhinus taken at 130 m and 123-127 m off North Carolina and South Carolina, respectively, the other three specimens were taken at insular locations off the Bahamas (two between 166-193 and one at 201 m). In fact, of examined specimens taken at insular locations in this region (n=26), all but five were collected at or beyond 70 m, and usually much deeper (Appendix 2). The depth range summarized in the present study is similar to that reported previously for this species. For example, Gutherz and Blackman (1970) noted captures of C. gymnorhinus in depths of 37-92 m, and one speci- men (USNM 203602), their deepest, was collected from about 201 m off the Bahamas (27°23'N, 78°35'W) (this specimen also represents our deepest record). Topp and Hoff (1972) reported that off west Florida this species is found at moderate depths on the continental shelf; 23 of 35 of their specimens were found at their deepest station (73 m), and all of their specimens were found deeper than 55 m. Tucker (1982: Table 1) listed a depth range of 37-201 m for adult C. gymnorhinus based on data from Gutherz and Blackman (1970). Saavedra-Diaz et al. (2000) reported that this species is found between 20 and 40 m in Colombian waters and between 37 and 139 m elsewhere. In nearly all other recent studies where information has been compiled for this species (Robins and Ray, 1986; Cervigon, 1996; Munroe, 2003; McEachran and Fechhelm, 2005; Lyczkowski-Shultz and Bond, 2006; Fahay, 2007), C. gymnorhinus are reported to be found on the continental shelf to depths of 200 m but more commonly taken in the shallower portion of this depth range (usually between 30 and 90 m). Size The largest C. gymnorhinus measured in this study are males of 52.4 and 52.1 mm (Fig. 7A), which is close to the reported maximum size (about 55 mm SL) observed for the species. A total of 131 males ranged from 20.3 to 52.4 mm (Fig. 7A), whereas females (n = 58) were 26.2-48.0 mm (Fig. 7B). Most males (122 of 131, 93%) were 30-50 mm, and only four were larger than 50 mm. For females, most (42 of 58, 74%) were 35-45 mm, and only one (48.0 mm) exceeded 45 mm. The size ranges for C. gymnorhinus taken along the eastern United States and nearby regions are compara- ble to those reported for this species from other parts of its geographic range. For example, Gutherz and Black- man (1970), who reported that C. gymnorhinus is the smallest species of Citharichthys, stated that maximum size for specimens they examined did not exceed 60 mm SL, when actually the largest of 34 specimens measured in their study (male, 52.5 mm SL) was smaller than this stated maximum size. Ten males examined in their study ranged in size from 41.9 to 52.5 mm SL, whereas four females were 32.9-41.8 mm SL. The largest of 47 individuals examined by Topp and Hoff (1972) was a 336 Fishery Bulletin 108(3) A Males 4 1 10 36 55 21 4 20.1- 25.1- 30.1- 35.1- 40.1- 45.1- 50.1- 25.0 30.0 35.0 40.0 45.0 50.0 55.0 Standard length (mm) B Females "D > T3 ns o o c <1) O <5 CL 25.1-30.0 30.1-35.0 35.1-40.0 40.1-45.0 45.1-50.0 Standard length (mm) Figure 7 Size (standard length) and maturity information for 189 Citharichthys gymnorhinus captured off the east coast of the United States and nearby regions. (A) Males (n=131). (B) Females (n = 58). Number above each column equals the number of individuals examined in that size range. male of 53.7 mm SL, and they indicated a maximum size for C. gymnorhinus of less than 55 mm SL. Topp and Hoff (1972) did not provide a complete size break- down for all specimens they examined but did give size ranges for these specimens. Males (n- 25) ranged in size from 24.9 to 53.7 mm SL; females (n = 6) were 36.1-47.0 mm SL. Cervigon (1996) reported that 55 mm SL was the maximum size observed for C. gymnorhinus taken off Venezuela. Fahay (2007) also considered C. gymno- rhinus to be a dwarf species that seldom exceeds 55 mm SL in length. Other literature reporting sizes for C. gymnorhi- nus larger than the maximum size observed in our study and the studies listed immediately above based their maximum sizes on an estimated value of 60 mm SL following that listed in Gutherz and Blackman (1970). These reports include those of Munroe (2003), McEachran and Fechhelm (2005), Robins and Ray (1986), and also FishBase (Froese and Pauly, 2010), which report a maximum size for C. gymnorhinus of 75 mm TL. This estimate of total length, corresponding to approximately 60 mm SL, is an overestimate of actual observations of maximum sizes recorded for the species from throughout its range. Size at maturity In other literature (where sizes of C. gymnorhinus are larger than the maximum size observed in our study and in the studies listed immediately above), maximum Munroe and Ross: Distribution and life history of Citharichthys gymnorhinus and C. cornutus 337 length was estimated at 60 mm SL as listed in Gutherz and Blackman (1970). Of 131 males examined in the present study, all but nine are mature with obvious discernible dimorphic characters typical of sexually mature individuals (Fig. 2A; and discussed above). All four males <25.0 mm are immature. The only male (29.0 mm) in the next larger size class is the smallest mature male in the study (Fig. 7A). Between 30 and 35 mm, half (5 of 10) of the males examined are immature, whereas five others are mature and display external features characteristic of adult males. At 35.1 mm and larger, all males examined are mature. Of 58 female C. gymnorhinus examined, all but six (26.2, 26.5, 27.5, 27.6, 28.0, and 30.1 mm) are mature (Fig. 7B). The smallest mature female was 30 mm, and eight others smaller than 35 mm are also mature. At sizes >35.1 mm SL, all females are mature. No published information on maturity schedules or size at maturity based on microscopic staging of gonads is available for C. gymnorhinus. Information on size at maturity for this species is based only on examination of external sexual dimorphisms for males and exter- nal examination of ovaries of females. For example, Gutherz and Blackman (1970) noted that the smallest mature male in their study was 42 mm SL. Small size at maturity compared to that of congeners was also noted for C. gymnorhinus in specimens taken in the eastern Gulf of Mexico off the west Florida shelf (Topp and Hoff, 1972). Among the six females examined by Topp and Hoff, ovaries of a 21.0-mm-SL female were maturing (observed macroscopically), whereas ovaries of a 30.2-mm-SL female were filled with ripe, spherical eggs. They also observed that males of the same size had sexually dimorphic features, indicating they were mature at sizes similar to those at which females reach maturity. Abundance Of 42 field collections containing C. gymnorhinus (Appen- dix 2), 16 comprised solitary specimens, 11 consisted of two or three specimens each, and 15 collections con- tained four or more specimens. The largest collections of C. gymnorhinus comprised 36 and 23 specimens taken in the Straits of Florida off Key West and off South Caro- lina, respectively. Other significant collections of this species are those containing 19, 16, and 15 specimens taken in single trawls in the Straits of Florida, off the eastern side of the Florida Peninsula, and South Caro- lina, respectively. Three other trawls made off North Carolina, South Carolina, and the British Virgin Islands contained eight, nine, and eight individuals, respec- tively. From the collections we examined and those listed in previous studies that provide detailed information on C. gymnorhinus (Gutherz and Blackman, 1970; Topp and Hoff, 1972), we believe that this species appar- ently is not taken anywhere in its geographic range in such large numbers as is C. cornutus (see above). The largest collections reported in Gutherz and Blackman (1970) and Topp and Hoff (1972) contained only 10 and 11 individuals, respectively, for trawl catches made in the Straits of Florida and off Venezuela, but most of their collections of C. gymnorhinus contained only five or fewer specimens. Wenner et al. (1979b; 1979c) listed collections of 15 and 12 specimens taken off the east coast of Florida (28°50.3'N, 80°07'W and 29°50.3'N, 80°07'W, respectively). General discussion These data represent the most comprehensive assess- ments of biological, ecological, and distributional infor- mation for C. cornutus and C. gymnorhinus. Data on geographic occurrences, bathymetric distributions, maxi- mum sizes, sizes at maturity, and depth of occurrence are provided for the majority of known specimens of both species that have been collected off the eastern United States. The combined information gleaned from a variety of mostly small collections of these species from this region, including data from specimens reported on in previously published studies and data associated with specimens vouchered in museum collections but for which no previously published information has been available, provides considerable insights into, and com- prehensive documentation of, the occurrence, distribu- tion, and natural history of these interesting flatfishes. This updated information, in turn, provides a baseline for evaluating any changes observed in their geographic and bathymetric distributions along the continental shelf off the eastern United States. In prevailing literature since the late 1960s, the northern extent of the geographic ranges for adult C. cornutus and C. gymnorhinus has been misreported, resulting in a long history of inaccurate distribution data for these species. Both species are residents on the continental shelf off the southeastern United States off North Carolina and South Carolina, and their persis- tent presence in this region is documented from nearly a century (C. gymnorhinus) to more than a century ago (C. cornutus). Occurrences of C. cornutus north of North Carolina appear to he irregular, based on the absence of this species in many of the fish community studies conducted in this region (e.g., Grosslein and Azarovitz, 1982; Colvocoresses and Musick, 1984), and based on its generally low frequency of occurrence in the NMFS- NEFSC groundfish surveys conducted annually in this area (J. Galbraith, personal commun.3). Gear selectiv- ity in most trawl surveys influences the frequency of occurrence of small-size species in their catches, and many small-size flatfishes, including C. cornutus and C. gymnorhinus , have often escaped capture or have been misidentified or overlooked in earlier surveys. These factors also likely contributed to the infrequency of reports on these species. 3 Galbraith, John. 2009. Woods Hole Laboratory, North- east Fisheries Science Center, National Marine Fisheries Service, National Oceanic and Atmospheric Administration, 166 Water Street, Woods Hole, MA 02543. 338 Fishery Bulletin 108(3) Recent captures of these species — C. cornutus from off Hudson Canyon (Fahay, 2007) and the shelf off New York (Steves et al., 1999) and C. gymnorhinus from off North Carolina (Quattrini and Ross, 2006; this study) in areas north of their perceived adult geographic rang- es— could have been misinterpreted as evidence for ex- panding geographic distributions (i.e. , poleward range extensions) in response to global warming because some marine fishes respond to oceanic warming with shifts in their latitudinal range (for further discussion on this topic in relation to marine fishes see Perry et ah, 2005; Nye et al., 2009). However, adult speci- mens captured nearly a century or more ago, as well as historical literature documenting occurrences of both adults and larvae of these species in these northern lo- cations, prove that these species were present in these areas historically. Similarly, purported new records of reef fishes off North Carolina were probably mistak- enly attributed to ocean warming (Parker and Dixon, 1998), when in fact many of those species were already known from the area (see comments in Quattrini and Ross, 2006). These examples clearly indicate a need for more careful research and the need for examination of historical data (published and unpublished) before in- voking climate-change hypotheses to explain observed distributional patterns of marine organisms, especially for species that are less well known. Available data in- dicate only sporadic and infrequent occurrence of adult C. cornutus on the continental shelf north of North Carolina, and no records exist of adult C. gymnorhinus from north of North Carolina. Changes in the distribu- tions of these diminutive flatfishes on the northwest Atlantic shelf in response to climatic factors may be signaled by northward extension in their geographic ranges, by increases in frequency of capture and preva- lence within northern portions of their ranges, or by increases in their biomass in deeper areas within their ranges, as has been observed for other North Atlantic fish species (Dulvy et al., 2008; Nye et al., 2009). The updated baseline information contained herein on the frequencies of occurrences, relative abundances, and detailed examination of bathymetric and geographic ranges of both flatfish species provides the basis for comparison with future evaluations of the responses of these species to changes occurring on the continental shelf of the western North Atlantic Ocean. Several factors likely contributed to inaccuracies in geographic distributions for these species appear- ing in recently published literature. Given that these flatfishes are not usually taken in abundance, nor are they commercially important, significant captures could be overlooked in studies of fish communities or commercial fisheries. Important distributional in- formation for both species also have escaped notice because incidental captures of these diminutive flat- fishes were often buried in tables or appendices and the significance of these records was not emphasized in the works documenting these captures. Confusion in identifying these species also occurs (e.g., Goode and Bean, 1895; Hildebrand, 1941) and misidentifica- tions in the literature and in museum collections and databases have contributed to oversights of important specimens. Furthermore, studies in which the ecol- ogy or distribution of these species was summarized (Gutherz, 1967; Robins and Ray, 1986; Munroe, 2003; McEachran and Fechhelm, 2005; Lyczkowski-Shultz and Bond, 2006; and others) did not contain earlier published records. Unfortunately, the incomplete, and sometimes inaccurate, information contained in these synthetic studies has subsequently been perpetuated in other publications. As biology and ecology disciplines increase the use and reliance on web-based and web- available references, historically important, but less available, less popular, or more obscure literature will be increasingly overlooked. We caution that sole reli- ance on popular citation sources for historical informa- tion, in many cases, is a poor substitute for first-hand examination of all of the pertinent original literature related to a subject. Additionally, regardless of what popularly cited literature sources report, or even what museum databases indicate on their websites, these are no substitutes for actually examining specimens and analyzing collection data associated with them. In our study, nine different references noted the occur- rence of C. cornutus on the continental shelf south of Rhode Island or off New Jersey and New York, yet all of these publications were disregarded in all but one contemporary literature source for this species. Moreover, museum specimens represent important archived sources of geographic and bathymetric infor- mation. Museum lots of both C. cornutus and C. gymno- rhinus collected over 100 and 92 years ago, respectively, voucher occurrences of adults of these species from the northernmost regions of their geographic ranges. These specimens, too, were overlooked by the same popularly cited studies mentioned above. Accurate records of geographic occurrence and bathy- metric distribution for species such as these two di- minutive flatfishes are necessary and important because they provide the background information required to evaluate future changes in their distributions or oc- currences in relation to large-scale changes in oceanic conditions on the continental shelf of the western North Atlantic. Changes in distributions of organisms are rel- evant for assessments of effects of climate change, and more information is needed on the regional dynamics of populations over time to better understand effects of various environmental conditions on their survival and persistence within an area (Nye et al., 2009; Tolan and Fisher, 2009; Wood et al., 2009). Changes in the spatiotemporal occurrences of small-size, noncommer- cial species often are overlooked or disregarded (Link, 2007). But changes in geographic and bathymetric dis- tributions and relative abundances for such species may provide additional evidence of larger-scale envi- ronmental changes impacting biological communities that inhabit the continental shelf off the eastern United States and elsewhere. Without proper understanding of historical data describing distributional patterns, it will be impossible to accurately document any changes that Munroe and Ross: Distribution and life history of Citharichthys gymnorhinus and C. cornutus 339 may occur, or be occurring, in the geographic or ecologi- cal distributions of such species that inhabit particular geographic regions. While assembling this information, we updated iden- tifications of museum specimens, located important specimens in a variety of fish collections, and cor- rected and updated incomplete geographic records for some museum specimens. For both C. cornutus and C. gymnorhinus , much still remains to be learned about their biology and distributions, especially for a popu- lation, or populations, on the continental shelf off the southeastern United States. Age and growth, reproduc- tion, habitat preferences, and trophodynamics remain virtually unknown for these diminutive species. Sadly, this situation is also true for most of the small fishes or those that are not economically important, and as management agencies move toward ecosystem-based, rather than single-species management approaches, da- ta on the whole community becomes even more critical. We urge agencies and academic communities to expand the scope of fishery studies to address this need. Additional material not included in appendices 1-2 C. gymnorhinus. Larvae: ARC 24522, 32°58'37"N, 77°51'53"W; ARC 24538, 33°24'07"N, 76°42'33"W; MCZ 77935, 41°33'N, 54°55'W. C. cornutus. Adults. AMNH 86095, 72 = 103; AMNH 86160, 71 = 43; AMNH 86101, 72 = 100; AMNH 84815, 77 = 63; AMNH 84857, 77=47; AMNH 85590, 72 = 172; AMNH 85529, 77 = 321. Acknowledgments We thank the following individuals who provided infor- mation or assisted with specimens, data, and informa- tion regarding collections or databases in their care: M. Fahay, J. Finan, J. Galbraith, P. Gerdes, A. Harold, K. Hartel, W. Kramer, J.D. Lyons, D. Machowski, J. Moore, K. Murphy, R. Robins, M. A. Rogers, W. Starnes, M. Underwood, R. Worthen, D. Wyanski, and M. Zokan. D. Steere, M. Rosen, S. Dawicki, and C. Struthers assisted with locating library references. J. Brown, D. Walker, and M. Vecchione provided assistance locating nautical charts; M. Rhode, M. Partyka, M. Carlson, and M. Nizinski assisted with figure preparations; S. Raredon provided photographs; A. Quattrini and L. Willis assisted with various aspects of this study. Partial support for recent sampling was provided by the U.S. Geological Survey and NOAA Office of Ocean Exploration (grants to S.W. Ross). Special thanks are extended to the Ecosystems Survey Branch, NMFS, Woods Hole, for providing specimens and data from their long-term synoptic surveys of the continental shelf off the eastern United States. We also extend our appreciation to all those who collected fishes that were used in this study. Literature cited Able, K. W. 1992. Checklist of New Jersey saltwater fishes. Bull. N.J. Acad. Sci. 37:1-11. Boschung, H. T. 1992. Catalogue of freshwater and marine fishes of Alabama. Bull. Ala. Mus. Nat. Hist. 14:1-266. Castro-Aguirre, J. L., H. Espinosa Perez, and J. J. Schmitter-Soto. 1999. Ictiofauna estuarino-lagunar y vicaria de Mexico, 705 p. Limusa, Noriega Editores. Mexico City. I In Spanish.] Cervigon, F. 1996. Los peces marinos de Venezuela, vol. IV, 254 p. Fund. Los Roques. Caracas. [In Spanish.] Colvocoresses, J. A., and J. A. Musick. 1984. Species associations and community composition of Middle Atlantic Bight continental shelf demersal fishes. Fish. Bull 82:295-313. Dulvy, N. K., S. I. Rogers, S. Jennings, S. Vanessa, S. R. Dye, and H. R. Skjoldal. 2008. Climate change and deepening of the North Sea fish assemblage: a biotic indicator of warming seas. J. Appl. Ecol. 45:1029-1039. Evermann, B. W., and M. C. Marsh. 1902. The fishes of Porto Rico. Bull. U.S. Fish Comm. 20:49-350. Fahay, M. P. 2007. Early stages of fishes in the western North Atlantic Ocean (Davis Strait, Southern Greenland and Flem- ish Cap to Cape Hatteras). Vol. two: Scorpaeniformes through Tetraodontiformes, p. 932-1696. N.W. Atl. Fish. Organ., Dartmouth, Nova Scotia. Figueiredo, J. L., and N. A. Menezes. 2000. Manual de peixes marinhos do sudeste do Brasil. VI. Teleostei (5), 116 p. Mus. Zool., Univ. Sao Paulo, Brasil. [In Portuguese.] 2003. Familia Paralichthyidae. In Catalogo das espe- cies de peixes marinhos do Brasil (N. A. Menezes, P. A. Buckup, J. L de Figueiredo, and R. L. Moura, eds.), p. 105- 106. Mus. Zool., Univ. Sao Paulo, Brasil. [In Portuguese.] Fowler, H. W. 1952. A list of fishes of New Jersey, with off-shore species. Proc. Acad. Nat. Sci. Phila. 104:89-151. Froese, R., and D. Pauly (eds.). 2000. FishBase 2000: concepts, design and data sources. ICLARM, Los Banos, Laguna, Philippines. http:www.fishbase.org (accessed January 2010). Goode, G. B. 1880. Description of seven new species of fishes from deep soundings on the southern New England coast, with diagnoses of two undescribed genera of flounders and a genus related to Merlucius. Proc. U.S. Natl. Mus. 3:337-350. Goode, G. B., and T. H. Bean. 1895. Oceanic Ichthyology. U.S. Natl. Mus. Spec. Bull. No. II (part I, text), 553 p. Grosslein, M. D., and T. R. Azarovitz. 1982. Fish Distribution. Marine EcoSystems Analysis [MESA] Program, New York Bight Project. NY Bight Atlas, Monogr. 15, 182 p. Grothues, T. M., R. K. Cowen, L. J. Pietrafesa, F. Bignami, G. L. Weatherly, and C. N. Flagg. 2002. Flux of larval fish around Cape Hatteras. Limnol. Oceanogr. 47:165-175. 340 Fishery Bulletin 108(3) Gunther, A. 1880. Zoology. 1. Report on the shore fishes pro- cured during the voyage of the H. M. S. Challenger, 1873-1876. In Report on the scientific results of the voyage of the H. M. S. Challenger during the years 1873-1876. Vol. 6, p. 1-82; pis. 1-32. Printed by Eyre and Spottiswoode, London. 1887. Zoology. Vol. 57. Report on the deep-sea fishes procured during the voyage of the H. M. S. Challenger, 1873-1876. In Report on the scientific results of the voyage of the H. M. S. Challenger during the years 1873-1876. Vol. 22, p. 1-335; pis. 1-72. Printed by Eyre and Spottiswoode, London. Gutherz, E. J. 1967. Field guide to the flatfishes of the family Bothidae in the western North Atlantic. U.S. Fish Wildl. Serv. Circ. 263, 47 p. 1969. Hermaphroditism in Citharichthys cornutus (Het- erosomata. Family Bothidae). Copeia 1969:352-356. Gutherz, E. J., and R. R. Blackman. 1970. Two new species of the flatfish genus Citharichthys (Bothidae) from the western North Atlantic. Copeia 1970:340-348. Hartel, K. E., C. P. Kenaley, J. K. Galbraith, and T. T. Sutton. 2008. Additional records of deep-sea fishes from off Greater New England. Northeast. Natural. 15:317-334. Hildebrand, S. F. 1941. An annotated list of salt and brackish water fishes, with a new name for a menhaden, found in North Carolina since the publication of “The Fishes of North Carolina” by Hugh M. Smith in 1907. Copeia 1941:220-232. Hoese, H. D., and R. H. Moore. 1977. Fishes of the Gulf of Mexico, Texas, Louisiana, and adjacent waters, 327 p. Texas A&M Univ. Press, College Station, TX. 1998. Fishes of the Gulf of Mexico: Texas, Louisiana, and adjacent waters, 2nd ed., 422 p. Texas A&M Univ. Press, College Station, TX. Jordan, D. S., and B. W. Evermann. 1898. The fishes of North and Middle America: A descrip- tive catalogue of the species of fish-like vertebrates found in the waters of North America, north of the Isthmus of Panama. Bull. U.S. Natl. Mus. 47 (Pt. III):2184— 3136. Jordan, D. S., and C. H. Gilbert. 1883. Synopsis of the fishes of North America. Bull. U.S. Natl. Mus. 5 (No. 16 ) : 1 — 1018 . Jordan, D. S., and D. K. Goss. 1889. A review of the flounders and soles (Pleuronectidae) of America and Europe. Ann. Rept. Commissioner Fish Fisher, for 1886. Rept. U.S. Fish Comm. 14:225-342. Link, J. S. 2007. Underappreciated species in ecology: “Ugly fish” in the Northwest Atlantic Ocean. Ecol. Appl. 17:2037- 2060. Longley, W. H., and S. F. Hildebrand. 1941. Systematic catalogue of the fishes of Tortugas, Florida. Papers Tortugas Lab. 34:331 p. Lozano-Vilano, M. L., M. E. Garcia-Ramirez, and S. Contreras- Balderas. 1993. Peces costeros y marinos del estado de Veracruz. In Biodiversidad marina y costera de Mexico (S. I. Salazar Vallejo, and N. E. Gonzalez, eds.), p. 576-595. Com. Nac. Biodiver, and Centro Invest. Quintana Roo, Mexico, 862 p. [In Spanish.] Lyczkowski- Shultz, J., and P. J. Bond. 2006. Paralichthyidae: Sand flounders. In Early Stages of Atlantic Fishes. An Identification Guide for the West- ern Central North Atlantic, vol. II. (W. J. Richards, ed.), p. 2291-2325. CRC Press, Boca Raton, FL. Marancik, K. E., L. M. Clough, and J. A. Hare. 2005. Cross-shelf and seasonal variation in larval fish assemblages on the southeast United States continental shelf off the coast of Georgia. Fish. Bull. 103:108- 129. McEachran, J. D., and J. D. Fechhelm. 2005. Fishes of the Gulf of Mexico. Vol. 2: Scorpaeni- formes to Tetraodontiformes, 1004 p. Univ. Texas Press. Austin, TX. Moore, J. A., K. E. Hartel, J. E. Craddock, and J. K. Galbraith. 2003. An annotated list of deepwater fishes from off the New England area, with new area records. Northeast. Natural. 10:159-248. Munroe, T. A. 2003. Family Paralichthyidae. In Living marine resources of the western Central Atlantic, vol. 3. (K. E. Carpenter, ed.), p. 1898-1921. FAO species identifica- tion guide for fishery purposes and Am. Soc. Ichthyol. Herpetol. Spec. Publ. 5. FAO, Rome. 2005. Systematic diversity of the Pleuronectiformes. Chapter 2, in Flatfishes: biology and exploitation, (R. N. Gibson, ed ), p. 10-41. Blackwell Science Ltd., Oxford. Norman, J. R. 1934. A systematic monograph of the flatfishes (Hetero- somata). Vol. 1. Psettodidae, Bothidae, Pleuronectidae, 459 p. Brit. Mus. (Nat. Hist.). London. Nye, J. A., J. S. Link, J. A. Hare, and W. J. Overholtz. 2009. Changing spatial distribution of fish stocks in relation to climate and population size on the northeast United States continental shelf. Mar. Ecol. Prog. Ser. 393:111-129. Parker, R. O., Jr., and R. L. Dixon. 1998. Changes in a North Carolina reef-fish commu- nity after 15 years of intense fishing: global-warming implications. Trans. Am. Fish. Soc. 127:908-920. Parr, A. E. 1931. A practical revision of the western Atlantic species of the genus Citharichthys (including Etropus ). Bull. Bingham Oceanogr. Coll. 4 (Art. l):l-24. Perry, A. L., P. J. Low, J. R. Ellis, and J. D. Reynolds. 2005. Climate change and distribution shifts in marine fishes. Science 308:1912-1915. Powell, A. B., D. G. Lindquist, and J. A. Hare. 2000. Larval and pelagic juvenile fishes collected with three types of gear in Gulf Stream and shelf waters in Onslow Bay, North Carolina, and comments on ichthyo- plankton distribution and hydrography. Fish. Bull. 98:427-438. Quattrini, A. M., and S. W. Ross. 2006. Fishes associated with North Carolina shelf-edge hardbottoms and initial assessment of a proposed marine protected area. Bull. Mar. Sci. 79:137—163. Robins, C. R., and G. C. Ray. 1986. A field guide to Atlantic Coast fishes of North America, 354 p. Peterson Field Guide Series. Houghton Mifflin Co. Boston. Saavedra-Diaz, L. M., A. Acero P., and G. R. Navas S. 2000. Lenguados de la familia Paralichthyidae (Pisces: Pleuronectiformes) conocidos del Caribe colombiano. Rev. Acad. Colomb. Cienc. 24:295-310. [In Spanish.] Munroe and Ross: Distribution and life history of Citharichthys gymnorhmus and C. cornutus 341 Scott, W. B., and M. G. Scott. 1988. Atlantic Fishes of Canada. Can. Bull. Fish. Aquat. Sci. 219, 731 p. Starck, W. A., II. 1968. A list of fishes of Alligator Reef, Florida with com- ments on the nature of the Florida reef fish fauna. Un- dersea Biol. 1:5-40. Staiger, J. C. 1970. The distribution of the benthic fishes found below two hundred meters in the Straits of Florida. Ph.D. diss., 219 p. Univ. Miami, Coral Gables, FL. Steves, B. P., R. K. Cowen, and M. H. Malchoff. 1999. Settlement and nursery habitats for demersal fishes on the continental shelf of the New York Bight. Fish. Bull. 98:167-188. Tolan, J. M., and M. Fisher. 2009. Biological response to changes in climate patterns: population increases of gray snapper (Lutjanus griseus) in Texas bays and estuaries. Fish. Bull. 107:36-44. Topp, R. W„ and F. H. Hoff, Jr. 1972. Flatfishes (Pleuronectiformes). Mem. Hourglass Cruises, Fla. Dept. Nat. Resour., St. Petersburg, FL. 4:1-135. Tucker, J. W., Jr. 1982. Larval development of Citharichthys cornutus, C. gymnorhinus, C. spilopterus, and Etropus crossotus (Bothidae), with notes on larval occurrence. Fish. Bull. 80:35-73. Vergara Rodriguez, R. 1974. Nuevos registros para la ictiofauna Cubana. II. Poeyana 134:1-7. [In Spanish] Walsh, H. J., K. E. Marancik, and J. A. Hare. 2006. Juvenile fish assemblages collected on unconsoli- dated sediments of the southeast United States conti- nental shelf. Fish. Bull. 104:256—277. Wenner, C. A., C. A. Barans, B. W. Stender, and F. H. Berry. 1979a. Results of MARMAP otter trawl investigations in the South Atlantic Bight. I. Fall, 1973. S.C. Mar. Resour. Cent., Tech. Rep. 33, 79 p. [Available from http://spinner.cofc.edu/~mrl/scdnrTechReportsl_50.html, accessed April 2010.] 1979b. Results of MARMAP otter trawl investigations in the South Atlantic Bight. II. Spring, 1974. S.C. Mar. Resour. Cent., Tech. Rep. 40, 78 p. [Available from http://spinner.cofc.edu/~mrl/scdnrTechReportsl 50.html, accessed April 2010.] 1979c. Results of MARMAP otter trawl investigations in the South Atlantic Bight. III. Summer, 1974. S.C. Mar. Resour. Cent., Tech. Rep. 41, 62 p. [Available from http://spinner.cofc.edu/~mrl/scdnrTechReportsl_50.html, accessed April 2010.1 1979d. Results of MARMAP otter trawl investigations in the South Atlantic Bight. IV. Winter-early spring, 1975. S.C. Mar. Resour. Cent., Tech. Rep. 44, 59 p. [Available from http://spinner.cofc.edu/~mrl/scdnrTechReportsl_ 50.html, accessed April 2010 ] 1980. Results of MARMAP otter trawl investigations in the South Atlantic Bight. V. Summer, 1975. S.C. Mar. Resour. Cent., Tech. Rep. 45, 57 p. [Available from http://spinner.cofc.edu/~mrl/scdnrTechReportsl_50.html, accessed April 2010.] Wood, A. J. M„ J. S. Collie, and J. A. Hare. 2009. A comparison between warm-water fish assem- blages of Narragansett Bay and those of Long Island Sound waters. Fish. Bull. 107:89-100. 342 Fishery Bulletin 108(3) Munroe and Ross: Distribution and life history of Citharichthys gymnorhinus and C. cornutus 343 344 Fishery Bulletin 108(3) Munroe and Ross: Distribution and life history of Citharichthys gymnorhinus and C. cornutus 345 346 Abstract — Preliminary validation of annual growth band deposition in vertebrae of great hammerhead shark (Sphyrna mokarran) was con- ducted by using bomb radiocarbon analysis. Adult specimens (n= 2) were collected and thin sections of verte- bral centra were removed for visual aging and use in radiocarbon assays. Vertebral band counts were used to estimate age, and year of formation was assigned to each growth band by subtracting estimated age from the year of capture. A total of 10 samples were extracted from growth bands and analyzed for A14 C. Calculated A14C values from dated bands were compared to known-age reference chronologies, and the resulting pat- terns indicated annual periodicity of growth bands up to a minimum age of 42 years. Trends in A14C across time in individual specimens indi- cated that vertebral radiocarbon is conserved through time but that habitat and diet may influence A14C levels in elasmobranchs. Although the age validation reported here must be considered preliminary because of the small sample size and narrow age range of individuals sampled, it represents the first confirmation of age in S. mokarran, further illustrat- ing the usefulness of bomb radiocar- bon analysis as a tool for life history studies in elasmobranchs. Manuscript submitted 22 December 2009. Manuscript accepted 19 May 2010. Fish. Bull. 108:346-351 (2010). The views and opinions expressed or implied in this article are those of the author (or authors) and do not necessarily reflect the position of the National Marine Fisheries Service, NOAA. Age validation of great hammerhead shark ( Sphyrna mokarran ), determined by bomb radiocarbon analysis Michelle S. Passerotti (contact author)1 John K. Carlson1 Andrew N. Piercy2 Steven E. Campana3 Email address for contact author: Michelle.Passerotti@noaa.gov 1 Southeast Fisheries Science Center National Marine Fisheries Service National Oceanic and Atmospheric Administration 3500 Delwood Beach Road Panama City, Florida 32408 2 Florida Program for Shark Research Florida Museum of Natural History University of Florida P.O. Box 117800 Gainesville, Florida 32611 3 Bedford Institute of Oceanography PO. Box 1006 Dartmouth, Nova Scotia, Canada B2Y 4A2 Validation of the periodicity of ver- tebral growth band deposition over the entire life span of a species is an important aspect of age estimation and growth determination in elasmo- branch fishes (Caillet and Goldman, 2004; Francis et al., 2007). Accurate age estimation is critical because it forms the basis for calculating growth and mortality rates, age at maturity, and estimates of longevity, all of which are essential for population assess- ment. The need for accurate age and growth estimates is especially great for many elasmobranch species, which tend to be data poor and highly vul- nerable to fishing pressure (Musick, 1999). Bomb radiocarbon dating has been successfully used to validate age estimates for several elasmobranch species (e.g., Campana et al., 2002; Kneebone et al., 2008; McPhie and Campana, 2009). The peak in atmo- spheric radiocarbon (14C) from test- ing nuclear bombs in the 1950s and 1960s is used as a marker that can be dated in the calcified structures of marine organisms. Bomb testing yielded synchronous known-age refer- ence chronologies in corals, bivalves, and fish otoliths worldwide (Campana, 1997; Druffel, 1989), which can be used to confirm the accuracy of age estimates for various marine species (Campana et al., 2002). The presence of a tracer over such a protracted time span makes bomb radiocarbon analy- sis highly suitable for age validation, especially for typically long-lived elas- mobranchs. The great hammerhead shark (Sphyrna mokarran) is a large (maxi- mum size of 550-610 cm total length [TL] ) cosmopolitan species found cir- cumtropically in both inshore and oceanic habitats to depths of over 80 meters (Compagno, 1984). Great hammerhead sharks tend to be reef- associated, but some populations un- dertake seasonal offshore migrations (Compagno, 1984). Life history in- formation for the great hammerhead shark is very limited; reports consist mostly of notes on their reproduction (Stevens and Lyle, 1989). There are no published age validations for S. mokarran. The vulnerability of great ham- merhead sharks to fishing pressure is potentially high given the tenden- cy of elasmobranchs to exhibit slow growth, late age at maturity, and Passerotti et al: Age validation of Sphyrna mokarran 347 low fecundity (Musick, 1999). Although not generally targeted in fisheries, S. mokarran are favored among incidentally caught species because their fins are highly valued due to their size and the density of their fin rays. In an assessment of the Hong Kong shark fin market, it was found that fins from hammerhead shark species were among the most valuable fin types in the market (Clarke et al., 2004; Abercrombie et al., 2005). Recently, concern has arisen in regard to populations of S. mo- karran worldwide because the International Union for Conservation of Nature (IUCN) assessed the species as endangered.1 These circumstances illustrate the need for validated age estimates of S. mokarran. Here, we present the preliminary results of bomb radiocarbon analysis as a novel and accurate method of age valida- tion for this species. Materials and methods Vertebrae for bomb radiocarbon age validation were taken from two S. mokarran specimens (SM-112 and SM-114) caught from commercial longline vessels off the Georgia coast in the U.S. south Atlantic between 2003 and 2004. Specimens were both male, measur- ing 300 cm and 276 cm fork length (FL), respectively. Although S. mokarran are frequently caught as bycatch in several commercial and recreational fisheries, it is difficult to obtain vertebral samples from individuals of sufficient age for the purposes of this study. Ideally, specimens would have vertebral tissue formed between 1955 and 1965, the years encompassing the period of initial increase in 14C (Campana et al., 2002; Piner et al., 2005). However, individuals living during this time period would at present be quite large (>300 cm FL) and the occurrence of specimens of this size are infrequent in catches available for sampling. The two specimens used in this study provided the only vertebral samples of appropriate age available to the authors at the time of this study. Vertebrae were collected either from the area under the dorsal fin or above the branchial chamber, stored on ice, and later frozen upon arrival at the labora- tory. Excess tissue was manually removed from thawed vertebrae, which were then soaked in varying concen- trations of sodium hypochlorite solution for 5-30 min- utes to remove remaining tissue. Cleaned vertebrae were rinsed in tap water and stored in 70% ethanol. Vertebral sections (1 mm thick) were prepared by a single longitudinal cut with paired blades separated by a spacer on an IsoMet low-speed diamond-bladed saw (Buehler, Lake Bluff, IL). Sections were immersed in ethanol and digitally photographed under a binocular microscope at 16-40 x magnification with reflected light. 1 Camhi, M. D., S. V. Valenti, S. V. Fordham, S. L. Fowler, and C. Gibson. 2009. The conservation status of pelagic sharks and rays: report of the IUCN shark specialist group pelagic shark Red List workshop, 78 p. IUCN Species Survival Commission Shark Specialist Group, Newbury, UK. Age interpretation was based on visual counts of paired growth increments (growth bands) from images en- hanced for contrast with Adobe Photoshop CS2 (Adobe Systems, Inc., Burlington, NJ), and interpretation was based on the criteria of Natanson et al. (2002). Vertebral tissue samples (n = 10 samples; 4-9 mg each) were extracted from multiple growth bands in the corpus calcareum region of each vertebral sec- tion. Extractions were performed under the binocular microscope with 16 x magnification. Extracted samples were isolated as solid pieces by using a Gesswein high- speed hand tool (Gesswein, Bridgeport, CT) fitted with steel bits <1 mm in diameter. The first-formed growth band (corresponding to the first year of growth) was extracted from each vertebra; individual growth bands corresponding to later years were also extracted. The samples from both specimens corresponding to the most recent growth (where growth bands were very narrow) consisted of 6-10 pooled growth bands. The presumed date of sample formation (i.e. growth band formation) was calculated as the year the shark was collected minus the growth band count from the birth band to the mid-point of the sample. After sonification in Super Q water and drying, the sample was weighed to the nearest 0.1 mg in preparation for 14C assay with accelerator mass spectrometry (AMS). AMS assays also provided A13C (%o) values, which were used to correct for isotopic fractionation effects. Radiocarbon values were subsequently reported as A14C, which is the per mil ( %c ) deviation of the sample from the radio- carbon concentration of 19th-century wood, corrected for sample decay before 1950 according to methods outlined by Stuiver and Polach (1977). To assign dates of formation to an unknown tissue sample, it is necessary that the A14C of the unknown sample be compared with a A14 C chronology based on known-age material (a reference chronology). To match the water mass characteristics of S. mokarran habi- tat, we used a reference chronology for Florida corals developed by Druffel (1989). This chronology would be expected to show A14C values comparable to those of the great hammerhead shark because of similarity of habitat. However, the carbon source for vertebrae is metabolic in origin unlike the dissolved inorganic carbon (DIC) source for coral (Campana et al., 2002). Therefore, we also used a reference chronology devel- oped from known-age porbeagle ( Lamna nasus) in the northwest Atlantic (Campana et al., 2002). The period of increase in 14C in this chronology would be expected to be very similar to that of great hammerhead sharks inhabiting the U.S. south Atlantic, although with very different absolute values owing to the different water mixing characteristics of the two regions. Results Based on annual growth band counts, the age esti- mate for each vertebra was 42 years for SM-112 and 36 years for SM-114, yielding birth years of 1961 348 Fishery Bulletin 108(3) Figure 1 Digital image (2x magnification) of vertebral section from a 36-year old male hammerhead shark ( Sphyrna mokarran) (SM-114), captured 14 January 2004. White dots denote single winter growth bands used to assign age. White line near base of centrum denotes birth band. and 1967, respectively (Fig. 1). These estimates fall within the range of dates useful for bomb radiocarbon analysis and indicate that the specimens were both born during the initial rise in 14C. Bomb radiocar- bon analysis yielded results for seven samples from SM-112 and three from SM-114 (Table 1). Values of <513C were relatively stable over the range of samples (mean=-11.0, standard deviation [SD] = 0.1) and were similar to those from other elasmobranch species, verifying a dietary (metabolic) carbon source (Fry, 1988; Campana, 1997; Campana et al., 2006). The mean standard deviation of the individual radiocarbon assays was about 5 %o. Values of z\14C in S. mokarran ranged from 18.6 to 148.3 units, reaching a maximum in the early 1970s (Fig. 2). The birth dates of the two sharks were not quite old enough for us to document the initial year of radiocarbon increase, which likely occurred before 1961. Given the available data, the trend in timing and magnitude of the 414C chronology for all of the S. mo- karran samples most closely resembled that of Florida coral. Timing of the period of increase and peak in 414C was also similar between S. mokarran and porbeagle chronologies, but there were large differences in abso- lute values. When trends in Z\14C for the two specimens were examined individually, a difference in trajectories was apparent between SM-112 and SM-114. Values from both specimens fell mostly along the curve of the coral chronology, with one exception. The sample Passerotti et al: Age validation of Sphyrna mokarran 349 Year of formation Figure 2 Bomb radiocarbon (A14 C) values over time in two great hammerhead shark ( Sphyrna mokarran) individuals (SM-112 and SM-114), compared with Florida coral and porbeagle (Lamna nasus) reference chronologies. LOESS (locally weighted regression) curves have been fitted to the Florida coral and porbeagle data series. Terminal points for SM-112 and SM-114 are bracketed to show years of formation for pooled growth bands included in each sample. Table 1 Summary of age data and bomb radiocarbon analysis for vertebral samples taken from great hammerhead sharks ( Sphyrna mokarran). Sample SM-112 was male, fork length = 300 cm, aged at 42 years, and captured on 23 July 2003. Sample SM-114 was male, fork length=276 cm, aged at 36 years, and captured on 14 January 2004. Year of formation =year of growth-band formation. Sample Year of formation Median year of formation Estimated age &3C A14C SM-112 1961 1961 0 -11.4 18.6 1962 1962 1 -11.4 43.9 1963 1963 2 -11.0 58.7 1964 1964 3 -10.5 84.4 1965 1965 4 -9.9 71.9 1970 1970 9 -9.2 97.7 1993-2003 1998 37 -10.7 99.7 SM-114 1967 1967 0 -12.0 117.2 1971 1971 4 -12.0 148.3 1997-2003 2000 32 -12.2 69.8 dated 1970 from SM-112 fell well below the curve of the coral chronology, whereas the most recent (pooled) sample from SM-112 fell more closely to the expected downward trajectory for Florida coral. Samples from SM-114 were nearly synchronous with the coral chro- nology extrapolated to the year 2000. Within-shark patterns in h14C accumulation were similar to those of the reference chronologies, confirming that car- bon recorded in the vertebrae is conserved through time. 350 Fishery Bulletin 108(3) Discussion Although a larger sample size would have been prefer- able in this study, the results of the bomb radiocarbon assays support the hypothesis of annual growth band deposition in S. mokarran vertebrae. Additionally, the similarity in the timing of increase and peak in A14 C between the great hammerhead shark samples and both reference chronologies indicates that the aging techniques employed in this study produce ages accu- rate to within a few years. It is generally accepted that the timing of the initial increase in 414C in relation to prebomb values is the most accurate dated marker for age validation (Campana et ah, 2008). Although there were no prebomb samples in our study (all z\14C values were above zero), the close alignment of values between S. mokarran samples and those from the coral chronol- ogy during the period of increase indicates that our ages were assigned correctly. If the specimens analyzed in this study had been under-aged, the entire great ham- merhead shark chronology would have been shifted to the right in relation to the coral chronology, and over- aging would have caused the reverse to be true. No such shifting was apparent. Differences in both the magnitude and timing of ra- diocarbon chronologies between vertebral samples and those from carbonate sources have been noted in pre- vious age studies of elasmobranchs. The difference in the magnitude of Au C values is largely attributable to the different carbon sources in carbonate (DIC uptake) compared to cartilaginous (dietary uptake) systems (Fry, 1988), but also to environmental factors such as habitat depth and the mixing rates of waters (Williams et al., 1987). This difference has been demonstrated in porbeagle (Campana et ah, 2002), shortfin mako ( Isurus oxyrinchus ) (Ardizzone et al., 2006), and white sharks (Carcharadon carcharias) (Kerr et al., 2006), as well as in two species of skates (McPhie and Campana, 2009), and can also be caused by the age of the carbon in prey items found at different depths, which can produce a delay in the radiocarbon chronology. In the case of S. mokarran, however, overall values of 414C followed those of Florida coral very closely, indicating little difference in timing of uptake between coral and vertebrae. The similarity in values of 414C is likely due to similarity in habitat for both the coral and the shark; S. mokarran are reef-associated for much of their lives (Compagno, 1984) and feed on reef-associated prey (Stevens and Lyle, 1989), which may assimilate carbon more quickly because of the well-mixed shallow habitat. Kneebone et al. (2008) found that young tiger sharks ( Galeocerdo cuvieri) exhibit similar patterns in 414C uptake, attrib- uting the pattern to a diet of small teleosts during the time that these sharks inhabit shallow nursery grounds. Campana et al. (2006) also found similar results in spines of spiny dogfish ( Squalus acanthias), in which carbon uptake into fin spines mirrored that of DIC uptake in otolith chronologies from the same region. Despite the apparent similarities between values of 414C in S. mokarran and Florida coral, there were some differences, such as the slight left shift in the first two samples from SM-112 and the depletion of 414C in the penultimate (1970) sample from SM-112, in relation to the rest of the chronology. The first two samples from SM-112, corresponding to formative years of 1961 and 1962, respectively, fell slightly left of the coral curve. Although a phase-shift to the right can be explained as a diet- or habitat-induced delay in carbon uptake, a shift to the left could indicate a slight over-aging of SM-112 of only 1-2 years, or the shift could be the re- sult of inclusion of material from more recently formed bands in the sample. In addition, the A14 C in the 1970 sample from SM-112 was depleted in comparison to the rest of the chronology and approached values more like those of porbeagle as opposed to coral. This singu- lar deviation could again be the result of an error in micromilling or could be the start of a more depleted trajectory for SM-112, reflecting an ontogenetic shift in habitat and diet. Although reef-associated for much of their lives, S. mokarran also undertake oceanic migra- tions through deeper water habitats (Compagno, 1984) that tend to be depleted in 414C. Consumption of prey from these habitats would result in depleted values of A14C in the vertebrae, as demonstrated in porbeagle and other deepwater sharks (Campana et al., 2002). Another possibility for this depletion in 414C is a shift in age of prey taken by SM-112; owing to its size this shark may have taken larger (and possibly older) prey. Obtaining additional 414C samples from both sharks would certainly clarify these results. This study confirms the longevity of great hammer- head sharks to an age of at least 42 years, although maximum reported lengths indicate that they may live well beyond this age. Further study on the life history of S. mokarran is needed to identify factors affecting individual patterns in 414C assimilation. Acknowledgments We thank observers of the National Marine Fisheries Service, NOAA, for obtaining samples from the directed shark longline fishery. W. Joyce provided technical assis- tance in sample processing. Funding was provided by the Southeast Fisheries Science Center, National Marine Fisheries Service, National Oceanic and Atmospheric Administration. Literature cited Abercrombie, D. L., S. C. Clarke, and M. S. Shivji. 2005. Global-scale genetic identification of hammer- head sharks: application to assessment of the interna- tional fin trade and law enforcement. Conserv. Genet. 6:775-788. Ardizzone, D., G. M. Cailliet, L. J. Natanson, A. H. Andrews, L. A. Kerr, and T. A. Brown. 2006. Application of bomb radiocarbon chronologies to shortfin mako ( Isurus oxyrinchus ) age validation. En- viron. Biol. Fishes 77:355-366. Passerotti et al: Age validation of Sphyrna mokarran 351 Cailliet, G. M., and K. J. Goldman. 2004. Age determination and validation in chondrich- thyan fishes. In The biology of sharks and their rela- tives (J. Carrier, J. A. Musick, and M. R. Heithaus, eds.), p. 399-447. CRC Press, Boca Raton, FL. Campana, S. E. 1997. Use of radiocarbon from nuclear fallout as a dated marker in the otoliths of haddock Melanogrammus aeglefinus. Mar. Ecol. Prog. Ser. 150:49-56. Campana, S. E., J. M. Casselman, and C. M. Jones. 2008. Bomb radiocarbon chronologies in the Arctic, with implications for the age validation of lake trout ( Salve - linus namaycush) and other Arctic species. Can. J. Fish. Aquat. Sci. 65:733—743. Campana, S. E., C. Jones, G. A. McFarlane, and S. Myklevoll. 2006. Bomb dating and age validation using the spines of spiny dogfish ( Squalus acanthias). Environ. Biol. Fishes 77:327-336. Campana, S. E., L. J. Natanson, and S. Myklevoll. 2002. Bomb dating and age determination of large pelagic sharks. Can. J. Fish. Aquat. Sci. 59:450-455. Clarke, S., M. K. McAllister, and C. G. J. Michielsens. 2004. Estimates of shark species composition and numbers associated with the shark fin trade based on Hong Kong auction data. J. Northwest Atl. Fish. Sci. 35:453-465. Compagno, L. J. V. 1984. FAO species catalogue. Vol. 4. Sharks of the world. An annotated and illustrated catalogue of shark species known to date. Part 2: Carcharhiniformes. FAO Fish. Synop. 125:251-655. FAO, Rome. Druffel, E. R. M. 1989. Decade time scale variability of ventilation in the north Atlantic: high-precision measurements of bomb radiocarbon in banded corals. J. Geophys. Res. 94:3271-3285. Francis, M. P, S. Campana, and C. M. Jones. 2007. Age underestimation in New Zealand porbeagle sharks {Lamna nasus ): is there an upper limit to ages that can be determined from shark vertebrae? Mar. Freshw. Res. 58:10-23. Fry, B. 1988. Food web structure on Georges Bank from stable C, N, and S isotopic compositions. Limnol. Oceanogr. 33:1182-1190. Kerr, L. A., A. H. Andrews, G. M. Cailliet, T. A. Brown, and K. H. Coale. 2006. Investigations of A 14C, 6 13C, and 6 15N in ver- tebrae of white shark ( Carcharodon carcharias ) from the eastern North Pacific Ocean. Environ. Biol. Fishes 77:337-353. Kneebone, J., L. J. Natanson, A. H. Andrews, and W. H. Howell. 2008. Using bomb radiocarbon analyses to validate age and growth estimates for the tiger shark, Galeo- cerdo cuvier, in the western North Atlantic. Mar. Biol. 154:423-434. McPhie, R. P, and S. E. Campana. 2009. Bomb dating and age determination of skates (family Rajidae) off the eastern coast of Canada. ICES J. Mar. Sci. 66:546-560. Musick, J. A. 1999. Criteria to define extinction risk in marine fishes. Fisheries 24:6-14. Natanson, L. J., J. J. Mello, and S. E. Campana. 2002. Validated age and growth of the porbeagle shark, Lamna nasus, in the western North Atlantic Ocean. Fish. Bull. 100:266-278. Piner, K. R., O. S. Hamel, J. L. Menkel, J. R. Wallace, and C. E. Hutchinson. 2005. Age validation of canary rockfish ( Sebastes pinniger) from off the Oregon coast (USA) using the bomb radiocarbon method. Can. J. Fish. Aquat. Sci. 62:1060-1066. Stevens, J. D., and J. M. Lyle. 1989. Biology of three hammerhead sharks (Eusphyra blochii, Sphyrna mokarran and S. lewini) from northern Australia. Aust. J. Mar. Freshw. Res. 40:129-146. Stuiver, M. and H. A. Polach. 1977. Reporting of C-14 data. Radiocarbon 19:355-363. Williams, P. M., E. R. M. Druffel, and K. L. Smith Jr. 1987. Dietary carbon sources for deep-sea organisms as inferred from their organic radiocarbon activities. Deep Sea Res. 34:253-266. 352 Abstract — We describe the applica- tion of two types of stereo camera systems in fisheries research, includ- ing the design, calibration, analysis techniques, and precision of the data obtained with these systems. The first is a stereo video system deployed by using a quick-responding winch with a live feed to provide species- and size- composition data adequate to produce acoustically based biomass estimates of rockfish. This system was tested on the eastern Bering Sea slope where rockfish were measured. Rockfish sizes were similar to those sampled with a bottom trawl and the relative error in multiple measurements of the same rockfish in multiple still-frame images was small. Measurement errors of up to 5.5% were found on a calibration target of known size. The second system consisted of a pair of still-image digital cameras mounted inside a midwater trawl. Processing of the stereo images allowed fish length, fish orientation in relation to the camera platform, and rela- tive distance of the fish to the trawl netting to be determined. The video system was useful for surveying fish in Alaska, but it could also be used broadly in other situations where it is difficult to obtain species-compo- sition or size-composition informa- tion. Likewise, the still-image system could be used for fisheries research to obtain data on size, position, and orientation of fish. Manuscript submitted 21 January 2010. Manuscript accepted 27 May 2010. Fish. Bull. 108:352-362 (2010). The views and opinions expressed or implied in this article are those of the author (or authors) and do not necessarily reflect the position of the National Marine Fisheries Service, NOAA. Use of stereo camera systems for assessment of rockfish abundance in untrawiable areas and for recording pollock behavior during midwater trawls Kresimir Williams (contact author) Christopher N. Rooper Rick Towler Email address for contact author: Kresimir.Williams@noaa.gov Alaska Fisheries Science Center National Marine Fisheries Service National Oceanic and Atmospheric Administration 7600 Sand Point Way NE Seattle, Washington 98115 For modern fisheries stock assess- ments, fisheries-independent data are necessary to estimate population abundances and population trends. For most marine species, fisheries- independent abundance estimates are primarily obtained from large-scale multispecies bottom trawl surveys (e.g., Gunderson and Sample, 1980) and from acoustic surveys of pelagic fish stocks (e.g., Karp and Walters, 1994). Although acoustic backscatter is used to measure fish abundance, midwater trawl samples are needed to determine the size and species composition of acoustically sampled fish populations. Both of these survey methods require physical sampling of trawl catches and such sampling can result in unrepresentative data in several ways. Bottom-trawl surveys are limited to the areas they can sample be- cause many research trawls are not constructed to efficiently fish over a rough or rugose seafloor. Thus, sur- veys with bottom trawls may not be appropriate for some species with affinities for untrawiable habitat or in survey areas where significant patches of untrawiable ground can be found (Zimmermann, 2003; Cordue, 2007). In Alaska, semipelagic species such as northern rockfish ( Sebastes polyspinis) and Pacific ocean perch (S. alutus) are an important part of the commercial catch, but they also show some affinity for untrawiable areas (Clausen and Heifetz, 2002; Rooper et al., 2007). In addition, inferences from spe- cies- and size-composition data ob- tained from trawl catches can be biased on account of trawl selectiv- ity. Trawls are generally designed to capture larger, market-size fish, and their design for this selected size results in the under-retention of juvenile size classes. In acoustic surveys of walleye pollock (Theragra chalcogramma ), biases in midwater trawl catches directly translate into biases in abundance estimates for areas where large and small fish are found (Godo et al., 1998). Selective retention of fish is a consequence of size and species-dependent fish be- havior during the trawling process. Observation of fish reactions to trawl gear is critical to understanding the behavioral mechanisms responsible for trawl selectivity and to develop future trawl gear for research. Here, we describe the use of stereo photography to sample rockfish in un- trawlable habitats using a drop unit with a stereo video camera (hereaf- ter termed “video-drop” camera), and to study fish behavior in midwater trawls using a trawl-mounted pair of still-frame stereo cameras (hereafter, termed “still-frame” camera). Stereo cameras have been successfully used to measure fish in controlled aquacul- ture settings (Ruff et al., 1995; Har- vey et al., 2003) and in open water Williams et al.: Use of stereo-camera systems in assessing rockfish abundance and pollock behavior 353 Table 1 Design, manufacturer, and cost (approximate estimates in U.S. dollars) for drop stereo-video camera and still-frame stereo- camera systems used for surveying untrawlable habitat and studying fish behavior in midwater research trawls. Both sys- tems were used in the field in July 2008 and July 2007, respectively. HID=high-intensity discharge; LED=light-emitting diode; UHMW = ultra high molecular weight plastic. System Component Design Manufacturer Cost Drop stereo- HID light HID Xenon lights, 12 V, 50 W Underwater Lights USA $814 video Video line driver Balanced line driver and transceiver Nitek $133 camera Conducting cable 4 conductor wire, 4.72 mm diameter Rochester Cable $1601 Sled frame Aluminum channel and tubing Local manufacture $2000 Winch and slip ring CSW-6 electronic win A.G.O. Environmental $11,268 Underwater housings cameras 5" diameter Local manufacture $729 Underwater housings lights — Local manufacture $729 LED sync — Ramsey Electronics $24 Underwater cable and connections — Teledyne Impulse $614 Batteries 4x12 V 4 Ah NiMH Energy sales Total video system cost $396 $18,308 Still-frame Strobe Oceanic 3000 Oceanic $990 stereo Cameras Canon Digital Rebel Xt (8Mp) Canon USA $1100 camera Lenses Canon EF 28 mm f/2.8 Canon USA $450 Microcontroller & circuitry — Local manufacture $150 Underwater housings and viewports 10" floats, 1.5" acrylic flat viewports Local manufacture $1400 Mounting frame UHMW plastic and aluminium stock Local manufacture $350 Underwater connections — Teledyne Impulse $650 Batteries Software 3x12 V 4 Ah NiMH Matlab V 7.6 Energy sales Total still-frame system cost Mathworks $297 $5387 (i.e., van Rooij and Videler, 1996; Shortis et al., 2009). The recent development of high-resolution digital cam- eras has vastly improved the performance and reduced the complexity of image-based sampling because high- quality digital images can be directly analyzed with image-processing software. In general, stereo methods provide highly precise measurements in comparison to single-camera-based photogrammetric methods (Har- vey et al., 2002). However, these systems necessitate maintaining a stable two-camera geometry and must be initially calibrated with targets of known sizes. Despite these constraints, stereo photography is widely used in optical-based sampling in a variety of marine studies. We demonstrate the precision of stereo-camera-based measurements, attainable from initial deployments in the field, in comparison with traditional survey mea- surements. The results show that stereo-based optical sampling is a viable method for augmenting bottom- trawl data for abundance estimations; the stereo cam- eras allow scientists to survey sampling areas that are unavailable to standard survey trawl gear. In addition, stereo cameras can be used to observe and quantify the behavior of fish in the process of being captured by trawl gear to further improve estimates of abundance because they allow scientists to determine the potential biases in trawl-based catch data. Materials and methods Sampling untrawlable areas with the video-drop camera system The design of the video-drop system was based on two key needs. Because rockfish are found in areas of high relief, the camera needed to have adequate protection for their electronic components and have the ability to maintain visual contact with the bottom through rough substrate areas. Therefore, essential to sampling with this camera system was the ability to live-view the video and the use of a quick-responding winch system that could be controlled by the operator aboard the research vessel. The specifications of the camera components are presented in Table 1. The winch used to deploy and retrieve the camera system and navigate the seafloor was a CSW-6 multi- purpose win (A.G.O. Environmental, Nanaimo, BC, Canada; Fig. 1A). The winch motor was a 3/4 horse- 354 Fishery Bulletin 108(3) Figure 1 Drop stereo-video camera system showing (A) the quick-responding winch, and (B) the locations of cameras and lights in the aluminum frame. (C) The still-frame camera system, which is deployed in a midwater trawl. (D) The camera housings (shown) were constructed from deep-water trawl floats which provided buoyancy and reduced the weight of the cameras. power Leeson wet-duty motor powered by 100 V AC. The winch speed ranged from 43 m/min (with a bare drum) to approximately 58 m/min (with a full drum). The approximate square-in area of the winch was 48 in2 (0.031 m2) and its weight was 155 lb (70.3 kg). The drum was 16 in (40.6 cm) in circumference and was filled with 1312 ft (400 m) of 3/16-in (4.72-mm) conducting wire. The wire had a breaking strength of 3300 lb (1497 kg) and was connected to the camera sled with a cable-grip. The video feed from the cameras was passed up a cable and through a four-conductor slip ring mounted on the winch and routed into a junction box where it was connected to a monitor for real-time viewing. The protective cage around the camera and lights was constructed of 1.5-in (3.81-cm) aluminum tubing, and the interior members of the frame were composed of 6- in (15.24-cm) aluminum channel (Fig. IB). A tail chain was attached to the rear of the ventral surface of the cage to drag along the seafloor to help keep the camera unit in contact with the seafloor and oriented forward during deployment. The tail chain was connected to the cage by a short piece of twine to act as a “weak link” in case the tail chain snagged on the seafloor. The underwater video was recorded by two identical Sony TR-900 camcorders (Sony Electronics Inc., San Diego, CA) located inside the camera housings. The cameras were capable of collecting 720 p progressive scan video images at a resolution of 720x480 pixels. The video was recorded to digital video tapes for a maximum of one hour per tape. Because the cable was too long (400 m) to transmit a standard video signal, it was transformed by using a video balun (in the camera housing) and a receiver (at the winch) to reconvert the video signal back to a viewable picture of the seafloor to use for real-time navigation. Cameras were placed in separate housings constructed of titanium tubing and that had a glass dome port (pressure-rated to 3000 m depth) covering the lens. The lens of each camera was keyed to its port to prevent the camera from being inserted into the housing in a position other than the exact keyed position and stabilized the relative position of the cameras from deployment to deployment — an important consideration for accurate Williams et al.: Use of stereo-camera systems in assessing rockfish abundance and pollock behavior 355 measurement of targets (Shortis et al., 2000). The housings were mounted side by side on the aluminum frame (Fig. IB). Illumination was provided by two lights mounted above the camera housings inside the aluminum frame (Fig. IB). The lights were 50-watt high-intensity dis- charge (HID) Xenon lights with 5300 lumen output and 3900 Kelvin color temperature. The lights were inserted into 3-in (7.62-cm) diameter titanium hous- ings and the entire light weighed 5 lb (2.27 kg). The lights were powered by a battery located in the camera housing and linked to the light housing by underwater connectors. Four rechargeable 4 Ah 12 V nickel-metal hydride batteries were connected in parallel to provide approximately 1.5 hour of light per deployment. Each light housing was mounted on an adjustable mount that allowed even illumination of the target. Observing fish behavior in a trawl with the still-frame system The still-frame system was designed to be light and small enough to be easily attached to the inside of a survey trawl without significantly changing the fishing activity of the net. The system also needed to provide adequate illumination and resolution in order to allow the fish inside the net to be observed at a range of up to 6 m as they passed though a midwater survey trawl 40 m ahead of the codend. A pair of Canon Rebel Xt 8 megapixel digital single-lens reflex cameras (Canon USA, Lake Success, NY) were used to capture fish images. Both cameras were outfitted with 4-gigabyte compact flash memory cards for storage of the images. A high-power wide-angle Xenon strobe (90°, 150 W/s) was used to illuminate the field of view. Three 4-Ah 12 V batteries were mounted in the strobe housing; two were connected to the strobes and the third was used to power the cameras. The cameras were mounted in separate housings made from 10-in (25-cm) diameter deep-water-rated (1800 m) trawl floats. Images were taken though a 25- mm thick flat acrylic viewport. The strobe and batteries were mounted in a third float housing (Fig. 1C). All three float housings were secured on a sled constructed of 25-mm thick plastic plate and aluminum rails for protection. The approximate weight of the complete as- sembly in air was 30 kg and was positively buoyant be- cause of the float housings. Quick-release trigger snaps were attached to the ends of the plastic mounting board for attachment to the inside of the trawl. The cameras were aimed across the trawl, perpendicular to the wa- ter flow to provide lateral views of fish passing by. The trigger on the camera shutter was controlled by using a microprocessor that was programmed for the study and that located in one of the camera housings. A two-axis tilt sensor was attached to the microprocessor board to allow measurements of fish tilt (deviation of snout-tail axis from the horizontal) and yaw (angle of fish head- ing in the horizontal plane) to be adjusted from being relative to the camera platform to being in absolute orientation. A pressure switch was used to activate the system once the depth exceeded 20 meters. Images were taken at intervals of 5 s to reduce the influence of light on fish behavior and to ensure that a new group of fish was observed in each frame. The system was capable of taking about 400 images or operating for 33 min of trawl time per deployment. Calibration of the two types of stereo cameras The same calibration procedure was used for both stereo- camera systems. The basic procedure required collecting images of a target plate with a printed 10x10 square checkerboard pattern of known dimensions (50x50 cm squares for the video-drop system, 100 x 100 cm for the still-frame system). This calibration was performed underwater. The video-drop system cage was suspended in the water while the research vessel was secured to the dock. The approximate depth of the camera was 1 m and the distance from the target was 2 m. The checkerboard target was lowered into the water along the vessel until it was plainly visible in both cameras. The target was then slowly moved horizontally and vertically through the field of view of both cameras. Up to 15 min of calibration video was collected by this method. For the still-frame system, an external trigger cable was attached to the assembly, and the system slowly moved about while capturing images of the fixed checkerboard plate. To calibrate the video-drop system, progressive scan video images were collected at 29.97 frames/s in each camera, and the beginning of the video feed from each camera was aligned by using a light-emitting diode (LED) synchronization light at the beginning of de- ployment. This process was repeated at the end of the deployment to confirm that the video frames were still aligned. For the calibration procedure, still frame im- ages were extracted from the aligned video at 1-s in- tervals with Adobe Premier software (Adobe Systems, Inc., San Jose, CA). Synchronization was not necessary for the still-frame system because the cameras were triggered simultaneously. Approximately 20 paired im- ages where the target checkerboard was visible in both cameras were randomly selected for the calibration of each camera system. The calibration parameters were estimated with the camera calibration toolbox in Matlab, a freely available software analysis toolbox built with Matlab computing language (Mathworks, Inc.; Bouget, 2008; Fig. 2). For each image pair, the position of the corner points of the checkerboard pattern were identified by clicking on the images and the location of these points in the still im- ages was computed by the calibration software to deter- mine the intrinsic parameters of each camera. Intrinsic parameters were used to correct the individual images for optical distortion resulting from the camera lenses. The checkerboard pattern allowed the software to auto- matically pinpoint exact corner locations based on the color contrast of the square boundaries, making the initial precision of the manual clicking less critical. 356 Fishery Bulletin 108(3) Figure 2 The checkerboard pattern used to calibrate and test the drop stereo-video camera and the still- frame stereo-camera systems. Images used for calibration were processed by using the camera calibration toolbox written for Matlab. An image taken by the still-frame stereo-camera system is shown with the user-selected extreme corners shown as white circles and the automated extrac- tion of all intermediate corners shown with white crosses. This procedure was repeated for up to 20 different images with both systems. Stereo calibration required that the checkerboard corners be identified in the same order in each of the synchronous image pairs to correctly match up the analogous corner points. These points, once corrected for optical distortion in individual cameras, were used to compute the epipolar geometry, by iteratively solving for the translation and rotation vectors that describe the relationship between the coordinate systems of the two cameras (Xu and Zhang, 1996). Once these matrices were estimated by the software, the three-dimensional position of a target point viewed in both cameras could be determined by triangulation. Fish measurements with the camera Fish lengths were measured by using stereo triangu- lation functions supplied with the camera calibration software package (Bouguet, 2008). For the video-drop system, images were extracted from the two video feeds at 1-s intervals. The images were synchronized at the beginning of each transect before deployment by using the LED synchronization light. The images were checked at the end of each transect to confirm that the cameras remained synchronized. Length measurements were obtained by identifying the pixel coordinates of corresponding pixel locations in the left and right camera still frames such as a fish snout and tail (Fig. 3). These points were used to solve for the three-dimensional coordinates of the points in the images by triangulation, by using the calibration-derived parameters. Once the three-di- mensional coordinates of the fish snout and tail were obtained, the length was measured as the simple Eu- clidian distance between the points in real space. This measurement method underestimated length for fish whose bodies were curved; however; fish in the video and still camera were almost exclusively seen with little or no curvature in their bodies and the few indi- viduals that were obviously strongly curved were not measured. Length data were collected by using a basic software application built with the Matlab computing language (Fig. 4; available from the authors upon re- quest), which incorporated the triangulation function supplied by the calibration toolbox. In addition to length measurements, the three-di- mensional coordinates extracted from the still-frame images provided data on the position and orientation of walleye Pollock in relation to the trawl (Fig. 5). These data were used to determine distances of pollock targets to trawl components for position of fish and to calculate tilt and yaw for orientation of fish. Data collections Field testing of the video-drop system was conducted 12-15 July 2008 at Zhemchug Ridges, located on the eastern Bering Sea shelf adjacent to Zhemchug Canyon where a sizable rockfish population is pres- ent in untrawlable and isolated rocky ridge area (Fig. 6; Rooper et al., in press). The camera system was deployed off the side of the vessel FV Vesteraalen by a winch suspended from a block attached to the vessel’s crane. The camera sled drifted with the prevailing current, while the camera winch operator kept the seafloor in view and avoided any obstacles using real- time navigation. Stereo video was collected over 11 transects, each ranging in length from 3.5 to 49.5 min and covering distances of 95 m to 1673 m. Observa- tions of trawl movements with the still-frame system were made during acoustic surveys of pollock in the eastern Bering Sea in June and July 2007 onboard the RV Oscar Dyson (Fig. 6). Testing of the calibrations for the two camera systems To test the video calibration five random still images was selected from the video-drop system of the checkerboard taken at the beginning and end of the study. Three intervals of 10 cm, 20 cm and 30 cm each were measured three times from the top to the bottom of the checker- board (?i = 3 for each interval) and averaged within each frame. The average from each frame multiplied by the interval combination was then tested in an analysis of variance to determine whether there were significant differences between measurements from the first and second measurement set. Williams et al.: Use of stereo-camera systems in assessing rockfish abundance and pollock behavior 357 fish length = distance ([Xh, Yh, Zh\, [Xv Yv Zt]) Method for determining fish length measurements by using stereo images. The three-dimen- sional coordinates of the fish head and tail (Xh,Yh,Zh; Xt,Yt,Zt > were determined by stereo- triangulation and by using the image-based coordinates from the image pairs (i.e., LXh,LY h; RXh,RYh). Fish length was estimated as the Euclidian distance between the three-dimensional points of the head and tail. Results Calibration The estimates of distance of the fish to the trawl deter- mined with the second calibration of the drop-video system were significantly larger and more variable than those from the first measurement set across all three intervals (Fig. 7). Differences between the mean measurements and known values in the second set ranged from 6.6% to 8.2%. However, the 95% confidence intervals for both sets included the actual values for the intervals in all cases, and the coefficients of varia- tion for the measurements ranged to 5.5% of the mean value, indicating that the length measurements were reasonably precise. A similar procedure was also per- formed with the still-frame system, but only a single set of validation measurements was made before the start of field operations. The results of this set closely matched that of the first set made with the video cam- eras (Fig. 7). Fish lengths determined with the video-drop system The adult rockfish observed in the video were northern rockfish (96.94%), unidentified adult rockfish ( Sebastes spp., 0.98%), adult Pacific ocean perch (0.49%), and dusky rockfish (S. ciliatus, 1.60%), whereas most of the juveniles that were identified to species were Pacific ocean perch (Rooper et al. in press). Some of the juve- nile rockfishes observed in the video were too small to identify to species. Individuals of each species group were randomly chosen to be measured in proportion with their abundance. Up to 200 randomly selected individual rockfish were measured in each transect, resulting in a total of 1489 length measurements. Rockfish were measured by using fork length only if both the tip of their snout and the end of the tail were plainly visible in both still images. If the randomly chosen rockfish could not be measured, the next available rockfish of the same species group that was deemed measureable was chosen. In a few cases, where the occurrence of a spe- cies group was very small (<5 individuals in a transect), none were measured. A random sample of 20 rockfish that were observed in successive still frames of both video cameras was used to determine measurement precision and to estimate distance of the fish from the camera. These fish were measured in up to four consecutive frames and their estimated length were compared by using linear regres- sion (Fig. 8). The percent difference between successive length measurements was not significantly related to the average fish length (P=0.28); in other words, there was no length-related bias in the measurements. The length data were also tested for a relationship with dis- tance from the camera by using linear regression. There was no bias in the measurements of fish for distance from the camera (P=0.29). The standard deviation of 358 Fishery Bulletin 108(3) Figure 4 Computerized display of the stereo processing options for the drop stereo-video with a custom-built application written in the Matlab computing language. Synchronous images were extracted from videos taken by two cameras and used to estimate fish length. Images were taken in Zhemchug Ridges in the eastern Bering Sea in July 2008. the percent difference in multiple measurements of the same fish was 0.076. Analysis of still-frame stereo images Ten deployments were made with the still-frame system, and -200 fish were measured per deployment. Catches consisted almost exclusively of walleye pollock (>99%). A comparison between the length frequencies derived from the stereo analysis (/z = 360) and physical measurements of fish captured in the codend (n=1260, Fig. 9) showed that optical sampling approximates the length-frequency distribution of fish caught, despite the smaller sample size for optical sampling. In addition to length measurements, the stereo analy- sis provided data on walleye pollock orientation and their relative position within the trawl. Quantitative descriptions of the distribution of tilt and yaw angles were easily calculated by using the same points in im- ages (head and tail) derived for fish lengths (Fig. 10). To calculate the position of fish within the trawl addi- tional corresponding points along the trawl panel were identified and their three-dimensional coordinates were determined by the triangulation process outlined above (Fig. 5). Discussion The potential of stereo cameras for measuring marine organisms has been shown in many studies (i.e., Shortis et al., 2000; Flarvey et al., 2003), but here we present a description of the complete implementation of stereo cameras, including equipment costs (Table 1), image analysis process, and expected precision in data from these systems. The two stereo-camera systems described here were studied for their potential to provide infor- mation to augment fisheries assessment surveys in Alaska. Specifically, the stereo-camera systems in our study provided species and length data for untrawlable regions located within bottom-trawl survey boundaries and provide a new method for studying the behavior of fish in a midwater trawl. Our main goal was to present field-tested methods to provide quantifiable image-based data for fisheries surveys and our results may help similar research with stereo-camera-based sampling systems. The video-drop system was useful for estimating rockfish size and species composition in field tests in Alaska. Error rates for size were on the order of 8.2% or less, which equates to about 2.5 cm for a 30-cm fish. Compared with other studies with error rates of -0.1% to 0.7% in stereo-video systems (Harvey et ah, 2002; Harvey et al., 2003; Shortis et al., 2009), the measure- ment error rate in our study was high. This rate repre- sents systematic error most likely caused by the need to remove cameras from the housing after each deploy- ment because a slight misalignment of the cameras in relation to the position at calibration would reduce the precision of the measurements. Ruff et al. (1995) report an achievable level of precision in measuring fish of 3.5%, based on repeat measurements of indi- viduals, which is also better than the 5.9% observed in our study. The error rates also compare well to the rates of 1-5% for measuring rigid items with parallel lasers (Rochet et al., 2006). However, only fish on or near which the parallel laser beams are projected can be measured. This restriction limits the measurement sample size. In contrast, any fish simultaneously viewed by both cameras in a stereo-camera system can be mea- Williams et al.: Use of stereo-camera systems in assessing rockfish abundance and pollock behavior 359 A raw image B processed image C 3D reconstruction Figure 5 Images of walleye pollock (Theragra chalcogramma ) from a still-frame stereo camera in a midwater trawl and a 3-D reconstruction of fish in relation to the trawl net. (A) Fish lengths were measured by enlarging the image of a fish and indicating the position of the snout and tail (shown as U) in both right and left raw images (only the left image is shown above). (B) The chosen fish endpoints are overlaid on the image as lines. In addition to estimates of fish length, stereo-processing allows the position of fish in relation to the trawl to be estimated. Additional points in the images can be determined by finding corresponding left-right image pixel coordinates (B, shown as crosses). (C) Following stereo- triangulation, a three-dimensional plot shows the fish targets as arrows and trawl mesh knots as dots. sured, and thus the number of fish that can be measured is larger from the same length transect. Improvements in the quality of the still-frame im- ages and in the collection of calibration data from a target at the beginning of each tran- sect may allow more precise measurements to be taken in future studies. Given our inability with other survey gears to determine fish size and species composition in un- trawlable habitats, the use of stereo cameras holds promise for stock assessments of rock- fish and other species. Stereo- camera-based sampling could also be used broadly wher- ever gears other than bottom trawls are needed to obtain species- or size-composition information. Lengths of rockfish derived from the video-drop system were generally comparable to trawl catch-based size distri- butions for the species exam- 180’0' 177°0'W 174°0'W 171°0'W 168°0'W Figure 6 Map of study areas in the eastern Bering Sea showing the location of field tests of the drop stereo video cameras for sampling untrawlable areas (black square) in July 2008 and for sampling fish behavior in a trawl (circles) with a still-frame stereo camera in July 2007. 360 Fishery Bulletin 108(3) ~o C 0 30 20 10 0 -10 -20 -30 4- * 4] ~ ! l -1 _T_ ■ _L T -I -p- video - 1 st measurement set video - 2nd measurement set still-image calibration 0.1 0.2 0.3 Distance measured (m) Figure 7 Results of measuring known distances on the check- erboard (see Fig. 2) during the first and second cali- brations of the drop stereo video-camera system and a still-frame stereo-camera system used for estimating fish length and studying behavior. Calibration is a necessary step in the use of stereo cameras to allow measurements to be made from images. Values repre- sent the percent deviation in measurements in relation to known values, including 95% confidence intervals based on five measurements. -20 -16 -12 -8 -4 0 4 8 12 16 20 Percent difference from mean Figure 8 Frequency of percent differences from multiple length measurements of n = 20 rockfish and the mean length of the fish. Individual fish were measured multiple times from a series of images extracted from video taken by a drop stereo-video camera. Measurements from each image were then compared to the mean measurement for that individual to estimate poten- tial measurement error. Rockfish observations were collected at Zhemchug Ridges, eastern Bering Sea in July 2008. A _| cpd Figure 9 Comparison between length frequencies of walleye pollock ( Tlieragra chal- cogramma ) estimated from (A) the images from a still-frame stereo camera (« = 360) within a midwater trawl, and length frequencies obtained from (B) fish captured in the codencl and directly measured (n = 1260). The smaller camera-based sample results are similar to the direct measurements in their overall size distribution, but there was less definition for larger fish (>20 cm). Data were pooled from three trawl samples taken in the eastern Bering Sea in July 2007. ined. Northern rockfish lengths from stereo-video images taken along transects at Zhemchug Ridg- es ranged from 9 to 41 cm (mean length = 30.0 cm). In three bot- tom-trawl surveys near the Aleu- tian Islands, Clausen and Heifetz (2002) found that the mean size of northern rockfish was 29.9 cm and ranged from 15 to 38 cm. Ju- venile Pacific ocean perch lengths (2.6 cm to 25.0 cm) were similar but ranged to smaller sizes than those found for the Aleutian Is- lands (8.3cm to 24.9 cm; Boldt and Rooper, 2009). Lengths of juvenile Pacific ocean perch obtained from stereo video were also similar to those in three experimental tows in the Zhemchug Ridges area in 2004 and 2007 (juveniles ranged from 10 cm to 25 cm). However, these lengths were measured from fish captured during bottom trawl hauls, where the incidence of smaller fish may have been due to reduced catchability of smaller individuals. Although these ob- servations are not meant to serve Williams et al.: Use of stereo-camera systems in assessing rockfish abundance and pollock behavior 361 Figure 10 Orientiation of walleye pollock ( Theragra clialcogramma) within a midwater trawl. A still-frame stereo camera was used to determine (A) the distribution of individual fish tilt angle (deviation of snout-tail axis from the horizontal), and (B) the yaw (angle of fish heading in the horizontal plane). Most fish were oriented horizontally and facing toward the trawl opening. Image data were collected in the eastern Bering Sea in July 2007 as part of a study of fish behavior within the trawl. as a quantitative comparison of trawl- and stereo-camera-derived size estimates, they demonstrate the similarities in information supplied by the two methods and the potential for stereo cameras to overcome some problems with the catchability of juvenile fish. The stereo camera was very use- ful for studying behavior of pollock in the trawl. The data show the possibility of performing a length- based analysis of behavior which will directly contribute to studies of gear selectivity and future de- signs of scientific trawl gear. Al- though a postsurvey calibration was not performed with the still- frame system, the cameras were securely fastened in the housings and were not removed during the entire data collection, thus main- taining intercamera spacing and angles. The agreement between the catch-based length measure- ments and the stereo-derived lengths provides direct validation of the stereo-derived measurements. The low sampling frequency of 1 frame per 5 s ensured minimal influ- ence of the artificial lighting from the cameras on behavior because the fish photographed had not been previously exposed to the light source. Recent development of high-resolution digital im- aging devices and an increased access to custom de- signed, freely available software tools have made ste- reo-camera methods easy to implement by research groups without direct expertise in the subject. The camera calibration toolbox (Bouguet, 2008) provided the basis for software development. Although the cur- rent analysis approach is still fairly time intensive, the volumes of data analyzed were not very large. In a routine application of stereo-video cameras in untraw- lable areas, additional levels of automated processing would likely be required because the quantity of video footage would substantial. For some aspects of the analysis, such as the matching of targets on the stereo cameras and the extraction of fish lengths, automation may be attainable, whereas automating more difficult tasks of isolating and identifying fish targets may not be feasible. Stereo photography will continue to be developed as survey tools are developed for monitoring fish stocks and thereby improving the quality of stock assessments of fishery resources in Alaska. Some challenges remain; for instance, the challenge of institutionalizing image- based sampling as a routine survey method for untraw- lable habitats. As a method of studying fish behavior in trawls, stereo cameras provide promising results by allowing three-dimensional reconstructions of the trawl environment. Acknowledgments The development and deployment of the camera and winch system was possible only through the assistance of G. McMurrin, G. Mundell, B. Lauth, G. Hoff, and especially S. McEntire. T. Cosgrove, K. Sjong, L. Mavar, and M. Booth of the FV Vesteraalen were instrumental in conducting the field tests. Calibration and deploy- ment of the still-frame camera system were significantly aided by D. Jones and A. McCarthy. This manuscript was reviewed by D. Somerton, G. Hoff, R. Lauth, and M. Wilkins. Literature cited Boldt, J. L., and C. N. Rooper. 2009. Abundance, condition, and diet of juvenile Pacific ocean perch ( Sebastes alutus) in the Aleutian Islands. Fish. Bull.:107:278-285. Bouguet, J. Y. 2008. Camera calibration toolbox for Matlab [online], [Available from http://vision.caltech.edu/bouguetj/calib_ doc/index. html (accessed September 2008)]. Clausen, D. M., and J. Heifetz. 2002. The northern rockfish, Sebastes polyspinis, in Alaska: commercial fishery, distribution, and biology. Mar. Fish. Rev. 64:1-28. Cordue, P. 2007. A note on non-random error structure in trawl survey abundance indices. ICES J. Mar. Sci. 64:1333- 1337. Godo, O. R., W. A. Karp, and A. Totland. 1998. Effects of trawl sampling variability on precision of acoustic abundance estimates of gadoids from the 362 Fishery Bulletin 108(3) Barents Sea and the Gulf of Alaska. ICES J. Mar. Sci. 55:86-94. Gunderson, D. R., and T. M. Sample. 1980. Distribution and abundance of rockfish off Wash- ington, Oregon and California during 1977. Mar. Fish. Rev. 42:2-16. Harvey, E. S., M. Cappo, M. R. Shortis, S. Robson, J. Buchanan, and P. Speare. 2003. The accuracy and precision of underwater measure- ments of length and maximum body depth of southern bluefin tuna (' Thunnus maccoyii ) with a stereo-video camera system. Fish. Res. 63:315-326. Harvey, E. S., D. Fletcher, and M. R. Shortis. 2002. Estimation of reef fish length by divers and by stereo-video: a first comparison of the accuracy and precision in the field on living fish under operational conditions. Fish. Res. 57:255-265. Karp, W. A., and G. E. Walters. 1994. Survey assessment of semi-pelagic Gadoids: the example of walleye pollock, Theragra chalcogramma , in the Eastern Bering Sea. Mar. Fish. Rev. 56:8-22. Rochet, M. J., J. F. Cadiou, and V. M. Trenkel. 2006. Precision and accuracy of fish length measurements obtained with two visual underwater methods. Fish. Bull. 104:1-9. Rooper, C. N., J. L. Boldt, and M. Zimmermann. 2007. An assessment of juvenile Pacific ocean perch ( Sebastes alutus) habitat use in a deepwater nursery. Es- tuar. Coast. Shelf Sci. 75:371-380. Rooper, C. N., G. H. Hoff, and A. DeRobertis. In press. Assessing habitat utilization and rockfish bio- mass on an isolated rocky ridge using acoustics and stereo image analysis. Can. J. Fish. Aquat. Sci. Ruff, B. P., J. A. Marchant, and A. R. Frost. 1995. Fish sizing and monitoring using stereo image analysis system applied to fish farming. Aquat. Engi- neer. 14:155-173. Shortis, M. R., S. Miller, E. S. Harvey, S. Robson. 2000. An analysis of the calibration stability and mea- surement accuracy of an underwater stereo-video system used for shellfish surveys. Geomatics Res. Aust. 73:1-24. Shortis, M. R., J. W.Seager, A. Williams, B. A. Barker, and M. Sherlock. 2009. Using stereo video for deep water benthic habitat surveys. Mar. Tech. Soc. J. 42:28-37. Van Rooij, J. M, and J. J. Videler. 1996. A simple field method for stereo-photographic length measurement of free swimming fish: merits and constraints. J. Exp. Mar. Biol. Ecol. 195:237-249. Xu G., and Z. Zhang. 1996. Epipolar geometry in stereo, motion, and object recognition: a unified approach, 336 p. Kluwer Aca- demic Pubis., Norwell, MA. Zimmermann, M. 2003. Calculation of untrawlable areas within the bound- aries of a bottom trawl survey. Can. J. Fish. Aquat. Sci. 60:657-669. 363 Best paper awards for 2009 The award for best publication of the year is given to authors who are employees of the National Marine Fish- eries Service, NOAA, and whose articles are judged to be the most noteworthy of those published in Fishery Bulletin and Marine Fisheries Review. Authors from the National Marine Fisheries Service are noted in bold below. The winners for Fishery Bulletin: First prize: Jeffrey J. Polovina, Melanie Abecassis, Evan A. Howell, and Phoebe Woodworth Increases in the relative abundance of mid-trophic level fishes concurrent with declines in apex predators in the subtropical North Pacific, 1996-2006. Fish. Bull. 107:523-531. Honorable mention: Richard Brill, Peter Bushnell, Leonie Smith, Coley Speaks, Rumya Sundaram, Eric Stroud, and John Wang The repulsive and feeding-deterrent effects of electro- positive metals on juvenile sandbar sharks (Carcharhi- nus plumbeus). Fish. Bull. 107:298-307. The winners for Marine Fisheries Review: Phil Clapham and Yulia Ivashchenko. A Whale of a Deception This article documents the untold numbers of whales taken by the Soviet whaling industry in violation of international quotas. The article was discussed in the “News of the Week” section of the journal Science'. “Mystery of the Missing Humpbacks Solved by Soviet Data.” Science, vol. 324:1132. May 29, 2009. Mar. Fish. Rev. 71< 1):44— 52. 364 Fishery Bulletin Guidelines for authors Manuscript Preparation Contributions published in Fishery Bulletin describe original research in marine fishery science, fishery engineering and economics, as well as the areas of marine environmental and ecological sciences (including modeling). Preference will be given to manuscripts that examine processes and underlying patterns. Descriptive reports, surveys, and observational papers may occa- sionally be published but should appeal to an audience outside the locale in which the study was conducted. Although all contributions are subject to peer review, responsibility for the contents of papers rests upon the authors and not on the editor or publisher. Submission of an article implies that the article is original and is not being considered for publication elsewhere. Articles may range from relatively short contributions (10-15 typed, double-spaced pages, tables and figures not included) to extensive contributions (20-30 typed pages). Manu- scripts must be written in English; authors whose native language is not English are strongly advised to have their manuscripts checked by English-speaking col- leagues before submission. Title page should include authors’ full names and mailing addresses and the senior author’s telephone, fax number, and e-mail address, and a list of key words to describe the contents of the manuscript. Abstract should be limited to 200 words (one-half typed page), state the main scope of the research, and emphasize the author’s conclusions and relevant findings. Do not review the methods of the study or list the contents of the paper. Because abstracts are circulated by abstract- ing agencies, it is important that they represent the research clearly and concisely. Text must be typed in 12 point Times New Roman font throughout. A brief introduction should convey the broad significance of the paper; the remainder of the paper should be divided into the following sections: Materials and methods, Results, Discussion (or Conclusions), and Acknowl- edgments. Headings within each section must be short, reflect a logical sequence, and follow the rules of multi- ple subdivision ( i.e. , there can be no subdivision without at least two items). The entire text should be intelligible to interdisciplinary readers; therefore, all acronyms, abbreviations, and technical terms should be written out in full the first time they are mentioned. Include FAO common names for species in the list of keywords and in the introduction. Regional common names may be used throughout the rest of the text if they are dif- ferent from FAO common names which can be found at http://www.fishbase.org/search.html. Follow the U.S. Government Printing Office Style Manual (2000 ed.) and Scientific Style and Format: the CSE Manual for Authors, Editors, and Publishers (7th ed.) for editorial style; for fish nomenclature follow the most current issue of the American Fisheries Society’s Common and Scien- tific Names of Fishes from the United States, Canada, and Mexico, 6th ed. Dates should be written as follows: 11 November 2000. Measurements should be expressed in metric units, e.g., 58 metric tons (t); if other units of measurement are used, please make this fact explicit to the reader. Write out the numbers zero through nine unless they form part of measurement units (e.g., nine fish but 9 mm). Refrain from using the shorthand slash (/), an ambiguous symbol, in the general text Literature cited comprises published works and those accepted for publication in peer-reviewed literature (in press). Follow the name and year system for citation format in the “Literature cited” section (that is say, citations should be listed alphabetically by the authors’ last names, and then by year if there is more than one citation with the same authorship). If there is a sequence of citations in the text, list chronologically: (Smith, 1932; Green, 1947; Smith and Jones, 1985). Abbrevia- tions of serials should conform to abbreviations given in the Serial Sources for the BIOSIS Previews Database. Authors are responsible for the accuracy and complete- ness of all citations. Literature citation format: Author (last name, followed by first-name initials). Year. Title of report or manuscript. Abbreviated title of the series to which it belongs. Always include number of pages. Cite all software and special equipment or chemical solutions used in the study, not in a footnote but within parentheses in the text (e.g., SAS, vers. 6.03, SAS Inst., Inc., Cary, NC). Tables are often overused in scientific papers; it is seldom necessary or even desirable to present all the data associated with a study. Tables should not be excessive in size and must be cited in numerical order in the text. Headings should be short but ample enough to allow the table to be intelligible on its own. All unusual symbols must be explained in the table legend. Other incidental comments may be footnoted with italic numeral footnote markers. Use asterisks to indicate probability in statistical data. Do not type table legends on a separate page; place them above the table data. Do not submit tables in photo mode. • Zeros should precede all decimal points for values less than one. • Sample size, n, should be italicized. • Capitalize the first letter of the first word in all labels within figures. • Do not use overly large font sizes in maps and for units of measurements along axes in figures. • Do not use bold fonts or bold lines in figures. • Do not place outline rules around graphs. • Do not use horizontal lines through graphs to indicate measurement units. • Use a comma in numbers of five digits or more (e.g. 13,000 but 3000). • Maps require a North arrow and degrees latitude- longitude (e.g., 170°E). Fishery Bulletin 108(3) 365 Figures include line illustrations, photographs (or slides), and computer-generated graphs and must be cited in numerical order in the text. Graphics should aid in the comprehension of the text, but they should be limited to presenting patterns rather than raw data. Figures should not exceed one figure for every four pages of text. Figures must be labeled with author’s name and number of the figure. Avoid placing labels vertically (except of y axis). Figure legends should explain all symbols and abbreviations and should be double-spaced on a separate page at the end of the manuscript. Color is allowed in figures to show morpho- logical differences among species (for species identifica- tion), to show stain reactions, and to show gradations in temperature contours within maps. Color is discouraged in graphs, and for the few instances where color may be allowed, the use of color will be determined by the Managing Editor. Failure to follow these guidelines and failure to correspond with editors in a timely manner will delay publication of a manuscript. Copyright law does not apply to Fishery Bulletin, which falls within the public domain. However, if an author reproduces any part of an article from Fishery Bulletin in his or her work, reference to source is consid- ered correct form (e.g., Source: Fish. Bull 97:105). Submission The Scientific Editorial Office encourages authors to submit their manuscripts as a single PDF (pre- ferred) or Word (zipped) document by e-mail to Fishery.Bulletin@noaa.gov. Please use the subject heading, “Fishery Bulletin manuscript submission”. Do not send encrypted files. Please provide names and contact information for 3-4 suggested reviewers. Commerce Department personnel should submit papers under a completed NOAA Form 25-700. Or you may send your manuscript on a compact disc in one of the above formats. For further details on electronic sub- mission, please contact the Scientific Editorial Office directly (see address below). Richard D. Brodeur, Ph.D. Scientific Editor, Fishery Bulletin Northwest Fisheries Science Center 2030 S. Marine Science Dr. Newport, Oregon 97365-5296 Once the manuscript has been accepted for publication, you will be asked to submit a final electronic copy of your manuscript. When requested, the text and tables should be submitted in Word or Word Rich Text Format. Figures should be sent as PDF files, Windows metafiles, tiff files, or EPS files. Send a copy of figures in the origi- nal software if conversion to any of these formats yields a degraded version. Questions? If you have questions regarding these guidelines, please contact the Managing Editor, Sharyn Matriotti, at Sharyn.Matriotti@noaa.gov Questions regarding manuscripts under review should be addressed to Richard Brodeur, Scientific Editor, at Rick.Brodeur@noaa.gov. Fishery Bulletin Subscription form Superintendent of Documents Publications Order Form *5178 I I YES, please send me the following publications: Subscriptions to Fishery Bulletin for $36.00 per year ($50.40 foreign) The total cost of my order is $ . Prices include regular domestic postage and handling and are subject to change. (Company or Personal Name) (Please type or print) (Additional address/attention line) (Street address) (City, State, ZIP Code) (Daytime phone including area code) (Purchase Order No.) Charge your order. ITS EASY! Please Choose Method of Payment: | | Check Payable to the Superintendent of Documents | | GPO Deposit Account ] VISA or MasterCard Account your orders | | | | | | | | | j -| I' ■ | -|- | | | | | | | (202) 512-2250 (Credit card expiration date) (Authorizing Signature) Mail To: Superintendent of Documents P.O. Box 371954, Pittsburgh, PA 15250-7954 Thank you for your order! Also available online at http://bookstore.gpo.gov/collections/fishery-bulletin SMITHSONIAN INSTITUTION LIBRARIES 3 9088 01 556 9767