ll .Ax F53 U.S. Department of Commerce Volume 109 Number 2 April 2011 Fishery Bulletin 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 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 109 Number 2 April 2011 Fishery Bulletin Contents Articles 139-146 Sanchez-Rubio, Guillermo, Harriet M. Perry, Patricia M. Biesiot, Donald R. Johnson, and Romuald N. Lipcius Climate-related hydrological regimes and their effects on abundance of juvenile blue crabs (Callinectes sapidus) in the northcentral Gulf of Mexico 147-161 Fry, Brian Mississippi River sustenance of brown shrimp ( Farfantepenaeus aztecus ) in Louisiana coastal waters 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. 162-169 Duffy-Anderson, Janet T., Deborah M. Blood, and Kathryn L. Mier Stage-specific vertical distribution of Alaska plaice ( Pleuronectes quadrituberculatus ) eggs in the eastern Bering Sea 170-185 Prista, Nuno, Norou Diawara, Maria Jose Costa, and Cynthia Jones Use of SARIMA models to assess data-poor fisheries: a case study with a sciaemd fishery off Portugal 186-197 Reum, Jonathan C. P., and Timothy E. Essington Season- and depth-dependent variability of a demersal fish assemblage in a large fiord estuary (Puget Sound, Washington) 198-216 He, Xi, Stephen Ralston, and Alec D. MacCall Interactions of age-dependent mortality and selectivity functions in age-based stock assessment models ii Fishery Bulletin 109(2) 217-231 Ralston, Stephen, Andre E. Punt, Owen S. Hamel, John D. DeVore, and Ramon J. Conser A meta-analytic approach to quantifying scientific uncertainty in stock assessments 232-242 Collins, Angela B., and Richard S. McBride Demographics by depth: spatially explicit life-history dynamics of a protogynous reef fish 243 Errata 244 Guidelines for authors Subscription form (inside back cover) 139 Climate-related hydrological regimes and their effects on abundance of juvenile blue crabs ( Ccillinectes sapidus) in the northcentraS Gulf of Mexico Guillermo Sanchez-Rubio (contact author)1 Harriet M. Perry1 Patricia M. Biesiot2 Donald R. Johnson1 Romuald N. Lipcius3 Email address for contact author: guillermo.sanchez@usm.edu 1 The University of Southern Mississippi Center for Fisheries Research and Development Gulf Coast Research Laboratory 703 East Beach Drive Ocean Springs, Mississippi 39564 2 The University of Southern Mississippi Department of Biological Science 118 College Drive Hattiesburg, Mississippi 39406 3 Virginia Institute of Marine Science The College of William and Mary Department of Fisheries Science P.O. Box 1346 Gloucester Point, Virginia 23062 Abstract — The abundance of juvenile blue crabs ( Callinectes sapidus ) in the northcentral Gulf of Mexico was inves- tigated in response to climate-related hydrological regimes. Two distinct periods of blue crab abundance (1, 1973-94 and 2, 1997-2005) were associated with two opposite climate- related hydrological regimes. Period 1 was characterized by high numbers of crabs, whereas period 2 was char- acterized by low numbers of crabs. The cold phase of the Atlantic Mul- tidecadal Oscillation (AMO) and high north-south wind momentum were associated with period 1. Hydrologi- cal conditions associated with phases of the AMO and North Atlantic Oscil- lation (NAO) in conjunction with the north-south wind momentum may favor blue crab productivity by influ- encing blue crab predation dynamics through the exclusion of predators. About 25% (22-28%) of the variability in blue crab abundance was explained by a north-south wind momentum in concert with either salinity, precipita- tion, or the Palmer drought severity index, or by a combination of the NAO and precipitation. Manuscript submitted 22 February 2010. Manuscript accepted 5 January 2011. Fish. Bull. 109:139-146 (2011). 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. Nonlinear oceanic-atmospheric oscil- lations have been linked to hydro- logical conditions in the continental United States. Individual and com- bined nonlinear oceanic-atmospheric oscillations, such as the Pacific Decadal Oscillation (PDO), Atlan- tic Multidecadal Oscillation (AMO), North Atlantic Oscillation (NAO), and El Nino Southern Oscillation (ENSO) have been shown to modulate Mis- sissippi, Atchafalaya, Pearl, and Pascagoula river flows in their lower basins (Sanchez-Rubio et ah, 2011). Discharge from the Mississippi and Atchafalaya rivers represents over 90% of the total river discharge in Louisiana (Perret et ah, 1971). The Pascagoula and Pearl rivers account for more than 90% of the freshwater discharge into the Mississippi Sound (Eleuterius, 1978). High river flows in northern Gulf of Mexico (GOM) estu- aries have been linked to increased commercial landings of blue crabs (Callinectes sapidus ) in Texas (More, 1969) and Florida (Wilber, 1994) and to both commercial landings and abundance of juvenile crabs (<40 mm carapace width [CW]) in Louisi- ana (Guillory 2000). River discharge enhances wetland nursery areas by increasing the geographic extent of marsh-edge habitat and by provid- ing nutrients that facilitate growth of vegetation. The quantity and quality of coastal marsh habitat have been linked to the successful production of blue crabs. Flooding events directly influence the degree of accessibility of marsh habitats (Rozas and Reed, 1993; Minello and Webb, 1997; Castel- lanos and Rozas, 2001). Vegetated and ephemeral structured habitats provide chemical cues for settlement, food, and refuge to juvenile crabs (Williams et al., 1990; Heck et ah, 2001; Rakocin- ski et ah, 2003). The blue crab is a conspicuous member of coastal ecosystems along the Atlantic and Gulf coasts and the species supports important recre- ational and commercial fisheries for both hard and soft crabs (Guillory et 140 Fishery Bulletin 109(2) Table 1 Meteorological and hydrological parameters and sources for juvenile blue crab (Callinectes sapidus) abundance data used in data analyses. Data for climate-related hydrological regimes, oceanic-atmospheric indices, and Mississippi and Pascagoula river flows were adopted from Sanchez-Rubio et al. (2011). Parameter Annual period Source Kessler east-west and north-south wind Sep-Aug momentum, ( dynes/cm2 )h http://cdo.ncdc.noaa. gov/qclcd/QCLCD?prior=N&state=MS&w ban=13820 (accessed Mar 2007). Coastal Louisiana and Mississippi Palmer drought severity index and precipitation, mm http://www7.ncdc.noaa.gov/CDO/CDODivisionalSelect.jsp (accessed Mar 2008). Louisiana coastal water level, m http://www.mvn.usace.army.mil/eng/edhd/watercon.htm (accessed Feb 2008). Trawl sampling salinity, ppt Louisiana Department of Wildlife and Fisheries, Gulf Coast Research Laboratory-Mississippi Department of Marine Resources Catch per unit of effort Jan-Dec al., 2001). The ability to predict adult population size and thus annual available harvest has been limited by an incomplete understanding of the impact of biotic and abiotic variables as they relate to recruitment and survival of juvenile blue crabs. Although the oceanic- atmospheric oscillations have been associated with the amount of Mississippi River and Pascagoula River dis- charge (Sanchez-Rubio et al., 2011), they have not been related to the periodicity of blue crab population levels in the northcentral GOM. The purpose of the present study is to examine the relationship between nonlinear oceanic-atmospheric oscillations and juvenile blue crab abundance in the northcentral GOM and to elucidate underlying mechanisms involved in that association. This article also addresses the relevance of this study for the management of blue crab in the northcentral GOM. Materials and methods Data acquisition Sanchez-Rubio et al. (2011) examined combinations of oceanic-atmospheric oscillations related to river flow in the northcentral GOM and determined that two regime occurred during the period covered in this study: I) the AMO (cold)-NAO (positive [=high water flow]) and II) the AMO (warm)-NAO (negative [Mow water flow]). These regimes were used to examine the relationship between climate and juvenile blue crab abundance. Individual oceanic-atmospheric indices and river-flow anomalies were also adopted from that study. Other meteorologi- cal (wind momentum) and hydrological (precipitation, Palmer drought severity index [PDSI] , water level, and salinity) data and the biological data (crab abundance) used in the present study are described and illustrated in Table 1 and Figure 1, respectively. The annual envi- ronmental data were calculated from September to August because that period incorporates the time of peak settlement of megalopae in the northern GOM and these megalopae are an important link in determining early year-class strength. Perry and Stuck (1982) noted that the large catches of blue crab megalopae in August and September were followed by an increased catch of juve- nile crabs (10.0 to 19.9 mm) in October or November in Mississippi estuaries. Thus the chosen period for exami- nation follows the dominant modal group responsible for year-class strength. Blue crabs recruit to trawls at ~30 mm CW and are abundant at this size in the winter. The January to December time frame covers the period of juvenile development and it is this period when year- class success is established. Daily coastal water-level data were obtained from two U.S. Army Corps of Engineers gauges along the Louisi- ana coast (Fig. 1: see Cocodrie [1969-2000], and Rigo- lets [1966-2005]). Daily water-level data from the Rigo- lets gauge were averaged to obtain monthly water-level values. The monthly water-level values were averaged to obtain an annual water-level data set for the years 1973-2005. An annual water level anomaly was cal- culated by subtracting the average value by year from the yearly values of water level. Hourly wind data were obtained from the National Climatic Data Center, Ash- ville, NC. Hourly records of wind speed and direction were taken from the Kessler Airport, Biloxi, MS, with an anemometer mounted at 10 m height. The direction of the winds was subdivided into winds from the east (67.5-112.5°), west (247.5-292.5°), north (337.5-22.5°), and south (157.5-202.5°). Wind stress values were cal- culated for the four directions in dynes/cm2. For each direction of the winds, the monthly average of wind stress was multiplied by the number of hours (wind momentum = [dynes/cm2]h) and then, annual values of wind momentum were calculated from 1973 through 2003. To compare climate-related periods of blue crab Sanchez-Rubio et al.: Climate-related hydrological regimes and their effects on abundance of |uvenile Callinectes sapidus 141 30°N 20.649, P<0.001) and precipitation (Pearson r>0.646, P<0.001) values were highly correlated among the four divisions and thus allowed calculation of regional annual ( 1967 — 2005 ) data for both variables. Long-term, fishery-independent, biological data were acquired from 47 stations in Louisiana (Louisiana De- partment of Wildlife and Fisheries) and four stations in Mississippi (Gulf Coast Research Laboratory and Mis- sissippi Department of Marine Resources). This region was divided into eight coastal study areas (CSAs): seven in Louisiana and one in Mississippi (Fig. 1). Louisiana data (CSAs I— VII ) cover the period 1967 to 2005 and samples were collected weekly from March to October and biweekly from November to February. Mississippi data (CSA VIII) extended from 1973 to 2005 and sam- ples were taken monthly. Both states, by agreement, use standard gear and sampling protocols: a 4.9-m otter trawl (1.9-cm bar mesh with a 6.35-mm mesh liner in the codend) pulled for 10 minutes. Crabs were counted and measured to the nearest carapace width (mm). Monthly surface salinities were calculated from trawl stations west (CSA III, V-VII) and east (CSA I and VIII) of the Mississippi River Delta. Monthly salini- ties were averaged to obtain single data sets of annual (1973-2004) salinity for each CSA. The annual salinity of each CSA was multiplied by the number of samples taken annually and the products for all CSAs were added and then divided by the total number of samples collected in the eight CSAs. The yearly regional salinity was a weighted average by sample size, which gives to the CSAs where few samples were collected less weight than those where large numbers of samples were taken in the calculation of the regional salinity data set. An annual weighted salinity anomaly was calculated by subtracting the average value by year from the yearly values of weighted salinity. The variability of salinity can be considered regionally, because two major riv- ers in the west (Mississippi and Atchafalaya rivers) and two in the east (Pearl and Pascagoula rivers) of the Mississippi River Delta are responsible for 90% of freshwater discharge to the northern GOM (Eleuterius, 1978; Perret et ah, 1971). Although the biological data for Louisiana cover the period 1967 to 2005, trawl sampling effort and areal 142 Fishery Bulletin 109(2) coverage among and within CSAs were variable from 1967 through 1981 and more equally distributed be- ginning in 1982. A regional data set of yearly overall abundance (all size classes) was constructed. The vast majority of the crabs collected were less than one year old. Crabs <50 mm CW represented 61.7% of the catch and crabs <90 mm CW represented 82%. To obtain a yearly catch per unit of effort (CPUE), the average catch by station in each study area was calculated by dividing the total catch by the total number of samples. The annual CPUE for each station within a study area was added and then divided by the number of stations to obtain a yearly CPUE for each of the eight CSAs. The annual CPUE of each CSA was multiplied by the number of samples taken annually and the products for all CSAs were added and then divided by the total num- ber of samples collected in the eight CSAs. The yearly regional CPUE was a weighted average by sample size, which gives the CSAs with few collected samples less weight than those with a large number of samples in the calculation of the regional CPUE. Climate-related hydrological regimes and crab abundance Over the period of the biological surveys (1967-2005), two climate-related hydrological regimes (1973-94 and 1997-2005) were identified (Sanchez-Rubio et ah, 2011). To evaluate the response of blue crab abundance to these hydrological regimes, a /-test was used. Relation- ships among crab abundance and oceanic-atmospheric oscillations and hydrological and meteorological param- eters were determined by using correlation analysis. To identify models of oceanic-atmospheric oscillations and meteorological and hydrological parameters that contribute to the variability in blue crab abundance in the northcentral GOM, multiple linear regression analysis (Statistical Package R, vers. 2.7.0, http://www.r- project.org/) was used. To find the best-fitting model, an Akaike’s information criterion (AIC; Akaike, 1981) and Bayesian information criterion (BIC; Raftery, 1996) were calculated for each model. To check model reliability, the models with the lowest BIC and AIC values were com- pared after having been corrected for small sample size (McQuarrie and Tsai, 1998). Multiple linear regression in SPSS (IBM, Somers, NY) was used on the selected models to determine their r2 values. Results Climate-related hydrological regimes and crab abundance Two long-term climate-influenced hydrological regimes were found to be related to two distinct periods of blue crab abundance in the northcentral GOM. Significance differences in blue crab mean abundance (/=3.196, P= 0.003; Fig. 2) were found within regimes that were associated by Sanchez-Rubio et al. (2011) with wet and dry conditions. During regime I (wet), there were higher catches of juvenile crabs than during regime II (dry). The regime with the highest blue crab abundances had a significantly higher mean of the north-south wind momentum (/=2.187, P= 0.038) and a lower mean of AMO (/=-7.276, P<0.001) than did the regime with low crab abundance (Table 2). Correlation analysis showed that blue crab abun- dance was positively correlated with the north-south wind momentum (Pearson /-= 0.406, P= 0.023) and PDSI (Pearson r=0.356, P=0.042) and was negatively related to salinity (Pearson ?•=(). 345, P=0.053). According to the regression models developed from AIC and BIC, the north-south wind momentum in concert with either salinity, precipitation, or the Palmer drought severity index, or the combination of the NAO and precipitation were influential in determining 22% to 28% of the vari- ability in blue crab abundance (Table 3). Figure 3 shows histograms of the variables that were associated with blue crab abundance by year. Discussion Early investigations into factors affecting population dynamics of blue crabs attempted to relate fluctuations in abundance to physiological tolerances to temperature and salinity. Livingston (1976) was among the first to Table 2 Juvenile blue crab (Callinectes sapidus) weighted catch-per-unit-of-effort data and climatological factors showing significant differences in mean values during two hydrological regimes in the northcentral Gulf of Mexico. AMO: Atlantic Multidecadal Oscillation and NAO: North Atlantic Oscillation. Climate-related hydrological regimes AMO cold-NAO positive AMO warm-NAO negative Average values 1973—94 1997—2005 Weighted catch per unit of effort 7.207 4.395 AMO -0.147 0.201 North— south wind momentum, (dynes/cm2)h 0.083 -0.082 Sanchez-Rubio et al.: Climate-related hydrological regimes and their effects on abundance of |uvenile Callinectes sapidus 143 Table 3 Model results determined with Akaike’s information criterion (AIC) and Bayesian information criterion (BIC) used to describe annual variability of the weighted catch per unit of effort (CPUE = average crab catch per ten minute trawl) of juvenile blue crabs (■ Callinectes sapidus) in the northcentral Gulf of Mexico. Response Sum of Model variables Explanatory variables df squares F value P(>F) Coeff. AIC BIC r2 1 Weighted North-south wind momentum 1 32.261 6.8453 0.01438 5.477 51.8 0.38 0.28 CPUE North Atlantic Oscillation 1 12.550 2.6629 0.11432 2.278 Precipitation 1 23.697 5.0282 0.03335 1.058 Intercept 5.616 2 North— south wind momentum 1 32.261 6.3966 0.01735 5.049 53.0 0.18 0.23 Weighted salinity 1 22.277 4.4169 0.04471 -0.3328 Intercept 6.0473 3 North-south wind momentum 1 32.261 6.3389 0.01781 5.5061 53.3 0.46 0.22 Precipitation 1 20.992 4.1248 0.05185 0.9928 Intercept 5.8755 4 North-south wind momentum 1 32.261 6.3234 0.01794 4.7625 53.4 0.54 0.22 Palmer drought severity index 1 20.642 4.0460 0.05399 0.5745 Intercept 5.8365 Climate-related hydrological regimes Figure 2 Weighted catch per unit of effort (CPUE = average crab catch per ten-minute trawl) of juvenile blue crabs (Callinectes sapidus) related to the two dominant climate-related hydrological regimes in the northcentral Gulf of Mexico between 1973 and 1994 (AMO cold-NAO positive: gray bar) and 1997-2005 (AMO warm-NAO negative: white bar). AMO=Atlantic Multidecadal Oscillation and NAO=North Atlantic Oscillation. Horizontal lines for each box plot indicate 5th, 25th, 50th (median), 75th, and 95th percentiles. suggest that the influence of salinity might be operating extrinsically by structuring the sur- rounding biotic community. Recent research indicates that predation affects abundance in the northern GOM. Studies on predator-prey interactions (Heck and Coen, 1995; Guillory and Prejean, 2001; Moksnes and Heck, 2006) and habitat selection and utilization (Williams et al., 1990; Morgan et al., 1996; Rakocinski et al., 2003) indicate that factors that increase or decrease refuge availability are also determi- nant in the establishment of population levels. Both inter- and intraspecific predation, operate to regulate abundance of juvenile blue crabs in the GOM (Guillory et al., 2001). A high diver- sity of predators, few predation-free refuges, and year round predation activity (i.e., a lack of seasonality in predation) all contribute to the high regional mortality of juvenile crabs (Heck and Coen, 1995). If predation is the primary determinant of population levels, then those factors that influence available refuge may ultimately control abundance. In the current study, the period of greatest crab abundance (climate-related hydrological regime I) was associated with a mean positive north-south wind momentum and a mean low value of AMO. Blue crab abundance was also positively correlated with the north-south component of wind momentum and PDSI and was negatively related to salinity. About 25% (22-28%) of the variability in blue crab abun- dance was explained by a north-south wind momentum in concert with either salinity, precipitation, or PDSI, or by the combination of NAO and precipitation. The AMO and NAO were found to be important drivers of climate-related features influencing long-term hydrolog- ical conditions across coastal Louisiana and Mississippi 144 Fishery Bulletin 109(2) (Sanchez-Rubio et al., 2011). Mississippi, Atchafalaya, Pearl, and Pascagoula river flows and blue crab abun- dance were higher during AMO cold and NAO posi- tive phases than during AMO warm and NAO negative phases. Other studies have linked blue crab abundance to river flow and salinity. Guillory (2000) noted juvenile blue crab abundance was positively related to river flow and negatively related to salinity in fishery-independent (crabs <40 mm CW) trawl survey data for Louisiana. High commercial landings of blue crabs were associated with increased river flow by More (1969) in Texas bays, Wilber (1994) in Apalachicola Bay, Florida, and Guillory (2000) in Louisiana estuaries. Hydrological conditions associated with phases of AMO and NAO in conjunction with the north-south wind may influence blue crab pre- dation dynamics through predator exclusion. Under an annual positive north-south wind regime with flooding rain events, greater availability of low-salinity habitats increases the survival of juvenile crabs by diminishing intra- and interspecific predation. Under an annual negative north-south wind regime (inshore water move- ment), low-salinity habitats are reduced and there is a greater suite of predators and an increased opportunity for predation. Although long-term climatological patterns influence the abundance of estuarine organisms, there is also evidence that an interannual oceanic-atmospheric os- cillation (ENSO) can affect population levels. In mic- rotidal Louisiana estuaries, ENSO-related hydrological conditions were found to influence the abundance of estuarine organisms over limited time periods (Childers et al., 1990). High (or low) rates of local precipitation and Mississippi River discharge were gener- ally associated with anomalously high (or low) marsh inundation, respectively, that coincided with ENSO warm (or ENSO cold) phases. The ENSO warm and cold phases generally coin- cided with the lowest abundance of organisms, and the ENSO neutral phase was related to high abundance. Sanchez-Rubio et al. (2011) found that the effect of ENSO phases on river discharge was most evident in the last climate- related hydrological regime (drought) in the Pascagoula River and flow from this river was significantly higher during ENSO warm and ENSO neutral phases than during the ENSO cold phase. Although the ENSO was found to affect river flow, the limited number of ENSO phases (warm, cold, neutral) occurring dur- ing the last hydrological regime precluded any meaningful analysis of the abundance of crabs in relation to ENSO events. Implications for management Management of any fishery requires some knowledge of the factors that contribute to year-class strength. Initial population levels are established by recruitment. In the northern GOM, recruitment success (measured as the number of megalopae at settlement) was found to be dependent on interannual variations in wind stress patterns coupled with basin-scale events, such as Loop Current spin-off eddies, generated during critical periods of larval devel- opment (Johnson and Perry, 1999; Perry et al., 2003). Seasonality of spawning coincided with climatological inner shelf water circulation pat- terns that transported larvae offshore initially but then acted to return them to shore at the appropriate developmental stage. Although annual temporal periodicity of settlement was similar, settlement was highly episodic and there were large annual variations in numbers of megalopae (Perry et al., 1995; Perry et al., 2003). Perry et al. (1998) noted that regard- 'd =5 a) O as Q- O) o OS -C X CD CD =3 TD O C ■5 ' >» CD Fn -4 J Ulnmn lOnnlnniinln nmnnnnnn n n Flnl L _ ■urn ILI1 ’ U‘-J' ' ' ~ UU1 1 1 ' _=Oo □□□ On CL —I— 1 Ql-I 1 yu 1 ^ 1 < ^uj u' 1—1 ~ E .9- E 60 -60 J nfln u •PpPj-pO. Inn flr euQ= i -i -l J I 1 1 I f~l r-ir-in n fl i ii — i i — il ll I -fl TT SO) — roihhO'-mio t^'P-r^r^-ooooooooooo><^'ON On On On On On On On On On On On ON Figure 3 Weighted catch per unit of effort (CPUE = average crab catch per ten minute trawl) of juvenile blue crabs ( Callinectes sapidus ) related to highly influential hydrological and meteorological factors (salinity, PDSI, precipitation, and north-south wind momentum) in the northcentral Gulf of Mexico within two climate- related hydrological regimes (AMO cold-NAO positive: gray bars, and AMO warm-NAO negative: white bars). AMO=Atlantic Multidecadal Oscillation and NAO=North Atlantic Oscillation. Sanchez-Rubio et al. : Climate-related hydrological regimes and their effects on abundance of juvenile Ca/linectes sapidus 145 less of the level of recruitment, by the time crabs reach ~30 mm CW, population abundance begins to level off and then decreases at a gradual rate. In that study, high numbers of megalopae and early-stage crabs did not result in proportionally elevated numbers of late- stage juveniles; instead, high and low recruitment years had similar population levels. They concluded that the northcentral GOM blue crab fishery was not recruitment limited and that year-class strength was dependent on juvenile survival. In the northcentral GOM, there have been significant declines in numbers of later stage juve- niles in trawl surveys; however, blue crabs at early life history stages collected in beam plankton trawls and seines do not exhibit similar trends (Riedel et al., 2010). Climate interacts with an ever-changing physiograph- ic landscape world-wide. Significant downward trends in abundance of juvenile blue crabs across the northern GOM have occurred over a period characterized by drought and unprecedented changes in habitat associ- ated with catastrophic storms and the cumulative con- sequences of man-made alterations to coastal wetlands (Riedel et al., 2010). Recruitment has been adequate and numbers of megalopae and early juveniles do not exhibit declines. Unlike the fishery in Chesapeake Bay, the fishery in the GOM does not suffer from overharvesting (Riedel et al., 2010). There is strong evidence that fish- ery sustainability is dependent upon juvenile survival. In the northcentral GOM, climate and hydrological fea- tures operate to structure available habitat in ways that affect juvenile survival of blue crabs. Whether the shift to a more favorable climate phase would reverse declin- ing trends is unknown because it is currently impossi- ble to quantitatively account for the influence of chang- ing habitats. The results of this work are a starting point toward understanding the complex relationship between climate, habitat, and fisheries productivity. Acknowledgments The authors would like to thank C. F. Rakocinski and R. R Riedel for their expert advice. We are very grate- ful to V. Guillory, K. Ibos, J. Adriance, P. Cook, and M. Harbison from the Louisiana Department of Wildlife and Fisheries and the personnel from the University of Southern Mississippi, Gulf Coast Research Labora- tory, Center for Fisheries Research and Development for providing us with the fishery independent data from the coastal study areas of Louisiana and Mississippi, respectively. Appreciation must also be expressed to M. G. Williams and C. A. Schloss of the University of Southern Mississippi, Gulf Coast Research Laboratory, Gunter Library. Literature cited Akaike, H. 1981. Likelihood of a model and information criteria. J. Econometrics 16:3-14. Castellanos, D. L., and L. P. Rozas. 2001. Nekton use of submerged aquatic vegetation, marsh, and shallow unvegetated bottom in the Atchafalaya River Delta, a Louisiana tidal freshwater ecosystem. Estu- aries 24:184-197. Childers, D. L., J. W. Day Jr., and R. A. Muller. 1990. Relating climatological forcing to coastal water levels in Louisiana estuaries and the potential impor- tance of El Nino-Southern Oscillation events. Clim. Res. 1:31-42. Eleuterius, C. K. 1978. Classification of Mississippi Sound as to estuary hydrological type. Gulf Res. Rep. 6:185-187. Guillory, V. 2000. Relationship of blue crab abundance to river dis- charge and salinity. Proc. Ann. Conf. SE Assoc. Fish Wild!. Agen. 54:213-220. Guillory, V., H. Perry, and S. VanderKooy. 2001. The blue crab fishery of the Gulf of Mexico, United States: a regional management plan. Gulf States Mar. Fish. Comm. 96, 308 p. GSMFC, Ocean Springs, MS. Guillory, V., and P. Prejean. 2001. Red drum predation on blue crabs ( Callinectes sapidus). In Proceedings of the blue crab mortality symposium (V. Guillory, H. Perry, and S. Vanderkooy, eds.), p. 93-104. Gulf States Mar. Fish. Comm. 90, Ocean Springs, MS. Heck, K. L. Jr., and L. D. Coen. 1995. Predation and the abundance of juvenile blue crabs: a comparison of selected East and Gulf Coast (USA) studies. Bull. Mar. Sci. 57:877-883. Heck, K. L. Jr., L. D. Coen, and S. G. Morgan. 2001. Pre- and post-settlement factors as determinants of juvenile blue crab Callinectes sapidus abundance: results from the north-central Gulf of Mexico. Mar. Ecol. Prog. Ser. 222:163-176. Johnson, D. R., and H. M. Perry. 1999. Blue crab larval dispersion and retention in the Mississippi Bight. Bull. Mar. Sci. 65:129-149. Livingston, R. J. 1976. Diurnal and seasonal fluctuations of organisms in a North Florida estuary. Estuar. Coast. Mar. Sci. 4:373-400. McQuarrie, A. D. R., and C. L. Tsai. 1998. Regression and time series model selection, 455 p. World Scientific Publ. Co., Singapore. Minello, T. J., and J. W. Webb Jr. 1997. Use of natural and created Spartina alterniflora salt marshes by fishery species and other aquatic fauna in Galveston Bay, TX, USA. Mar. Ecol. Prog. Ser. 151:165-179. Moksnes, P. O., and J. L. Heck Jr. 2006. Relative importance of habitat selection and pre- dation for the distribution of blue crab megalopae and young juveniles. Mar. Ecol. Prog. Ser. 308:165-181. More, W. R. 1969. A contribution to the biology of the blue crab ( Cal- linectes sapidus Rathbun) in Texas, with a description of the fishery. TX Parks Wildl. Dep., Tech. Ser. 1:1-31. Morgan, S. G., R. K. Zimmer-Faust, K. L. Heck Jr., and L. D. Coen. 1996. Population regulation of blue crabs Callinectes sapidus in the northern Gulf of Mexico: postlarval supply. Mar. Ecol. Prog. Ser. 133:73-88. Perret, W. S„ W. R. Latapie, J. F. Pollard, W. R. Mock, G. B. Adkins, W. J. Gaidry, and C. J. White. 1971. Fishes and invertebrates collected in trawl and 146 Fishery Bulletin 109(2) seine samples in Louisiana estuaries. In Cooperative Gulf of Mexico inventory and study phase IV, biology, section I, p. 39-105. LA Wildl. Fish. Comm., Baton Rouge, LA. Perry, H. M., C. K. Eleuterius, C. B. Trigg, and J. R. Warren. 1995. Settlement patterns of Callinectes sapidus mega- lopae in Mississippi Sound: 1991, 1992. Bull. Mar. Sci. 57:821-833. Perry, H. M., D. R. Johnson, K. M. Larsen, C. B. Trigg, and F. Vukovich. 1998. Blue crab larval dispersion and retention in the Mississippi Bight: testing the hypothesis. Bull. Mar. Sci. 72:331-346. Perry, H. M., and K. C. Stuck. 1982. The life history of the blue crab in Mississippi with notes on larval distribution. Gulf States Mar. Fish. Comm. 7:17-22. Perry, H., J. Warren, C. Trigg, and T. VanDevender. 1998. The blue crab fishery in Mississippi. J. Shellfish Res. 17:425-433. Raftery, A. E. 1996. Approximate Bayes factors and accounting for model uncertainty in generalized linear models. Bio- metrika 83:251—266. Rakocinski, C. F., H. M. Perry, M. A. Abney, and K. M. Larsen. 2003. Soft sediment recruitment dynamics of early blue crab stages in Mississippi Sound. Bull. Mar. Sci. 72:393-408. Riedel, R., H. Perry, L. Hartman, and S. Heath. 2010. Population trends of demersal species from inshore waters of Mississippi and Alabama. Final Rep. MS-AL Sea Grant Consort, 89 p. [Available from Mississippi- Alabama Sea Grant, 703 East Beach Drive, Ocean Springs, MS.] Rozas, L. P., and D. J. Reed. 1993. Nekton use of marsh-surface habitats in Louisi- ana (USA) deltaic salt marshes undergoing submer- gence. Mar. Ecol. Prog. Ser. 96:147-157. Sanchez-Rubio, G., H. M. Perry, P. M. Biesiot, D. R. Johnson, and R. N. Lipcius. 2011. Oceanic-atmospheric modes of variability and their influence on Mississippi River and Pascagoula River flows. J. Hydrol. 396:72-81. Wilber, D. H. 1994. The influence of Apalachicola River flows on blue crab, Callinectes sapidus, in north Florida. Fish. Bull. 92:180-188. Williams, A. H., L. Coen, and M. Stoelting. 1900. Seasonal abundance, distribution, and habitat selection of juvenile Callinectes sapidus (Rathbun) in the northern Gulf of Mexico. J. Exp. Mar. Biol. Ecol. 137:165-183. 147 Abstract — Brown shrimp (Farfan- tepenaeus aztecus) are abundant along the Louisiana coast, a coast- line that is heavily influenced by one of the world’s largest rivers, the Mississippi River. Stable carbon, nitrogen, and sulfur (CNS) isotopes of shrimp and their proventriculus (stomach) contents were assayed to trace riverine support of estuarine- dependent brown shrimp. Extensive inshore and offshore collections were made in the Louisiana coastal zone during 1999-2006 to document shrimp movement patterns across the bay and shelf region. Results showed an unexpectedly strong role for nursery areas in the river delta in supporting the offshore fishery, with about 46% of immigrants to off- shore regions arriving from riverine marshes. Strong river influences also were evident offshore, where cluster analysis of combined CNS isotope data showed three regional station groups related to river inputs. Two nearer-river mid-shelf station groups showed isotope values indicating river fertilization and productivity responses in the benthic shrimp food web, and a deeper offshore station group to the south and west showed much less river influence. At several mid-shelf stations where hypoxia is common, shrimp were anomalously 15N depleted versus their diets, and this 515N difference or mismatch may be useful in monitoring shrimp move- ment responses to hypoxia. Manuscript submitted 10 July 2010. Manuscript accepted 07 January 2011. Fish. Bull. 109:147-161 (2011). 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. Mississippi River sustenance of brown shrimp ( Farfantepenaeus aztecus ) in Louisiana coastal waters Brian Fry Email address: bfry@lsu.edu Department of Oceanography and Coastal Sciences Louisiana State University Baton Rouge, Louisiana 70803 Brown shrimp ( Farfantepenaeus az- tecus) have an estuarine-depen- dent life history that is well known (Gaidry and White1). Adults spawn offshore, and postlarvae enter bays to settle as benthic juveniles. The juveniles typically reside in bays for 1-3 months until they reach about 70-100 mm total length, then leave for offshore shelf areas where they may double in length before complet- ing a largely annual life cycle. Both estuarine and offshore phases of this life cycle have been studied in detail; recent shrimp studies in estuaries have focused on loss of marsh nurs- ery habitats (Peterson and Turner, 1994; Haas et ah, 2004), and offshore studies have focused on bottom water hypoxia that can impede shrimp migrations and decrease overall habi- tat area for brown shrimp (Craig et ah, 2005). The brown shrimp fishery of Loui- siana is one of the largest fisheries in the United States and occurs down- stream of Mississippi River inflows that fertilize the Louisiana coastal zone (Moore et ah, 1970). Nutrient loading from the Mississippi River has increased at least 2—4 times in recent decades in contrast to histori- cal background levels (Turner and Rabalais, 1991; Turner et ah, 2007), and this increase in fertilization of the coastal zone may be affecting off- shore shrimp dynamics. In this study, 1 Gaidry, W. J., Ill, and C. J. White. 1973. Investigations of commercially important penaeid shrimp in Louisiana estuaries. LA Wildl. Fish. Comm. Tech. Bull. #8, 154 p. LA Wildl. Fish. Comm., New Orleans, LA. stable isotopes were used to trace how the river currently supports brown shrimp populations because isotopes are increasingly used to trace link- ages between riverine nutrients and coastal fisheries (Schlacher et ah, 2005; Leaky et al., 2008). The Mississippi River is one of the world’s largest rivers in terms of catchment size, total discharge, and sediment load (Deegan et ah, 1986; Rabalais et ah, 1996). Most of the river flows into northern Gulf of Mexico in the Bird’s Foot Delta south of New Orleans, and also into Fourleague Bay west of New Orleans where the Atchafalaya River carries 30% of the river flow into the Gulf of Mexico. During spring and early summer months when brown shrimp are found in coastal bays, most flow of the river is to the west along the coast and has typically high produc- tivity and high chlorophyll levels in the shallow offshore waters within 5—10 km of the barrier islands (Walk- er and Rabalais, 2006). Eddies force some river water into bays where phytoplankton use nitrates from the river. Tides subsequently export pro- ductive phytoplankton to the Gulf of Mexico (Das et al., 2009). The high nutrients and strong water column mixing create conditions for high shrimp productivity that are similar to those observed in shrimp aquacul- ture ponds, but the coasts and bays are more open and allow extensive brown shrimp migrations at 10-100 km scales. In these open systems, it can be difficult to trace the connec- tions between life history stages and populations that are important for managing fisheries. 148 Fishery Bulletin 109(2) Previous studies of stable isotopes for offshore brown shrimp were conducted mostly along the south Texas shelf, showing that isotope “tags” allow some estimates about those estuarine habitats that are most important in supporting offshore populations (Fry, 1981, 1983). In particular, seagrass meadows produced shrimp with high 513C values and small shrimp entering the off- shore fishery often had these distinctive isotope tags, indicating a strong linkage between inshore seagrass meadows and offshore populations (Fry, 1981). Offshore shrimp populations in Texas waters and the deeper Gulf of Mexico had very uniform carbon isotope values within a 2 %o range, consistent with relatively uniform average isotopic compositions of phytoplankton and phy- todetritus that support offshore benthic food webs (Fry, 1983; Fry et al., 1984). Estuarine shrimp had a much greater (>5x) diversity of isotope values, reflecting the much more diverse set of food types supporting benthic food webs in estuaries (Fry, 1981), but shrimp arriving offshore as immigrants gradually lost these divergent estuarine labels and their isotope values converged to relatively uniform offshore values. Experiments showed that this change in isotope label was due to shrimp replacing their old estuarine biomass during normal metabolism, while also acquiring new biomass from offshore foods (Fry and Arnold, 1982). Calculations indi- cated that a 2-4x increase in mass was generally suffi- cient to lose the estuarine isotope tags for rapidly grow- ing shrimp that had switched to a new diet (Fry and Arnold, 1982; Fry, 2006). This tag loss could occur in 1-3 weeks for smaller-size (<125 mm) shrimp that grow at rates near 1 mm/day and occurs over a longer (3-8 week) period for any larger immigrant animals that grow more slowly offshore, but all immigrants gradually become residents as they acquire the distinctive offshore isotope tags. Experimental and field results thus both indicated that this type of food-related “disappearing” isotope tag had a relatively short life for rapidly grow- ing brown shrimp, but work with the isotope tags was nonetheless interesting because shrimp acquire the isotope tags naturally without handling or stress, all shrimp are tagged instead of just a few, and the isotope tags provide information about origins that is very dif- ficult to obtain otherwise (Fry 1981; 1983; 2008). These initial studies and much subsequent research has shown that isotopes can be used as tracers, tags, or labels for studying animal diets, origins, and movements (Hobson and Wassenaar, 2008; West et al., 2010). Given that previous studies of shelf areas off Texas and deeper waters of the Gulf of Mexico show uniform carbon isotope values in areas that lack strong river inputs, a comparative examination of isotopes in Loui- siana waters was undertaken in the present study to identify river impacts on brown shrimp origins and diets. A first goal was to test origins of shrimp along the Louisiana coast. Seagrass meadows that are hot spots of shrimp abundance in Texas waters are largely lacking along the Louisiana coastline owing to turbid waters, but Louisiana brown shrimp are nonetheless common in open bays and areas near salt marshes. Brown shrimp are especially abundant in Barataria and Terrebonne bays along the central Louisiana coast and these bays are sampled regularly by personnel of the Louisiana Department of Wildlife and Fisheries (LADWF) to help set various opening and closing dates for shrimp fishing seasons. These bays have relatively little input from the Mississippi River but are often considered the major estuarine source regions for Loui- siana shrimp production (Gaidry and White1). However, during the course of this study brown shrimp were found to be also abundant in delta marshes in the Bird’s Foot Delta around the mouth of the Mississippi River, just to the east of the central coast and Barataria and Terrebonne bays. To test whether bays of the central coast or riverine marshes were more important shrimp source areas for the offshore fishery, small shrimp were collected as they arrived as immigrants to the offshore system and tested for their isotope tags. A second goal of this study was to test for a distinctive riverprint or isotope landscape (“isoscape”; West et al., 2010) by map- ping offshore shrimp isotopes to trace river subsidies to benthic food webs. The Mississippi River supplies most (>90%) of the freshwater in the Gulf of Mexico so that any river-related signals could be expected to be stronger in areas closer to the river. Combinations of C, N, and S stable isotope measure- ments were investigated as possible tracers of river influences. Carbon isotopes were used to investigate bay origins and linkages to offshore productivity, with low 813C values (<-18%©) generally indicating estua- rine origins, and highest offshore values (near — 15%o) indicating higher phytoplankton productivity at the base of the food web (Fry, 1981; Fry and Wainright, 1991). For nitrogen isotopes, studies of nitrates in the Mississippi River show a relatively high average value near 8 %o (Fry and Allen, 2003), so that estua- rine food webs incorporating nitrates became enriched in 15N, a bottom-up labeling of whole food webs also observed in other human-impacted systems (Schlacher et al., 2005). Higher 815N was expected for shrimp from river-influenced delta marshes than for shrimp from Barataria and Terrebonne bays that have much smaller river inputs. Sulfur isotopes also can provide an interesting label when high productivity in the wa- ter column leads to more organic matter settling to the seafloor and more sulfate reduction in benthic sedi- ments (Peterson and Howarth, 1987). Pelagic plants and animals have high S34S values near the +21%c value of marine sulfate (Rees et al., 1978; Peterson et al., 1985), but sulfides that are produced in sediments from sulfate reduction have low 834S values and enter benthic food webs, resulting in lower 834S values of 5-15%e for animals such as estuarine brown shrimp (Fry, 2008). Geochemical studies in the northern Gulf of Mexico indicate that most sedimentary sulfides are bound with iron (Lin and Morse, 1991), but it is still possible that some of these sulfides are used by benthic bacteria and enter the organic food web, so that lowest shrimp S34S values might be expected for eutrophic river-influenced areas. Fry: Sustenance of Farfantepenaeus aztecus in Louisiana coastal waters 149 Figure 1 Study area along the Louisiana-Texas coast, northern Gulf of Mexico. River inflows important in this study are the Mis- sissippi River at the Bird’s Foot Delta, the Atchafalaya River along the central coast, and the Houston Ship Channel that flows into upper Galveston Bay. GB = Galveston Bay, TB=Terrebonne Bay, BB = Barataria Bay, BFD = Bird’s Foot Delta. The north-south dividing line between BB and BFD marks the zero-km reference used in Figure 3. SEAMAP=Southeast Area Monitoring and Assessment Program. Isotope studies are complementary to taxonomy-based studies of diets and generally show contributions from plants and bacteria in supporting food webs rather than details of predator-prey interactions (Fry, 2006). Taxonomic work was not part of this study but isotope data were collected for the proventriculus (stomach) contents of brown shrimp to help map river support of the benthic food web. The CNS isotope studies reported add to an extensive literature about shrimp isotopic variation in food webs of the Gulf of Mexico (Fry 1981, 1983, 2008; Fry et ah, 1984, 2008) and also comple- ment recent studies of stable isotope studies of fish in the northwestern Gulf of Mexico (Roelke and Cifuentes, 1997; Senn et al., 2010). Materials and methods Samples were collected from several locations in the northern Gulf of Mexico, from Galveston Bay in the west to the Bird’s Foot Delta in the east (Fig. 1). Most samples from Louisiana bays were collected during spring (April and May) brown shrimp trawl surveys conducted by LADWF in 1999 and 2005 in Barataria and Terrebonne bays. Additional shrimp were col- lected with seines during June 2006 in Terrebonne Bay and in the Bird’s Foot Delta. Offshore animals were collected during the National Marine Fisheries Service June- July summer SEAMAP (Southeast Area Monitoring and Assessment Program) surveys in 2005 and 2006. Offshore station depths declined gradually from 10 m inshore near barrier islands to the 200-m isobath at about 28°N (Fig. 1) and included interme- diate mid-shelf areas regularly affected by summer hypoxia (Rabalais et ah, 2002). Offshore isobaths run approximately parallel to the coast through most of the study region. Shrimp were placed on ice and frozen soon after col- lection for 6-24 months until further processing. A few Galveston Bay samples were analyzed that were col- lected in previous studies in the 1990s and preserved in formalin (Rozas and Zimmerman, 2000). Preserva- tion in formalin has been shown to influence C and N isotope composition, but not S isotope values (Edwards et al., 2002). Accordingly, isotope values reported here for the Galveston Bay samples have been adjusted by 150 Fishery Bulletin 109(2) +1.1 %o for S13C and -0.5 %c for S15N to account for the effects of formalin (Edwards et al., 2002). In the laboratory, shrimp were thawed, total length and blotted wet mass were measured, and white muscle tissue was dissected from the tail area. Muscle tissue was cleaned by rinsing it under running tap water, then soaking the tissue in deionized water in glass vi- als for 15-60 minutes to remove saltwater. The water used for soaking was discarded, tissues were dried at 60°C, and then pulverized with a Wig-L-Bug automated grinder (Dentsply International, York, PA). Proventricu- lus contents were obtained by dissection, acidified with 10% HC1, centrifuged, and the stomach contents pellet was kept, and the acid was discarded. To further rinse and remove acid and traces of seawater, the pellet was resuspended with 20 mL of deionized water and then centrifuged again. This rinsing process was repeated three times before final drying of the pellet at 60°C. Shrimp were analyzed as individuals, but proventricu- lus contents were pooled by station to obtain enough material for analysis. Samples were analyzed according to established procedures for stable C, N, and S isotopic determinations (Fry, 2007, 2008). These procedures in- volve combustion of samples to C02, N2, and S02 gases in an elemental analyzer, followed by chromatographic separation and measurement of these gases with an isotope ratio mass spectrometer. Results are reported in d notation as a %o difference from standards accord- ing to the formula 8 (in %c)= (RSAMPLE/i?STANDARD - D '1000, where standards are PDB (PeeDee Belemnite) limestone for 813C, nitrogen gas in air for S15N, and CDT (Canyon Diable troilite) for S34S, and corresponding R values are 13C/12C, 15N/14N, or 34S/32S (Fry, 2007). Possible continued digestion during long-term storage in freezers and repeated washing of acidified proven- triculus samples undoubtedly removed some labile organic matter from the proventriculus samples, but samples were treated similarly and used for between-sample and between- station comparisons. Shrimp tissues had low C/N ratios of 3. 3-3.7 that were consistent with a mostly protein composition with little lipid content, and consequently no corrections were made to the carbon isotope data for lipid contri- butions (Fry and Allen, 2003; Post et al., 2007). Mean values are given with standard errors of the mean (SE), unless otherwise stated. Statistical comparisons among multiple means were made by using Fisher’s least significant difference method with significant differences indicated when P<0.05. Cluster analysis was done with the program Statgraphics Plus vers. 5.1 (Statpoint Technologies, Warrenton, VA). Results Analysis of CNS isotope values for 969 offshore brown shrimp showed that isotope values of larger animals generally converged to a narrow range that was considered representative of offshore resident animals (Fig. 2). The number of samples was not equal for large and small shrimp (Fig. 2) because of the irregular avail- ability of samples, but the pattern of conver- gence to much narrower isotope ranges for large animals was the expected pattern and the same as that observed in previous extensive studies of shrimp and fish in the northern Gulf of Mexico (Fry, 1981, 1983). In many cases, smaller shrimp had these same convergent isotope values likely due to early migration from estuaries and rapid growth on offshore diets (Fry, 1981). Divergent isotope values were more interesting, especially for S15N, where small animals had values both above and below the values for larger shrimp 15 -! 0 20 30 40 80 100 Wet mass (g) Figure 2 Stable isotope carbon, nitrogen, and sulfur (CNS) compositions (813C, 515N, 534S; in units of%e) of brown shrimp (Farfantepenaeus aztecus) collected offshore in the Gulf of Mexico in summers of 2005 and 2006. Fry: Sustenance of Farfantepenaeus aztecus in Louisiana coastal waters 151 -150 -75 0 -75 Distance from western edge of the delta (km) 150 Figure 3 Average §13C and 515N values (in units of %o) for inshore brown shrimp ( Farfantepenaeus aztecus) collected in Terrebonne Bay (-100 km), three regions of Barataria Bay (west bay at -60 km, central bay at -30 km, and east bay at 0 km), and in the Bird’s Foot Delta region (5-110 km). The north-south dividing line between Barataria Bay and Bird’s Foot Delta shown in Figure 1 marks the zero-km reference used here. Values are means ±standard error from Table 1. Squares represent 515N, circles 513C. (Fig. 2). The spread in 515N values for small shrimp was an indication that different estuarine source regions might be involved, source regions with higher and lower 815N values than the offshore values. In contrast, smaller immigrant shrimp with S13C and 834S values divergent from those of the largest offshore animals had values mostly lower than the offshore values, so that estuarine source regions seemed likely to be similar in 813C and S34S values. Shrimp were collected over several years to test these ideas about possible isotopic differ- ences among estuarine source regions. Surveys of inshore Louisiana bays showed that shrimp from the Bird’s Foot Delta had a combination of relatively high 815N values and low 813C values in contrast to shrimp from Terrebonne and Barataria bays (Fig. 3). Highest S15N values were reached in the central delta and extended along the eastern side of the delta. Stations along the northwest side of the delta at the margin of Barataria Bay showed the beginnings of an increase in S15N, but the coordinated pattern of higher S15N and lower 813C developed about 20 km farther east of this margin (Fig. 3, Table 1). This same dual isotope pattern of high S15N and low S13C values was also found in another river-influenced bay system, at a station sampled in upper Galveston Bay near inflows from the Houston Ship Channel (Table 1, station Upper Galves- ton Bay vs. other Galveston Bay stations). Both shrimp size and isotope information were used to estimate immigrant origins in offshore populations. First, shrimp were selected that were relatively small (125 mm or less, 13 g wet mass or less). These shrimp were closest in size to shrimp collected in inshore bays, where the inshore shrimp averaged 85 mm and 4.7 g wet mass, whereas the <125 mm shrimp collected offshore averaged 109 mm and 9 g wet mass. It was the <125 mm offshore shrimp that were expected to have arrived most recently offshore and therefore best reflect prior feeding in inshore bays (Fry, 1981; Fry and Arnold, 1982), and it was these smaller animals that accounted for most of the variation in the offshore isotope values (Fig. 2). Secondly, the C and N isotope information for large offshore shrimp was used to set bounds or cut-off values expected for resident shrimp that had grown for longer periods of time offshore and had time to equilibrate with the offshore diets. As with approaches used earlier (Fry, 1981, 1983), data for larg- est shrimp were used as a second criterion to define isotope ranges characteristic for offshore residents, and shrimp >175 mm (>35 g wet mass) ranged from -15.3%c to -18.4%c for 813C and from 9.1%c to 12.2%o for S15N. Overall, offshore residents were defined as >125 mm shrimp with isotope values between -15.3%e to -18.4%c for 813C and between 9.1%c to 12.2%c for 815N. Isotope values for residents fell within the boxes in Figures 4 and 5. Inshore studies showed that riverine shrimp from the Bird’s Foot Delta generally had higher 815N and lower S13C than resident offshore animals (Fig. 4). Inshore shrimp from Barataria and Terrebonne bays had more diverse isotope values, but always had S15N values less than 11.6%c (Fig. 4). Based on these isotope dis- tributions, two types of immigrant shrimp to offshore systems were identified: riverine immigrants with high 815N and low 813C (>11.6%e and <— 18.4%c, respectively; solid squares in Fig. 5) and bay immigrants with lower 815N plus 813C values that were outside the range of offshore resident values (triangles in Fig. 5). A last group of shrimp was considered likely to resident (open diamonds in Fig. 5, see Discussion section). Over the two years of summer collections, 406 shrimp were collected offshore that were <125 mm, and accord- ing to the above isotope-based criteria for distinguish- ing immigrants and residents, 185 of these shrimp were classified as residents at the time of collection and 221 were immigrants. About 46% of these immigrants had a riverine origin and 54% had a bay origin. The fraction of riverine immigrants was very similar in the two years, 48% in 2005 and 45% in 2006. The <125 mm immigrants were present mostly as mixed popu- lations (bay+riverine+residents) along the inner and mid-shelf, and riverine shrimp were dominant (50% or greater of the <125 mm shrimp) at stations nearest the Bird’s Foot Delta and along the central coast at stations to the south and west of the Atchafalaya River (Fig. 6). The dual isotope label present in shrimp from the Bird’s Foot Delta, as high 815N and low 813C, was also present in proventriculus contents, i.e., both delta shrimp tissues (Fig. 4) and shrimp diets (Fig. 7) had relatively high 815N and low S13C values. The question was whether this riverine dual label would persist in offshore foods, so that animals feeding offshore might acquire this riverine dual label offshore and thus be 152 Fishery Bulletin 109(2) Fry: Sustenance of Farfantepenaeus aztecus in Louisiana coastal waters 153 misclassified as immigrants from deltaic regions. In a test of this idea, proventric- ulus contents from near-delta offshore areas (open squares and open triangles in Fig. 7A) were sampled but generally did not show this riverine dual isotope label, i.e., 15 of 16 near-delta samples did not have the dual isotope delta label, but were relatively enriched in 13C and followed the same isotope trend as that found for other samples collected from deeper offshore areas (Fig. 7B). Only one offshore proventriculus sample collected very close to the shore west of the Atcha- falaya Delta had the dual-label riverine combination of low high 515N and low 513C (see open triangles with arrows, Fig. 7, A and B). Because the inshore bay and delta regions contained geographic isoscape distinctions that were useful in follow- ing shrimp movements, the offshore data for residents also were examined for possible geographic patterns. Clus- ter analysis was used to identify sepa- rate groups by using multivariate data for 48 stations sampled in 2006 where measurements included CNS isotope values for proventriculus samples and parallel CNS isotope values for muscle samples. For the cluster analysis, the muscle averages were compiled by us- ing only larger animals (>125 mm total length) that, as above, had C and N iso- tope values within the range of largest (>175 mm) resident animals, and there- fore were classified as offshore residents. The resulting cluster analysis identified three general regional offshore groups of shrimp: two mid-shelf groups inshore and closer to the river, and one offshore group farther away from the river to the south and west (Fig. 8). The two mid- shelf groups were mostly in or near the area identified by Rabalais et al. (2002) as regularly affected by summer hypoxia and linked to inputs from the Missis- sippi River (Fig. 8, polygon), whereas the offshore group was largely on the southwest side of this region, away from river inputs (Fig. 8). The offshore group was significantly different in average isotope values from the inshore group in all cases for the mid-shelf transition group and in all but one case for the mid-shelf hypoxic group (Table 2). Relative to this offshore group, the mid-shelf groups both showed signifi- cant enrichment in proventriculus 15N and 13C, and depletion in 34S (Fig. 9, 16 i 4 H i i i i i i i i i i i i i ~ n -26 -21 -16 -11 813C Figure 5 813C and 815N values (in units of %o) for smaller brown shrimp ( Farfan- tepenaeus aztecus ) (<125 mm total length) collected offshore and that had recently arrived from inshore estuaries. Shrimp were classified into three groups by considering the combined §13C and 815N data: riverine shrimp (gray squares), bay shrimp (triangles), residents (x’s and diamonds, with diamonds indicating likely residents of the hypoxic zone, see Discussion section). The boxed values indicate the range of values observed in largest (>175 mm) offshore resident shrimp. 154 Fishery Bulletin 109(2) + Bay immigrants present 1000m A y 50 —\ 1 km 0 Riverine immigrants present 1 Riverine immigrants 50% or more w 94° W 93° W 92°' W 91° W 90° W 89° W Figure 6 Offshore locations in the Gulf of Mexico where smaller (<125 mm) brown shrimp (Farfantepenaeus aztecus) were captured as immigrants from estuaries. Symbols indicate inferred origins of these shrimp. Table 2), a triple isotope label associated with higher primary productivity (see Discussion section). These proventriculus isotope labels were strongest in the more inshore of the two groups (Fig. 9), namely the mid-shelf hypoxic group, and S15N values in the pro- ventriculus contents were significantly higher in this group than in the offshore and mid-shelf transition group (Table 2). When all proventriculus samples were considered together, C and S isotopes were sig- nificantly (P<0.01) and linearly correlated with N isotopes (Fig. 10), consistent with mixing between two food sources across the shelf. The correlation of S and C isotopes for these samples also was significant (P<0.01, data not shown). The mid-shelf and offshore station groups (Fig. 8) differed in their patterns of trophic enrichment factors (TEFs) (the difference between isotopes measured in consumers and their diets, i.e., TEF=muscle S- proven- triculus 5). Average TEFs for the most offshore group were close to expected (Peterson and Fry, 1987) at 2.8%c and 0.2%e, respectively, for 815N and 534S, but rel- atively high at 5.2 %c for S13C (Table 2). Inshore groups differed significantly from these offshore TEF values, notably with significantly lower nitrogen isotope TEF values for the mid-shelf groups (Table 2). For the more inshore of the two mid-shelf groups, average nitrogen isotope TEF values were unexpectedly negative (-2.2%10.7%e and therefore were higher than the average values for offshore resident shrimp (Table 2, Fig. 8, circled points). Discussion There are many reasons to expect strong river sup- port of brown shrimp production, ranging from the riverine construction of inshore habitats by natural long-term delta-building processes to more recent river and nutrient-enhanced primary productivity of the off- shore ecosystem (Deegan et al., 1986; Bierman et ah, 1994; Green et al., 2008). Summer surface salinities are 20-33 psu across most of the study area owing to the enormous freshwater inputs from the Mississippi River, so that Louisiana brown shrimp exist in a river- influenced marine ecosystem. The river water affects the isotope biogeochemistry of receiving waters, adding nitrates with high 815N and dissolved inorganic carbon with low 813C (Fry and Allen, 2003). Primary productiv- ity and the wider shrimp food webs seemed to respond to these basal isotope changes in a straightforward way, with shrimp having high S15N and low S13C in the Bird’s Foot Delta region that was most influenced by the river. This same pattern of a riverine dual isotope label may be fairly general in human-influenced estuaries and was observed, for example, in shrimp from Galveston Bay at a low salinity station in the upper bay (Table 1) influenced by freshwater inflows from the urban Houston Ship Channel. The more negative average S13C values of -18.5%c or less that characterize these systems seem to develop for brown shrimp in planktonic bays of the Gulf of Mexico when salinities are <20 psu (Fry, 1981, 1983). There is less local information for the Gulf of Mexico about the determinants of spatial 815N Fry: Sustenance of Farfantepenaeus aztecus in Louisiana coastal waters 155 20-, Z To 15- 10- B ° & A A DnO pMj o o° o ° ° o ^Q0 ° ^ <$> o O CL, A O C# ° 0°J? O o ♦ ^ □ □ o o~ ~ o o □A O 4 -25 -24 -23 -22 5 13 C' -21 -20 -19 Figure 7 813C and 815N values (in units of %o) of proventriculus (gut content) samples collected in 2006. (A) Locations of sample collections. (B) Isotope values of samples. Arrows indicate the location (A) and isotope value (B) of the single offshore proventriculus sample that had isotope values similar to those of riverine samples (black diamonds) in the Bird’s Foot Delta, as presented in the Results section. patterns, but it is noteworthy that in the Bird’s Foot Delta region, the high 815N values for brown shrimp extended over a greater distance than did the the low 813C values (Fig. 3). Stated another way, C isotope values returned to marine values more quickly than did N isotope values at both ends of the delta (Fig. 3). This difference between the N and C isotope patterns is expected because river N nutrient concentrations are very high and dominate freshwater-marine mixing dynamics (Fry, 2002). In contrast, river and marine sources have fairly similar inorganic carbon concentra- tions, so that riverine signals are diluted much more quickly for C than for N. Riverine shrimp with the unique CN isotope tags of high S15N and low S13C accounted for about half of the recent immigrants arriving offshore. There is little independent tagging information that could validate this isotope estimate. LADWF does not sample low- salinity habitats in the Bird’s Foot Delta and instead focuses on routine sampling of Barataria and Terre- bonne bays of the central Louisiana coast to help set the periods for shrimp season openings and closings. River-influenced areas along the Bird’s Foot Delta are not given a special focus by LADWF, but the isotope estimates from the present study may indicate that they deserve more focus in future work. This may be 156 Fishery Bulletin 109(2) 29° N- Gulf of Mexico 30m n 28° N- 200m 1000m ^ 0 50 1 1 1 km □ □ W □ Offshore □ Mid-shelf transition B Mid-shelf hypoxia Proventriculus 15N-enriched 95° W 94° W 93° W 92° W 91° W 90° W 89° W 30° N Figure 8 Station groupings from cluster analysis of samples from 2006. Circled points had 15N-enriched proventriculus 815N >10.7 %o, greater than average values for offshore resident brown shrimp ( Farfantepenaeus aztecus). The polygon indicates the area of the long-term hypoxic zone documented by Rabalais et al. (2002), where hypoxia is present in >25% of summer surveys. Symbols indicate station groupings identified by cluster analysis. especially true because of loss of inshore habitat in the Bird’s Foot Delta (Britsch and Dunbar, 1993). Es- timates for riverine shrimp contributions to offshore fisheries were very similar for 2005 and 2006, with 2005 having average river discharge and 2006 having about 60% average discharge (http://www.mvn. usace. Figure 9 Trends in relation to depth for stable isotope averages (±1 standard error) in proventriculus contents of brown shrimp ( Farfantepenaeus aztecus ) collected from the Gulf of Mexico in 2006 (see also Table 2). army.mil/cgi-bin/watercontrol, accessed July 2010). The similar riverine contributions in the two different years may mean that it is the long-term structure of the deltaic marsh-bay platform rather than the annual river inputs that is more important for the shrimp sup- ply to the offshore fishery. There are various reasons why the isotope es- timates presented here could be overestimates for the contributions of riverine shrimp to off- shore populations. For example, inshore sampling showed that both riverine and bay shrimp popu- lations produce some shrimp that have the same isotope values as resident offshore shrimp (Fig. 4). The isotope accounting done here thus underes- timates the contributions of the inshore popula- tions, and if this underestimate is more severe for bay than riverine shrimp, this would lead to the apparent strong contribution of riverine shrimp. In the extreme, if all of the <125 mm offshore shrimp with isotope values inside the resident box of Figure 5 were actually misclassified and instead were all bay shrimp, the contribution of riverine shrimp would decline from 46% to 25%. Further research should include samples nearer the mouths of bays to check whether most animals leaving bays already have isotope values classi- fied here as resident offshore shrimp, but in the end, even a 25% contribution of riverine shrimp is probably noteworthy for management purposes. Fry: Sustenance of Farfantepenaeus aztecus in Louisiana coastal waters 157 20 co ■'t OO 10 Offshore Mid-shelf oligotrophic * hypoxic -15 o CD to -25 Figure 10 Covariation of proventriculus 534S and 513C with d15N, from brown shrimp ( Farfantepenaeus aztecus ) samples collected from the Gulf of Mexico from 2006. Equations for top and bottom lines are respec- tively 534S=-0.5(815N)+18.8 with r2=0.36, and 513C = 0.38(515N)-25.32 with r2=0.64. The relationship between 534S and 513C (not shown) was 513C = 0.26(834S)-18.17 with r2 = 0.21. All isotope values have units of %c. 10 16 515N Offshore phytoplankton productivity studies in Missis- sippi River plumes generally show 13C-enriched values for particulate organic matter formed in inshore and mid-shelf regions (Fry and Wainright, 1991; Fry un- publ. data), so that the dual isotope tag of low S13C and high 515N used to source riverine shrimp seemed largely confined to estuaries and was only rarely pres- ent offshore (Fig. 7). It is also possible that bays in adjacent Texas and other northern Gulf states supply some shrimp to the offshore Louisiana system. But those shrimp would likely have been offshore for extended periods of time and therefore would have been counted as residents in this study. Thus possible contributions from other states should have little effect on the estimates given above for bay and riverine contributions to Louisiana shrimp stocks. It also was evident that estuarine conditions prevail offshore in this river-influenced shelf ecosystem that is often considered an offshore estuary. Isotopic composi- tions of shrimp and proventriculus contents followed the same triple isotope gradients involving high §13C, high S15N, and low 834S nearest the river, vs. low 813C, low 815N, and high 834S offshore. These gradients were largely aligned with other offshore features associated with the river, notably finfish biomass (Moore et ah, 1970) and hypoxia (Rabalais et ah, 2002). It is possible that some of these isotope gradients reflect normal depth-related onshore-offshore productivity gradients not associated with rivers, and this idea should be ad- dressed in future comparative work involving continen- tal shelf systems with little river influence. Initial data for the Texas shelf have shown some cases of onshore- 158 Fishery Bulletin 109(2) offshore and seasonal C isotope changes in shrimp (Fry et al., 1984), but N and S isotope changes have not yet been systematically investigated. Comparative work might also focus on other river-influenced shelf sys- tems and one recent study of the Thames River estuary (Leakey et al., 2008) has shown the same triple isotope riverine-offshore gradients observed in this study for the Mississippi River. Because of these similar results for the Thames River, and because the pattern of iso- tope signals is consistent with a riverine source with high S15N, it seems likely that the Mississippi River is forcing many of the isotope signals observed on the Louisiana shelf. The three regional shelf groups shown in Figures 8 and 9 were identified by cluster analysis by using three proventriculus isotope variables and three mus- cle isotope variables. These six variables were used in concert for two reasons. First, the separated proven- triculus and muscle results each gave strong mid-shelf vs. offshore patterns (see significant differences among averages in Table 2), and therefore results could be legitimately combined for a stronger overall assess- ment. Second, the proventriculus and muscle samples show somewhat different aspects of shrimp biology and available diets, with proventriculus samples providing stronger local information and muscle samples provid- ing stronger time-integrated samples. On the negative side, the proventriculus samples are often the leftovers after digestion and can include inorganic sediment grains with pyritic sulfides (Howarth, 1979, 1984), whereas muscle samples are taken from shrimp that are mobile and may reflect diets from another place. Because there were both positive and negative aspects to using the separated proventriculus and muscle iso- tope data, the combined data (Table 2) were used to reach a balanced overall assessment in the cluster analysis. The offshore C isotopes showed a broad pattern of river influence across the inshore and middle shelf, consistent with wide dispersal of carbon from river- influenced planktonic primary producers. The riverine influence was expressed as higher S13C values in a mid- shelf maximum standing out against a background of lower 813C values both in shallower bays and in deeper offshore waters (Tables 1 and 2, Fig. 9). Higher S13C values are found associated especially with high pro- ductivity and diatom blooms (Fry and Wainright, 1991; Fry, 1996) — conditions that regularly occur on the Loui- siana shelf that is affected by Mississippi River inputs (Rabalais et al., 1996; Dagg et al., 2007; Green et al., 2008). The deeper shelf to the south and west had lower 813C consistent with lower primary productivity (Fry and Boyd, 2010). The offshore S isotopes were perhaps the most ex- pected results, showing a consistent onshore-offshore gradient in both proventriculus and muscle 834S values (Table 2, Fig. 9). These gradients likely originate with river-induced organic carbon gradients in primary pro- ductivity that subsequently fuel benthic sulfate reduc- tion and sulfide production in underlying sediments. The exact mechanism of sulfide incorporation into ben- thic food webs is still unknown but is likely the use of sulfides by bacteria growing in bottom sediments. Hy- poxia may increase aspects of sulfide cycling, especially by decreasing the importance of oxygenic decomposition reactions while increasing the importance of anaerobic reactions such as sulfate reduction and sulfide produc- tion. Hypoxia also may decrease oxidation reactions that consume sulfide, and decreased infaunal activity and decreased bioirrigation in sediments may also oc- cur when bottom waters become hypoxic (Eldridge and Morse, 2008). In sum, hypoxic conditions may promote more anaerobic conditions, more sulfide production and accumulation, and stronger bacterial uptake of sulfides into benthic food webs. The N isotope results were quite surprising in the very high S15N values (up to 15.2%c) found for some proventriculus content samples in the mid-shelf hypoxic region — values that were higher than those for shrimp muscle. Ongoing studies show no large 15N enrichment in particulate organic nitrogen samples collected in the water column in the offshore region, where values average 6-8%o (Wissel et al., 2005; Fry, unpubl. data). In the absence of a planktonic origin, the source of the high S15N values likely is in the benthos. Brown shrimp are benthic carnivores that consume polychaetes and meiofauna (McTigue and Zimmerman, 1998; Fry et al., 2003), and offshore brown shrimp generally rely on a benthic food web with bacterial contributions. York et al. (2010) have speculated that nitrogen cycling in the benthos is leading to high S15N values of benthic bac- teria, perhaps with some bacterial use of 15N-enriched ammonium left over from nitrification or annamox re- actions. Such processes are likely ubiquitous in shelf sediments, but details that are still to be elucidated could make these processes much more dominant in the low-oxygen mid-shelf hypoxic region. High S15N values were also found in inshore shrimp and proventriculus contents from the Bird’s Foot Delta region (Table 2), and the common denominator leading to these high 815N values is likely eutrophic deposition of large amounts of organic phytodetritus to the benthos. 815N values >15%c have also been observed in Mississippi River zebra mus- sels during summer, where high animal 815N values have been correlated with low ammonium concentra- tions in the river (Fry and Allen, 2003). Whatever the mechanism underlying the high S15N values, it was evident that the proventriculus 815N val- ues were often higher than those of offshore shrimp eat- ing these foods (Table 2). Previous work with estuarine brown shrimp has shown that brown shrimp normally have positive trophic enrichment factors (TEFs) aver- aging about 2.3%o higher than proventriculus contents 815N (Fry et al., 2003), and a similar average TEF of 2.8%c was observed for the most offshore shrimp of the present study (Table 2). The observed opposite pattern of negative TEF values for some mid-shelf shrimp likely means that these shrimp have not spent the several weeks (that can be calculated from diet turnover dy- namics) (Fry and Arnold, 1982) that they would need Fry: Sustenance of Farfantepenaeus aztecus in Louisiana coasta! waters 159 in the hypoxic zone to come to equilibrium with the 15N-enriched foods. This idea is reasonable given recent fisheries studies that show hypoxia is often displacing brown shrimp populations to areas of higher bottom- water oxygen (Craig and Crowder, 2005; Craig et al., 2005). In future studies, the disequilibrium or mis- match between shrimp and proventriculus 815N may help identify areas that do not continuously sustain brown shrimp populations. Areas where proventricu- lus 815N is higher than shrimp muscle 815N may be less suitable habitat that can be visited only briefly by brown shrimp. The isotope signals in diets of brown shrimp and their prey are built up over several weeks, so that the isotope measurements may provide longer- term information about shrimp use of hypoxic areas than do trawls that provide a more instantaneous snap- shot of how brown shrimp are using an area (Craig et al., 2005). However, occasional feeding in the hypoxic area should lead to somewhat elevated 815N values, so that higher 815N could develop in offshore resident shrimp. Several offshore shrimp were observed with high 815N that could indicate some feeding in the hypoxic zone. These animals also had high 813C values (less negative than -18%c; open diamonds in Fig. 5) expected for off- shore residents rather than for migrants from inshore regions, and were accordingly classified as offshore residents for purposes of estimating movement and inshore contributions to offshore fisheries. An interesting feature of this study was that offshore brown shrimp diets appeared to be linear mixtures between two sources, and variation in the source con- tributions accounted for most of the isotopic variation across the shelf (Fig. 10). The nature of these sources is not completely clear and may involve multiple factors. For example, high S15N values may reflect both a high value of Mississippi River nitrate at the base of coastal food webs (Fry and Allen, 2003; Wissel and Fry, 2005), and the presence of high trophic level consumers in the proventriculus contents. Conversely, low 815N may reflect relatively low values for offshore marine nitrate and prey from low trophic levels. Unfortunately, isotope values for specific prey taxa have not yet been measured for this shelf ecosystem, and therefore trophic-level ef- fects for isotopes cannot be directly evaluated. Nonethe- less, some inferences can be made from the measured isotope data for shrimp and their proventriculus con- tents, as follows. The farthest offshore animals had high S34S values (Table 2) characteristic of mostly plankton-derived sulfur in the diet, with little contribution of benthic sulfides. These high 834S values are consistent with relatively oligotrophic conditions across the deeper shelf, and with lower 834S values indicating more eu- trophic conditions inshore. Carbon isotope TEFs be- tween offshore shrimp and proventriculus contents were surprisingly large at 3.2-5.2%c (Table 2), espe- cially compared to the general expectation that the carbon isotope TEF is near 0 %c for animals and their diets (Peterson and Fry, 1987) and compared to a measured carbon isotope TEF near 1 %o for estuarine Louisiana brown shrimp (Fry et al., 2003). The off- shore shrimp muscle 813C values are fairly constant near -17.3%c, so that it is the very negative proven- triculus values that lead to the large observed TEFs. Nonetheless, the proventriculus 813C values are near the long-term -22%c value associated with offshore marine primary production (Fry and Sherr, 1984), and may represent a realistic marine background value. If this is the case, then mass balance calculations would indicate that the labile foods near -17.3%e that are be- ing assimilated out of the -22%c marine background are likely a small part of the proventriculus contents. A consistent picture for the C and S results is that background, low-productivity pelagic conditions deter- mine the food availability at the offshore stations, but labile fractions that are depleted in 34S and enriched in 13C are increasingly found in the proventriculus contents at the more eutrophic inshore stations (Fig. 10). These ideas need further study with taxonomic analyses of prey and with further studies on isotope changes during assimilation of offshore foods (Fry et al., 1984). In conclusion, further studies of both CNS isotopes and proventriculus contents in offshore brown shrimp could supplement annual summer water quality assess- ments of hypoxia and help determine hypoxia effects on living resources. Brown shrimp transit the mid-shelf hypoxic areas and isotopes in shrimp caught offshore show strong spatial signals that likely vary between years with high and low river flow. Isotope signals have been used as early warning indicators of the effects of eutrophication in coastal bays (McClelland et al., 1997), and it is possible that monitoring shrimp isotopes may help assess the effects of hypoxia on Louisiana shrimp populations. Adding benthic shrimp isoscape monitoring to ongoing water quality monitoring programs generally may be helpful for understanding changes in fisher- ies productivity and animal movements in this and other river-influenced marine ecosystems (Leakey et al., 2008). Acknowledgments K. Johnson and B. Pellegrin of the NOAA Pascagoula Laboratory kindly provided the offshore shrimp samples from SEAMAP cruises. G. Peterson assisted in shrimp collections made in the Bird’s Foot Delta region. LSU undergraduates E. Ecker, E. Gallagher, M. Grant, T. Pasqua, R. Sylvestri, and J. Wheatley helped with labo- ratory preparation of samples and data entry. J. Lentz drafted the maps used in this article. Brittany Graham read an initial draft of this manuscript and made helpful comments. This work was supported in part by Louisiana Sea Grant Projects R/CEH-13 and R-EFH-07, NOAA Multistress award NA 16OP2670, NOAA Coastal Ocean Program grant NA06NOS4780141, and NOAA grant 412 NA06OAR4320264 06111039 to the Northern Gulf Institute. This is N-GOMEX contribution 138. 160 Fishery Bulletin 109(2) Literature cited Bierman, V. L., S. C. Hinz, D -W. Zhu, W. J. Wiseman, N. N. Rabalais, and R. E. Turner. 1994. A preliminary mass balance model of primary pro- ductivity and dissolved oxygen in the Mississippi River plume/inner Gulf shelf region. Estuaries 17:886-899. Britsch, L. D., and J. B. Dunbar. 1993. Land loss rates: Louisiana coastal plain. J. Coastal Res. 9:324-338. Craig, J. K., and L. B. Crowder. 2005. Hypoxia-induced habitat shifts and energetic con- sequences in Atlantic croaker and brown shrimp on the Gulf of Mexico shelf. Mar. Ecol. Progr. Ser. 294:79-94. Craig, J. K., L. B. Crowder, and T. A. Henwood. 2005. Spatial distribution of brown shrimp ( Farfante - penaeus aztecus) on the northwestern Gulf of Mexico shelf: effects of abundance and hypoxia. Can. J. Fish. Aquat. Sci. 62:1295-1308. Das, A., D. Justic, and E. Swenson. 2009. Modeling estuarine-shelf exchanges in a deltaic estuary: Implications for coastal carbon budgets and hypoxia. Ecol. Model. 221:978-985. Dagg M., J. Ammerman, R. Amon, W. Gardner, R. Green, and S. Lohrenz. 2007. A review of water column processes influencing hypoxia in the northern Gulf of Mexico. Estuaries Coasts 30:735-752. Deegan, L. A., J. W. Day Jr., J. G. Gosselink, A. Yanez-Arancibia, G. Soberon-Chavez, and P. Sanchez-Gil. 1986. Relationships among physical characteristics, vegetation distribution and fisheries yield in Gulf of Mexico estuaries. In Estuarine variability (D. A. Wolfe, ed.), p. 83-100. Academic Press, New York. Edwards, M. S., T. F. Turner, and Z. D. Sharp. 2002. Short- and long term effects of fixation and preser- vation on stable isotope values (513C, 515N, d34S) of fluid- preserved museum specimens. Copeia 2002:1106-1112. Eldridge, P. M., and J. W. Morse. 2008. Origins and temporal scales of hypoxia on the Louisiana shelf: Importance of benthic and sub- pycnocline water metabolism. Mar. Chem. 108:159-171. Fry, B. 1981. Natural stable carbon isotope tag traces Texas shrimp migrations. Fish. Bull. 79:337-345. 1983. Fish and shrimp migrations in the northern Gulf of Mexico analyzed using stable C, N, and S isotope ratios. Fish. Bull. 81:789-801. 1996. 13C/12C fractionation by marine diatoms. Mar. Ecol. Progr. Ser. 134:283-294. 2002. Conservative mixing of stable isotopes across estuarine salinity gradients: A conceptual framework for monitoring watershed influences on downstream fisheries production. Estuaries 25:264-271. 2006. Stable isotope ecology, 308 p. Springer, New York. 2007. Coupled N, C, and S isotope measurements using a dual column GC system. Rapid Comm. Mass Spectr. 21:750-756. 2008. Importance of open bays as nurseries for Louisiana brown shrimp. Estuaries Coasts 31:776-789. Fry, B., and Y. Allen. 2003. Stable isotopes in zebra mussels as bioindica- tors of river-watershed linkages. Rivers Res. Appl. 19:683-696. Fry, B., R. K. Andersen, L. Entzeroth, J. L. Bird, and P. L. Parker. 1984. 13C enrichment and oceanic food web structure in the northwestern Gulf of Mexico. Contrib. Mar. Sci. 27:49-63. Fry, B., and C. K. Arnold. 1982. Rapid 13C/12C turnover during growth of brown shrimp (Penaeus aztecus). Oecologia (Berlin) 54:200- 204. Fry, B., D. Baltz, M. Benfield, J. Fleeger, A. Gace, H. A. Haas, and Z. Quinones. 2003. Chemical indicators of movement and residency for brown shrimp (Farfantepenaeus aztecus) in coastal Louisiana marshscapes. Estuaries 26:82-97. Fry, B., and B. Boyd. 2010. Oxygen concentration and isotope studies of pro- ductivity and respiration on the Louisiana continental shelf, July 2007. In Earth, life, and isotopes (N. Ohk- ouchi, I. Tayasu, and K. Koba, eds.), p. 223-247. Kyoto University Press, Japan. Fry, B., and E. Sherr. 1984. <513C measurements as indicators of carbon flow in marine and freshwater ecosystems. Contrib. Mar. Sci. 27:13-47. Fry, B., and S. C. Wainright. 1991. Diatom sources of 13C-rich carbon in marine food webs. Mar. Ecol. Progr. Ser. 76:149-157. Green, R. E., G. A. Breed, M. J. Dagg, and S. E. Lohrenz. 2008. Modeling the response of primary production and sedimentation to variable nitrate loading in the Mis- sissippi River plume. Cont. Shelf Res. 28:1451-1465. Haas, H. L, K. A. Rose, B. Fry, T. J. Minello, and L. Rozas. 2004. Brown shrimp on the edge: linking habitat to sur- vival using an individual-based simulation model. Ecol. Appl. 14:1232-1247. Hobson, K. A., and L. I. Wassenaar, eds. 2008. Tracking animal migrations with stable isotopes. Terrestrial Ecology Series, vol. 1, 160 p. Elsevier Sci- ence, Maryland, MO. Howarth, R. W. 1979. Pyrite — its rapid formation in a salt marsh and its importance in ecosystem metabolism. Science 203:49-51. 1984. The ecological significance of sulfur in the energy dynamics of salt marsh and coastal marine sedi- ments. Biogeochem. 1:5-27. Leakey, C. D. B., M. J. Attrill, S. Jennings, and M. F. Fitzsimons. 2008. Stable isotopes in juvenile marine fishes and their invertebrate prey from the Thames Estuary, UK, and adjacent coastal regions. Estuar. Coast. Shelf Sci. 77:513-522. Lin, S., and J. W. Morse. 1991. Sulfate reduction and iron sulphide mineral formation in Gulf of Mexico anoxic sediments. Am. J. Sci. 291:55-89. McClelland, J. W., I. Valiela, and B. Michener. 1997. Nitrogen-stable isotope signatures in estuarine food webs: A record of increasing urbanization in coastal watersheds. Limnol. Oceanogr. 42:930-937. McTigue, T. A., and R. J. Zimmerman. 1998. The use of infauna by juvenile Penaeus aztecus Ives and Pe?iaeus setiferus (Linnaeus). Estuaries 21:160-175. Moore, D., H. A. Brusher, and L. Trent. 1970. Relative abundance, seasonal distribution, and species composition of demersal fishes off Louisiana and Texas, 1962-1963. Contrib. Mar. Sci. 15:45-70. Peterson, B. J., and B. Fry. 1987. Stable isotopes in ecosystem studies. Ann. Rev. Ecol. System. 18:293—320. Fry: Sustenance of Farfantepenaeus aztecus in Louisiana coastal waters 161 Peterson, B. J., and R. W. Howarth. 1987. Sulfur, carbon, and nitrogen isotopes used to trace organic matter flow in the salt-marsh estuaries of Sapelo Island, Georgia. Limnol. Oceanogr. 32:1195- 1213. Peterson, B. J., R. W. Howarth, and R. H. Garritt. 1985. Multiple stable isotopes used to trace the flow of organic matter in estuarine food webs. Science 227:1361-1363. Peterson, G. W., and R. E. Turner. 1994. The value of salt marsh edge vs. interior as a habitat for fish and decapod crustaceans in a Louisiana tidal marsh. Estuaries 17:235-262. Post, D. M., C. A Layman, D. A., Arrington, G. Takimoto, J. Quattrochi, and C. G. Montana. 2007. Getting to the fat of the matter: models, methods and assumptions for dealing with lipids in stable isotope analysis. Oecologia 152:179-189. Rabalais, N. N., R. E. Turner, and W. J. Wiseman, Jr. 2002. Hypoxia in the Gulf of Mexico, a.k.a. “The Dead Zone”. Annu. Rev. Ecol. Syst. 33:235—263. Rabalais, N. N., W. J. Wiseman, R. E. Turner, B. K. SenGupta, and Q. Dortch. 1996. Nutrient changes in the Mississippi River and system responses on the adjacent continental shelf. Estuaries 19:386-407. Rees, C. E., W. J. Jenkins, and J. Monster. 1978. Sulfur isotopic composition of ocean water sulfate. Geochim. Cosmochim. Acta 42:377-381. Roelke, L. A., and L. A. Cifuentes. 1997. Use of stable isotopes to assess groups of king mackerel, Scomberomorus cavalla, in the Gulf of Mexico and southeastern Florida. Fish. Bull. 95:540- 551. Rozas, L. P., and R. J. Zimmerman. 2000. Small-scale patterns of nekton use among marsh and adjacent shallow nonvegetated areas of the Galveston Bay Estuary, Texas (USA). Mar. Ecol. Progr. Ser. 193: 217-239. Schlacher, T. A., B. Liddell, T. F. Gaston, and M. Schlacher- Hoenlinger. 2005. Fish track wastewater pollution to estuaries. Oeco- logia 144:570-584. Senn, D. B., E. J. Chesney, J. D. Blum, M. S. Bank, A. Age, and J. P. Shine. 2010. Stable isotope (N, C, Hg) study of methylmercury sources and trophic transfer in the Northern Gulf of Mexico. Environ. Sci. Tech. 44:1630-1637. Turner, R. E., and N. N. Rabalais. 1991. Changes in Mississippi River water quality this century and implications for coastal food webs. BioScience 41:140-147. Turner, R. E., N. N. Rabalais, R. B. Alexander, G. Mclsaac, and R. W. Howarth. 2007. Characterization of nutrient, organic carbon, and sediment loads and concentrations from the Mississippi River into the Northern Gulf of Mexico. Estuaries Coasts 30:773-790. Walker, N., and N. N. Rabalais. 2006. Relationships among satellite chlorophyll a, river inputs, and hypoxia on the Louisiana continental shelf. Gulf of Mexico. Estuaries Coasts 29:1081-1093. West, J. B., G. J. Bowen, T. E. Dawson, and K. P. Tu, eds. 2010. Isoscapes: Understanding movement, pattern, and process on Earth through isotope mapping, 487p. Springer Science + Business Media B.V., New York and Dordrecht, The Netherlands. Wissel, B., and B. Fry. 2005. Tracing Mississippi River influences in estuarine food webs of coastal Louisiana. Oecologia 144:659-672. Wissel, B., A. Gage, and B. Fry. 2005. Tracing river influences on phytoplankton dynam- ics in two Louisiana estuaries. Ecology 86:2251-2762. York, J. K., G. Tomasky, I. Valiela, and A. E. Giblin. 2010. Isotopic approach to determining the fate of ammonium regenerated from sediments in a eutrophic sub-estuary of Waquoit Bay, MA. Estuaries Coasts 33:1069-1079. 162 Stage-specific vertical distribution of Alaska plaice ( Pleuronectes quadrituberculatus ) eggs in tSie eastern Bering Sea Kathryn L. Mier Email address for contact author Janet.Duffy-Anderson@noaa.gov Resource Assessment and Conservation Engineering Division Recruitment Processes Program Alaska Fisheries Science Center National Marine Fisheries Service National Oceanic and Atmospheric Administration 7600 Sand Point Way NE Seattle, Washington 98115-6349 Abstract — The stage-specific distri- bution of Alaska plaice (Pleuronectes quadrituberculatus ) eggs in the south- eastern Bering Sea was examined with collections made in mid-May in 2002, 2003, 2005, and 2006. Eggs in the early stages of development were found primarily offshore of the 40 -m isobath. Eggs in the middle and late stages of development were found inshore and offshore of the 40-m iso- bath. There was some evidence that early-stage eggs occur deeper in the water column than late-stage eggs, although year-to-year variability in that trend was observed. Most eggs were in the later stages of develop- ment; therefore the majority of spawn- ing is estimated to have occurred a few weeks before collection — probably April — and may be highly synchro- nized among local spawning areas. Results indicate that sampling with continuous underway fish egg collec- tors (CUFES) should be supplemented with sampling of the entire water column to ensure adequate samples of all egg stages of Alaska plaice. Data presented offer new information on the stage-dependent horizontal and vertical distribution of Alaska plaice eggs in the Bering Sea and provide further evidence that the early life history stages of this species are vul- nerable to near-surface variations in hydrographical conditions and climate forcing. Manuscript submitted 6 July 2010. Manuscript accepted 19 January 2011. Fish. Bull 109:162-169 (2011). 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. Janet T. Duffy-Anderson (contact author) Deborah M. Blood Knowledge of the vertical distribu- tions of fish eggs and larvae is impor- tant to understanding how wind and currents may affect early life stages. The eggs of several pleuronectid spe- cies in the North Pacific are positively buoyant (Pearcy, 1962; Bailey et ah, 2005). Retention of pelagic eggs at the top of the water column exposes them to wind mixing that ensures an adequate oxygen supply for the developing egg, but also increases susceptibility of the eggs to stochastic wind events and adverse advection. Consistent baroclinic flows below the wind-mixed layer facilitate retention of developing eggs, but also could expose eggs to anoxia and increased predation. Vertical position often changes with development, and stage- dependent ascension can occur slowly throughout the developmental period, or quickly once eggs reach a critical stage. Alaska plaice (Pleu?'otiectes quadri- tuberculatus) is one of the major shal- low water flatfishes in the Bering Sea; however, there is not a significant fishery for the species. Alaska plaice are primarily harvested as bycatch in fisheries targeting other, more lucra- tive groundfishes, and a large por- tion of the Alaska plaice biomass is discarded. Adult Alaska plaice spawn in spring over the middle Bering Sea shelf at depths of 50-100 m, and egg and larval stages are pelagic (Bailey et al., 2003). Previous work (Duffy- Anderson et al., 2010) has shown that Alaska plaice larvae occur in the up- per 20 m of the water column, but vertical patterns of egg distribution have not been determined. The continuous underway fish egg sampler (CUFES; Checkley et al., 1997) is a tested collection system used in sampling near-surface eggs from a fixed depth (3 m) and has the advantage of being able to sample eggs in adverse weather conditions when tows with nets are not possi- ble (Checkley et al., 2000; Lo et al., 2001). The CUFES has been used in other regions to sample and estimate the densities of marine fish eggs in the water column (Dopolo et al., 2005; Pepin et al., 2005). Accurate derivation of depth-integrated egg densities from near-surface estimates requires a complete understanding of patterns of vertical egg distribu- tion with depth and development, but this information is not available for a number of fish species in the Bering Sea, including Alaska plaice. The goals of the present study were 1) to determine the developmental stage of Alaska plaice eggs collected from depth-discrete tows conducted in the eastern Bering Sea; 2) to exam- ine the vertical distribution of staged eggs; and 3) to determine whether the CUFES could be a suitable sampler of Alaska plaice eggs in the Bering Sea. Duffy-Anderson et at: Stage-specific vertical distribution of Pleuronectes qucidrituberculatus eggs 163 175°0'0"W 170°0'0"W 165°0'0"W 160°0'0"W 155WW Figure 1 Map of study region (inset) and stations where eggs of Alaska plaice (Pleuronectes quadritubercula- tus) were collected with the multiple opening and closing net and environmental sampling system (X) and neuston net (•) tows in the eastern Bering Sea during 2002-06. The polygon delineates the sampling area. Materials and methods Alaska plaice eggs were obtained from a series of fisheries research cruises conducted by the Alaska Fisheries Science Center (AFSC) along the Alaska Peninsula in the eastern Bering Sea in 2002, 2003, 2005, and 2006 (Fig. 1, Table 1). Eggs were collected from surface waters (<0.5 m depth) with a Sameoto neuston net with 505-pm mesh (Sameoto and Jaro- szynski, 1969; Jump et al., 2008). Depth-discrete sampling was conducted by using a 1-m2 multiple opening and closing net and environmental sampling system (MOCNESS; Wiebe et al., 1976) with 505-pm mesh nets in 2003 and 2005. Depth intervals sampled were 0-10 m, 10-20 m, 20-30 m, 30-40 m, 40-50 m, and >60 m. Oceanographic variables were collected simultaneously with all sampling, and these data have been published elsewhere (Duffy-Anderson et al., 2010). All eggs were fixed in 5% formalin, sorted, identi- fied, and enumerated at the Plankton Sorting and Identification Center in Szczecin, Poland. Egg identi- Table 1 Number of Alaska plaice ( Pleuronectes quadritubercu- latus) eggs collected and staged by year and gear type. MOCNESS =multiple opening and closing net and envi- ronmental sampling system. Year Cruise dates Gears used Number of eggs collected and staged 2002 12-21 May Neuston net 16 MOCNESS 2 2003 17-24 May Neuston net 26 MOCNESS 353 2005 10-20 May Neuston net 295 MOCNESS 867 2006 8-19 May Neuston net 615 fications were verified, and Alaska plaice eggs were measured and staged at the Alaska Fisheries Science Center in Seattle, WA. 164 Fishery Bulletin 109(2) Developmental stages of eggs Stages of Alaska plaice eggs were determined according to developmental criteria standardized and described by Blood et al. (1994) for walleye pollock (Theragra chal- cogramma). Staging criteria for walleye pollock are the standard criteria used to determine northern marine teleost egg development and were easily adapted to egg development in Alaska plaice. Modifications were made to the protocol in order to accommodate additional embryonic growth in Alaska plaice. For the present study, the last stage (21) in the egg development schedule for walleye pollock is defined as that stage when the tail tip reaches the back of the embryo’s head and two stages are added: stage 22, when the tail tip extends to 14 of the circumference of the egg beyond the snout; and stage 23, when the tail tip extends to V2 of the circumference of the egg beyond the snout. Eggs were then binned into three broader categories that comprised 1) early-stage eggs (E: stages 1-12), which includes all stages before the closing of the blas- topore; 2) middle-stage eggs (M: stages 13-15), when the blastopore is closed, the margin of the tail is de- fined, and the tail bud is thick, but the margin remains attached to the yolk; and 3) late-stage eggs (L: 16-23), when the tail lifts away from yolk and lengthens and encircles the top half of the yolk to extend just beyond the head of the embryo. The late-stage period was fur- ther subdivided into three stages after determining that the majority of eggs were in one of the following stages of development; early-late stage (EL: stages 16-18); middle-late stage (ML: 19-21); and late-late stage (LL: 22-23). Collections with the neuston net Eggs were collected from surface waters (<1 m depth) across the southern Bering Sea shelf in the vicinity of the Alaska Peninsula. Collections were made over the basin, outer, middle, and inner shelves. Eggs from each tow were staged and the hypothesis that there were differences in geographic (horizontal) distribution with stage of development was examined by using Cramer- von Mises tests. Vertical distributions of eggs determined with MOCNESS tows The hypothesis that vertical patterns in egg abundance vary with depth and developmental stage was evaluated by using data collected from MOCNESS tows. A fourth root transformation was used to improve the normality of the data. In 2003, a series of MOCNESS tows were conducted at a single station, whereas in 2005, MOC- NESS tows were conducted at multiple stations over the Bering Sea shelf. A general linear model ANOVA was used for each cruise by using haul as a blocking factor in 2003, and station as a blocking factor in 2005. If significant differences with depth were found, the analysis was followed with Fisher’s least significant dif- ference comparisons (Milliken and Johnson, 1992). We also checked for autocorrelation in 2003 (Durbin Watson statistic) because in that year samples were taken over time at a single station. Results Stages of developing eggs More than 2100 eggs were examined for this work (Table 1), of which more than 950 eggs were collected from neuston nets and the remainder from MOCNESS tows. The earliest developmental stage observed was stage 5 (32 cells) (Table 2); we estimated these eggs would be about 1 day old at in situ temperatures (4°C). Egg development studies of walleye pollock and arrowtooth flounder ( Atheresthes stomias) have established that the time required to reach stage 5 is 18-28 hours at about 3°C (Blood, 2002; Blood et ah, 2007), and we assumed similar rates for Alaska plaice eggs. The presence of stage-5 eggs indicates that residual spawning occurred in mid-May; however, the vast majority of the eggs were late-stage eggs (Table 2), indicating that most spawn- ing occurred a few weeks before sampling (Pertseva- Ostroumova, 1961). Most eggs collected were at stage 22. Hatching occurs at stage 23; the eyes of embryos are fully pigmented and numbers of eggs are greatly reduced in contrast to stage 22. Collections with the neuston net Eggs collected from neustonic surface collections repre- sented all stages of development (Fig. 2). Results of all pairwise Cramer-von Mises tests for differences in spa- tial distributions showed that there were no significant differences between the geographic distributions of any stages of larvae, except between the geographic distri- butions of EL and ML (P= 0.002). However, there were some trends that could be discerned. Earliest stage eggs collected in the neuston layer appeared to concentrate offshore of the 40-m isobath, over bottom depths ranging from 40 to 75 m. Eggs in middle stages of development appeared to spread shoreward toward shallower depths, but catches of eggs in both the early and middle stages of development were comparatively low. Late-stage eggs occurred over depths ranging from 40 to 100 m. Vertical distributions of eggs determined with MOCNESS tows Vertical distributions of Alaska plaice eggs showed dif- ferences in depth distribution with ontogenetic stage (Fig. 3), and differences between years. In 2003, there was no significant effect of haul and therefore no auto- correlation (Durbin Watson statistic =1.91; effect of haul), and eggs were generally distributed throughout the water column. There was only one egg collected in the early stage and it was located in the deepest depth stratum (40-50 m). There were no collections of eggs Duffy-Anderson et al.: Stage-specific vertical distribution of Pleuronectes quadrituberculatus eggs 165 Table 2 Results of staging Alaska plaice (Pleuronectes quadrituberculatus) eggs and the percentage of eggs in each developmental stage bin by gear type (early: stages 1-12; middle: stages 13-15; late: stages 16-23). The late stage was subdivided into three cat- egories (early late: stages 16-18; middle late: stages 19-21; and late late: stages 22-23). Stages were adapted from Blood et al. (1994) and two developmental stages were added for this study. MOCNESS=multiple opening and closing net and environmental sampling system. Percentage of Number of eggs/stage total eggs/stage Neuston net MOCNESS Developmental (neuston net (neuston net (percentage of (percentage of stage and MOCNESS) and MOCNESS) eggs collected) eggs collected) Early 1 0 0 2 0 0 3 0 0 4 0 0 5 3 0.14 6 7 0.32 7 20 0.92 8 2 0.09 9 20 0.92 10 7 0.32 11 14 0.64 12 10 0.46 Total 8.1 0.5 Middle 13 10 0.46 14 23 1.06 15 51 2.35 Total 8.4 0.4 Early late 16 58 2.67 17 158 7.27 18 67 3.08 Total 23.0 5.2 Middle late 19 209 9.61 20 230 10.58 21 285 13.11 Total 30.2 35.7 Late late 22 727 33.44 23 272 12.51 Total 30.3 58.2 in the middle stages of development (stages 13-15). Among late-stage eggs in 2003, the densities of egg abundances were depressed in the near surface waters (0-10 m) and significantly so for ML and LL stages (Table 3). In general however, most eggs were collected from above the mixed layer (<30 m). In 2005, a different pattern emerged; late-stage eggs consistently occurred at shallower depths than did earlier stages (Fig. 3), and the majority were collected in near-surface waters. Statistical examination revealed that more eggs were collected from depths 0-10 m and 10-20 m than in any of the deeper depths. Considered collectively, approximately 34% of the catch in the two years occurred in the top 10 m of the water column, 24% occurred between 10 and 20 m, 18% between 20 and 30 m, 11% between 30 and 40 m, 8% between 40 and 50 m, and 5% between 50 and 60 m depth. Discussion This is the first study to describe stage-dependent verti- cal and horizontal distribution of Alaska plaice eggs. Our data indicate that spawning occurs offshore of the 40-m isobath, and likely near-bottom, confirming hypotheses outlined in Bailey et al. (2003). Eggs occur throughout the water column, but many eggs occur in the upper water column (<30 m depth). The vast majority of eggs collected in the present study were in the later stages of development, and we estimate that the majority of spawning occurred a few weeks before collection. Maxi- 166 Fishery Bulletin 109(2) mum larval abundance in the Bering Sea occurs in May (Duffy-Anderson et al., 2010) and probably reflects peak hatching of eggs spawned in April and a relatively high degree of spawning synchrony. The geographic sampling area over which these eggs were collected was small compared to the potential spawning area over the Bering Sea continental shelf, and available evidence indicates that Alaska plaice do spawn over a large portion of the middle domain of the continental shelf (Zhang et al., 1998). The wind-mixed layer in the Bering Sea generally extends to 25-30 m in spring (Stabeno et al., 2001), and our data reveal that Alaska plaice eggs primarily occur above or within this layer. As such, the eggs are vulnerable to the sto- chastic effects of wind activity, which could disperse them widely over the shelf, especially in early spring months ( March-April) when the likelihood of storm events is high. However, prevailing winds over the shelf in late spring-summer are southwesterly and would therefore transport late-stage eggs and newly hatched larvae from the middle shelf toward nursery areas along the Alaska mainland coast. Indeed, previous work has shown that Alaska plaice larvae are relatively rare over the continental shelf (Bailey et al., 2003), lending credence to the idea of shoreward transport of older egg stages and hatched larvae. It should be noted that retention in near-surface layers is also likely to promote faster rates of egg development because temperatures in the upper water column are 1— 3°C warmer than those at depth over the middle shelf during spring. Alaska plaice eggs do occur in near-surface waters, making them accessible to CUFES system, but many eggs also occur below the depths sampled with the CUFES. Therefore, abundance determined from eggs caught with the CUFES system may be underestimat- ed-— particularly the abundance of early stages that might be deeper in the water column. This observation has been made elsewhere (Lo et al., 2001; Dopolo et al., 2005), and at least in the case of Alaska plaice, we recommend that sampling with the CUFES system be supplemented with sampling of the entire water column to ensure adequate sampling of eggs at all stages of development. Moreover, sampling earlier in the spring, in March-April, for earlier stages is encouraged. Acknowledgments Thanks to the officers and crew of the NOAA ship RV Miller Freeman. Comments by T. Smart, C. Jump, J. Napp, and three anonymous reviewers improved the manuscript. Funds were provided by the North Pacific Duffy-Anderson et al.: Stage-specific vertical distribution of Pleuronectes quadrituberculatus eggs 167 A o-io 10-20 20-30 30-40 40-50 50-60 B 0 1 2 3 0-10 E 10-20 .1 20-30 | 30-40 § 40-50 50-60 ^ 0 1 2 3 0-10 10-20 20-30 30-40 40-50 50-60 0 5 10 15 20 25 2003 X 2005 2003 D 2005 Mean abundance (catch per 1000 m3) Mean abundance (catch per 1000 m3) Figure 3 Mean abundance (± standard deviation) of Alaska plaice < Pleuronectes quadrituberculatus ) eggs by developmental stage and depth. The symbol X indicates that no sample was taken at that depth. Data are derived from catches with the mul- tiple opening and closing net and environmental sampling system in 2003 and 2005. (A) Early stage, (B) middle stage, (C) early-late stage, (D) middle-late stage, (E) late-late stage. Table 3 Statistical comparisons Off differences in abundances of Alaska plaice ( Pleuronectes quadrituberculatus) eggs in by depth bin. Significance is noted at P<0.05. ns=not significant. ML=middle late stage. LL=late late stage. 2003 and 2005 ML 2003 0—10 m 10-20 m 20—30 m 30-40 m 40-50 m 50-60 m 60+ m 0-10 m ns 10-20 m 0.004 ns 20—30 m <0.001 ns ns 30-40 m 0.035 ns ns ns 40-50 m 0.001 ns ns ns ns 50-60 m ns ns ns ns ns ns 60+ m ns ns ns ns ns ns ns continued 168 Fishery Bulletin 109(2) Table 3 (continued) LL 2003 0-10 m 10-20 m 20-30 m 30-40 m 40-50 m 50-60 m 60+ m 0-10 m ns 10-20 m ns ns 20-30 m 0.002 0.017 ns 30-40 m ns ns ns ns 40-50 m 0.041 ns ns ns ns 50-60 m ns ns ns ns ns ns 60+ m ns ns ns ns ns ns ns ML 2005 0-10 ill 10-20 m 20-30 m 30-40 m 40-50 m 50-60 nr 60+ m 0-10 m ns 10-20 m ns ns 20-30 m ns 0.009 ns 30-40 m ns 0.017 ns ns 40-50 m 0.005 0.001 ns ns ns 50-60 m ns 0.022 ns ns ns ns 60+ m ns 0.032 ns ns ns ns ns LL 2005 0-10 m 10-20 m 20-30 m 30-40 m 40—50 m 50-60 m 60+ m 0-10 m ns 10-20 m ns ns 20-30 m 0.013 ns ns 30-40 m 0.003 0.022 ns ns 40-50 m <0.001 <0.001 0.017 ns ns 50-60 m ns ns ns ns ns ns 60+ m .007 0.03 ns ns ns ns ns and Climate Regimes and Ecosystem Productivity Blood, D. M„ A . C. Matarese, and M. S. Busby. (NPCREP) program and the Alaska Fisheries Science Center’s Resource Assessment Conservation and Engi- neering Division (NOAA). This research is contribution EcoFOCI-N736 to NOAA’s Ecosystems and Fisheries- Oceanography Coordinated Investigations program. Literature cited 2007. Spawning, egg development, and early life history dynamics of arrowtooth flounder (Atheresthes stomias) in the Gulf of Alaska. NOAA Prof. Paper NMFS 7, 28 p. Blood, D. M., A. C. Matarese, and M. M. Yoklavich. 1994. Embryonic development of walleye pollock, Ther- agra chalcogramma , from Shelikof Strait, Gulf of Alaska. Fish. Bull. 92:207-222. Checkley, D. M., Jr, J. R. Hunter, L. Motos, and C.D. van der Lingen. Bailey, K. M., E. Brown, and J. T. Duffy-Anderson. 2003. Aspects of distribution, transport, and recruitment of Alaska plaice (Pleuronectes quadrituberculatus) in the Gulf of Alaska and eastern Bering Sea: comparison of marginal and central populations. J. Sea Res. 50:87-95. Bailey, K. M., H. Nakata, and H. W. Van der Veer. 2005. The planktonic stages of flatfishes: physical and biological interactions in transport processes. In Flat- fishes: biology and exploitation (R. N. Gibson, ed.), p. 94-119. Blackwell Science Publ., Oxford, U.K. Blood, D. M. 2002. Low temperature incubation of walleye pollock ( Theragra chalcogramma ) from the southeast Bering Sea and Shelikof Strait, Gulf of Alaska. Deep Sea Res. II 49:6095-6108. 2000. Report of a workshop on the use of the continuous underway fish egg sampler (CUFES) for mapping spawn- ing habitats of pelagic fish. GLOBEC Rep. 14:1-65. Checkley, D. M., P. B. Ortner, L. R. Settle, and S. R. Cummings. 1997. A continuous, underway fish egg sampler. Fish. Oceanogr. 6:58-73. Dopolo, M. T., C. D. van der Lingen, and C. L. Moloney. 2005. Stage-dependent vertical distribution of pelagic fish eggs on the western Agulhas Bank, South Africa. Afr. J. Mar. Sci. 27:249-256. Duffy-Anderson, J. T., M. Doyle, K. L. Mier, P. Stabeno, and T. Wilderbuer. 2010. Early life ecology of Alaska plaice ( Pleuronectes quadrituberculatus ) in the eastern Bering Sea: distribu- tion, transport pathways, and effects of hydrography. J. Sea Res. 64:3-14. Duffy-Anderson et al.: Stage-specific vertical distribution of Pleuronectes quadntuberculatus eggs 169 Jump, C. M., J. T. Duffy-Anderson, and K. L. Mier. 2008. Comparison of neustonic ichthyoplankton samplers in the Gulf of Alaska. Fish. Res. 89:222-229. Lo, N. C. H., J. R. Hunter, and R. Charter. 2001. Use of a continuous egg sampler for ichthyo- plankton surveys: application to estimation of daily egg production of Pacific sardine ( Sardinops sagax) off California. Fish Bull. 99:554—571. Milliken, G. A. and D. E. Johnson. 1992. Analysis of messy data. Vol. I: design experiments, 473 p. Chapman and Hall, London. Pearcy, W. G. 1962. Distribution and origin of demersal eggs within the order Pleuronectiformes. J. Conseil 27:232 — 235. Pepin, P., P. V. R. Snelgrove, and K. P. Carter. 2005. Accuracy and precision of the continuous underway fish egg sampler (CUFES) and bongo nets: a comparison of three species of temperate fish. Fish. Oceanogr. 14:432-447. Pertseva-Ostroumova, T. A. 1961. The reproduction and development of far eastern flounders. Akad. Nauk SSSR Inst. Okeanol., 484 p. [Transl. by Fish. Res. Board Can., 1967, Transl. Ser. 856.] Sameoto, D. D., and L. O. Jaroszynski. 1969. Otter surface trawl: a new neuston net. J. Fish. Res. Board Can. 26:2240-2244. Stabeno, P. J., N. A. Bond, N. B. Kachel, S. A. Salo, and J. D. Schumacher. 2001. On the temporal variability of the physical envi- ronment over the south-eastern Bering Sea. Fish. Oceanogr. 10:81-98. Wiebe, P. H., K. H. Burt, S. H. Boyd, and A. W. Morton. 1976. A multiple opening/closing net and environmental sensing system for sampling zooplankton. J. Mar. Res. 34:313-326. Zhang, C. I., T. K. Wilderbuer, and G. E. Walters. 1998. Biological characteristics and fishery assess- ment of Alaska plaice, Pleuronectes quadrituberculatus , in the eastern Bering Sea. Mar. Fish. Rev. 60:16-27. 170 Abstract — Research on assessment and monitoring methods has primar- ily focused on fisheries with long multivariate data sets. Less research exists on methods applicable to data- poor fisheries with univariate data sets with a small sample size. In this study, we examine the capabilities of seasonal autoregressive integrated moving average (SARIMA) models to fit, forecast, and monitor the landings of such data-poor fisheries. We use a European fishery on meagre (Sciaeni- dae : Argyrosomus regius), where only a short time series of landings was available to model (;; = 60 months), as our case-study. We show that despite the limited sample size, a SARIMA model could be found that adequately fitted and forecasted the time series of meagre landings (12-month fore- casts; mean error: 3.5 tons (t); annual absolute percentage error: 15.4%). We derive model-based prediction inter- vals and show how they can be used to detect problematic situations in the fishery. Our results indicate that over the course of one year the meagre landings remained within the predic- tion limits of the model and therefore indicated no need for urgent man- agement intervention. We discuss the information that SARIMA model structure conveys on the meagre life- cycle and fishery, the methodological requirements of SARIMA forecasting of data-poor fisheries landings, and the capabilities SARIMA models pres- ent within current efforts to monitor the world’s data-poorest resources. Manuscript submitted 8 March 2010. Manuscript accepted 20 January 2011. Fish. Bull. 109:170-185(2011). 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 SARIMA models to assess data-poor fisheries: a case study with a sciaenid fishery off Portugal Nuno Prista (contact author)1-2 Norou Diawara3 Maria Jose Costa1-2 Cynthia Jones4 Email address for contact author: nmprista@ipimar.pt 1 Centro de Oceanografia Faculdade de Ciencias da Universidade de Lisboa Campo Grande, 1749-016 Lisboa, Portugal Present address for contact author: Unidade de Recursos Marinhos e Sustentabilidade Instituto Nacional de Recursos Biologicos (INRB, I.P./IPIMAR) Avenida de Brasilia, 1449-006 Lisboa, Portugal 2 Departamento de Biologia Animal Faculdade de Ciencias da Universidade de Lisboa Campo Grande, 1749-016 Lisboa, Portugal 3 Department of Mathematics and Statistics Old Dominion University Norfolk, Virginia 23529-0077 4 Center for Quantitative Fisheries Ecology Old Dominion University 800W 46th Street Norfolk, Virginia 23508-2099 Research, assessment, and manage- ment have traditionally focused on fisheries with the greatest landings and revenues (Scandol, 2005; Vas- concellos and Cochrane, 2005). Such fisheries are generally data-rich and have available the funds and exper- tise required to complete stock assess- ments and provide state-of-the-art advice to management. However, that is not the case for the vast majority of fisheries worldwide, which remain subjected to limited (if any) assess- ment and management (Vasconcel- los and Cochrane, 2005). The latter have been collectively termed “data- poor fisheries” and are character- ized by a low diversity and quantity of data, limitations in funding and expertise, and an overall shortage of assessment methods (Mahon, 1997; Scandol, 2005). Among the world’s data-poorest fisheries are nearly all fisheries in developing countries, but also most fisheries in developed countries, namely the smaller-scale or less valuable commercial and rec- reational ones (NRC, 1998; Berkes et al., 2001; EEA, 2005; Vasconcellos and Cochrane, 2005; Worm et al., 2009; OSPAR, 2010; ICES1). Assessment of data-poor fisheries requires a significantly different ap- proach from their data-rich counter- parts. For data-poor fisheries, many deterministic multivariate stock as- sessment models cannot be used (e.g., NRC, 1998) and more pragmatic as- sessment methods must be put in place, particularly when fishery-in- dependent data are not available and fishing effort cannot be quantified (Berkes et al., 2001; Scandol, 2003; ICES1). In many countries, the most readily available fisheries data are commercial landings because of their 1 ICES (International Council for the Exploration of the Sea). 2008. Report of the study group on management strat- egies (SGMAS), 74 p. ICES CM 2008/ ACOM:24, Copenhagen, Denmark. Prista et al: Use of SARIMA models to assess data-poor fisheries 171 connection to the economy and business (Vasconcellos and Cochrane, 2005). Commercial landings result from complex interactions between the environment, the fish- ing fleet, and the stocks, and therefore do not directly reflect the status of exploited populations. However, landing records contain valuable information that can be useful to managers if routine monitoring, rather than stock assessment, is established as a manage- ment objective (Scandol, 2003). In fact, even if they provide suboptimal indications on the status of the stocks, statistical analyses of landings can lead to the timely detection of phenomena such as sudden increases in fishing effort or marked population declines that could otherwise remain undetected (Caddy, 1999). Such detection is important — particularly within multispe- cies, budget-limited, management contexts — because it allows the prioritization of research and management actions toward the subset of fisheries and stocks most likely to be depleted (Scandol, 2003). Autoregressive integrated moving-average (ARIMA) models are simple time series models that can be used to fit and forecast univariate data such as fisheries landings. With ARIMA models data are assumed to be the output of a stochastic process, generated by un- known causes, from which future values can be pre- dicted as a linear combination of past observations and estimates of current and past random shocks to the system (Box et al., 2008). In fisheries, ARIMA models (and their seasonal multiplicative version, SARIMA) have a long record of successful application that extends from modeling (e.g., Hare and Francis, 1994; Fogarty and Miller, 2004) to short-term forecasting of a variety of variables and resources for both data-rich and data- poor fisheries (Table 1). Specifically, SARIMA models, which are applicable to many already-available land- ings data sets, have been found to provide both annual and monthly forecasts that are comparable to, or even better than forecasts from many multivariate models, including some with fishing effort among the predictors (Stergiou et al., 1997). The good record, flexibility, and simplicity of SARI- MA models have made them natural candidates for the modeling of data-poor fisheries (Rothschild et al., 1996). However, to date, SARIMA models in fisheries have only been applied in detail on relatively long time series (>120 months) (Table 1), and a single study has provided a few (but not detailed) results from shorter series (Lloret et al., 2000). Such emphasis of previous SARIMA modeling on long time series finds little sup- port in statistical literature where 50 months is gener- ally regarded as the minimum sample size for model application (e.g., Pankratz, 1983; Chatfield, 1996a). Ad- ditionally, most literature to date has focused on SARI- MA models as tools to generate accurate forecasts of future landings. However, in addition to good forecast- ing, these models also possess significant capabilities for monitoring landings that have remained unexplored. These capabilities become apparent when SARIMA models are approached from a statistical process-control perspective and it is made known that SARIMA model forecasts include the assumption of persistence (through time) of the process that generated the data (Box et al., 2008; Mesnil and Petitgas, 2009). Briefly, good land- ing forecasts are only attainable as long as significant changes do not take place in the fishery; therefore large forecast errors can be regarded as indications that can be changes in the fishery process took place that may require management intervention (Pajuelo and Lorenzo, 1995; Georgakarakos et al., 2006; Box et al., 2008). In this study, we report the first detailed applica- tion of SARIMA models for monitoring of data-poor fisheries landings. We use data from a previously un- assessed Portuguese fishery on meagre (Sciaenidae: Argyrosomus regius ) as our example. The meagre is a valuable top predator from European coastal wa- ters but its stocks have not been analytically assessed because of limitations in data, personnel, and fund- ing existing at the national level. At the time of our analysis only a short time series of monthly landings (60 months) was available for this fishery, a situation that replicates conditions found in many other data- poor fisheries worldwide. We show that the short time series was not a problem for SARIMA modeling and forecasting and that prediction intervals from SARI- MA models can be used to provide this fishery with basic monitoring. We suggest that SARIMA models should be more widely considered to extend the cover- age of monitoring to all exploited marine resources. Materials and methods Meagre (Argyrosomus regius) and its fisheries Meagre is one of the world’s largest and most valuable sciaenids (up to 180 cm, 50 kg, and with a US$ 15 per kg exvessel price). It ranges from France to Senegal, and the largest fisheries take place off Mauritania, Morocco, and Egypt. In Europe, the meagre constitutes a prized trophy-fish for anglers and an important income for small-scale commercial fishermen along the Atlantic shores of France, Spain, and Portugal. Its biology and life cycle remain scarcely documented, but recent concerns about the overexploitation of juveniles and interests in aquaculture production have sparked some research. Currently, the fish is known to be fairly long-lived (up to 44 yr) (Prista et al., 2009), to present fast juvenile growth (Morales-Nin et al., 2010) and to spawn at 3-4 yr old (N. Prista, unpubl. data). Data on adult growth and reproduction have not been published, but preliminary reports indicate a life-cycle characterized by fast growth, high fecundity, and a long reproductive span, and that the estuaries of the Gironde (France), Tagus (Portugal), and Guadalquivir (SW Spain) rivers constitute the main spawning habitats (Quemener, 2002; Prista et al.2; N. 2 Prista, N., C. M. Jones, J. L. Costa, and M. J. Costa. 2008. Inferring fish movements from small-scale fisheries data: the case of Argyrosomus regius (Sciaenidae) in Por- tugal, 19 p. ICES CM 2008/K-19, Copenhagen, Denmark. 172 Fishery Bulletin 109(2) Prista, unpubl. data). Marked seasonal variations in landings linked to juvenile and adult migrations have been identified in local fisheries (Quero and Vayne, 1987; Prista et al.2). Overall, adults are thought to come inshore from spring to early summer to spawn but their overwin- tering grounds are still unknown; juveniles are thought Table 1 Primary fisheries literature that present seasonal autoregressive integrated moving-average models. Only studies with quan- titative forecast results are displayed. “No.”=the number of series, “Freq”=the sampling frequency (W=weekly, M=monthly, A= annual), “n” is the sample size of the fitting period, “F”=number of forecasts, “models” indicates the type of models compared, and “PI” indicates if prediction intervals were presented (yes, no). “/” separates annual and monthly data sets when both were analyzed, “sp” = species, “nsp groups” = nonspecific groups, “rel.” = relative, “CPUE”=catch per unit of effort, “LPUE”=landings per unit of effort. Reference Species Variable No. Freq n F Models3 PI Saila et al. (1980) Jasus edwardsii CPUE 1 M 144 12 1,5 n Mendelssohn (1981) Katsuwonus pelamis catch/effort 1 M 180 12 12 n Fogarty (1988) Homarus americanus catch/CPUE 3/1 A/M 41-58/216 1/12 12 n Jeffries et al. (1989) Pseudopleuronectes americanus rel. abundance 2/3 A/M 27/156;324 2/12 — y Stergiou (1989) Sardina pilchardus catch 1 M 204 12 — n Noakes et al. (1990) Oncorhynchus nerka total returns 2 A 24 8 1,10,12,19,20 n Stergiou (1990a) Engraulis encrasicolus catch 1 M 252 24 — n Stergiou (1990b) Mullidae catch 1 M 252 24 — n Campbell et al. (1991) Homarus americanus catch 4 A 61-97 10 12 n Molinet et al. (1991) Penaeus spp., Lutjanus synagris landings/LPUE 2 M 132;180 24 — n Stergiou (1991) Trachurus sp. catch 1 M 252 12 1,8 n Tsai and Chai (1992) Morone saxatilis harvest 1 A 27 4 3,4,12 n Pajuelo and Lorenzo 1 nsp group catch 1 M 131 24 — y (1995) Stergiou and Christou 4 sp; 12 nsp groups catch 16 A 24 2 1-9 n (1996) Stergiou et al. (1997) 4 sp; 12 nsp groups catch 16 M 288 24 1-5, 7-9 n Park (1998) Theragra chalcogramma landings 1 M 264 24 — n Lloret et al. (2000)6 30 sp; 36 nsp groups catch 66 M 51-200 12 — y Georgakarakos et al. (2002, 2006) Loligo vulgaris, Todarodes sagittatus landings 2 M 174 12 11,15,16 y Pierce and Boyle Loligo forbesi LPUE 1 A/M 27/324 3/36 3, 12 y (2003) Stergiou et al. (2003) Xiphias gladius catch 1 M 180 12 8,13 n Zhou (2003) Oncorhynchus tshawytscha spawner density 2 A 11 4 1, 15 n Hanson et al. (2006) Bi'evoortia tyrannus, B. patronus landings 2 A 57;63 10 3,14,15 n Koutroumanidis et al. (2006) E. encrasicolus, Merluccius merluccius, Sarda sarda landings 3 M 216;252 12 17,18 n Czerwinski et al. Hippoglossus stenolepis CPUE 1 W 107 31 15 n (2007) Tsitsika et al. (2007) Total pelagic production E. encrasicolus, S. pilchardus, T. trachurus CPUE 4 M 180 12 11 y a Models compared: l=naive, 2=linear regression (LR), 3=multiple LR, 4=multiple LR with correlated errors, 5=harmonic LR, 6=Fox surplus- yield, 7=model combination, 8=exponential, 9=vector autoregressive, 10=periodic autoregressive, ll=multivariate ARIMA, 12= transfer function noise, 13=census method II (X-ll), U.S. Dep. Commer., 14 = state space models, 15=artificial neural networks, 16=Bayesian dynamic modeling, 17=genetic modeling for optimal forecasting, 18=fuzzy expected intervals, 19 = stock-recruitment, 20 = sibling. b The Lloret et al. (2000) study includes 12 series with 51-64 months. Prista et al: Use of SARIMA models to assess data-poor fisheries 173 Figure 1 Time series of monthly meagre (Argyrosomus regius) landings, in tons, in the Lisboa region of the Portuguese coast (May 2002 to April 2008). The dashed vertical line is the forecast origin (April 2007) and separates the fitting period (May 2002 to April 2007, left) from the hold-out period (May 2007 to April 2008, right). (A) Raw data. (B) Log10-transformed mean-centered data. to use estuaries as nursery areas during the warmer months and overwinter in adjoining coastal grounds (Quero and Vayne, 1987; Quemener, 2002; Prista et al.2; N. Prista, unpubl. data). Recently, substantial conservation risks have been identified in European meagre fisheries that are related to the overexploitation of juvenile and adults schools in estuaries and nearby coastal areas (Quemener, 2002; Prista et al.2). To protect juveniles, precau- tionary management measures have been put in place (namely minimum landing size regulations) but the ac- tual status of the meagre stocks was never assessed. This lack of assess- ment mainly results from a lack of suf- ficient multivariate time-series data and because national assessment pri- orities, funding, and expertise are gen- erally allocated to the largest national and transnational fisheries instead of the less-significant, albeit numerous and regionally important, ones. The fish being largely absent from routine fishery-independent surveys (Quero and Vayne, 1987; F. Cardador, personal commun.3) and difficulties related to its sampling at port and the estima- tion of fishing effort (Prista et al.2’4) further contribute to its unassessed status. In this type of setting, if simple methods are not put in place that can, at least, detect the most alarming sig- nals in the landings data it is likely that stock collapses can occur without being detected. Data set and data transformations The Lisboa region in Central West Portugal (hence- forth termed “Lisboa region”) (38°25'N to 38°59'N lat., ~9°15'W long.) is the main fishing area for meagre off the Iberian Peninsula (between 29% and 45% of annual landings of meagre, all gears combined, in 2001-05). In this region, most of the catch is associated with the Tagus estuary and its adjoining coastal area. The catch derives essentially from a small-scale artisanal fleet in which gillnets, trammel nets, and longlines are used to catch meagre during its spawning and nursery season (Prista et al.2). To minimize overfishing of juvenile fish, a minimum landing size of 42 cm was established in 2002 that complements an array of other gear-related 3 Cardador, Fatima. 2008. INRB, I.P./IPIMAR, Av. Brasilia, 1449-006 Lisboa, Portugal. 4 Prista, N., J. L. Costa, M. J. Costa, and C. M. Jones. 2007. New methodology for studying large valuable fish in data poor situations: commercial mark-recapture of meagre Argyrosomus regius in the southern coast of Portugal, 18 p. and effort-related management regulations that are not specific to meagre. To test SARIMA models in the monitoring of the Lisboa meagre landings, we obtained a time series of meagre monthly landings from the Portuguese General- Directorate for Fisheries and Aquaculture (DGPA). The landings data resulted from mandatory reports of fish sales obtained at all ports of the Lisboa region ( A7= 14 ) from May 2002 to April 2008 (i.e., 72 monthly values) as part of a routine data collection program (Fig. 1). We used the first 60 months to fit the SARIMA models and the last 12 months as a hold-out period to evaluate fore- casting performance and to monitor the fishery. Some previous data were available on this fishery, but those data were found to be unreliable because of contamina- tion with landings from Portuguese vessels operating off North African waters. No significant management interventions occurred on the fishery during the course of our study. Before fitting a SARIMA model, the time series must be checked for violations of the weak stationarity as- sumption of the models (Brockwell and Davis, 2002; Box et al., 2008). In SARIMA models, trend and seasonal nonstationarities are handled directly by the model 174 Fishery Bulletin 109(2) Table 2 Candidate set of seasonal autoregressive integrated moving-average models. The “rule” column displays the mathematical expression used to determine the autoregressive components ( p ) and moving-average components ( q ) of the candidate models. “Max AR term” and “Max MA term” columns display the maximum autoregressive (AR) and moving-average (MA) lags included in the model equations, with respect to the original (xt) and 12-month differenced log10-transformed mean-centered data (wt='V112yt=V112 (log10xf-4.022)), respectively. Model structure No. of models Rule Max AR term Max MA term (p,0,q)x(0,l,0)12 325 q<25— p; p<24 Wt-24> Xt-36 Zt-12 (p,0,<7)x(l,l,0)12 91 q<13-p; p<12 wt-24 > xt-36 Zt-12 (p,0,<7)x(0,l,l)12 91 q<13-p; p<12 wt-12 ’ xt-24 Zt-24 (p,0,q)x(l,l,l)12 1 II © II o wt-12’ xt-24 Zt-12 structure so that only the nonstationarity of variance needs to be addressed before model fitting. The meagre time series (xt, t= 1, ... ,60) was seasonal and exhibited no trend (Fig. 1A), but annual variance-mean plots in- dicated an increase in variance with the series mean. To correct this, we evaluated Box-Cox transformations (Box and Cox, 1964) and found that a log10 transforma- tion successfully stabilized the variance of the series. Accordingly, we log-transformed the data, subtracted its mean, and then used the mean-centered log-trans- formed data set ( yt , t= 1, ... ,60) as input to the SARIMA analyses (Fig. IB). Data modeling We fitted SARIMA models to the meagre data using a semi-automated approach based on a combination of the Box-Jenkins method with small-sample, bias-corrected Akaike information criteria (AICc) model selection (Roth- schild et ah, 1996; Brockwell and Davis, 2002). This approach involved three major steps: 1) selection of the candidate model set; 2) estimation of the model and determination of AICc; and 3) a diagnostic check. Details on the notation and model selection procedures used to fit SARIMA models to short time series are given in Appendices 1 and 2. Selection of the candidate model set was carried out by first analyzing sample estimates of the autocor- relation function (ACF) and partial autocorrelation function (PACF) in order to select the three major orders of the SARIMA models: d, D, and S. In the meagre case, we concluded that a configuration with d= 0, D = 1, and S = 12 should be adopted (see Results section). Consequently, a SARIMA(p,0,q)x(P,l,Q)12 was selected as the basic model structure of the candidate set, with p, q, P, and Q left to vary. There is no a priori method to determine the maximum value that p, q, P , and Q can take, but the maximum orders of the models are obviously restricted by sample size. In our analysis, we conditioned p, q, P, and Q to the upper boundary max(p+q+SP+SQ) = 24 and p+q< 12 (Table 2), which caused the maximum possible term of any SARIMA model to be xt_36 and the maximum possible number of parameters to be 13. We found this procedure to provide a good compromise between model complexity and the convergence of estimation algorithms. Model estimation was carried out by using maximum likelihood methods, after conditional sum of squares estimation of the starting values (Brockwell and Da- vis, 2002). Given the large number of models requiring estimation (Table 2), we developed a semi-automated software routine in R, vers. 2.5.1 (R Development Core Team, 2007) that estimated the models and output their AICf values. This routine used several functions incorporated in the R packages “stats” (R Development Core Team, 2007), “tseries” (Trapletti and Hornik, 2007), and “FinTS” (Graves, 2008). After estimation, the model with the minimum AICc. was selected for further analysis. Diagnostic checks on the AICc-selected model involved the following steps: 1) verification of the resemblance of residuals to white noise (ACF plots, Ljung-Box test, cumulative periodogram test); 2) tests on the normality of residuals (Jarque-Bera and Shapiro-Wilks tests); and 3) confirmation of model stationarity, invertibility, and parameter redundancy (Shapiro et ah, 1968; Ljung and Box, 1978; Jarque and Bera, 1987; Box et al., 2008). All tests were carried out at a significance level of a=0.05. The variance explained by the model was determined as 1 - d2 / c2 (Stergiou, 1990a). Forecasts and model performance We evaluated 12 months of model forecasts, using the last month of the fitting data set as the forecast origin (i.e., April 2007). Forecasts were obtained in the mean- centered transformed scale ( yh , h = 1,...,12) and in the original scale of the data (xh, h- 1,...,12), after correcting for back-transformation bias (Pankratz, 1983). SARIMA model performance was assessed by comparing A-step forecasts [xh and yh ) with monthly landings observed between May 2007 and April 2008 ( xh and yh). This was done by evaluating monthly forecast errors (e.g., eh- xh - xh) and then considering a set of accuracy measures: 1) annual root mean-square error (RMSE); 2) mean error (ME); 3) absolute percent error (APE/;); 4) mean absolute percent error (MAPE); and 5) annual percent Prista et al: Use of SARIMA models to assess data-poor fisheries 175 error (PE) (Mendelssohn, 1981; Hyndman and Koehler, 2006). From these, RMSE was evaluated in the trans- formed scale to allow its comparison to a , and all others were computed in the more user-friendly original scale of the data. Additionally, we compared the forecasting performance of the SARIMA model against two simple naive forecasting models (naive model 1 or NM1, and naive model 2 or NM2) (Noakes et al., 1990; Stergiou et al., 1997). The latter represented ad hoc forecasting models likely to be used in data-poor fisheries with short time series of landings: with NM1, future landings were assumed to be equal to the landings registered in the previous year; and with NM2, future landings were assumed to be equal to the average monthly landings registered in the fitting period. We also evaluated the Kitanidis and Bras (1980) coefficient of persistence (P) that summarizes forecasting results by comparing them with those of a naive model where landings at time t+ 1 are assumed equal to landings at time t. This coefficient takes values smaller than or equal to 1, with P=1 representing perfect model forecasts. Monitoring of fisheries SARIMA models predict the future on the assumption that the statistical properties of the process generating the data remain the same over time (Box et al., 2008). When framed within the perspective of statistical pro- cess control (e.g., Scandol, 2005; Box et al., 2008; Mesnil and Petitgas, 2009), this characteristic allows the pre- dictions of well-developed SARIMA models to be used as “guidelines” to monitor future observations. When a SARIMA model is found that appropriately fits the landings data, a significant departure of its forecasts from future observations can be seen as an indication that changes in the underlying fishery process have occurred (=out-of-control situation). In contrast, if such a significant departure does not take place, then there is no indication for such changes (= in-control situation). From a data-poor fisheries perspective, such a distinction means that if funding is limited and multiple fisheries require assessment, research and management efforts should be allocated to fisheries displaying out-of-control decreasing trends in production rather than to fisher- ies that remain stable or display in-control increasing trends (Scandol, 2003, 2005). The distinction between in-control and out-of-control landings requires a set of detection limits. To date, process-control detection limits for fisheries indicators have been derived mostly from historical reference da- ta (Scandol, 2003; Mesnil and Petitgas 2009; Petitgas, 2009). However, most fisheries have only a few years of collected data and consequently historical limits are difficult to estimate. In such situations, model- based detection limits like the prediction intervals (Pis) of SARIMA models (Chatfield, 1993; Box et al., 2008) provide easy-to-compute detection limits that explicitly take into account the correlation structure of the data. SARIMA Pis resemble confidence inter- vals for model forecasts and consist of upper and lower boundaries that encompass a 1-a probability region for future forecasts (Chatfield, 1993). Their main use is to convey the uncertainty around forecasts (De Gooijer and Hyndman, 2006). However, because pre- diction intervals encompass only future observations, as long as no structural changes take place in the underlying process (Chatfield, 1993), their boundar- ies can be used to monitor univariate data such as fisheries landings. To date, the prediction intervals (Pis) from SARIMA models have seldom been reported in fisheries literature and, when they have, with little detail and discussion (Table 1). To monitor the landings of the meagre fish- ery we used two types of Pis: single step Pis ( PISS ^ ) and multistep Pis (PIms/,). Single step Pis refer to a single monthly forecast (e.g., h = 3) and are useful for determining whether a specific monthly observation is an outlier at a given significance level a. Multistep Pis encompass a 1-a prediction region that is a simultane- ous PI for all observations registered up to a certain /?-step and are useful in detecting systematic depar- tures from historical patterns. We calculated PISS h as y/> ± tdf,a/2'JPMSEh where PMSEh is the expected mean squared prediction error at step h and df=N-DS-d-r (Chatfield, 1993; Harvey, 1989). In the calculation of multistep Pis, we used a conservative approach based on a first-order Bonferroni inequality, whereby PIms h is given as yh ± tdf a/2h J PMSEh and joint prediction in- tervals of, at least, 1-a around the point forecasts are obtained (Chan et al., 2004). Results Data modeling Large autocorrelations were recorded for lags 1, 2, 11, 12, 23, and 24 with values 0.68, 0.32, 0.44, 0.46, 0.28 and 0.31, respectively (Fig. 2). The sharp decrease in autocorrelation values after lag 2 (0.07 at lag 3) indi- cated no evidence of a long-term trend; consequently, there was no need to include a first-lag difference term in the SARIMA model structure (d= 0). In contrast, large autocorrelation values were registered at annual lags (and its multiples) which indicated the need to include a 12-month difference term in the models (S=12, D = 1) (Fig. 2). The ACF and PACF plots of the differenced series provided further support for these conclusions (Fig. 2). Accordingly, a SARIMA(p,0,q)x(P,l,Q)12 was selected as the basic structure of the SARIMA candi- date set. Out of all models in the candidate set, a SARI- MA(0,0,5)x(l,l,0)12 was selected as the best model for the meagre data (-2 In (L) = - 26.32, n- 48, r=7, AICc=-9.52). This model had the following equation: (1+0.65, 10|R12) vli235= d+0.63, 19|£+0.56, 15)R2 + 0.51, 17|B3 + oT93,181R4+ 0^60, 21}B5)zt , with a noise variance estimate of &= 0.025 and 176 Fishery Bulletin 109(2) Yt v] y« v yt v12 vX2y, Lag Lag Lag Lag yt v1 y» v12y, v;v;2y. Lag Lag Lag Lag Figure 2 Sample autocorrelation function (ACF) and partial autocorrelation function (PACF) of the transformed meagre ( Argyro - somus regius) landings. ACF/PACF plots for log10-transformed mean-centered data ( yt , far left), lag-1 differenced series (Vjyf), lag-12 differenced series (Vi^), and lag-1 and lag-12 differenced series (viVi23V> far right) are displayed. Horizontal dashed lines represent the 95% confidence limits valid under the null hypothesis of white noise error structure. where yt - the mean-centered log-transformed meagre series (i.e., y?=log10x,-4.022) and the values in { ) are the standard errors of the estimates. Diagnostic checks indicated that the SARIMA model was stationary and invertible and did not have redun- dant parameters. The residuals were white noise (Ljung-Box Q = 3.35, P-value>0.05) and passed asymp- totic normality tests (Shapiro-Wilk W=0.97, P-value >0.05; Jarque-Bera LM= 4.91, P-value >0.05) indicating the model fitted the data and errors were normally distributed. The model explained 78.2% of the variance of the series. The final process equation selected for the meagre data was log10X,= 0.351og10A^12+0.651og10A?_24+Z,+0.63Z,_1 +0.56Z(._2+ 0.51Zt_3+0.93Z,_4+0.60Zf_5, where Zf ~ N (0, 0.025). Prista et al: Use of SARIMA models to assess data-poor fisheries 177 B Apr May Jun Jul Aug Sep Oct Nov Dec Jan Feb Mar Apr ( Fo ) ( 1 ) (2) (3) (4) (5) (6) (7) (8) ( 9 ) ( 10 ) ( 1 1 ) ( 12 ) Apr May Jun Ju! Aug Sep Oct Nov Dec Jan Feb Mar Apr ( Fo ) ( 1 ) (2) (3) (4) (5) (6) (7) (8) ( 9 ) ( 10 ) ( 1 1 ) ( 12 ) D "O c= 03 Apr May Jun Jul Aug Sep Oct Nov Dec Jan Feb Mar Apr ( Fo ) ( 1 ) (2) (3) (4) (5) (6) (7) (8) ( 9 ) ( 10 ) ( 1 1 ) ( 12 ) Apr May Jun Jul Aug Sep Oct Nov Dec Jan Feb Mar Apr ( Fo ) ( 1 ) (2) (3) (4) (5) (6) (7) (8) ( 9 ) ( 10 ) ( 1 1 ) ( 12 ) Figure 3 Forecasts and forecast prediction intervals (Pis) of meagre ( Argyrosomus regius) landings. The dashed vertical line is the forecast origin (“Fo”, April 2007). The gray circles and line represent the monthly forecasts. The black circles and line represent observed monthly landings. The dashed gray lines represent the upper and lower 75%, 95%, and 99% prediction intervals. (A and B) Single step prediction intervals (PIssA) of transformed centered landings and back-transformed landings, respectively in C and D. Multistep prediction intervals (PIms/,) of transformed centered landings and back-transformed landings, respectively. Model forecasts and performance The model forecasts presented two local maxima (May 2007 and September 2007) followed by a four-month period of low landings (December 2007 through March 2008) and an increase in the last month (April 2008) (Fig. 3, Table 3). This pattern in forecasts matched the one in observed landings and the only deviations were that the actual maxima took place one to two months later and the winter trough was sharper than that pre- dicted by the model (Fig. 3). RMSE during the hold-out period (0.234) was =1.5 times the RMSE of the fitting period. Eight of the 12 forecasts registered negative errors, but the low ME and PE indicated that under- estimation was minor in global terms. APE was large in August, September, December, and April, reflecting the delay in cessation of the 2007 fishing season and the hastening of the 2008 fishing season. Maximum APE coincided with the lowest landings (February), and the minimum APE with the first month forecasted (May) (Table 3). MAPE was 40.3%, reflecting the lagged seasonality and the low landings observed during the winter period. 178 Fishery Bulletin 109(2) Table 3 Forecasts of meagre ( Argyrosomus regius) landings (May 2007 to April 2008). Observed landings (xh), forecasted landings (x,), monthly forecast errors (eh), monthly absolute percent error (APE,,), mean error (ME), and mean absolute percent error (MAPE) are displayed for the two naive models (NM1 and NM2) and the seasonal autoregressive integrated moving-average model (SAR). Annual root mean-square error of the mean-centered transformed data (RMSE) and annual percent error (PE) for NM1, NM2 and SAR were 0.261 and 30.2%, 0.285 and 38.9%, and 0.234 and 15.4%, respectively. Month Step ( h ) Obs ( xh ) Forecasts (x,) Forecast errors ( e, ) APE, NM1 NM2 SAR NM1 NM2 SAR NM1 NM2 SAR May-07 1 37.1 29.9 21.0 36.4 -7.2 -16.1 -0.7 19.4 43.5 1.8 Jun-07 2 41.5 27.2 18.1 26.6 -14.3 -23.4 -14.9 34.4 56.5 35.8 Jul-07 3 23.0 17.9 14.7 26.1 -5.2 -8.3 +3.1 22.4 36.2 13.3 Aug- 07 4 15.7 25.9 18.4 25.8 +10.2 +2.8 +10.1 65.3 17.6 64.7 Sep-07 5 20.8 24.2 26.3 31.4 +3.4 +5.5 +10.6 16.3 26.2 51.1 Oct-07 6 30.6 15.3 21.9 23.0 -15.2 -8.7 -7.6 49.8 28.5 24.9 Nov-07 7 32.9 10.2 13.3 19.0 -22.7 -19.6 -13.9 69.0 59.5 42.2 Dec-07 8 16.1 6.8 6.8 6.0 -9.3 -9.2 -10.1 57.7 57.5 62.8 Jan-08 9 7.5 5.0 4.8 5.7 -2.5 -2.7 -1.8 32.8 35.7 24.5 Feb- 08 10 3.2 5.4 5.2 6.1 +2.1 +2.0 +2.9 66.6 61.9 90.7 Mar-08 11 8.0 5.8 4.1 6.5 -2.2 -3.9 -1.5 27.3 48.6 19.0 Apr- 08 12 34.1 15.2 10.8 16.3 -18.9 -23.4 -17.9 55.5 68.4 52.4 Mean 1:12 22.5 15.7 13.8 19.1 -6.8 -8.8 -3.5 43.1 45.0 40.3 Sum 1:12 270.5 188.8 165.4 228.9 -81.7 -105.1 -41.6 — — — As with SARIMA forecasts, naive model predictions also lagged observed values by one or two months. How- ever, the SARIMA forecasts registered the best perfor- mance in all accuracy measures, resulting in a 10% to 18% reduction in RMSE, 49% to 60% reduction in ME, 6% to 10% reduction in MAPE, and =15% reduction in PE (Table 3). The coefficient of persistence of the SARIMA model was also better (P=0.46) than the one registered by NM1 ( P= 0. 23 ) and NM2 (P=0.03). Monitoring of fisheries During the hold-out period, observed landings remained entirely within the 95% prediction intervals of the SARIMA forecasts (Fig. 3), indicating that the observed forecast errors were within the range of values expected from random variability. Consequently the time series for meagre landings may be described as having remained in-control during the forecasting period. The Pis were symmetrical in the log-transformed scale (Fig. 3, A and C), but asymmetrical in the original scale of the data (Fig. 3, B and D). This pattern was expected from predic- tions of log-transformed data and indicates that sudden increases in monthly landings (positive forecast errors) are considered “more acceptable” than sudden decreases (negative forecast errors). Individual forecast errors that could have signaled an alarm ranged from 4.3 to 23.0 t (negative errors) to 13.5-68.3 t (positive errors). In relative terms, alarms would have been triggered by a higher than 54-75% drop, or by a higher than 105-238% increase, in monthly landings (Table 4). Compared to monthly Pis, multistep Pis were wider as a result of the increasing number of comparisons performed (Table 4). Even so, it is noticeable that such widening took place mainly on their upper boundary, and only a 12% increase was observed on their lower boundary. Discussion Interpretation of the models Univariate SARIMA models based on landings do not have explanatory variables, but several studies have found the mathematical formulation in the models to correlate well with fish life history and fleet dynamics (Stergiou, 1990b; Stergiou et al., 1997; Lloret et al., 2000). In Europe, adult and juvenile meagre are thought to perform spring— summer migrations to major estuar- ies, remaining there until mid-summer (adults) and autumn (juveniles). These migrations are well known to local fishermen that actively target the meagre schools while they reside in estuarine grounds (Quero and Vayne, 1987; Prista et al.2). Such interactions between fish migrations and directed fishing effort are likely the cause of the strong seasonal component of the SARIMA model because target effort tends to intensify the natu- ral seasonal signal generated by fish migrating through a fishery (Lloret et al., 2000; Prista et al. 2008). In the case of central Portugal, such intensification is likely modulated at an interannual level by the expectations created for local fishermen by catches obtained in pre- Prista et al: Use of SARIMA models to assess data-poor fisheries 179 Table 4 Prediction intervals of meagre f Argyrosomus regius) landings (May 2007 to April 2008). Point forecasts ( x, ) and 95% boundar- ies of the single step (PISS,) and multistep (PImsA) prediction intervals are displayed. The prediction boundaries are given as absolute errors ( | | ) and absolute percent errors (APE,) in each monthly forecast step ( h ). In each cell, the left and right values represent the lower and upper boundaries, respectively. Month Step (h) K Kl APE, Kl APE, May-07 1 36.4 19.7-38.4 54-105 19.7-38.4 54-105 Jun-07 2 26.6 16.2-35.8 61-135 17.5-45.0 66-169 Jul-07 3 26.1 16.9-40.5 65-155 18.8-58.0 72-222 Aug- 07 4 25.8 17.3-43.7 67-169 19.6-68.8 76-266 Sep-07 5 31.4 23.0-68.3 73-217 25.9-120.0 82-382 Oct-07 6 23.0 17.3-54.7 75-238 19.5-103.6 85-451 Nov-07 7 19.0 14.3-45.2 75-238 16.2-89.7 85-472 Dec-07 8 6.0 4.5-14.2 75-238 5.1-29.4 86-491 Jan-08 9 5.7 4.3-13.5 75-238 4.9-28.8 86-509 Feb-08 10 6.1 4.6-14.6 75-238 5.3-32.2 87-525 Mar-08 11 6.5 4.9-15.5 75-238 5.7-35.1 87-539 Apr-08 12 16.3 12.3-38.7 75-238 14.2-89.9 87-553 ceding years (represented in the seasonal autoregressive term) and, at an intra-annual level, by random environ- mental and anthropogenic perturbations occurring on the fishery system (represented in the set of nonseasonal moving-average terms). Model fit and forecast performance The univariate SARIMA model presented a good fit to the short time series of meagre landings, explaining most of its variance and adequately modeling the sea- sonality and correlation structure of the data. Similar results were obtained in other studies of short and long time series: up to 68% (Lloret et al., 2000, series <64 months), 75% (Saila et al., 1980), 77% (Stergiou et al., 2003), 84-96% (Stergiou, 1989, 1991; Stergiou et al., 1997), and 93% (Pajuelo and Lorenzo, 1995). Taken together, these results indicate that SARIMA models should be adequate for data sets of monthly landings in general, and not just those with larger sample sizes. Bearing in mind that the minimum series length usu- ally stated for SARIMA model fitting is 50 (Pankratz, 1983; Chatfield, 1996b), such generalized applicability may make SARIMA models particularly useful for fish- eries with less reliable historical records or where only recently landings have been sampled. In addition to a good fit, the SARIMA model also pro- vided good short-term forecasts of meagre landings. The fact that all observed values were located within the predicted intervals of the model, and that naive fore- casts presented similarly lagged seasonality, indicates that the main forecast errors more likely resulted from natural variations in the timing of fish migrations and fishing seasons (Quero and Vayne, 1987; Prista et al.2) or from specifics of SARIMA forecasts and accuracy measures (namely, correlation and APE sensitivity to near-zero observations) (Hyndman and Koehler, 2006; Box et al., 2008) than from model misspecification. At the annual level, the 15% error achieved is comparable to results previously obtained in larger data sets and well within the 10-20% range considered acceptable for market-planning and fisheries management (e.g., Mendelssohn, 1981; Pajuelo and Lorenzo, 1995; Hanson et al., 2006). Additionally, SARIMA forecasts clearly outperformed naive forcasting in all accuracy metrics, underscoring the large benefits of using these models instead of simpler alternatives (Saila et al., 1980; Ster- giou, 1991; Stergiou et al., 1997). Considered together with the overall good forecasting performance reported by Lloret et al. (2000) in their shorter series, these re- sults build confidence that SARIMA models are useful for forecasting short time series of landings and thus can substantially contribute to the planning and man- agement of many data-poor fisheries. Use of SARIMA models to forecast landings of data-poor fisheries SARIMA models forecast future landings by directly handling the seasonality and autocorrelation structure of the data and assuming the continuity over time of past time series behavior (Box et al., 2008). These models are known to be well adapted to forecast highly seasonal and autocorrelated data (Stergiou et al., 1997; Georgakarakos et al., 2006). Additionally, some authors have reported better SARIMA forecasting performances in fisheries with lower interannual variability, namely those that target benthic and demersal long-lived spe- 180 Fishery Bulletin 109(2) cies (Lloret et al., 2000). The data for meagre are autocorrelated and present a relatively stable seasonal pattern. Also, the meagre is long-lived and a targeted fish in central Portugal (Prista et ah, 2009; Prista et al.2). Therefore, it is possible such features contributed to the good forecasts obtained from the SARIMA model. However, we note that the landings of many short-lived pelagic species and species with variable seasonal pat- terns have also been well forecasted with SARIMA models (Stergiou, 1990a; Stergiou et al., 1997; Geor- gakarakos et al., 2006; Tsitsika et al., 2007) and that the meagre landings also display substantial annual and monthly stochasticity Therefore, such general pat- terns should not be considered as strict limitations to SARIMA forecasting. More importantly, we note that SARIMA models can forecast well only if they have been adequately identified and estimated, and always under the assumption that the future is behaving like the past (Chatfield, 1993). Consequently, factors like data quality, presence of outliers, and model selection criteria are also very important for model performance. We discuss these next. The quality of the input data for SARIMA models is determined mainly by the temporal stability of the statistical properties of the fisheries process and the consistency of its sampling over time. Consequently, although accuracy is required for some model appli- cations (e.g., Zhou, 2003), data inaccuracies do not necessarily undermine SARIMA forecasts as long as factors such as fishing practices, regulatory measures, or data collection practices can be assumed to remain constant. When dealing with shorter series, a care- ful check whether these assumptions hold becomes particularly important because model identification and estimation are very dependent on the few obser- vations available (Hyndman and Kostenko, 2007) and statistical techniques used to incorporate the effects of process changes in the models (e.g., Fogarty and Miller, 2004) are difficult to implement. In the case of meagre, the use of a short and recent time series better supported the assumption that data collection procedures, fishing techniques, fishery regulations, unreported landings, discards, and law enforcement practices did not change over time. In contrast, it is probable that these assumptions were not met in some less successful applications of the model to longer time series (e.g., Park, 1998). Outliers are known to cause trouble in time series model identification, estimation, and forecasts — an ef- fect that is amplified in shorter time series (Chatfield, 1993; Trfvez and Nievas, 1998). The effects of outliers on forecasting performance are most disastrous when they occur near the forecasting origin because there they not only condition model structure and parameter estimates but are directly incorporated into the fore- casts (Chatfield, 1993). The meagre data set presented no apparent outliers and this likely contributed to the good fit and forecasting performance achieved. If outli- ers were present, specific modeling techniques could have been used to estimate their influence, smooth them, or incorporate them into the model (e.g., Chen and Liu, 1993; Lloret et al., 2000). We note, however, that any outlier during the hold-out period could still have changed our perception of model performance, even if it did not compromise the overall adequacy of the SARIMA model to forecast the landings. In time series analysis, adequate model specification is considered the most important driver of forecasting accuracy (Chatfield, 1996b). The difficulties of specify- ing an appropriate model increase for data sets with lower information content, such as those of highly vari- able short time series from more complex processes (Hyndman and Kostenko, 2007; Appendix 2). To date, fisheries applications of SARIMA models have essen- tially relied on Box-Jenkins (BJ) model selection pro- cedures to specify a model, and models with p <2 and q <2 have generally been selected (e.g., Mendelssohn, 1981; Pajuelo and Lorenzo, 1995; Lloret et al., 2000). Compared to these, the model for meagre seems over- parameterized, but we note that all of its parameters are statistically significant and that the low RMSE/i)rec to RMSE/(f ratio indicates an excellent correspondence between fit and forecasting performances (Chatfield, 1996b). In fact, although reduced model parameteriza- tion is considered beneficial to accuracy in forecast- ing, the most important aspect of time series analysis is not the number of parameters, but the degree to which the model approximates the statistical process underlying the data and whether or not it achieves the forecasting objectives (Chatfield, 1996b; Burnham and Anderson, 2002). In the case of meagre, had Box- Jenkins procedures been used, the selected models would be simpler and would still adequately fit the data: (l,0,0)x(l,l,0)12 or (0,0,l)x(0,l,l)12. However, they would have performed worse than our AICc-selected model in most performance metrics (RMSE: 0.245 and 0.302, APE: 1.7-92.7% and 20.6-72.4%, MAPE: 44.1% and 44.0%, PE: 13.7% and 31.7%, respectively). These results show the impact that different model selec- tion techniques may have on forecasting performance with SARIMA models and stress the importance of considering objective data-driven criteria like AICf for circumventing the subjectivities of model selection in smaller data sets (Hurvich and Tsai, 1989; Burnham and Anderson, 2002). Conclusions Use of SARIMA models in monitoring fisheries From a strictly forecasting perspective, SARIMA models have often been criticized for the excessive reliance on past time series behavior and their difficulty in predict- ing future structural changes (Georgakarakos et al., 2002; Koutroumanidis et al., 2006). Our results show that these drawbacks can become major advantages when SARIMA models are used for monitoring fisher- ies. At present, none of the European meagre fisheries is subjected to routine analytical assessment. By fitting Prista et al: Use of SARIMA models to assess data-poor fisheries 181 SARIMA models to already available landings data we were able to carry out a first baseline evaluation of one such fishery, using limited funds and minimal time. Our study provides a first example of how SARIMA models can be used to monitor data-poor fisheries. In the case of meagre, the data displayed no trend and the 95% SARIMA prediction intervals fully encom- passed all monthly landings, thus indicating a stable “in-control” fishery. Note that by stating this, at no point do we suggest that the meagre fishery is sustain- able long-term because landings do not necessarily reflect stock abundance and our study was limited in time. We suggest only that, since no motive for alarm exists in landings data, and because funds, personnel, and expertise are limited at the national level, atten- tion should be allocated to fisheries that, contrary to the meagre, display decreasing trends or out-of-control situations. Similar types of pragmatic reasoning are generally of great help to fisheries managers handling multiple data-poor fishery scenarios because they help them prioritize management actions for the subset of “problematic” resources in a statistically sound way (Scandol, 2003, 2005). Underlying the usefulness of SARIMA models in monitoring the meagre fishery and other data-poor fisheries is the use of prediction intervals as refer- ence points to signal alarming trends or sudden level shifts in the fisheries process (Caddy, 1999; Scandol, 2003; Mesnil and Petitgas, 2009). SARIMA Pis have been previously reported in the literature (Table 1), but their use in monitoring was not explored or formal- ized. These intervals are currently the focus of much statistical research on how to deal with their tendency toward “over-optimism,” i.e., the fact that nominal 95% prediction intervals generally contain less than 95% of future observations (Chatfield, 1993). Fortunately, from a fisheries conservation perspective such over-optimism does not constitute a major problem because narrower Pis will be more sensitive to changes in the fisheries process. Statistical process control (SPC) monitoring of uni- variate fisheries indicators has become the focus of in- creased research attention (Scandol, 2003, 2005; Mesnil and Petitgas, 2009; Petitgas, 2009; ICES1). The use of SARIMA Pis is similar to that of SPC control-charts, which makes them interesting candidates for the simul- taneous monitoring of multiple fisheries and fisheries indicators (Caddy, 1999; Scandol, 2005; Petitgas, 2009). For such cases, SARIMA Pis offer the advantage of be- ing model-based and do not require extensive historical reference data. They are also free from the assumption of statistical independence that frequently troubles the estimation of SPC detection limits (Mesnil and Petitgas, 2009). The simulation framework proposed by Scandol (2003, 2005) for SPC charts provides a means whereby SARIMA Pis can be calibrated toward specific detec- tion rates and management goals. Such calibration was beyond the objectives our study but constitutes an interesting research route for those in charge of more holistic fisheries management. SARIMA models in assessments of data-poor fisheries Formal stock assessment has traditionally been consid- ered as the starting point of any fisheries assessment (Mahon, 1997; Berkes et al., 2001). Such an approach is highly desirable but will not be implemented easily, nor quickly, in the many existing data-poor fisheries ( Vasconcellos and Cochrane, 2005). In fact, NRC (1998) estimated that 16% of U.S. stocks are not subjected to assessment; and the European Environmental Agency (EE A, 2005) estimated that, depending on the region considered, 20-90% of commercial stocks exploited in the Northeast Atlantic and Mediterranean are not routinely assessed. These figures are much worse in developing countries and when discard and bycatch species are included in the estimates (Vasconcellos and Cochrane, 2005). Addressing such situations requires increased focus on alternative stock indicators and assessment methods that can be used to monitor more fisheries by using available (or easily obtainable) data, funds, and human resources (e.g., Caddy, 1999; Scandol, 2005; Mesnil and Petitgas, 2009; OSPAR, 2010; ICES1). Uni- variate time series models fitted to landings data may be, for some time longer, the best possible approach to extend assessment and management coverage to many of these unassessed resources. SARIMA modeling and process-control schemes do not constitute alternatives to analytical stock assess- ment models. Rather, whenever possible, they should be seen as statistical tools to support expert judgment, funding allocation, and management decisions in the most data-limited and assessment-limited settings (Scandol, 2003; 2005). SARIMA modeling and model- based monitoring have a range of characteristics that make them worthy of future exploration in data-poor contexts. Among these are their appropriateness to nu- merous resources and variables, their strong statistical background and ecological plausibility, their good fore- casting performance and easy-to-estimate detection lim- its, and their applicability to both long and short time series. Furthermore, SARIMA models can also be used to model the nonspecific groupings that dominate many landings data sets, or can be upgraded if multivariate data become available (Stergiou et al., 1997; Vascon- cellos and Cochrane, 2005). Finally, the availability of SARIMA models in open-source software packages and their routine use in sectors other than fisheries (e.g., sales, economics, engineering) (Brockwell and Davis, 2002; Box et al., 2008) may be decisive advantages in budget-limited and expertise-limited countries. Acknowledgments Funding for this work was provided by a “Fundagao para a Ciencia e a Tecnologia” (FCT) grant BD/12550/2003 to N. Prista and by research project CORV (DGPA-Mare: FEDER— 22-05— 01-FDR— 00036). We thank Direcgao Geral das Pescas e Aquicultura (DGPA) for providing the meagre data set. We thank D. S. Stoffer and D. R. 182 Fishery Bulletin 109(2) Anderson for suggestions on the use of the “arima” func- tion and AICc model selection, respectively. We further thank M. F. Lane, J. L. Costa, J. J. Schaffler, and J. R. Ashford for commenting on earlier drafts of this manu- script. We thank the three anonymous reviewers for their constructive comments on this manuscript. Literature cited Akaike, H. 1974. A new look at the statistical model identifica- tion. IEEE T. Autom. Contr. 19:716-723. Berkes, F., R. Mahon, P. McConney, R. Pollnac, and R. Pomeroy. 2001. Managing small-scale fisheries: alternative direc- tions and methods, 320 p. IDRC Books, Ottawa, Canada. Box, G., and D. Cox. 1964. An analysis of transformations. J. R. Stat. Soc. Ser. B 26:211-243. Box, G. E. P., G. M. Jenkins, and G. C. Reinsel. 2008. Time series: forecasting and control, 4th ed., 784 p. John Wiley & Sons, Hoboken, NJ. Brockwell, P., and R. Davis. 2002. Introduction to time series and forecasting, 2nd ed., 469 p. Springer, New York. Burnham, K. P., and D. R. Anderson. 2002. Model selection and multi-model inference: a practical information-theoretic approach, 2nd ed., 488 p. Springer, New York. Caddy, J. F. 1999. Deciding on precautionary management measures for a stock based on a suite of limit reference points (LRPs) as a basis for a multi-LRP harvest law. North- west Atl. Fish. Org. Sci. Coune. Stud. 32:55-68. Campbell, A., D. J. Noakes, and R. W. Elner. 1991. Temperature and lobster, Homarus americanus, yield relationships. Can. J. Fish. Aquat. Sci. 48:2073- 2082. Chan, W„ S. Cheung, and K. Wu. 2004. Multiple forecasts with autoregressive time series models: case studies. Math. Comput. Simul. 64:421-430. Chatfield, C. 1993. Calculating interval forecasts. J. Bus. Econ. Stat. 11:121-135. 1996a. The analysis of time series: an introduction, 5th ed., 283 p. Chapman & Hall/CRC, Boca Raton, FL. 1996b. Model uncertainty and forecast accuracy. J. Forecast. 15:495-508. Chen, C., and L. -M. Liu. 1993. Joint estimation of model parameters and outlier effects in time series. J. Am. Stat. Assoc. 88:284-297. Czerwinski, I. A., J. C. Gutierrez-Estrada, and J. A. Hernando- Casal. 2007. Short-term forecasting of halibut C-PUE: linear and non-linear univariate approaches. Fish. Res. 86:120-128. De Gooijer, J. G., B. Abraham, A. Gould, and L. Robinson. 1985. Methods for determining the order of an autore- gressive-moving average process: a survey. Int. Stat. Rev. 53:301-329. De Gooijer, J. G., and R. J. Hyndman. 2006. 25 years of time series forecasting. Int. J. Fore- cast. 22:442-473. EEA (European Environmental Agency). 2005. The European environment — state and outlook 2005, 570 p. European Environmental Agency, Copenhagen, Denmark. Fogarty, M. J. 1988. Time series models of the Maine lobster fishery: the effect of temperature. Can. J. Fish. Aquat. Sci. 45:1145-1153. Fogarty, M. J., and T. J. Miller. 2004. Impact of a change in reporting systems in the Maryland blue crab fishery. Fish. Res. 68:37-43. Georgakarakos, S., J. Haralabous, V. Valavanis, C. Arvanitidis, D. Koutsoubas, and A. Kapantagakis. 2002. Loliginid and ommastrephid stock prediction in Greek waters using time series analysis tech- niques. Bull. Mar. Sci. 71:269-287. Georgakarakos, S., D. Koutsoubas, and V. Valavanis. 2006. Time series analysis and forecasting techniques applied on loliginid and ommastrephid landings in Greek waters. Fish. Res. 78:55-71. Graves, S. 2008. Companion to Tsay (2005) analysis of financial time series. R package vers. 0.10-12. URL: http:// CRAN.R-project.org., accessed October 2008. Hanson, R J., D. S. Vaughan, and S. Narayan. 2006. Forecasting annual harvests of Atlantic and Gulf menhaden. N. Am. J. Fish. Manag. 26:753-764. Hare, S. R., and R. C. Francis. 1994. Climate change and salmon production in the Northeast Pacific Ocean. In Climate change and north- ern fish populations (R. J. Beamish, ed.). Can. Spec. Publ. Fish. Aquat. Sci. 121:357-372. Harvey, A. C. 1989. Forecasting, structural time series models and the Kalman filter, 572 p. Cambridge Univ. Press, Cambridge. Hurvich, C. M., and C. -L. Tsai. 1989. Regression and time series model selection in small samples. Biometrika 76:297-307. Hyndman, R. J., and A. B. Koehler. 2006. Another look at measures of forecast accuracy. Int. J. Forecast. 22:679-688. Hyndman, R. J., and A. V. Kostenko. 2007. Minimum sample size requirements for seasonal forecasting models. Foresight 6:12-15. Jarque, C., and A. Bera. 1987. Efficient tests for normality, homoscedascity and serial independence of regression residuals. Econ. Lett. 6:255-259. Jeffries, P, A. Keller, and S. Hale. 1989. Predicting winter flounder ( Pseudopleuronectes americanus) catches by time series analysis. Can. J. Fish. Aquat. Sci. 46:650—659. Kitanidis, P. K., and R. L. Bras. 1980. Real-time forecasting with a conceptual hydrologic model 2: applications and results. Water Resour. Res. 16:1034-1044. Koutroumanidis, T., L. Iliadis, and G. K. Sylaios. 2006. Time-series modeling of fishery landings using ARIMA models and fuzzy expected intervals soft- ware. Environ. Model. Softw. 21:1711-1721. Ljung, G., and G. Box. 1978. On a measure of lack of fit in time series models. Biometrika 65:297-303. Lloret, J., J. Lleonart, and I. Sole. 2000. Time series modeling of landings in Northwest Mediterranean Sea. ICES J. Mar. Sci. 57:171-184. Prista et al: Use of SARIMA models to assess data-poor fisheries 183 Mahon, R. 1997. Does fisheries science serve the needs of managers of small stocks in developing countries? Can. J. Fish. Aquat. Sci. 54:2207-2213. Mendelssohn, R. 1981. Using Box-Jenkins models to forecast fishery dynam- ics: identification, estimation, and checking. Fish. Bull. 78:887-896. Mesnil, B., and P. Petitgas. 2009. Detection of changes in time-series of indicators using CUSUM control charts. Aquat. Living Resour. 22:187-192. Molinet, R., M. T. Badaracco, and J. J. Salaya. 1991. Time series analysis for the shrimp and snapper fisheries in Golfo Triste, Venezuela. Sci. Mar. 55:427- 437. Morales-Nin, B., A. Grau, S. Perez-Mayol, and M. M. Gil. 2010. Marking otoliths, age validation and growth of Argyrosomus regi us juveniles (Sciaenidae). Fish. Res, 106:76-80. Noakes, D. J., D. W. Welch, M. Henderson, and E. Mansfield. 1990. A comparison of preseason forecasting methods for returns of two British Columbia sockeye salmon stocks. N. Am. J. Fish. Manag. 10:46-57. NRC (National Research Council). 1998. Improving fish stock assessments, 177 p. National Academy Press, Washington, D.C. OSPAR (Oslo and Paris Commission). 2010. Quality status report 2010, 176 p. OSPAR Com- mission, London, UK. Pajuelo, J. G., and J. M. Lorenzo. 1995. Analysis and forecasting of the demersal fishery of the Canary Islands using an ARIMA model. Sci. Mar. 59:155-164. Pankratz, A. 1983. Forecasting with univariate Box-Jenkins models: concepts and cases, 576 p. John Wiley & Sons, Hobo- ken, NJ. Park, H. -H. 1998. Analysis and prediction of walleye pollock ( Ther - agra chalcogramma) landings in Korea by time series analysis. Fish. Res. 38:1-7. Petitgas, P. 2009. The CUSUM out-of-control table to monitor changes in fish stock status using many indicators. Aquat. Living Resour. 22:201-206. Pierce, G. J., and P. R. Boyle. 2003. Empirical modelling of interannual trends in abun- dance of squid (Loligo forbesi in Scottish waters. Fish. Res. 59:305-326. Prista, N., J. L. Costa, M. J. Costa, and C. M. Jones. 2009. Age determination in meagre Argyrosomus regius. Relat. Cient. Tec. Inst. Invest. Pescas Mar: Serie Digital 49:1-54. URL: http://ipimar-inrb.ipimar.pt/pdf/Reln49. pdf, accessed September 2010. Quemener, L. 2002. Le maigre commun ( Argyrosomus regius ) — biologie, peche, marche et potentiel aquacole, 32 p. IFREMER, Plouzane, France. [In French.] Quero, J.-C., and J. -J. Vayne. 1987. Le maigre, Argyrosomus regius (Asso, 1801) (Pisces, Perciformes, Sciaenidae) du Golfe de Gascogne et des eaux plus septentrionales. Rev. Trav. Inst. Peches Marit. 49:35-66 [In French.] R Development Core Team. 2007. R: A language and environment for statistical com- puting. R Foundation for Statistical Computing. L1RL: http://www.R-project.org, accessed August 2007. Rothschild, B. J., S. G. Smith, and H. Li. 1996. The application of time series analysis to fish- eries population assessment and modeling. In Stock assessment: quantitative methods and applications for small-scale fisheries (V. F. Gallucci, S. B. Saila, D. J. Gustafson, and B. J. Rothschild, eds), p. 354-402. CRC Press, Boca Raton, FL. Saila, S. B., M. Wigbout, and R. J. Lermit. 1980. Comparison of some time series models for the analysis of fisheries data. ICES J. Mar. Sci. 39:44-52. Scandol, J. 2003. Use of cumulative sum (CUSUM) control charts of landed catch in the management of fisheries. Fish. Res. 64:19-36. 2005. Use of quality control methods to monitor the status of fish stocks. In Fisheries assessment and management in data-limited situations (G. H. Kruse, V. F. Gallucci, D. E. Hay, R. I. Perry, R. M. Peterman, T. C. Shirley, P. D. Spencer, B. Wilson, and D. Woodby, eds.), p. 213-233. Alaska Sea Grant College Program, Univ. Alaska, Fairbanks, AK. Shapiro, S., M. Wilk, and H. Chen. 1968. A comparative study of various tests for normality. J. Am. Stat. Assoc. 63:1343-1372. Shumway, R. H., and D. S. Stoffer. 2006. Time series analysis and its applications: with R examples, 2nd ed., 575 p. Springer, New York. Stergiou, K. I. 1989. Modelling and forecasting the fishery for pilchard Sardina pilchardus in Greek waters using ARIMA time- series models. ICES J. Mar. Sci. 46:16-23. 1990a. A seasonal autoregressive model of the anchovy Engraulis encrasicolus fishery in the eastern Mediter- ranean. Fish. Bull. 88:411-414. 1990b. Prediction of the Mullidae fishery in the east- ern Mediterranean 24 months in advance. Fish. Res. 9:67-74. 1991. Short-term fisheries forecasting: a comparison of smoothing, ARIMA and regression techniques. J. Appl. Ichthyol. 7:193-204. Stergiou, K. I., and E. D. Christou. 1996. Modelling and forecasting annual fisheries catches: comparison of regression, univariate and multivariate time series methods. Fish. Res. 25:105-138. Stergiou, K. I., E. D. Christou, and G. Petrakis. 1997. Modelling and forecasting monthly fisheries catches: comparison of regression, univariate and mul- tivariate time series methods. Fish. Res. 29:55-95. Stergiou, K. I., G. Tserpes, and P. Peristeraki. 2003. Modelling and forecasting monthly swordfish catches in the Eastern Mediterranean. Sci. Mar. 67:283-290. Trapletti, A., and K. Hornik. 2007. Tseries: time series analysis and computational finance. R package vers. 0.10-12. URL: http:// CRAN.R-project.org, accessed December 2007. Trivez, F. J., and J. Nievas. 1998. Analyzing the effects of level shifts and temporary changes on the identification of ARIMA models. J. Appl. Stat. 25:409-424. Tsai, C. -F., and A. -L. Chai. 1992. Short-term forecasting of the striped bass Morone saxatilis commercial harvest in the Maryland portion of Chesapeake Bay. Fish. Res. 15:67-82. 184 Fishery Bulletin 109(2) Tsitsika, E. V., C. D. Maravelias, and J. Haralabous. 2007. Modeling and forecasting pelagic fish production using univariate and multivariate ARIMA models. Fish. Sci. 73:979-988. Vasconcellos, M., and K. Cochrane. 2005. Overview of world status of data-limited fisheries: inferences from landing statistics. In Fisheries assess- ment and management in data-limited situations (G. H. Kruse, V. F. Gallucci, D. E. Hay, R. I. Perry, R. M. Peterman, T. C. Shirley, P. D. Spencer, B. Wilson, and D. Woodby, eds.), p. 1-20. Alaska Sea Grant College Program, Univ. Alaska, Fairbanks, AK. Worm, K., R. Hilborn, J. K. Baum, T. A. Branch, J. S. Collie, C. Costello, M. J. Fogarty, E. A. Fulton, J. A. Hutchings, S. Jennings, O. P. Jensen, H. K. Lotze, P. M. Mace, T. R. McClanahan, C. Minto, S. R. Palumbi, A. M. Parma, D. Ricard, A. A. Rosenberg, R. Watson, and D. Zeller. 2009. Rebuilding global fisheries. Science 325:578-585. Zhou, S. 2003. Application of artificial neural networks for fore- casting salmon escapement. N. Am. J. Fish. Manag. 23:48-59. Appendix 1 ARIMA and SARIMA models An extensive review of ARIMA and SARIMA models can be found in, e.g., Box et al. (2008) and Brockwell and Davis (2002). A mean-centered time series xt can be modeled as an ARIMA(p,d,q), where p, d, q are non- negative integers, if it can be adequately fitted with the process equation (\>(B)(\-B)dXt= Q(B)Zt , where for a time interval T, (Xt) teT is a sequence of random variables, B is a backshift differencing opera- tor BhXt=Xt_h (h nonnegative integer), (l-B)dXt- ^dxXt is stationary, (j)(B) and 0(B) are linear filters defined as (B)= 1- ^ B - 1) required for Xt to become stationary. This differencing involves the loss of d observations in the series. The SARIMA (p,d,q)x(P,D,Q)s models, where P, D, Q , and S are nonnegative integers, extend the modeling ca- pabilities of ARIMA(p,d,q) models to seasonal processes. The SARIMA process equation is given by (B)&(Bs)(l-B)da-Bs)DXt=O(B)0(Bs)Zt , where Xt, Zt, 100), ACF/PACF estimates have lower variability and are more likely to approximate the theoretical ACF/PACF estimates of the underly- ing process. In such cases, less subjectivity exists in identification of the model. However, when sample size is small, the interpretation of ACF/PACF patterns be- comes increasingly confounded by the large variance of the sample estimates, particularly at larger lags (>n/4) (Box et al., 2008). This variability substantially increases the subjectivity of the model identification Prista et al: Use of SARIMA models to assess data-poor fisheries 185 stage of the BJ method and is the main issue to be dealt with when analyzing shorter time series. AIC approach To circumvent the subjectivity of the identification of the model with the BJ method and to aid in the determination of the final orders of the ARMA processes a wide variety of model selection criteria have been developed (De Gooijer et ah, 1985). The most frequently used are the Akaike information criteria (AIC) (Akaike, 1974) and the small-sample, bias-corrected equivalent, AICc (Hurvich and Tsai, 1989). Contrary to the Box- Jenkins method, AIC/AICc selection of a model involves the a priori estimation by maximum likelihood methods of a set of model struc- tures (here termed the candidate set). This estimation is followed by the determination of the AIC/AICc values for each individual model. The model with minimum AIC/AICc is then selected as the model that is closest to the statistical process “generating” the data. In SARIMA models, AIC is calculated as AIC = -21n(L)+2r , where ln(L) is the log-likelihood of the model, r=p+ q+P+Q+1, and the AICc, is given by AICc=-21n(L)+2r-t-2r(r+l)/(tt-r-l) , where n=N-DS-d is the number observations used to fit the model. AIC/AICc constitute objective methods to achieve model parsimony through a trade-off between the variance explained by the model and penalty terms caused by excessive model parameters. Both of them are well founded in the principles of information and likeli- hood theory and have been applied extensively in time series, fisheries, and ecological literature (e.g., Brock- well and Davis, 2002; Burnham and Anderson, 2002; Hanson et al., 2006). Burnham and Anderson (2002) suggest AICc is used when n/r <40, which prompts the consideration of this small-sample, bias-corrected ver- sion of AIC in studies of short time series. 186 Season- and depth-dependent variability of a demersal fish assemblage in a large fjord estuary (Puget Sound, Washington) Jonathan C. P. Reum (contact author) Timothy E. Essington Email address for contact author: reumj@u.washington.edu School of Aquatic and Fishery Sciences Box 355020 University of Washington Seattle, Washington 98195 Abstract — Fjord estuaries are com- mon along the northeast Pacific coast- line, but little information is available on fish assemblage structure and its spatiotemporal variability. Here, we examined changes in diversity met- rics, species biomasses, and biomass spectra (the distribution of biomass across body size classes) over three seasons (fall, winter, summer) and at multiple depths (20 to 160 m) in Puget Sound, Washington, a deep and highly urbanized fjord estuary on the U.S. west coast. Our results indicate that this fish assemblage is dominated by cartilaginous species (spotted ratfish I Hydrolagus colliei] and spiny dog- fish [Squalus acanthias ]) and there- fore differs fundamentally from fish assemblages found in shallower estu- aries in the northeast Pacific. Diver- sity was greatest in shallow waters (<40 m), where the assemblage was composed primarily of flatfishes and sculpins, and lowest in deep waters (>80 m) that are more common in Puget Sound and that are dominated by spotted ratfish and seasonally (fall and summer) by spiny dogfish. Strong depth-dependent variation in the demersal fish assemblage may be a general feature of deep fjord estuar- ies and indicates pronounced spatial variability in the food web. Future comparisons with less impacted fjords may offer insight into whether carti- laginous species naturally dominate these systems or only do so under conditions related to human-caused ecosystem degradation. Information on species distributions is critical for marine spatial planning and for modeling energy flows in coastal food webs. The data presented here will aid these endeavors and highlight areas for future research in this important yet understudied system. Manuscript submitted 6 July 2010. Manuscript accepted 3 February 2011. Fish. Bull. 109:186-197 (2011). 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. Estuaries are highly productive habi- tats that support a diversity of spe- cies, but have suffered because of growing human demands (Kennish, 2002; Lotze et al., 2006). Recognition of declining fish, marine mammal, and seabird populations and the need to consider the multiplicity of caus- ative agents for these declines and their ecological consequences have prompted an interest in adopting eco- system-based approaches for manage- ment, whereby knowledge of ecological interactions, such as trophodynamic control and competition is used to inform policy decisions (Pikitch et al., 2004). Our ability to implement more holistic management approaches, how- ever, can be limited by a lack of basic information on the distribution and abundance of species that a system comprises and how these vary over time and space. This information is particularly lacking for fjord estuar- ies that are common features along the northeast Pacific coastline. Fjord estuaries differ from other estuar- ies by possessing a deep inner basin that is separated from continental shelf waters by a shallow sill near the mouth of the estuary. Although some of these ecosystems are remote and show little sign of degradation, commercial and recreational fishing, aquaculture, shoreline development, pollution, and logging are degrading a growing number of them. Puget Sound, WA, is the south- ernmost fjord estuary in the north- east Pacific and supports major ur- ban centers with a combined human population of 4 million (PSAT1). Over the past 150 years Puget Sound has been commercially fished and sub- ject to increasing rates of habitat loss, eutrophication, pollution, and, more recently, acidification. Presently, commercial fishing for groundfish is not permitted, but some species (e.g., Sebastes spp., lingcod [Ophiodon elon- gates]) are targeted by recreational fishermen. Although an ecosystem- based approach is clearly relevant for Puget Sound, there is a paucity of published information on how the demersal fishes of Puget Sound use different habitats, and thus a need for studies on assemblage structure. Identifying major patterns of as- semblage variability along different habitat gradients has practical impli- cations not only for modeling energy flows, but for devising monitoring schemes that can adequately quantify interannual changes in population abundance (Greenstreet et al., 1997; Thompson and Mapstone, 2002). Al- though project and agency reports provide some descriptive analyses on Puget Sound fish communities, peer- reviewed literature on demersal fish distributions from other deep fjords in the northeastern Pacific are rare. 1 PSAT (Puget Sound Action Team). 2007. State of the Sound. 2007. Publication no. PSAT 07-01, 96 p. Office of the Gov- ernor, Olympia, State of Washington. [Available at: http://www.psp.wa.gov/ documents.php, accessed February 2011.) Reum and Essington: Season- and depth-dependent variability of a demersal fish assemblage in a fjord estuary 187 Here we evaluated depth- and season-related changes in the demersal fish assemblage struc- ture over a 10-month period in the Central Basin — the largest and deepest of four sub- basins within Puget Sound. We anticipated strong spatial gradients in community com- position on the basis of typical patterns of de- mersal fish assemblages in other ecosystems (e.g., Mueter and Norcross, 1999) and also hypothesized that species might exhibit sea- sonal patterns of habitat use that would be reflected in community structure. First, we evaluated differences in assemblage diversity metrics across depths and seasons. Second, we performed taxon-based multivariate analyses on the fish assemblage and explicitly tested whether assemblage structure was related to season and depth. Finally, we performed a size-based analysis, examining variability in the distribution of biomass across body size classes (biomass spectra) to identify whether the prevalence of small-body or large-body in- dividuals also changed with depth and season. The value of examining biomass spectra is grounded in the observation that trophic level generally increases with body size in aquatic systems (Kerr, 1974; Jennings et al., 2001), and the relative importance of different body size classes to the overall flow of energy in the food web can be revealed by biomass spectra (Haedrich and Merrett, 1992). The patterns revealed by each of these approaches were sub- sequently compared. Materials and methods 48°0'0"N 47°00'0"N Straight of Georgia Possession Sound North CTD rl, Olympic Peninsula Kitsap Peninsula South CTO _Tacoma Tacoma Narrows Southern Basin Puget Sound WA 1 23°0'0"W 1 22°00'0"W Data collection The demersal fish assemblage was sampled by bottom trawl during 18-22 October 2004, 10-14 March, and 7-11 July 2005 along the eastern coastline of the larg- est subbasin within Puget Sound, the Central Basin (Fig. 1). The Central Basin is separated from the Strait of Georgia and Whidbey Basin by a sill (-60 m depth) at Admiralty Inlet and by Possession Sound to the north, respectively, and from Southern Basin by a sill at Tacoma Narrows (Fig. 1). We identified six sampling stations located in low-relief regions and amenable to bottom trawls spaced 5-10 km apart (high relief, hard bottom habitats are not common in this region of Puget Sound). At each station, we sampled four sites at depths of 20, 40, 80, and 160 m. In October we sampled stations 1-4, sampling all four depths. To increase the spatial coverage of the survey in March, we visited stations 1, and 3—6, but only sampled at 40 and 160 m. Lastly, in July we visited stations 1, 3, 4, 5, and 6, and sampled all four depths. In this manner we were able to increase the spatial coverage of the survey while maintaining overlap with the previous season. We modified our survey when Figure 1 Demersal fish assemblages were sampled with bottom trawls at six stations (indicated by the circled numbers) in the Central Basin of Puget Sound, WA. At each station, depths at 20, 40, 80, and 160 m were sampled. Sampling occurred in October 2004 and March and July 2005. Isobaths at 40 and 160 m are indicated. Sampling sites where conductivity-tempera- ture-density instruments (CTDs) were deployed by the King County Puget Sound Marine Monitoring Program to determine temperature and salinity are denoted by square symbols. necessary because of limitations in boat and crew time and because of the absence of prior information on which we could base our survey. Sampling was performed during daylight hours with a 400-mesh Eastern otter bottom trawl lined with 3.2-cm mesh in the codend. The net had a head rope and foot rope of 21.4 m and 28.7 m, respectively, and the gear was towed across the seafloor for 400 — 500 m at 2.5 knots. Catch was sorted to species, weighed, enumerat- ed, and subsampled for length composition of fish. Catch was standardized to biomass density (g/m2) by using measurements of the area swept by the net. The area 188 Fishery Bulletin 109(2) swept was calculated by multiplying the tow distance by the net opening width, the latter calculated from an empirical relationship between depth and net width. To compare differences in salinity and temperature among depths, months, and latitude we obtained data from the King County Puget Sound Marine Monitor- ing Program which monthly monitors water quality at two stations in the northern and southern regions of the survey area (Fig. 1). Data were collected with a conductivity-temperature-density instrument (CTD) consisting of an SBE 3 temperature sensor, SBE 4 con- ductivity sensor, and SBE 29 pressure sensor (SeaBird Electroics Inc., Bellevue, WA) and were binned at 0.5- m intervals. CTD sampling occurred within 10 days of trawl sampling. Statistical analysis More than 200 species of fish have been documented in Puget Sound, but many of these are rare or sparsely dis- tributed such that an intensive sampling effort would be required to sufficiently describe the distribution patterns of all species. Instead, we focused our research on the commonly occurring species that accounted for the bulk of the demersal fish biomass and therefore represent the most significant fish in the food web. To calculate diver- sity metrics for comparisons, rare species that occurred in fewer than 10% of the sampled trawls were removed from the data set; the exclusion of rare species permitted a coarse-scale evaluation of differences in the common components of the assemblage. Differences in species richness (IV) and diversity (Shannon-Wiener diversity index, H'; Krebs, 1989) across depths and among months were examined by using two-way analysis of variance (ANOVA). Standard two-way ANOVA requires that treatment levels be fully replicated across both main fac- tors (in this case, depth and month). Because we lacked samples from depths of 20 m and 40 m in March, we performed two sets of tests. In the first set we included samples from all four depths, but only from October and July. In the second, we included samples from all three months, but only from 40 and 160 m. Initial examination of the data indicated normal or near-normal distribu- tions, therefore data transformations were not called for because ANOVA is robust to minor departures from normality (Zar, 1984). In instances where either of the main factors was significant, Tukey’s honestly signifi- cant difference (HSD) tests were used to identify which depth and month levels differed. In the above analysis, both N and H' are simple and widely used measures of diversity that describe the number or relative biomass of species at each sample but ignore similarity in species composition among samples. In contrast, canonical correspondence analy- ses (CCA) are used to explicitly evaluate multivariate patterns of species biomass among sample sites. Es- sentially, CCA is a multivariate extension of multiple regression where species and sites are simultaneously ordinated in a manner that maximizes the variance related to a set of explanatory (constraining) variables (ter Braak, 1986). As with multiple regression, the inertia, or variance explained, by a given model can be determined and the significance of the explanatory variable tested (Legendre and Legendre, 1998). In in- stances where two or more sets of explanatory variables are of interest (in this case, month and depth), partial CCA can be employed to isolate the effect of each vari- able (Legendre and Legendre, 1998). The technique is analogous to partial regression, where the response variable (species biomass) is first constrained by one of the explanatory variables (either month or depth, expressed as factors with dummy variable coding). The resulting residuals are then constrained by the second explanatory variable. Effectively, the first explanatory variable is treated as a confounding variable and its effect is “cleansed” from the data set. The assemblage pattern related solely to the second explanatory vari- able can then be isolated and explored (Legendre and Legendre, 1998). An advantage of CCA over standard univariate tests of species biomass (e.g., ANOVA) is that the method simultaneously depicts the strength and direction of species responses to predictor variables by the position and spread of species in ordinate space. The approach therefore offers insights into species as- sociations that are not readily obtained by univariate methods (Legendre and Legendre, 1998). We applied partial CCA to the data set alternating depth and month as the confounding and explicit ex- planatory variables, respectively. We recognized that the habitat of most fishes changes with body size (Wer- ner and Gilliam, 1984) and therefore we divided species with abundant, small size classes (individuals less than 30% of maximum recorded total length [TL] that oc- curred in at least 10% of the sites sampled) into small and large size classes (greater than 30% of maximum recorded TL were categorized as large) and we treated them as distinct species in the analysis. An exception was made for spiny dogfish, which possessed a bimodal size distribution that separated at approximately 500 mm or 47% of the maximum recorded TL. Owing to the lack of samples from 20- and 80-m depths in March, we performed partial CCA (one each for depth and month), using 1) samples collected from all three months, but from 40 and 160 m only; and 2) for all four depths from October and July only for a total of four partial CCA tests. We identify the data included in each univariate and multivariate analysis by labeling tests as “all-months” or “October-t- July,” respectively. The analysis was split to avoid ambiguity that may have arisen from performing partial CCA on data lacking full treatment replication across factor levels (Anderson and Gribble, 1998). To increase the robustness of each CCA only species that occurred in at least 20% of the sites sampled were included. The variance explained by a global CCA model (month+depth) was used in conjunction with results from the partial CCA to identify variance components that were uniquely and jointly explained by each predic- tor (Borcard et al., 1992; Anderson and Gribble, 1998). The significance of each predictor in the global and Reum and Essington: Season- and depth-dependent variability of a demersal fish assemblage in a fjord estuary 189 partial CCA models was tested by comparing a pseudo F statistic to a null distribution generated by permut- ing the species-site matrix 5000 times (Legendre and Legendre, 1998). The analysis of biomass spectra paralleled our analy- ses of the biomass of assemblage species. Length-fre- quency information for each species was used to divide species biomass into log2 body size classes. The total sum of biomass (all species) in each size class was then normalized by dividing the biomass by the antilog body size interval of each respective size class, as is com- monly done in analyses of biomass spectra (Kerr, 1974; Sprules et al., 1983). In subsequent partial CCA, the size class biomassxsite matrix was treated as the re- sponse variable. To determine whether the analyses should account for spatial autocorrelation, we tested residuals result- ing from the global species biomass and biomass spec- tra CCA models for spatial dependence, using a multi- scale direct ordination technique. The method entails performing a constrained ordination on the global mod- el CCA residuals with an explanatory matrix that is coded for geographic distance (Wagner, 2004). We used a grain-size equivalent to 0.1° latitude that resuled in four distance classes. Results were nonsignificant for species biomass and biomass spectra models, indicating that spatial structure at the assemblage level was not detectable. We therefore disregarded spatial dimensions and pooled samples by depth and season in subsequent analyses. During the exploratory stage of our analysis, univariate and multivariate tests that were performed with biomass densities resulted in conclusions similar to those obtained with numerical densities. We chose to limit our analyses to biomass densities to avoid re- dundancy and because the importance of a species to energy flow in a food web is more readily (although not perfectly) approximated by information on its biomass. We considered all statistical tests significant at the P=0.05 level. Results We captured 23,100 individual fish in our survey that represented 62 species from 23 different families. Of these, 32 species occurred in more than 10% of the trawls. In general, small size classes (<30% of maxi- mum recorded length) for individual species were rare and only English sole ( Parophrys vetulus), and spiny dogfish ( Squalus acanthias ) were abundant enough for inclusion in the analysis as separate size classes. Water conditions were relatively homogenous throughout the area surveyed. Temperature and salinity values dif- fered by less than 1°C and 0.2, respectively, between the northern and southern CTD stations for each depth and month combination (Table 1). Differences in tem- perature and salinity among depths were small in Octo- ber and March but were more evident in July; waters 20 m deep were warmer and fresher by 1.3° and 0.8°C, respectively (Table 1). Table 1 Temperature (°C) and salinity measurements obtained with conductivity-temperature-density (CTD) casts from the northern (N) and southern (S) regions of the area sur- veyed in Central Basin, Puget Sound. Data were obtained within ten days of trawl sampling. Although stations at 20 and 80 m were not sampled in March, water properties are provided to aid comparisons among sampling months. Temperature Salinity Month Depth (m) N S N S October 20 12.1 12.1 30.3 30.5 40 11.9 12.0 30.5 30.6 80 11.7 11.8 30.6 30.7 160 11.4 11.5 30.7 30.8 March 20 8.7 8.5 29.6 29.5 40 8.7 8.5 29.7 29.6 80 8.7 8.4 29.9 29.9 160 OO bo 8.3 30.1 29.8 July 20 12.8 13.0 29.6 29.6 40 12.1 12.1 29.8 29.8 80 11.5 11.4 30.0 30.0 160 11.5 11.4 30.4 30.4 Overall, spotted ratfish ( Hydrolagus colliei ), spiny dogfish, and flatfish were the dominant taxonomic groups in the survey. Biomass patterns observed at each depth and month combination are depicted in Fig- ure 2. Shallow waters (20 and 40 m) were dominated by flatfishes, which composed between 64% and 83% of the fish assemblage biomass in all three months. In deep water, assemblage biomass was nearly double that found in shallow waters (80 and 160 m; Fig. 2). In to- tal, spotted ratfish composed approximately 80% of the fish assemblage at 160 m in all three months (Fig. 2). Spiny dogfish were found primarily at depths of 80 and 160 in October, were nearly absent from the survey in March (two individuals were captured), and present at all depths in July, with the highest biomasses occurring at depths of 80 and 160 m (Fig. 2). Diversity metrics Variation in species richness (N) was observed across depths in both October+July (ANOVA, F|3 32j = 3. 9, P=0.01) and all-months tests (ANOVA, F(1 26] = 103.7, P<0.001) where N at 40 m was higher than at 160 m (Fig. 3). Post hoc analyses of the October+July test indi- cated that N at 160 m was significantly lower than at 80 m, but that both of these depths did not differ from N at 20 and 40 m (Fig. 3). Similar patterns were observed for species diversity (H'), with significant differences across depth for both the October+July and all-months tests (ANOVA, P[132]=18.4, P<0.001 and P(1 26]=137.3, P<0.001, respectively; Fig. 3). In both all-months and 190 Fishery Bulletin 109(2) October+July tests H' at 160 m was significantly lower than that observed at shallower depths (Fig. 3). N did not differ significantly across months (ANO- VA, F[132]=1.7, P=0.18 and F(1 26] = 0.3847, P=0.68 in October+July and all-months tests, respectively) and there was no significant interaction between depth and month (ANOVA, F[3 32] = 2.3, P=0.08 and P[2 26) = 1.4, P=0.26 in October+July and all-months tests, respec- tively). In contrast, H' in the October+July test differed significantly by month (ANOVA, P|132]=13.4, PcO.OOl), as well as the interaction between depth and month (ANOVA, F(3 32|=:8.3275, P<0.001). Overall, H' was high- er in October than in July and the interaction term appears to be related to a strong seasonal change in H' at 80 m (Fig. 3). Month and the interaction between month and depth were not significant in the test for all-months (ANOVA, F^ 26]=2.0, P=0.15 and F(1 26]=1.7, P=0.2, respectively). Taxon-based analysis Depth was a significant predictor of assemblage struc- ture in all-months and in October+July CCA tests, and explained 44% (F(1 27j = 19.1, P<0.001) and 34% (F[3 33] = 6.1, PcO.OOl) of the variances, respectively. Season explained a smaller proportion of the vari- ance (5%) in the October+July CCA test (F(1 33]=2.7, P<0.05) and was not a significant predictor in the all-months test (F^ 27]=1.9, P<0.18). Temporal shifts in depth distributions were negligible at the assemblage level; the joint variance explained by season and depth was zero for October+July and all-months analyses. The resulting tri-plots for each partial CCA (based on weighted averages of the species scores) simultaneously depict the centroid of the sites cod- ed for the constraining variables and the position (eigenvectors) of the species forming the response matrix. Examination of the depth partial CCA tri-plots for October+July and all-months analy- ses, indicated that the first CCA axis primarily separated shallow (20 and 40 m) and deep (80 and 160 m) fish communities (Fig. 4). The spread of the variable centroids indicates the relative dif- ferences in species composition among the respec- tive depth and month factors (distant centroids are more dissimilar in species composition than close centroids). Species centered near the origin of the tri-plot have little to no association with the predictor variables included in the analyses, but those furthest from the origin have higher loadings, the strongest associations with the CCA axis, and contribute the most to differentiating sites that separate along the same axis. For the October+July samples, the second CCA axis also separated the 80- and 160-m fish assemblages (Fig. 4). Partial CCA results for the October+July and all-months samples confirmed that 160 m was dominated by spotted ratfish and small spiny dog- fish, but that Pacific hake (Merluccius productus), rex sole ( Glyptocephalus zachirus), and dover sole (Microstomus pacificus ) also typified that depth (Fig. 4). Furthermore, species that were associ- ated with both 80 and 160 m included large spiny dogfish, and quillback rockfish (Sebastes maliger), brown rockfish (Sebastes auriculatus), blackbelly eelpout (Lycodes pacificus ), Pacific tomcod (Mi- crogadus proximus ), walleye pollock ( Theragra chalcogramma) , shiner perch ( Cymatogaster ag- gregate), pile perch ( Rhacochilus vacca ), black tip poacher (Xeneretmus latifrons), slender sole ( Lyop - setta exilis), and plainfin midshipman ( Porichthys Reum and Essington: Season- and depth-dependent variability of a demersal fish assemblage in a fjord estuary 191 □October H March H July Figure 3 Average species richness (N\ upper panel) for (A) October and July samples and (B) October, March, and July (all-months) by depth sampled. Average species diversity (H'\ lower panel) for October and July samples (C) and October, March and July ( D ) . Lower case letters above each depth category denote depths that do not differ on the basis of post hoc Tukey honestly significant difference tests. (A) and (C) correspond to October+July two-way ANOVA tests and (B) and (D) correspond to all-months two-way ANOVA tests. H' differed by month only for the October+July. Note: depths at 20 and 80 m were not sampled in March. Error bars indicate standard deviation. For October, March, and July, 4, 4, and 5 sites were sampled at each depth, respectively. notatus ; Fig. 4). In shallow waters (20 and 40 m) several flatfishes dominated including small and large English sole, rock sole ( Lepidopsetta bilineata), C-0 sole ( Pleuronichthys coenosus ), Pacific sanddab ( Citharichthys sordidus ), and sand sole iPsettichthys melanostictus) among other species (Fig. 4). In the October+July analyses, tem- poral changes in assemblage structure were largely driven by species that were found primarily at depths of 80 m. The biomass of shiner perch, pile perch, walleye pollock, and Pacific tomcod was highest in October and the biomass of large spiny dogfish and slender sole was highest in July (Fig. 4). At depths of 20 and 40 m, the biomass of small English sole and C-0 sole was highest in Octo- ber and the biomass of staghorn sculpin (Leptocottus armatus ) and sand sole was highest in July. The biomass of species that typified the 160-m depths changed little between October and July (e.g., spotted ratfish, dover sole, rex sole, and Pacific hake; Fig. 4). In the all-months analysis, species with higher biomass in March included C-0 sole, sand sole, great sculpin, and rock sole (Fig. 4). Size-based analysis Body sizes encountered in the survey spanned eleven size classes, ranging from 2 to 2048 g. Overall, biomass spec- tra were nonlinear in appearance and approximately parabolic for most depth and season combinations (Fig. 5). For that reason metrics describing linear biomass size spec- tra (intercept, slope) were not estimated. Deep waters were dominated by individuals larger than 32 g, whereas shallow waters contained relatively more individuals that were less than 128 g (Fig. 5). Temporal differences were most apparent at depths of 80 m where biomass was concentrated in body size classes greater than 128 g in July and at 40 m that contained peak biomass levels in the 16-g body size class in March. Overall, 28% and 29% of the biomass spectra variance was associated with depth in both all-months (F^ 10] = 5.4, P<0.001) and October+July tests (F(3 10) = 2.3, PcO.001), respectively. Month explained a smaller but significant proportion of variance in the October+July test (11%; F,x 10,=2.1, PcO.001) and was not a significant predictor in the all seasons test (P|133!=2.1, P=0.09). Variance explained jointly by season and depth was again zero. In the October+July analyses, the first CCA axis ac- counted for 20.9% of the total variation and separated the shallow (20 and 40 m) and deep assemblages (160 m). The second axis accounted for 10.3% of the variance and distinguished 80 m from the other depths. The largest size class (2048 g) was associated with depths of 80 m, and the next three largest size classes (256, 512, and 1024 g) were associated with depths of 160 m (Fig. 6). In contrast, the smallest size classes (4, 8, and 16 g) were affiliated with depths of 20 and 40 m. The remaining intermediate body size classes were near the origin of the ordination plot and not closely associated with any of the depths. The analysis of all-months tests reiterated these patterns (Fig. 6). Tracking the arrival of dogfish, the size classes with the strongest temporal responses were also the largest size classes (1024 and 2048 g) which exhibited higher abundances in July (Fig. 6). In October, biomass in the smallest size classes (2, 8, and 16 g) was relatively higher. Discussion Fjord systems such as Puget Sound typically possess steep bathymetries and deep basins that result in deep- water habitat relatively close to shore. As expected, the demersal fish assemblage in Puget Sound varied 192 Fishery Bulletin 109(2) 2“ 1 - CM < O o 0. PTC**BTP PHR* : SIP* BFE m • WBP* SL^ *WEP SPg| 480 m PIP* PFP* BRF, SHP* *QRF *BBE #SPDs 20 m pc^ m en!^ tG>NS' RKS^^SFS COS^SPS DOV •REX LNS RAT **phk 160 m SAS* GRS -3 -1 0 1 CCA 1 (4.9%) CCA 1 (43.6%) CCA 1 (6.7%) Figure 4 Species-based partial canonical correspondence analysis (CCA) tri-plots for fall+summer and all-seasons analyses in which depth (A, B) and season (C, D) are the constraining variables, respectively. In the tri-plots, the site centroids coded for each constraining factor are indicated by triangles. The proxim- ity of factor centroids in ordination space to one another corresponds to their respective similarity in average species composition. Species that have strong loadings (large eigenvectors) are distant from the tri-plot origin and contribute the most to differentiating factor centroids that separate along the same axis. The eigenvalue or variance associated with each axis is indicated by parentheses. In cases were the constraining factor consisted of only two groups as in (B) and (C) only one CCA axis was generated. To aid interpretation two dimensional plots were generated by plotting the CCA (x-axis) against the residual axis (y-axis) derived from standard correspondence analysis (CA) performed on the remaining community variance. Species codes: BFE = bigfin eelpout ( Lycodes cortezianus)\ BBE=blackbelly eelpout (Lycodes pacificus)', BTP=blacktip poacher ( Xeneretmus latifrons)', BRF=brown rockfish ( Sebastes auriculatus ); C0S = C-0 sole (Pleuronichthys coenosus)', DOV=Dover sole ( Micros - tomus pacificus ); ENS1 and ENSs for English sole ( Parophrys vetulus) large and small, respectively; GRS=breat sculpin ( Myoxocephalus polyaca n thoceph alus); LNS=longnose skate (Raja rhino ); PHK=Pacific hake (Merluccius productus ); PHR = Pacific herring ( Clupea pallasii pallasii ); PSD = Pacific sanddab (Citharichthys sordidus)\ PTC = Pacific tomcod (Microgadus proximus)', PIP=pile perch (Rhacochilus vacca)\ PFP=plainfin midshipman (Porichthys notatus ); QRF=quillback rockfish ( Sebastes maliger ); REX=rex sole (Glyptocephalus zachirus ); RKS=rock sole (Lepidopsetta bilineata); RBS=roughback sculpin (Chitonotus pugetensis)\ SAS = sand sole ( Psettichthys melanostictus); RAT=spotted ratfish (Hydrolagus colliei ); SGP=sturgeon poacher ( Podothecus accipenserinus ); SFS = sailfin sculpin (Nautich- thys oculofasciatus ); SIP=shiner perch (Cymatogaster aggregata)\ SLS = slender sole (Lyopsetta exilis)', SPS = speckled sanddab (Citharichthys stigmaeus ); SPD1 and SPDs = spiny dogfish (Squalus acanthias) large and small, respectively; SHP=staghorn sculpin ( Leptocottus armatus ); WEP=walleye pollock ( Theragra chalcogramma ); WBP=whitebarred prickleback ( Poroclinus rothrocki). Reum and Essington: Season- and depth-dependent variability of a demersal fish assemblage in a fjord estuary 193 October 20 m Spiny dogfish ( Squaius acanthias) g Spotted ratfish (Hydrolag us colliel) □ Flatfish ■ Other 0.4 0.3 0.2 0.1 0 July 20 m bd n 0.4 1 n October 40 m i 1 1 j — i — j — i March 40 m n Un 7 9 11 I October 80 m 11 0.4 0.31 0.2 0.1 0 0.4 0.3 5 7 9 11 July 40 m y i=n — — i 9 11 1 0.2- , , , ov "SS — i n July 80 m 1 3 5 7 9 11 October 1 60 m 1 3 5 7 9 11 July 1 60 m 0 -I — 1 — ' i F3 1 3 5 7 9 11 Log2 body mass (g) Figure 5 Average normalized biomass spectra of the demersal fish assemblage from the Central Basin of Puget Sound sampled with bottom trawl at different depths from October 2004 and March and July 2005. substantially across depths and among seasons, but was unusual in that chondrichthyans (spotted ratfish and spiny dogfish) made up a majority of the fish bio- mass. Fish assemblages in estuaries on the outer coast of Washington, Oregon, and California are dominated by teleosts (e.g., Armor and Herrgesell, 1985; Bottom and Jones, 1990; De Ben et al., 1990), and this finding shows that demersal fish assemblages in Puget Sound differ fundamentally from shallower estuarine systems. Total assemblage biomass at depths of 80 and 160 m principally comprised spotted ratfish year round and spiny dogfish seasonally. Deepwater habitat is wide- spread in the Central Basin (average depth: 120 m; Burns, 1985), making these two species the dominant species, in biomass, in the area surveyed. An abundance of spotted ratfish has also been noted in deep waters immediately adjoining Puget Sound (Palsson et al.2), but what remains poorly understood is whether spotted ratfish are common in northeastern Pacific fjords or whether they are superabundant only in Puget Sound. Spotted ratfish are primarily benthic feeders, feeding on polychaetes and bivalves and occasionally on small fishes (Quinn et al., 1980), which indicates that they potentially are competitors to benthic feeding flatfish- es (Reum and Essington, 2008). In addition, they are preyed upon by large elasmobranchs (e.g., sixgill sharks [Hexanchus griseus], spiny dogfish). There is presently insufficient information to evaluate whether the high abundance of spotted ratfish is a secondary effect of shifts in food web structure, or whether other envi- 2 Palsson, W. A., S. Hoffmann, P. Clarke, and J. Beam. 2003. Results from the 2001 transboundary trawl survey of the southern Strait of Georgia, San Juan Archipelago and adjacent waters, 117 p. Washington State Dept. Fish and Wildlife, Olympia, WA. 194 Fishery Bulletin 109(2) CCA1 (10.9%) Figure 6 Partial canonical correspondence analysis (CCA) tri-plots of biomass spectra of fall+summer and all seasons analyses in which depth is the constraining variable for (A) and (B), respec- tively. For all seasons, only depth was significantly related to variation in the biomass spectra (C). In the tri-plots, the site centroids coded for each constraining factor are indicated by triangles. The proximity of factor centroids in ordination space to one another corresponds to their respective similarity in distribution of biomass across body size classes. Body size classes are indicated by circles. Those that have strong loadings (large eigenvectors) are distant from the tri-plot origin and contribute the most to differentiating factor centroids that separate along the same axis. The eigenvalue or variance associated with each axis is indicated within parentheses. In cases were the constraining factor consisted of only two groups, as in (B) and (C), only one CCA axis was generated. To aid interpretation two-dimensional plots were generated by plotting the CCA (*-axis) against the residual axis (y-axis) derived from standard correspondence analysis (CA) performed on the remaining community variance. ron mental features of Puget Sound promote their high abundance. Historical abundance records of spotted ratfish in Puget Sound are available in agency and proj- ect reports and indicate that they have been a common component of the food web at least since the 1930s, but it is difficult to extract trends from these data because of the lack of standardization of sample sites and gears. Future comparisons with less impacted fjords may of- fer insight into whether cartilaginous species naturally dominate in these systems or do so only under condi- tions related to human-caused ecosystem degradation. Despite our poor understanding of long-term changes in the Puget Sound fish community, the high abundance of ratfish in the area surveyed indicate that they likely constitute a significant node in the Puget Sound food web and are therefore deserving of further study. Spiny dogfish was the most abundant demersal pisci- vore and may have a particularly important influence on assemblage structure in Puget Sound and other coastal ecosystems. Spiny dogfish have a diverse diet, feeding on Pacific herring, flatfish, spotted ratfish, sal- monids, as wells as a wide range of benthic and pelagic invertebrates (Reum and Essington, 2008; Beamish and Sweeting, 2009). Large mobile predators such as spiny dogfish play an important role by linking differ- ent communities through predation and may stabilize Reum and Essington: Season- and depth-dependent variability of a demersal fish assemblage in a fjord estuary 195 communities by dampening oscillations in prey popula- tions through behavioral mechanisms such as switching prey (McCann et al., 2005). Moreover, understanding how large predators use different habitats is impor- tant for estimating prey mortality rates and system energy flows (Bax, 1998). Information on spiny dogfish movement patterns in Puget Sound is limited to tag- ging studies from the 1940s and 1970s. Results from these studies indicate that approximately 70% of the spiny dogfish population resides in the Sound and the adjoining Strait of Georgia for multiple years, while the remaining dogfish are transient (McFarlane and King, 2003). Spiny dogfish on the outer Washington and Brit- ish Columbia coasts are known to exhibit seasonal lati- tudinal migrations, but seasonal movements of dogfish tagged within Puget Sound and the Strait of Georgia are less clear (McFarlane and King, 2009; Taylor and Gallucci, 2009) and remain the subject of ongoing re- search. It is unknown whether spiny dogfish migrate in the winter to other subbasins within Puget Sound, to habitats shallower than 20 m, or whether they simply feed higher in the water column, disassociating with the benthos. There is evidence that catch rates of spiny dogfish in monitoring surveys have declined since the mid 1980s (Palsson, 2009) and that growth and size-at- maturity have also undergone substantial shifts (Taylor and Galluci, 2009). Characterizing movement patterns and population dynamics will clarify the impact of spiny dogfish on nearshore food webs. As demonstrated in Puget Sound, fjordal fish as- semblages vary markedly in space and time, and this variation has practical implications for modeling energy flows and interspecific interactions such as predation and competition. Food web models can foster ecosystem based management because they offer a framework for summarizing system knowledge and permit the simu- lation of alternative management scenarios. However, food web models for Puget Sound (as well as any other temperate system) parameterized by using fish abun- dances from one season alone may misrepresent the importance of different feeding modes and mischarac- terize patterns of trophic links in the fish assemblage (Greenstreet et al., 1997). In addition, stark differences between deep and shallow assemblages in terms of total biomass and species composition indicate strong spatial structuring in the likelihood and intensity of interactions among species, which may be a more gen- eral feature of fjord estuaries with similar deep basin bathymetrics. The data presented here, when coupled with diet information, provide a basis for determining the parameters of trophodynamic models that account for spatial and temporal variability in the fish assem- blage (e.g., Pauly et al., 2000). The use of three separate methods for analyzing dif- ferences in the demersal fish assemblage allowed us to investigate whether our interpretation of assemblage variation differed depending on the method. All three methods indicated significant differences between shal- low (20, 40, and 80 m) and deep waters (160 m) across seasons, with the exception of N, where values from 20 and 40 m waters did not differ from values in 160 m waters. We note however, that variation in N may have been artificially reduced by our exclusion of rare spe- cies from the analysis. In contrast, species diversity, H\ which takes into account the relative biomass of each species, should be more robust to the exclusion of rare species. Results from taxon-based CCA complemented results from the diversity metrics by highlighting those species that co-varied with depth and simultaneously depicted site similarity in ordination space. Lastly, size-based analyses revealed differences in assemblage structure based on biomass spectra, a macroecological descriptor of assemblage structure (Jennings, 2005). Because body size is correlated with trophic level in aquatic systems (Kerr 1974; Jennings et al., 2001), dif- ferences in biomass spectra among assemblages may re- veal fundamental differences in trophic structure (Rice, 2000; Sweeting et al., 2009). Significant depth related differences in the biomass spectra paralleled results from our taxon-based analyses and offer evidence that food web structure likely varies with depth. Combined, these approaches offer alternative prisms through which to view the demersal fish assemblage and mutually confirm important differences in assemblage structure. As with other marine fish surveys, our results are partly contingent on the effectiveness of the sampling gear and are premised on the assumption that catch- ability varies little among species. To maintain compa- rability with past demersal fish studies in Puget Sound we intentionally sampled using trawl equipment with specifications nearly identical to those used in previ- ous agency surveys in the region. The use of bottom trawls, however, also meant that species associated with rocky reef habitats would be excluded from our survey because the gear was suitable only for trawling in soft-bottom habitats. Moreover, large species such as six gill shark, which typically exceed 2 m in length in Puget Sound, were missed from the survey altogether. For future comparisons with other fjord systems, an effort should be made to sample regions with similar soft bottom habitats. We note that the data presented here reflect daytime distribution patterns, and diel movements of species may potentially connect deep and shallow communities. Another limitation of the present study is that our results span only a single time period and cover only a single basin in Puget Sound. Thus, we do not presume that the patterns described here will necessarily hold for all regions and be stable across time. Indeed, ample evidence from marine ecosystems points to the dynamic nature of community structure (Anderson and Piatt, 1999). Future research may very well improve our estimates of diversity and increase our understanding of temporal shifts in the Puget Sound fish assemblage. Although depth and season were clearly important in explaining community structure and species abun- dances, roughly one-half of the variation in these met- rics was not related to depth or season. Differences in temperature and salinity between the northern and southern CTD stations were relatively slight and differ- 196 Fishery Bulletin 109(2) I ences among other parameters, such as oxygen concen- tration and turbidity (data not shown), showed similar patterns. Because water conditions vary little in the region surveyed they are unlikely to explain site-spe- cific variability. Other unmeasured variables, however, such as substrate composition, sediment contamination, or shoreline and land cover characteristics in regions adjoining each sample site may potentially explain ad- ditional variance in the assemblage. We note, however, that spatial autocorrelation was not detected in the data set and indicates that spatially structured environmen- tal variables, such as those that covary with coastal urbanization, are unlikely to explain much additional variation at least at the spatial scales embraced by our survey (Borcard et al., 1992). Our research provides the first published assessment of seasonal variability in assemblage structure across multiple depths in the Puget Sound demersal fish as- semblage and offers insight into general features of deep fjord systems. We found strong structuring of the assemblage by depth and smaller, although important, differences across seasons. These shifts were manifest in simple assemblage metrics and in multivariate taxon- based and size-based analyses. The identification of these patterns, in turn, identifies priorities for future investigators that will further our understanding of the demersal assemblage and the forces that act to shape it. Notably, we identified species that may seasonally modify the Puget Sound food web in significant ways (e.g., spiny dogfish) and we confirmed the findings of other researchers who have identified spotted ratfish as one of the most abundant fishes in the Puget Sound region (Quinnel and Schmitt3). Remarkably, research directed toward uncovering the life history and ecol- ogy of spotted ratfish has been limited (but see Quinn et al., 1980; Barnett et al. 2009). Furthermore, key species in the Puget Sound food web may be those that link habitats through movement and foraging activi- ties. Understanding diel and seasonal-scale movement patterns will greatly improve our understanding of the Puget Sound fish assemblage. Acknowledgments Financial support was provided for J. Reum by the Vincet Liguori Fellowship and funding for boat time was provided by the University of Washington Research Royalties Fund. We are grateful to A. Beaubreaux, M. Hunsicker, J. Murphy, K. Marshall, M. Anderson, and D. English for assistance with field work. Comments from M. Hunsicker, C. Harvey, G. Williams, and three anonymous reviewers greatly improved earlier versions of this article. 3 Quinnell, S., and C. Schmitt. 1991. Abundance of Puget Sound demersal fishes: 1987 research trawl survey results, 240 p. Washington State Dept. Fish and Wildlife, Olympia, WA. Literature cited Armor, C. and P. L. Herrgesell. 1985. Distribution and abundance of fishes in the San Francisco Bay estuary between 1980 and 1982. Hydro- biol. 129:211-227 Anderson, M. J., and N. A. Gribble. 1998. Partitioning the variation among spatial, temporal and environmental components in a multivariate data set. Aust. J. Ecol. 23:158-167 Anderson, P. J., and J. F. Piatt. 1999. Community reorganization in the Gulf of Alaska following ocean climate regime shift. Mar. Ecol. Prog. Ser. 189:117-123. Barnett, L. A. K, R. L. Earley, D. A. Ebert, and G. M. Cailliet. 2009. Maturity, fecundity, and reproductive cycle of the spotted ratfish, Hydrolagus colliei. Mar. Biol. 156:301-316 Bax, N. J. 1998. The significance and prediction of predation in marine fisheries. ICES J. Mar. Sci. 55:997—1030 Beamish R. J., and R. M. Sweeting. 2009. Spiny dogfish in the pelagic waters of the Strait of Georgia and Puget Sound. In Biology and management of dogfish sharks (V. F. Gallucci, G. A. McFarlane, and G. G. Bargmann, eds.), p. 101-118. Am. Fish. Soc., Bethesda, MD. Borcard, D., P. Legendre, and P. Drapeau. 1992. Partialling out the spatial component of ecological variation. Ecology 73:1045-1055. Bottom, D. L. and K. K. Jones. 1990. Species composition, distribution, and invertebrate prey of fish assemblages in the Columbia River Estu- ary. Prog. Oceanog. 25:243-270. Burns, R. 1985. The shape and form of Puget Sound, 100 p. Wash- ington Sea Grant Program, Seattle, WA. De Ben, W. A., W. D. Clothier, G. R. Ditsworth, and D. J. Baumgartner. 1990. Spatiotemporal fluctuations in the distribution and abundance of demersal fish and epibenthic crustaceans in Yaquina Bay, Oregon. Estuaries 13:469-478. Greenstreet, S. P. R., A. D. Bryant, N. Broekhuizen, S. J. Hall, and M. R. Heath. 1997. Seasonal variation in the consumption of food by fish in the North Sea and implications for food web dynamics. ICES J. Mar. Sci. 54:243-266. Haedrich, R. L., and N. R. Merrett. 1992. Production/biomass ratios, size frequencies, and biomass spectra in deep-sea demersal fishes. In Deep- sea food chains and the global carbon cycle (G. T. Rowe and V. Pariente, eds.), p. 157-182. Kluwer Academic Publ., Amsterdam. Jennings, S. 2005. Size-based analyses of aquatic food webs. In Aquatic Food Webs: An Ecosystems Approach (A. Bel- grano, A., U. M, Scharler, J. Dunne, and R. E. Ulano- wicz, eds.), p. 86-97. Oxford Univ. Press, New York. Jennings, S., J. K. Pinnegar, N. V. C. Polunin, and T. W. Boon. 2001. Weak cross-species relationships between body size and trophic level belie powerful size-based tro- phic structuring in fish communities. J Anim. Ecol. 70:934-944. Kennish, M. J. 2002. Environmental threats and environmental future of estuaries. Environ. Conserv. 29:78-107. Reum and Essington: Season- and depth-dependent variability of a demersal fish assemblage in a fjord estuary 197 Kerr, S. R. 1974. Theory of size distribution in ecological communi- ties. J. Fish. Res. Board Can. 31:1859—1862. Krebs, C. J. 1989. Ecological methodology, 654 p. Harper Collins Publishers, Inc., New York. Legendre, P., and L. Legendre. 1998. Numerical ecology, 2nd ed., 853 p. Elsevier Sci- ence, Amsterdam. Lotze, H. K., H. S. Lenihan, B. J. Bourque, R. H. Bradbury, R. G. Cooke, M. C. Kay, S. M. McCann, K. S., J. B. Rasmus- sen, and J. Umbanhowar. 2005. The dynamics of spatially coupled food webs. Ecol. Lett. 8:513-523. McFarlane, G. A., and J. R. King. 2003. Migration patterns of spiny dogfish (Squalus acanthias) in the North Pacific Ocean. Fish. Bull. 101:358-367. 2009. Movement patterns of spiny dogfish within the Strait of Georgia. In Biology and management of dogfish sharks (V. F. Gallucci, G. A. McFarlane, and G. G. Barg- mann, eds.), p. 77-88. Am. Fish. Soc., Bethesda, MD. Mueter, F. J. and B. L. Norcross. 1999. Linking community structure of small demersal fishes around Kodiak Island, Alaska, to environmental variables. Mar. Ecol. Prog. Ser. 190:37-51. Palsson W. A. 2009. The status of spiny dogfish in Puget Sound. In Biology and management of dogfish sharks (V. F. Gal- lucci, G. A. McFarlane, and G. G. Bargmann, eds.), p. 53-66. Am. Fish. Soc., Bethesda, MD. Pauly, D., V. Christensen, S., and C. J. Walters. 2000. Ecopath, Ecosim and Ecospace as tools for evalu- ating ecosystem impact of fisheries. ICES J. Mar. Sci. 57:697-706. Pikitch, E. K., C. Santora, E. A. Babcock, A. Bakun, R. Bonfil, D. O. Conover, P. Dayton, P. Doukakis, D. Fluharty, B. Hene- man, E. D. Houde, J. Link, P. A. Livingston, M. Mangel, M. K. McAllister, J. Pope, and K. J. Sainsbury. 2004. Ecosystem-based fishery management. Science 305:346-347. Quinn, T. P., B. C. Miller, and C. Wingert. 1980. Depth distribution and seasonal and diel move- ments of ratfish, Hydrolagus colliei, in Puget Sound, Washington. Fish. Bull. 78: 816—821. Reum, J. C. P., and T. E. Essington. 2008. Seasonal variation in guild structure of the Puget Sound demersal fish community. Estuar. Coasts 31:790-801. Rice, J. C. 2000. Evaluating fishery impacts using metrics of com- munity structure. ICES J. Mar. Sci. 57:682-688. Sprules, W. G., J. M. Casselman, and B. J. Shuter. 1983. Size distribution of pelagic particles in lakes. Can. J. Fish. Aquat. Sci. 40:1761-1769. Sweeting, C. J., F. Badalamenti, G. D’Anna, C. Pipitone, and N. V. C. Polunin. 2009. Steeper biomass spectra of demersal fish commu- nities after trawler exclusion in Sicily. ICES J. Mar. Sci. 66:195-202. Taylor, I. G. and V. F. Gallucci. 2009. Unconfounding the effects of climate and den- sity- dependence using 60 years of data on spiny dog- fish. Can. J. Fish. Aquat. Sci. 66:351—366. ter Braak, C. J. F. 1986. Canonical correspondence analysis: a new eigen- vector technique for multivariate direct gradient analy- sis. Ecology 67:1167-1179. Thompson, A. A., and B. D. Mapstone. 2002. Intra- versus inter-annual variation in counts of reef fishes and interpretations of long-term monitoring studies. Mar. Ecol. Prog. Ser. 232:247-257. Wagner, H. H. 2004. Direct multi-scale ordination with canonical cor- respondence analysis. Ecology 85:342-351. Werner, E. E., and J. F. Gilliam. 1984. The ontogenetic niche and species interactions in size structured populations. Ann. Rev. Ecol. Syst. 15:393-425. Zar, J. H. 1984. Biostatistical analysis, 2nd ed., 718 p. Prentice Hall, Inc., Englewood Cliffs, NJ. 198 Interactions of age-dependent mortality and selectivity functions in age-based stock assessment models Xi He (contact author) Stephen Ralston Alec D. MacCall Abstract — The natural mortality rate (M) of fish varies with size and age, although it is often assumed to be constant in stock assessments. Mis- specification of M may bias important assessment quantities. We simulated fishery data, using an age-based pop- ulation model, and then conducted stock assessments on the simulated data. Results were compared to known values. Misspecification of M had a negligible effect on the estimation of relative stock depletion; however, misspecification of M had a large effect on the estimation of parame- ters describing the stock recruitment relationship, age-specific selectivity, and catchability. If high M occurs in juvenile and old fish, but is misspeci- fied in the assessment model, virgin biomass and catchability are often poorly estimated. In addition, stock recruitment relationships are often very difficult to estimate, and steep- ness values are commonly estimated at the upper bound (1.0) and over- fishing limits tend to be biased low. Natural mortality can be estimated in assessment models if M is constant across ages or if selectivity is asymp- totic. However if M is higher in old fish and selectivity is dome-shaped, M and the selectivity cannot both be adequately estimated because of strong interactions between M and selectivity. Manuscript submitted 18 August 2010. Manuscript accepted 17 February 2011. Fish. Bull. 109:198-216 (2011). 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. Email address for contact author: xi.he@noaa.gov Fisheries Ecology Division Southwest Fisheries Science Center National Marine Fisheries Service National Oceanic and Atmospheric Administration 110 Shaffer Road Santa Cruz, California 95060 Correctly specifying the instanta- neous rate of natural mortality CM) in stock assessment models is important because misspecification may lead to over- or underestimates of criti- cal assessment quantities, including stock depletion, maximum sustain- able yield (MSY), virgin biomass, and density dependence (Lapointe et al., 1989; Thompson, 1994; Mertz and Myers, 1997; Punt and Walker, 1998; Clark, 1999; Wang et al., 2006). It is widely believed that natural mortality varies with age or size; young (small) fish have higher natural mortality rates due to higher predation risks, disease, or starvation (Lorenzen, 1996), whereas older (larger) fish may have increased natural mortality with senescence or because of cumulative reproductive stress (Mangel, 2003; Moustahfid et al., 2009). In spite of the widely held percep- tion that natural mortality varies considerably with age, most stock assessment models assume that M is constant for all ages, mainly be- cause there are insufficient data with which to estimate natural mortal- ity on an age-specific basis. Anoth- er reason for assuming constant M in stock assessment models is that natural mortality is typically highly correlated with other key parame- ters, including stock recruitment and selectivity parameters (Lapointe et al., 1992; Thompson, 1994; Schnute and Richards, 1995; Fu and Quinn, 2000), quantities that are often quite difficult to estimate with accuracy (Maunder et al.1). In previous studies a variety of approaches have been developed to estimate natural mortality, includ- ing the use of maximum observed age (Hoenig, 1983) and life-history parameters (Alverson and Carney, 1975; Gunderson, 1980; Myers and Doyle, 1983; Roff, 1992; Jensen, 1996; Gunderson, 1997). In other studies, life-history data and envi- ronmental variables have been com- bined to establish empirical relation- ships to predict natural mortality (Pauly, 1980; Gislason et al., 2010). These studies have provided esti- mates of natural mortality that can be useful for stock assessments but these estimates may not be suffi- cient for species-specific stock as- sessment because of bias (e.g., only a subset of possible life histories was considered). Other studies have shown that unless species-specific data were collected before exploi- tation of the species, estimates of 1 Maunder, M. N„ H. H. Lee, and K. R. Piner. 2010. A review of natural mor- tality, its estimation, and use in fisher- ies stock assessment. Unpubl. manuscr, 35 p. Inter-American Tropical Tuna Commission, 8604 La Jolla Shores Drive, La Jolla, CA. He et at: Interactions of age-dependent mortality and selectivity functions in age-based stock assessment models 199 natural mortality are impractical, if not impossible, to derive from fishery or survey data, because of the interaction between fishing and natural mortality (Vetter, 1988; Quinn and Deriso, 1999). Clark (1999) examined the effects of incorrectly specifying M for a simple age-structured stock assessment and concluded that errors in M mainly affect estimates of fishing mortality and abundance, but not estimates of age- specific selectivity. In most regions of the world, where statistical stock assessments of single species provide the ba- sis for management advice (Worm et al., 2009), it is commonplace to assume a constant natural mortal- ity rate for all exploitable ages or sizes (or for both). Moreover, natural mortality is also typically assumed to be constant over time and identical among regions (Punt, 2003; Yin and Sampson, 2004; PFMC, 2008). Uncertainty in the use of constant natural mortal- ity in these assessment models is usually evaluated by an approach that is similar to likelihood profiles, where M values are changed and other parameters are fixed. However, this approach is highly dependent on the specific model structure and parameter set- tings being evaluated. For example, if stock recruit- ment relationships or selectivity functions are fixed in an assessment model, likelihood profile methods on natural mortality can provide only the validity of the model fitted to fixed values of natural mortality and not the validity of the model for its interactions with other model parameters. In this study, we compare stock assessment re- sults among simulated populations with different natural mortality schedules. The simulation data were generated with an age-based population model characterized by exploitation from a single fishery with a constant selectivity pattern over an extended period of time, representing somewhat ideal condi- tions. Simulations were crafted to reflect conditions in the U.S. west coast groundfish fishery — the source of most available fishery data for the last 40 years, a period when fishing intensity was high in the early years and has been low in recent years. In the simu- lation operating model, two different natural mortal- ity patterns were used: 1) constant natural mortality for all ages; and 2) elevated values of M in both juvenile and old fish. The data, along with sampling errors, were input into the assessment models. In the assessment model, natural mortality was assumed to be known and constant for all ages, estimated to be constant for all ages, or was estimated to follow an age-specific pattern. Estimated quantities from the assessments were then compared with the simulation models that generated the data. Important assess- ment results, e.g., stock depletion and stock-recruit relationships, were then compared to evaluate the effect of misspecification of natural mortality and se- lectivity on stock assessment estimates. In addition, results from the assessment models were compared with and without an informative parameter prior for spawner-recruit steepness parameters. Methods Simulation model The simulation or operating model in this study was an age-structured population model with a max age of 30 years. The last age was an age plus group. The Beverton-Holt stock-recruitment function was used to model stock recruitment. In particular, the “steep- ness” parameterization of Mace and Doonan2 with /; = 0.6 was used (see also Dorn, 2002). Recruitment variability was lognormal with aR set equal to 0.5 and lognormal survey variability set to equal 0.25. Specifics of the simulated population dynamics are presented in the Appendix. Base values for biologi- cal, fishery, and modeling parameters are presented in Table 1. Because of variability in recruitment and catchability, the model was run for 260 years, with the first 200 years as a “burn-in” period with no fish- ing to minimize the effect of initial conditions in the model. Only the last 40 years of data were provided for the assessment model. Biological parameters, including growth, fecundity, and the length-weight relationship were patterned after widow rockfish ( Sebastes entomelas) off the U.S. west coast (He et al., 2009). Although widow rockfish shows differences between the sexes in biological pa- rameters, the same values were used for both sexes to simplify the model. We modeled two different functional types of age- dependent natural mortality ( M ) in the simulations including 1) constant natural mortality for all ages (0.15/yr); and 2) high M in both juvenile and old fish (Table 2; Fig. 1). The annual sample size for age com- positions was 500 for all simulations — a size that ensured that informative age composition data were available to the assessment models. We used only one fishery in the operating model, and catches began in the last 40 years of the simula- tions. Fishing mortalities (F) were modeled as propor- tions of Fmsy, which varied over time. During the first 20 years, fishing at FMSY occurred, and in the last 20 years fishing mortality was 10% of FMSY. Two types of fishery selectivity patterns were simulated, i.e., simple asymptotic logistic and double normal curves (Fig. 2). The later was moderately dome-shaped and is implemented in the stock synthesis model (Methot, 2009a), a widely used stock assessment model. In all simulations the ascending limbs of the two selectiv- ity curves were constrained to be similar, i.e., 50% of individuals were selected at age 8. The above specific values and patterns in M, F, and selectivity were based on typical life history patterns of fish, and fishing patterns, off the U.S. west coast (e.g., those of widow rockfish). 2 Mace, P. M., and I. J. Doonan. 1988. A generalized bioeco- nomic simulation model for fish population dynamics. New Zealand Fishery Assessment Res. Doc., 88/4, 21 p. MAF Fisheries, Greta Point, Wellington, New Zealand. 200 Fishery Bulletin 109(2) Table 1 Biological, fishery, and modeling parameters used in the simulation model to evaluate interactions between mortality and selec- tivity. See Appendix for equations, definitions of parameters, and symbols. Parameters are the same for both sexes. True values are the base parameter values used in the simulation models. Lower and upper bounds are boundary limits used in the stock assessment program. NA=not applicable. Parameter Symbol True value Estimated in assessment model Lower and upper bounds Unit and note Minimum age amin 0 No NA Year Maximum age amax 30 No NA Age plus group Virgin recruitment R0 10 Yes 0.1, 30 Log scale Recruitment steepness h 0.6 Yes 0.2, 1.0 Annual recruit deviation R\ 0 Yes -5, 5 Log scale, 76 years Growth K 0.14 No NA Per year Growth 50.54 No NA cm Growth *0 -2.68 No NA Year Length-weight h 5.45e-6 No NA Length-weight h 3.2878 No NA Kg/cm Natural mortality M 0.15 Yes or no 0.01, 1 Per year, varied, see text Logistic selectivity 'll 8 Yes 0, 50 50% selectivity at age 8 Logistic selectivity 42 5 Yes 0, 50 Width for 95% selection Double normal selectivity 'li 13 Yes -507, 533 See Appendix Double normal selectivity n2 -2 Yes -82, 80 See Appendix Double normal selectivity n3 3.5 Yes -136, 143 See Appendix Double normal selectivity n4 2.6 Yes -101,106 See Appendix Double normal selectivity n5 -5 Yes -205, 195 See Appendix Double normal selectivity 46 0.65 Yes -25, 26 See Appendix Catchability — survey of juveniles , 160 £ 120 I 80 £ 40 0 Percentage of difference: virgin spawning output 200 >, 160 £ 120 | 80 £ 40 0 Percentage of difference: depletion Figure 4 Frequency plots of estimated differences of virgin spawning outputs (B0) and depletions between simulation and stock synthesis (SS3) assessment models from run 1. The differences are percentages of differences between simulation and assessment divided by true simulation values. Values equaled to zero indicate no differences between simulation and assessment models. population trajectories were very dif- ferent between the simulation and assessment models for run 2 (top right panel, Fig. 6), and stock recruitment parameters ( B0 h) were poorly esti- mated, with B() being lower and h being higher in the assessment models than those in the simulation models. The estimated catchability coefficients in the assessment models were higher than those in the simulation models. Estimated catchability coefficients for juvenile fish (q2) were especially high (>3.6 versus the correct value of 1.0) for run 2. This result occurred also for all other scenarios in which juve- nile natural mortalities were misspeci- fied in assessment models (Table 3). However, estimated selectivity func- tions matched fairly well between the simulation and estimation models (top row, Fig. 7). Performance of the stock assessment models in this setting was very good; 100% of the runs finished successfully and MGC values were sat- isfied (Table 4). Similar results were obtained if se- lectivity functions were double normal and were correctly specified in the He et at: Interactions of age-dependent mortality and selectivity functions in age-based stock assessment models 205 Table 5 Estimated natural mortalities (M) with 2.5% and 97.5% quantiles in parentheses for runs 13 to 16. A single M for all ages is estimated in runs 13 and 15, and four M values (break points) are estimated in runs 14 and 16. See the Methods section for how these four M values were assigned to each age group. Run no. M: m2 M, m4 13 0.150(0.139, 0.161) 14 0.448 (0.377, 0.513) 0.148 (0.108, 0.193) 0.147 (0.119, 0.169) 0.359 (0.321, 0.404) 15 0.148 (0.138, 0.163) 16 0.285 (0.131, 0.450) 0.169 (0.068, 0.258) 0.139 (0.080, 0.225) 0.074 (0.010, 0.389) Age Figure 5 Estimated selectivity functions from simulation (Sim) and stock synthesis (SS3) assessment models for run 1. Dashed lines are 2.5% and 97.5% quantiles from assessment model outputs. assessment models (runs 3 and 4). Population depletion, as well as other stock assessment pa- rameters, was well estimated if M was constant in both simulation and assessment models (run 3, second row in Fig. 6 and Table 3). The esti- mated double normal selectivity functions in the assessment model also matched well with that in the simulation model (run 3 in Fig. 7). Estimated population depletions were also matched reason- ably well, even with misspecified natural mortali- ties, but the estimated OFL statistics were about 10% negatively biased (run 4, Table 3). However, if natural mortality was higher for younger and older age classes in the simulation models but was constant in the assessment models, the estimated population trajectories were different, with the estimated jB0 biased high (run 4 in Fig. 6; Table 3). Selectivity functions matched fairly well in the ascending limb between the simulation and as- sessment models but failed to match the descend- ing limb of the selectivity curve (run 4, Fig. 7). Convergence of the estimation model was poor in this setting. In runs 3 and 4, 86.2% and 78.6% of 500 SS3 runs finished successfully, respectively, whereas only 81.3% and 48.2% of 500 SS3 runs produced satisfactory MGC values (Table 4). If selectivity functions were logistic in the simulation models but were double normal in the assessment mod- els and M was correctly specified (runs 5 and 6, Table 2) , most of the estimated parameters from the stock assessment models were close to those in the simulation models, generally less than 10% of differences (Table 3) . However, when natural mortality in the simulation model varied, but was assumed constant in the assess- ment model, the estimated catchability coefficient for the juvenile survey (q 2) was positively biased (run 6, Table 3). Time series of estimated spawning output matched reasonably well (runs 5 and 6, Fig. 6), and the estimated selectivity function showed a negative bias for old fish (runs 5 and 6, Fig. 7). Convergence performance was poor (Table 4); less than 83.3% of runs finished and only 53.0% of runs satisfied the MGC criterion (Table 5). If selectivity functions were double normal in the sim- ulation model but were misspecified as logistic functions in the estimation model (runs 7 and 8, Table 2), the curves fits were very poor, as expected (last row in Fig. 7 and Table 3). Spawning output was poorly estimated; all estimated spawning outputs were lower than those in the simulation models in the early years (last row in Fig. 6). If natural mortality was incorrectly specified in the assessment models (run 8), estimated parameters from the stock assessment models were strongly biased (Table 3). This bias included high correlations between the two stock recruitment parameters (B0 and /?), and positive biases in both catchability coefficients (ql and q2) (Table 3). Convergence of the assessment models, however, was very good. The percentages of runs finish- ing successfully and satisfying the MGC criterion were 100% (Table 4). If no prior for h was used in the assessment models and natural mortality was assumed to be constant (runs 9 tol2), the results in general were very similar to those from runs 1 to 4, where a prior on h was used (Figs. 8 and 9; Table 3). However, an important ex- 206 Fishery Bulletin 109(2) CT> CT> <1) 50 25 1960 50 25 1960 50 25 0 1960 50 - 25 1960 M 1 18 10 1990 1990 1990 1990 Year 18 10 18 10 18 10 2020 MJO I960 1990 2020 4 v4 \ V, 1960 ’ 1990 2020 6 « - V - ■ ^*'**«sL"* • 1960 1990 2020 8 - *'* 1960 1990 2020 Year Figure 6 Time series of spawning outputs from simulation and stock synthesis assessment models (runs 1 to 8). The run number is shown in the upper right of each graph. See Table 2 for model and parameter setup for all runs. Solid lines are median simulation outputs and thick dashed lines are median assessment outputs. Thin dashed lines are 2.5% and 97.5% of quantiles from simulation outputs. Specifica- tions for each panel are: M = constant M and MJO = high M in juvenile and old fish; L/L=logistic selectivity in both simulation and assessment models; D/D = double normal selectivity in both simulation and assessment models; D/L = double normal selectivity in simulation model, but logistic selectivity in assessment model; and L/D = logistic selectivity in simulation model, but double normal selectivity in assessment model. ception was that a high percentage of runs estimated steepness at the upper bound of (/i = 1.0). If logistic selectivity functions were used and natural mortali- ties were correctly specified in both simulation and assessment models (run 9), there were still 16% of runs with h at the upper bound (Fig. 10). If logistic selectivity functions were used but natural mortality was incorrectly specified assessment models (run 10), He et al.: Interactions of age-dependent mortality and selectivity functions in age-based stock assessment models 207 M MJO Age Age Figure 7 True selectivity from simulation models and estimated selectivity from stock syn- thesis assessment models (run 1 to run 8). The run number are shown in the upper left of each graph. Solid lines are true selectivity and thick dashed lines are median selectivity from assessment models. Thin dashed lines are 2.5% and 97.5% quantiles from assessment outputs. Specifications for each panel are: M = constant M and MJO=high M in juvenile and old fish; L/L=logistic selectivity in both simu- lation and assessment models; D/D = double normal selectivity in both simulation and assessment models; D/L = double normal selectivity in simulation model, but logistic selectivity in assessment model; and L/D = logistic selectivity in simulation model, but double normal selectivity in assessment model. there were close to 90% of runs with /i = 1.0 (Fig. 10). If double normal selectivity functions were used and natural mortality was constant in both simulation and assessment models (run 11), 17% of runs settled on the upper bound (h = 1.0) (Fig. 10). Results were similar even when natural mortality was high for both juvenile and old fish in the simulation model but was assumed to be constant in the assessment model 208 Fishery Bulletin 109(2) M MJO —. Q co ^ cn Q o 25 CD a. c n 50 25 50 25 10 18 10 10 1990 1990 2020 12 2020 Year Year Figure 8 Time series of spawning outputs from simulation and stock synthesis assessment models (runs 9 to 16). The run number is shown in the upper right of each graph. See Table 2 for model and parameter setup for all runs Solid lines are median simulation outputs and thick dashed lines are median assessment outputs. Thin dashed lines are 2.5% and 97.5% of quantiles from simulation outputs. Specifica- tions for each panel are: M = constant M and MJO=high M in juvenile and old fish; L/L=logistic selectivity in both simulation and assessment models; D/D = double normal selectivity in both simulation and assessment models; D/L = double normal selectivity in simulation model, but logistic selectivity in assessment model; and L/D = logistic selectivity in simulation model, but double normal selectivity in assessment model. (Fig. 10). However, selectivity was poorly fitted for old fish (panel 12, Fig. 9). Percentages of runs that finished and that had satisfactory MGC rates were similar to those runs (runs 1 to 4) with the same selectivity and M specifications but with h priors included (Table 4). If no priors for h were used and natural mortality was estimated in the assessment models (runs 13 to He et at: Interactions of age-dependent mortality and selectivity functions in age-based stock assessment models 209 Figure 9 True selectivity from simulation models and estimated selectivity from stock syn- thesis assessment models (run 9 to run 16). The run number is shown in the upper left of each graph. Solid lines are true selectivity and thick dashed lines are median selectivity from assessment models. Thin dashed lines are 2.5% and 97.5% quantiles from assessment outputs. Specifications for each panel are: M = constant M and MJO=high M in juvenile and old fish; L/L=logistic selectivity in both simu- lation and assessment models; D/D = double normal selectivity in both simulation and assessment models; D/L = double normal selectivity in simulation model, but logistic selectivity in assessment model; and L/D=logistic selectivity in simulation model, but double normal selectivity in assessment model. 16), the results varied. For runs 13 and 15, in which a single natural mortality was used for all ages and was estimated in the assessment models, key assess- ment outputs, including spawning outputs, selectivity, and distributions of estimated h values, were very similar to runs 9 and 11 (Tables 3 and 4; Figs. 8-10). Estimated values of natural mortality (M) were also very close to the true underlying M values (Table 5). 210 Fishery Bulletin 109(2) M MJO 1 20 9 450 10 2j 60 250 0 ■4!llili!lllllllllli.lt....i.n. 0 0.2 0.6 1.0 0.2 0.6 1.0 120 120 1 1 12 o CD a/a 60 o 0 .lllllllllllllllllll.lllll.-l_ ..1. 0 ■■■iillllililiiiih.ii...i C O’ 3.2 0.6 1.0 0'2 0.6 1.0 <1> it 120 - 13 120 14 ^ 60 60 0 ...lllllllllllllllllillil...i.l 0 0.2 0.6 1.0 0.2 0.6 1.0 120 15 120 16 D/D CD o 60 0 .lllllllllllllllllll..l.iii ■!„ 0 linllllllllllllllllilliiiii.il. «.. 0.2 0.6 1.0 0.2 0.6 1.0 Steepness (h) Steepness (h) Figure 10 Frequency plots of estimated steepness (h) in stock synthesis assessment model for runs 9 to 16. The run number is shown in the upper left of each graph. True steepness value is 0.6. No prior for /; was used in the assessment models. Specifica- tions for each panel are: M=constant M and MJO=high M in juvenile and old fish; L/L=logistic selectivity in both simulation and assessment models; D/D = double normal selectivity in both simulation and assessment models; D/L = double normal selectivity in simulation model but logistic selectivity in assessment model; and L/D = logistic selectivity in simulation model, out double normal selectivity in assessment model. For run 14, which had the same model configuration as run 10, except that four natural mortality values were used in both the simulation and estimation models, the assessment outputs matched well with those in the simulation model (Table 3; Figs. 8 and 9). Estimated natural mortalities also matched reasonably well with He et al.: Interactions of age-dependent mortality and selectivity functions in age-based stock assessment models 211 the true values (Table 5). For run 16, which had the same model configuration as run 12 except that four natural mortalities were used in both models, the assessment outputs matched very poorly with those in the simulation model (Table 3; Figs. 8 and 9). Spawning outputs of all years, including B0, estimated by the assessment model were much higher than those in the simulation model (Table 3; panel 16 in Fig. 8), and selectiv- ity was poorly fitted for old fish (panel 16, Fig. 9). Estimated natural mortalities for old fish (M4) showed a bi-modal distribution (Fig. 11), and there were strong interactions between es- timated M4 and selectivity for old fish (e.g., fish at age 30) (Fig. 12). There was a high proportion of cases (394 out of 500, Fig. 12) where M was estimated to be very small (mean=0.03). Patterns of convergence performance, between runs 9 to 12 and between runs 13 to 16, were very similar to those runs between runs 1 to 4 because these runs had the same setup for selectivities (Table 4). Discussion 0.0 0.06 0.12 0.18 0.24 0.30 0.36 0.42 Natural mortality for 25+ year-old fish Figure 11 Frequency plots of estimated natural mortality for 25+ year- old fish (M4) from run 16. True M4 values ranged from 0.25 to 0.5, and no prior for steepness (h) was used in the assess- ment models. 1.0- ° o — i — i — i — i — i — i — i — i — i — i — i — i — i — i — i — i — i — i — i — i — i — i — i — i — 0.0 0.06 0.12 0.18 0.24 0.30 0.36 0.42 Natural mortality for 25+ year-old fish Figure 12 Estimated selectivity at age 30 versus natural mortality for 25+ year-old fish (M4) from run 16. True M4 values ranged from 0.25 to 0.5 and no prior for steepness (/?) was used in the assessment models. Outputs were plotted in two separated groups based on the estimated M4 values. The first group had estimated M4 values <0.28 (solid dots) and the second had estimated M4 values >0.28. Means on the graph are mean selectivities for age-30 fish. Our research has shown that the assumption of a constant natural mortality for all ages when natural mortality is actually elevated in young and old fish can lead to inaccurate estimates of many important population and management quantities. The manner in which selectivity is modeled is also very important in determining which assessment parameters are poorly estimated and how these interact with one another in the model. In general, population depletion was well esti- mated, even when mortality and selectivity were incorrectly specified in assessment models because population depletion is a robust indicator of popu- lation status. This is mainly because depletion is estimated as the ratio of two quantities (termi- nal spawning output divided by virgin spawning output), both of which exhibit similar relative bi- ases. Estimates of another management variable, i.e., the OFL, were consistently biased, although 95% quantile intervals overlapped zero for some runs. These results indicate that OFL may be a more precautionary management indicator than population depletion. However, more research is needed to compare these two indicators because OFL depends on FMSY and biomass in the termi- nal year and estimates of these two quantities were strongly influenced by how natural mortality, selectivity and other population parameters were modeled in the assessment. Our results show that catchability coefficients for juvenile and adult surveys can be strongly positively biased if natural mortalities are higher in young and old fish but are misspecified in the estimation model, even when selectivity is correctly specified. If juvenile natural mortality is higher than that for adult fish, but is assumed to be the same as that for adult fish, catch- ability coefficients for juveniles from surveys of prere- cruits are poorly estimated. In many stock assessments, these coefficients are unknown and are often very small numbers because relative abundance is measured in 212 Fishery Bulletin 109(2) surveys. In this case there is no logical way within the estimation model to identify these poorly estimated pa- rameters. In cases where survey indices are derived to measure absolute population abundances, catchability coefficients for these survey indices could be estimated to be greater than unity because of misspecified natural mortality. The best to model selectivity in stock assessment models poses great challenges. This is especially true in modeling the selectivity of old fish. That is, one must decide if asymptotic (i.e., logistic) or dome-shaped (i.e., double normal) selectivity should be used. In most cas- es, no field or experimental data exist to support the choice of which form of selectivity is appropriate. As shown in this study, decreased selectivity in old fish can erroneously be attributed to increased natural mor- tality in old fish, and stock assessment models cannot resolve this error. In addition, the available sampling data for old fish in either age or length compositions are typically rare and render the estimation of the descending portions of selectivity curves imprecise and uncertain. Moreover, misspecifications of selectivity for old fish can still lead to moderately incorrect estimates of population status and management parameters (runs 7, 8, 11, and 12). Such incorrect estimates can have a greater effect on population status if old fish have higher weight-specific fecundity than young spawners, as is the case in many rockfish species along the U.S. west coast (Dick, 2009). Double normal selectivity has been widely used in recent stock assessments where the SS3 program was used. It has six parameters and is a very flexible selec- tivity function that can model a wide range of shapes for fishery selectivity (Methot, 2009a). Our study shows that double normal selectivity can sometimes lead to “unstable” estimations in stock assessment models. That is, the model may fail to converge properly, even in the absence of model specification error (runs 3 to 6, and runs 11 and 12). In the case of run 3, in which double normal selectivity is used in both the simulation and the assessment model, and natural mortality is also cor- rectly specified, model runs succeeded only 86% of the time and the MGC criterion was satisfied only 81% of the time. This finding further highlights the difficulty in estimating the descending portion of a dome-shaped selectivity curve and the uncertainty in estimating se- lectivity parameters. Unstable descending curves have also been observed in some recent west coast groundfish assessments, where selectivities for the last age (length) group drops to a very small value (He et ah, 2009). Fur- ther study on the stability of double normal selectivity may be needed to address this issue. We also conducted additional runs, in which high natural mortalities were simulated only for juvenile fish, and only for old fish, but were assumed to be con- stant in assessment models. The results showed that if high natural mortalities in juvenile fish existed but were misspecified in the assessment model, catchability coefficients for surveys of juveniles would be estimated to be much higher in assessment models (from 2.5 to 3.6 as compared to the true value of 1.0). Other assessment results for runs with high natural mortalities in juve- nile fish, however, were very similar to runs presented in this paper. If only high natural mortalities for old fish existed but were misspecified in the assessment model, assessment results would also be very similar to those of runs presented in this study with no biases in estimates for catchability coefficients for surveys of juveniles. This conclusion would indicate that effects of misspecifications of natural mortalities for juvenile and old fish on assessment results are mostly independent of each other. Natural mortality has rarely been treated as an esti- mable parameter and has often been set as a constant in stock assessment models. Our study shows that, given informative age composition data, natural mor- tality can be estimated if M is constant across ages or selectivity is asymptotic. However, if M is high in both juvenile and old fish, and selectivity is dome-shaped, estimates of M for old fish are very unreliable because that parameter strongly interacts with selectivity. Be- cause we examined only limited scenarios of data and model configurations, further and more detailed studies are needed to fully explore the feasibility and benefits of estimating natural mortality for fishery stock as- sessments. Stock assessment models in this study were fitted to data from simulation models with known model structure and error variance. In all simulation runs, the stock-recruitment function variability parameter was fixed (S^O. 5) and is relatively small compared to that of some stock assessments of the U.S. westcoast groundfish (Field et al., 2009; He et al., 2009; Wallace and Hamel, 2009). Given that the simulation data were much “better” than those available for most stock as- sessments, we found that it is still difficult to estimate the stock-recruitment relationship. As shown in runs 9 to 12, in which no priors for steepness (h) were given to the model, steepness was often estimated to be near or at the upper bound of 1.0 (Fig. 10), as has been found in other studies (Magnusson and Hilborn, 2007; Haltuch et al., 2008). This finding indicates that it is very difficult, if not impossible, to accurately estimate stock productivity in practice, where other uncertain- ties, such as model structures or lack of recruitment surveys, may further confound this issue (Haltuch et ah, 2009). Test runs on the simulation and assessment models with much longer time periods (300-year runs with 200 years of fishing down and data outputs to as- sessment models) show that estimates of stock recruit- ment relationships were reasonably close to true values. But this long period of data collection is generally not available for stock assessments. In many previous stud- ies, the difficulty in estimating stock recruitment rela- tionships has been emphasized, and sufficient biologi- cal information and fisheries data, which are lacking in many fisheries, are required to achieve reasonable estimation for stock recruitment relationships (Myers et ah, 1995; Rose et al., 2001; Magnusson and Hilborn, 2007; Conn et ah, 2010). Further studies on how or if He et at: Interactions of age-dependent mortality and selectivity functions in age-based stock assessment models 213 stock-recruitment relationships can be estimated at given levels of recruitment variability, data availabil- ity, and stock contracts are needed to provide general guidelines for estimating stock-recruitment relationship in assessment models. We believe that even with informative data and a correctly specified estimation model, there are strong interactions between natural mortality and fishery se- lectivity in stock assessment models. Misspecification of both parameters can lead to poor estimates of im- portant population and fishery parameters, which in turn can produce under- or over-estimates of important management quantities, such as stock depletion and OFL. Improvement in the assessment modeling ap- proach itself may not resolve these problems because of the interdependence of mortality, selectivity, and stock recruitment functions within the models. Uncer- tainty analysis of stock assessment models on age- or length-specific mortality and selectivity is also needed and should be included for assessments of model perfor- mance and for management of assessed stocks. Acknowledgments We thank E. Dick and J. Field for helpful comments on earlier versions of the manuscript. We also thank three anonymous reviewers for very helpful and constructive comments on earlier versions of the manuscript. Literature cited Alverson, D. L., and M. J. Carney. 1975. A graphic review of the growth and decay of popu- lation cohorts. J. Cons. Int. Explor. Mer 36:133-143. Clark, W. G. 1999. Effects of an erroneous natural mortality rate on a simple age-structured stock assessment. Can. J. Fish. Aquat. Sci. 56:1721-1731. Conn, P. B, E. H. Williams, and K. W. Shertzer. 2010. When can we reliably estimate the productivity of fish stocks. Can. J. Fish. Aquat. Sci. 67:511-523. Dick, E. J. 2009. Modeling the reproductive potential of rockfishes ( Sebastes spp.). Ph.D. diss., 353 p. Univ. California, Santa Cruz, CA. Dorn, M. W. 2002. Advice on west coast rockfish harvest rates from Bayesian meta-analysis of stock-recruit relation- ships. N. Am. J. Fish. Manag. 22:280-300. Field, C. F., D. J. Dick, D. Pearson, and A. D. MacCall. 2009. Status of bocaccio, Sebastes paucispinis, in the Conception, Monterey and Eureka INPFC area for 2009. Status of the Pacific coast groundfish through 2009, stock assessment and fishery evaluation: stock assessment, STAR panel report, and rebuilding analysis, 255 p. [Available from Pacific Fishery Management Council, 7700 NE Ambassador Place, Suite 101, Port- land, OR 97220-1384.] Fu, C., and T. J. Quinn II. 2000. Estimability of natural mortality and other popu- lation parameters in a length based model: Pandalus borealis in Kachemak Bay, Alaska. Can. J. Fish. Aquat. Sci. 39:1195-1207. Gislason, H., N. Daan, J. C. Rice, and J. G. Pope. 2010. Size, growth, temperature and the natural mortal- ity of marine fish. Fish Fish. 11:149-158. Gunderson, D. R. 1980. Using r-K selection theory to predict natural mor- tality. Can. J. Fish. Aquat. Sci. 37:2266-2271. 1997. Trade-off between reproductive effort and adult survival in oviparous and viviparous fishes. Can. J. Fish. Aquat. Sci. 54:990-998. Haltuch, M. A., A. E. Punt, and M. W. Dorn. 2008. Evaluating alternative estimators of fishery man- agement reference points. Fish. Res. 94:290-303. 2009. Evaluating the estimation of fishery management reference points in a variable environment. Fish. Res. 100:42-56. He, X, D. E. Pearson, D. J. Dick, J. C. Field, S. Ralston, and A. D. MacCall. 2009. Status of the widow rockfish resource in 2009. Status of the Pacific coast groundfish through 2009, stock assessment and fishery evaluation: stock assess- ment, STAR panel report, and rebuilding analysis, 187 p. [Available from Pacific Fishery Management Council, 7700 NE Ambassador Place, Suite 101, Port- land, OR 97220-1384.] Hoening, J. M. 1983. Empirical use of longevity data to estimate total mortality rates. Fish. Bull. 82:898-903. Jensen, A. L. 1996. Beverton and Holt life history invariants result from optimal trade-off of reproduction and survival. Can. J. Fish. Aquat. Sci. 53:820-822. Lapointe, M. F., R. M. Peterman, and A. D. MacCall. 1989. Trends in fishing mortality rate along with errors in natural mortality rate can cause spurious time trends in fish stock abundances estimated by virtual population analysis (VPA). Can. J. Fish. Aquat. Sci. 46:2129-2139. Lapointe, M. F., R. M. Peterman, and B. J. Rothschild. 1992. Variable natural mortality rates inflate variance of recruitments estimated from virtual population analysis (VPA). Can. J. Fish. Aquat. Sci. 49:2020-2027. Lorenzen, K. 1996. The relationship between body weight and natu- ral mortality in juvenile and adult fish: a comparison of natural ecosystems and aquaculture. J. Fish Biol. 49:627-647. Magnusson, A., and R. Hilborn. 2007. What makes fisheries data informative? Fish Fish. 8:337-358. Mangel, M. 2003. Environment and longevity: the demography of the growth rate. In Life span: evolutionary, ecological, and demographic perspectives, Supplement to population and development review (J. R. Carey and S. Tuljapurkar, eds.), p. 57-70. Population Council, New York. Mertz, G., and R. A. Myers. 1997. Influence of errors in natural mortality estimates in cohort analysis. Can. J. Fish. Aquat. Sci. 54:1608-1612. Methot, R. D. 2009a. User manual for stock synthesis, model vers. 3.04b, 143 p. Northwest Fisheries Science Center, NOAA Fisheries, 2725 Montlake Boulevard East, Seattle, WA. 2009b. Stock assessment: operational models in support of fisheries management. In The future of fisheries 214 Fishery Bulletin 109(2) science in North America, vol. 31 (R. J. Beamish and B. J. Rothschild, eds.), p. 137-165. Fish & Fisheries Series, Springer Science, Hoboken, NJ. Moustahfid, H., J. S. Link, W. J. Overholtz, and M. C Tyrrell. 2009. The advantage of explicitly incorporating preda- tion mortality into age-structured stock assessment models: an application for Atlantic mackerel. ICES J. Mar. Sci. 66:445-454. Myers, R. A., N. J. Barrowman, J. A. Hutchings, and A. A. Rosenberg. 1995. Population dynamics of exploited fish stocks at low population levels. Science 269:1106-1108. Myers, R. A., and R. W. Doyle. 1983. Predicting natural mortality rates and repro- duction-mortality trade-offs from fish life history data. Can. J. Fish. Aquat. Sci. 40:612-620. PFMC (Pacific Fishery Management Council). 2008. Status of the Pacific coast groundfish fish- ery through 2008, stock assessment and fishery evalua- tion: stock assessments, STAR panel reports, and rebuild- ing analyses, 13 p. [Available from Pacific Fishery Management Council, 7700 NE Ambassador Place, Suite 101, Portland, OR 97220-1384.] Punt, A. E. 2003. Evaluating the efficacy of managing west coast groundfish resources through simulation. Fish. Bull. 101:860-873. Punt, A. E. and T. I. Walker. 1998. Stock assessment and risk analysis for the school shark (Galeorhinus galeus) off southern Australia. Mar. Freshw. Res. 49:719-731. Pauly, D. 1980. On the interrelationships between natural mor- tality, growth parameters, and mean environmental temperature in 175 stocks. J. Cons. Int. Explor. Mer 39:175-192. Quinn, T. J., II, and R. B. Deriso. 1999. Quantitative fish dynamics, 542 p. Oxford Univ. Press. New York. Roff. D. A. 1992. The evolution of life histories, 535 p. Chapman and Hall, New York. Rose, K. A., J. H. Cowan, K. O. Winemiller, R. A. Myers, and R. Hilborn. 2001. Compensatory density dependence in fish popu- lations: importance, controversy, understanding and prognosis. Fish Fish. 2:293-327. Schnute, J. T. and L. J. Richards. 1995. The influence of error on population estimates from catch-age models. Can. J. Fish. Aquat. Sci. 52:2063-2077. Thompson, G. G. 1994. Confounding of gear selectivity and the natural mortality rate in cases where the former is a non- monotone function of age. Can. J. Fish. Aquat. Sci. 51:2654-2664. Vetter, E. F. 1988. Estimation of natural mortality in fish stocks: a review. Fish. Bull. 86:25—43. Wallace, J. R., and O. S. Hamel. 2009. Status and future prospects for the darkblotched rockfish resources in waters off Washington, Oregon, and California as updated in 2009. Status of the Pacific coast groundfish through 2009, stock assessment and fishery evaluation: stock assessment, STAR panel report, and rebuilding analysis, 179 p. [Available from Pacific Fishery Management Council, 7700 NE Ambassador Place, Suite 101, Portland, OR 97220-1384.] Wang, S. P, C. L. Sun, A. E. Punt, and S. Z. Yeh. 2006. Application of the sex-specific age-structured assessment method for swordfish, Xiphias gladius, in the North Pacific Ocean. Fish. Res. 84:282-300. Worm, B., R. Hilborn, J. K. Baum, T. A. Branch, J. S. Collie, C. Costello, M. J. Fogarty, E. A. Fulton, J. A. Hutchings, S. Jennings, O. P. Jensen, H. K. Lotze, P. M. Mace, T. R. McCla- nahan, C. Minto, S. R. Palumbi, A. M. Parma, D. Ricard, A. A. Rosenberg, R. Watson, and D. Zeller. 2009. Rebuilding global fisheries. Science 325:578-585. Yin, Y., and D. B. Sampson. 2004. Bias and precision of estimates from an age-struc- tured stock assessment program in relation to stock and data characteristics. N. Am. J. Fish. Manag. 24: 865-879. Appendix: Description of simulation model and data errors The population is age-structured and is assumed to be subject to one fishery with constant selectivity over years. There are two survey indices. Recruits vary over years and there are sampling errors in surveys, catches, and age-composition data. where R0 = initial (virgin) recruitment; amin = age of recruitment (minimum age in model); amax = maximum age, including age-plus groups; and Mx a = natural mortality for sex x, at age a, which is constant across years, and can be con- stant or vary by age, depending on the model setup. Initial condition and cohort growth Population dynamics Initial conditions of the population are numbers of fish at sex x, at age a, and at the first model year (y=0), which is given by the equation N x,0 ,a [ 0.5R0 |-^x,0,a-l -Mr „ e if a — a. ifamin 2). The standard deviation ( a C CD E c n c n CD c n c n 03 C 'sz I 0.50 - 0.40 - 0.30 - 0.20 - 0.10 - shortspine thornyhead o o.oo -I 1 1 1 1 1 1 i 0.00 0.20 0.40 0.60 0.80 1.00 1.20 1.40 Among-assessment CV Figure 4 Relationship between the coefficients of variation (CV) calculated from biomass variation over multiple full stock assessments (x axis) and the CV based on the measurement error of the most recent analysis ( y axis). 5 PFMC and NMFS. 2010. Proposed harvest specifications and management measures for the 2011-2012 Pacific Coast groundfish fishery and Amendment 16-5 to the Pacific Coast Groundfish Fishery Management Plan to update existing rebuilding plans and adopt a rebuilding plan for petrale sole: Draft environmental impact statement including Regulatory Impact Review and Initial Regulatory Flexibility Analy- sis. Pacific Fishery Management Council, Portland, OR (submitted to NOAA Fisheries Service), June 2010. 228 Fishery Bulletin 109(2) Residual Figure 5 Composite distributions of log-deviations from the mean, pooled for four meta- analytic groupings (rockfish, roundfish, flatfish, and coastal pelagic species). remains an active field of research. This fact is reflected by the range and changes over time in the ways that risk and uncertainty have been represented in fisheries assessments (see reviews by Francis and Shotton, 1997, and Patterson et ah, 2001). Our characterization of uncertainty does not include variability attributable to sources other than terminal- year biomass, which would tend to lead to negatively biased estimates of scientific uncertainty. Procedures for incorporating forecast uncertainty have been de- veloped (Shertzer et al., 2008) and could be blended with our approach. Likewise, there is fertile ground to be explored with respect to uncertainty in FMSY. Dorn (2002), for example, has developed a Bayesian prior for rockfish spawner-recruit steepness (h) that expresses uncertainty in estimates of stock productivity (see also Brooks et ah, 2010). Because steepness maps almost directly onto FMSY over a diverse range of groundfish life history patterns (Punt et al., 2008), a distribution of fishing mortality rates could be developed by math- ematical composition of these functions, conditioned on the form of the stock-recruitment relationship. We assumed that estimates of FMSY have negligible error and, as a consequence, uncertainty in OFL arises only from uncertainty in biomass estimates — an obvious simplification. Although we elected to characterize uncertain- ty by analyzing variability in biomass estimates from historical stock assessments, an alternative approach might be to use decision table results, which are a required element in groundfish stock assessments. Specifically, the PFMC terms of ref- erence for groundfish assessments2 require the development of a decision table for use in charac- terizing uncertainly in stock assessments. The guidance states the following: Once a base model has been bracketed on either side by alternative model scenarios, which cap- ture the overall degree of uncertainty within the assessment, a 2-way decision table analysis (states-of-nature versus management action) is the preferred way to present the repercussions of uncertainty to management. An attempt should be made to develop alternative model scenarios such that the base model is considered twice as likely as the alternative models, i.e., the ratio of probabilities should be 25:50:25 for the low stock size alternative, the base model, and the high stock size alternative. Ralston et al.: A meta-analytic approach to quantifying scientific uncertainty in stock assessments 229 It is therefore possible, in theory, to express uncer- tainty regarding biomass in a quantitative manner by appropriately weighting different states of nature presented in groundfish decision tables, which are derived through the collective expert opinion of the analytical team, the review panel, and the Scientific and Statistical Committee. A preliminary analysis of this approach has been completed, although a comprehensive analysis was not possible because of incomplete data in stock assessment documents. In particular, statistical weights for all three states of nature that are defined in the decision analysis (low, base, and high) have not always been explicitly expressed. When a characterization of the relative probabilities of the various states of nature under consideration is lacking, decision tables provide a type of risk assessment, but they are inadequate for risk management ( sensu Francis and Shotton, 1997). Still, in three of the nine cases examined, variances from decision tables were greater than a CV=37%. We view these preliminary findings as promising and recommend that a thorough analysis of statisti- cally weighted states of nature be considered as an alternative approach to characterization of scientific uncertainty in groundfish stock assessments. Figure 7 Relationship between the probability of overfishing (P*) and an appropriate buffer between the allowable biologi- cal catch (ABC) and the overfishing level (OFL), based on varying amounts of uncertainty ( <7=0.36, 0.72, and 1.44) assigned to different stock assessment tiers (l=data-rich, 2 = data-moderate, and 3 = data-poor), respectively. Conclusions Present and future management approaches for setting catch limits This analysis was prepared in response to a pressing management need that arose from the requirements of the reauthorized MSA to implement ACLs by 2011. It is revealing to consider the ultimate impact of accounting for scientific uncertainty when setting catch limits at the PFMC. For all assessed groundfish species Table 4 provides the ABC and ACL as a percentage of the esti- mated Fmsy harvest level (OFL) as they were adopted by the PFMC in June 2010 (see footnote 5). Note that groundfish stocks are classified into three tiers based on the amount and quality of the information that is avail- able for assessment modeling: tier-1 stocks are those for which there is data-rich information; tier-2 stocks are those for which there is data-moderate information; and tier-3 stocks are those for which there is data-poor infor- mation. Moreover, there was a consensus that scientific uncertainty cannot be lower for stocks that are more data limited. Hence, because o=0.36 was derived from 81 tier-1 stock assessments, the Scientific and Statisti- cal Committee recommended, and the PFMC elected to adopt, proxy estimates of uncertainty equal to double (0.72) and quadruple (1.44) a for tier-2 and tier-3 stocks, respectively. This framework then provided a basis for separate ABC control rules for each tier. The PFMC then elected to adopt a P*=0.45 for all tier-1 stocks and, with certain exceptions, P*= 0.40 for tier-2 and tier-3 stocks. Hence the scientific uncertainty buffers for the Council’s data-rich stocks amounted to setting the ABC 4% below the OFL. Similarly, the adjustment for tier-2 stocks (<7=0.72 and P*=0.40) was a 17% reduction (ABC = 83% of the OFL) (Fig. 7). The differences between ACLs and ABCs shown in Table 4 reflect a variety of other factors, including 1) requirements for rebuilding overfished stocks; 2) harvest control rules for prevent- ing stocks from becoming overfished; 3) socioeconomic considerations; 4) bycatch concerns for depleted species; 5) ecological considerations, and other factors. Parsing scientific uncertainty in estimates of OFL from these other considerations was a particular chal- lenge for the PFMC. Before the implementation of the new harvest specification framework recommended in the revised National Standard Guidelines, which were compelled by the reauthorized MSA, scientific and man- agement uncertainties were considered jointly in set- ting optimum yields below the MSY harvest level. We have shown that quantifying scientific uncertainty in estimating exploitable biomass across multiple assess- ments through meta-analysis is a reasonable first ap- proximation for explicitly accounting for uncertainty in preventing overfishing. Although all sources of error may not be have been considered with this approach, it was a helpful first step in the PFMC process. Im- portantly, with this approach the role of the Scientific and Statistical Committee in quantifying scientific un- certainty (by determining o, a purely technical issue) and the role of the PFMC in deciding a preferred level of risk aversion to overfishing (by choosing P*, which is a policy decision), are both duly respected. Coupling these two independent actions will help determine the ABC harvest level in a manner that is responsive to the mandates of the reauthorized MSA. 230 Fishery Bulletin 109(2) Table 4 Relative reductions from the overfishing limit (OFL) due to accounting for scientific and management uncertainty in setting 2011 groundfish allowable biological catches (ABCs) and annual catch limits (ACLs) at the Pacific Fishery Management Council (stocks in bold are overfished and their ACLs are based on rebuilding analyses). Tier l=data rich; tier 2=data moderate; tier 3= data poor. Stock/ Complex Tier ABC -r OFL ACL -r OFL Bocaccio ( Sebastes paucispinis) 1 96% 36% Canary rockfish ( Sebastes pinniger ) 1 96% 17% Cowcod ( Sebastes levis) 2/3 76% 30% Darkblotched rockfish ( Sebastes crameri) 1 96% 59% Pacific ocean perch ( Sebastes alutus ) 1 96% 18% Widow rockfish ( Sebastes entomelas) 1 96% 12% Yelloweye rockfish ( Sebastes ruberrimus ) 1 96% 42% Petrale sole (Eopsetta jordani) 1 96% 96% Lingcod (OR & WA) (Ophiodon elongatus) 1 96% 96% Lingcod (CA) 2 83% 83% Pacific cod (Gadus rnacrocephalus ) 3 69% 50% Sablefish (Anoplopoma fimbria) 1 96% 77% Shortbelly rockfish (Sebastes jordani) 2 83% 1% Chilipepper (S 40°10') (Sebastes goodei) 1 96% 96% Splitnose rockfish (S 40°10') (Sebastes diploproa) 1 96% 96% Yellowtail rockfish (N 40°10') (Sebastes flavidus) 1 96% 96% Shortspine thornyhead ( Sebastolobus alascanus ) 1 96% 83% Longspine thornyhead (Sebastolobus altivelis ) 2 83% 70% Black rockfish (WA) (Sebastes melanops) 1 96% 96% Black rockfish (OR-CA) 1 96% 82% California scorpionfish (Scorpaena guttata) 1 96% 96% Cabezon (CA) (Scorpaenichthys marmoratus ) 1 96% 96% Cabezon (OR) 1 96% 96% Dover sole (Microstomus pacificus ) 1 96% 56% English sole (Parophrys vetulus) 1 96% 96% Arrowtooth flounder (Atheresthes stomias) 2 83% 83% Starry flounder (Platichthys stellatus) 2 83% 75% Longnose skate (Raja rhina ) 1 96% 43% Minor Nearshore rockfish North (species complex) 3 85% 85% Minor Shelf rockfish North (species complex) 3 88% 44% Minor Slope rockfish North (species complex) 3 91% 79% Minor Nearshore rockfish South (species complex) 3 87% 87% Minor Shelf rockfish South (species complex) 3 84% 32% Minor Slope rockfish South (species complex) 3 92% 69% Other Flatfish (species complex) 3 69% 48% Other Fish (various) (species complex) 3 69% 50% Acknowledgments This study was generated, reviewed, and endorsed by members of the Pacific Fishery Management Coun- cil’s Scientific and Statistical Committee, whom we would like to thank and acknowledge for their critical suggestions. In particular, members of the groundfish and coastal pelagic species subcommittees were most engaged in developing the analytical approach, including T. Barnes, L. Botsford, M. Dorn, S. Heppell, T. Jagielo, T. Tsou, and V. Wespestad. In addition, a number of people mined the primary PFMC stock assessment literature and provided us with a stock-specific time series of population abundance for use in the meta-analysis. In particular, we would like to offer our thanks to J. Cope (cabezon), P. Crone (Pacific mackerel), J. Field (bocaccio), M. Haltuch (petrale sole), X. He (widow rockfish), K. Hill (Pacific sardine), I. Stewart (canary rockfish, Pacific whiting, and yelloweye rockfish), and J. Wallace (yel- lowtail rockfish). Lastly, we appreciate the constructive reviews of this work that were provided by J. Cope, E.J. Dick, and A. MacCall. We also acknowledge the helpful comments of two anonymous reviewers and E. Williams who evaluated the manuscript for the journal. Ralston et at: A meta-analytic approach to quantifying scientific uncertainty in stock assessments 231 Literature cited Brandon, J., and P. R. Wade. 2006. Assessment of the Bering-Chukchi-Beaufort Seas stock of bowhead whales using Bayesian model averag- ing. J. Cet. Res. Manag. 8:225-239. Brodziak, J., and K. Piner. 2010. Model averaging and probable status of North Pacific striped marlin, Tetrapturus audax. Can. J. Fish. Aquat. Sci. 67:793-805. Brooks, E. N., J. E. Powers, and E. Cortes. 2010. Analytical reference points for age-structured models: application to data-poor fisheries. ICES J. Mar. Sci. 67:165-175. Caddy, J. F., and R. McGarvey. 1996. Targets or limits for management of fisheries? N. Am. J. Fish. Manag. 16:479-487. Cochran, W. G. 1977. Sampling techniques, 3rd ed., 429 p. John Wiley & Sons, New York. Dorn, M. W. 2002. Advice on west coast rockfish harvest rates from Bayesian meta-analysis of stock-recruit relation- ships. N. Am. J. Fish. Manag. 22:280-300. FAO (United Nations Food and Agriculture Organization). 1996. Precautionary approach to capture fisheries and species introductions. Technical guidelines for respon- sible fisheries, no. 2, 54 p. FAO, Rome. Federal Register. 2009. Magnuson-Stevens Act Provisions; Annual Catch Limits; National Standard Guidelines; final rule, vol. 74, no. 11, January 16, p. 3178-3213. GPO, Wash- ington, D.C. Francis, R. I. C. C., and R. Shotton. 1997. “Risk” in fisheries management: a review. Can. J. Fish. Aquat. Sci. 54:1699-1715. Hilborn, R., and C. J. Walters. 1992. Quantitative fisheries stock assessment — choice, dynamics, and uncertainty, 570 p. Kluwer Academic Pubis., Boston. IWC (International Whaling Commission). 1999. The report of the Scientific Committee, annex N. The revised management procedure (RMP) for baleen whales. J. Cet. Res. Manag. 1 (suppl.):251-258. Kolody, D., T. Polacheck, M. Basson, and C. Davies. 2008. Salvaged pearls: lessons learned from a flounder- ing attempt to develop a management procedure for southern bluefin tuna. Fish. Res. 94:339-350. Johnson, N. L., and S. Kotz. 1970. Continuous univariate distributions, part 1, 300 p. Distributions in statistics. John Wiley & Sons, New York. Methot, R. D. 2000. Technical description of the stock synthesis assess- ment program. NOAA Tech. Memo. NMFS-NWFSC-43, 46 p. Patterson, K., J. Cook, C. Darby, S. Gavaris, L. Kell, P. Lewy, B. Mesnil, A. Punt, V. Restrepo, D. W. Skagen, and G. Stefansson. 2001. Estimating uncertainty in fish stock assessment and forecasting. Fish Fish. 2:125-157. Pawitan, Y. 2001. In all likelihood — statistical modelling and infer- ence using likelihood, 528 p. Oxford Univ. Press, Oxford. PFMC (Pacific Fishery Management Council). 2010. FMP amendment addresses National Standard 1 — annual catch limits, accountability measures. Pacific Council News 34(1):5. [Available at: http://www.pcoun- cil.org/wp-content/uploads/Spring_2010_Newsletter. pdf.) Poole, D., G. H. Givens, and A. E. Raftery. 1997. A proposed stock assessment method and its appli- cation to bowhead whales, Balaena mysticetus. Fish. Bull. 97:144-152. Prager, M. H., C. E. Porch, K. W. Shertzer, and J. F. Caddy. 2003. Targets and limits for management of fisheries: a simple probability-based approach. N. Am. J. Fish. Manag. 23:349-361. Pribac, F., A. E. Punt, B. L. Taylor and T. I. Walker. 2005. Using length, age and tagging data in a stock assessment of a length selective fishery for gummy shark (Mustelus antarcticus). J. Northwest Atl. Fish. Sci. 35L:267-290. Punt, A. E., and. G. Donovan. 2007. Developing management procedures that are robust to uncertainty: lessons from the International Whaling Commission. ICES J. Mar. Sci. 64:603-612. Punt, A. E., M. W. Dorn, and M. A. Haltuch. 2008. Evaluation of threshold management strategies for groundfish off the U.S. west coast. Fish. Res. 94:251-266. Punt, A. E., and R. Hilborn. 1997. Fisheries stock assessment and decision analy- sis: the Bayesian approach. Rev. Fish Biol. Fish. 7:35-63. Quinn, T. J., II, and R. B. Deriso. 1999. Quantitative fish dynamics, 524 p. Oxford Univ. Press, New York. Ralston, S. 2002. West coast groundfish harvest policy. N. Am. J. Fish. Manag. 22:249-250. Restrepo, V. R., J. M. Hoenig, J. E. Powers, J. W. Baird, and S. C. Turner. 1992. A simple simulation approach to risk and cost analysis, with applications to swordfish and cod fisheries. Fish. Bull. 90:736-748. Shertzer, K. W., M. H. Prager, and E. H. Williams. 2008. A probability-based approach to setting annual catch levels. Fish. Bull. 106:225—232. Stewart, I. J., and O. S. Hamel. 2010. Stock assessment of Pacific hake, Merluccius pro- ductus, (a.k.a. whiting) in U. S. and Canadian waters in 2010, 290 p. [Available from Pacific Fishery Manage- ment Council, 7700 NE Ambassador Place, Suite 101, Portland, OR, 97220-1384.1 Wade. P. R. 1998. Calculating limits to the allowable human-caused mortality of cetaceans and pinnipeds. Mar. Mamm. Sci. 14:1-37. 232 Abstract — Distribution and demo- graphics of the hogfish ( Laclinolai - mus maximus) were investigated by using a combined approach of in situ observations and life history analyses. Presence, density, size, age, and size and age at sex change all varied with depth in the eastern Gulf of Mexico. Hogfish (64-774 mm fork length and 0-19 years old) were observed year- round and were most common over complex, natural hard bottom habi- tat. As depth increased, the presence and density of hogfish decreased, but mean size and age increased. Size at age was smaller nearshore (<30 m). Length and age at sex change of nearshore hogfish were half those of offshore hogfish and were coinci- dent with the minimum legal size limit. Fishing pressure is presum- ably greater nearshore and presents a confounding source of increased mortality; however, a strong red tide occurred the year before this study began and likely also affected near- shore demographics. Nevertheless, these data indicate ontogenetic migra- tion and escapement of fast-growing fish to offshore habitat, both of which should reduce the likelihood of fish- ing-induced evolution. Data regarding the hogfish fishery are limited and regionally dependent, which has con- founded previous stock assessments; however, the spatially explicit vital rates reported herein can be applied to future monitoring efforts. Manuscript submitted 29 September 2010. Manuscript accepted 24 February 2011. Fish. Bull. 109:232-242. 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. Demographics by depth: spatially explicit life-history dynamics of a protogynous reef fish Angela B. Collins (contact author)1 Richard S. McBride2 Email address for contact author: angela.collins@myfwc.com 1 Florida Fish and Wildlife Conservation Commission Fish and Wildlife Research Institute 100 8th Avenue SE Saint Petersburg, Florida 33701 2 Northeast Fisheries Science Center National Marine Fisheries Service National Oceanic and Atmospheric Administration 166 Water Street Woods Hole, Massachusetts 02543 Protogynous species require special management considerations when fish- ing reduces the probability of survival to the male phase. Selective harvest- ing of males may skew the sex ratio and reduce the reproductive capacity of a population by increasing the prob- ability of sperm limitation (Hamilton et al., 2007). Also, selective removal of a particular sex or size class over many generations can have evolution- ary consequences, including slower growth rates, reduced size at matu- ration, and earlier sexual transfor- mation (Harris and McGovern, 1997; Adams et ah, 2000; Brule et al., 2003; Heppell et al., 2006). However, pro- togyny does not automatically imply elevated vulnerability to fishing if the population is able to compensate for reduced male survival (e.g., by earlier transition to the male phase). This ability to compensate is most likely to occur in species in which sex change is socially or environmentally mediated rather than constrained to a certain size or age (Alonzo and Mangel, 2005). Therefore, to predict stock dynam- ics and a species’ response to fishing pressure, it is important not only to establish whether sex change occurs, but also to quantify the mechanisms that influence sex change and charac- terize the related demographics. We synthesized data from in situ observations and life history collec- tions to evaluate factors that could potentially influence the presence, density, and demographics of a reef fish. The hogfish (Labridae: Lachno- lairnus maximus ), which occurs from temperate to tropical waters of the western North Atlantic Ocean, Gulf of Mexico, and Caribbean Sea, was chosen for this study for several rea- sons. It is an economically important reef fish (for a list of total U.S. fish- ery landings and their estimated val- ues see: www.st.nmfs.noaa.gov/stl/ commercial/index. html, accessed Feb- ruary 2011), and a better understand- ing of its ecology will assist manag- ers in evaluating regulatory options. The principal fishing method for this species is spearfishing (McBride and Richardson, 2007), which presents an opportunity to evaluate the effect of a single fishery sector with fewer confounding effects from other fish- ery sectors (e.g., hook-and-line). Hog- fish can exceed 800 mm fork length (FL), weigh more than 10 kg, and live as long as 23 years (McBride and Richardson, 2007). These life- history characteristics allow a wide latitude for measuring differences in size and age. Finally, they are mo- nandric, protogynous hermaphrodites (all fish begin life as female and can eventually change sex to male) (Mc- Bride and Johnson, 2007) that form harems, with a single male control- ling 2-15 females (Davis, 1976; Colin, 1982; Claro et al., 1989). This mat- ing system allowed for investigation of the effects of fishing, habitat, and other environmental variables on sex change and social structure. Collins and McBride: Spatially explicit life-history dynamics of a protogynous reef fish 233 Figure 1 Study location in the central eastern Gulf of Mexico. Dive sites are indicated by dots (431 sites) and were surveyed for hogfish ( Lachnolaimus maximus) between 2005 and 2007. Hogfish were harvested randomly from dive sites during scuba surveys. Bathymetry contours are isobaths and are labeled to 100 m; the 30-m isobath is bold and separates nearshore (<30 m) from offshore (>30 m) sites. Fishery regulations for hogfish were first implemented in 1994. The minimum size limit (305 mm FL) corresponded with the mini- mum length at sex change (Da- vis, 1976) and was established to protect spawning fish. However, concerns about the effectiveness of this size limit emerged when further research demonstrated that median size at sex change was significantly larger (-380 mm FL; McBride et ah, 2008). Continual removal of the domi- nant male can impact the repro- ductive capacity of a population (Bannerot et al., 1987; Sluka and Sullivan, 1998). Under heavy fishing pressure, constant disrup- tion of hogfish spawning harems could be problematic because sev- eral months are required to com- plete sex change (McBride and Johnson, 2007) and new males have lower reproductive success (Munoz et al., 2010). A stock as- sessment in 2003 (Ault et al.1) stated that hogfish were under- going overfishing in the U.S., but these findings were disputed because of concerns that catch and effort data were inadequate (Kingsley2). Under such condi- tions, demographic data may provide the only basis for setting management refer- ence points (Brooks et al., 2010) and evaluating future monitoring strategies. Data were collected through cooperation with the spearfishing community, and revealed abrupt, cross- shelf patterns in hogfish demographics. These findings highlight interactions between fishing operations and the environment on reef fish populations, specifically demonstrating that sex change mechanisms can be spatially explicit and that refuges may exist for larger spawners that survive long enough to reach offshore habitats. 1 Ault, J. S., S. G. Smith, G. A. Diaz, and E. Frank- lin. 2003. Florida hogfish fishery stock assessment. Final report to Florida Fish & Wildlife Conservation Commission, 89 p. [Available from NOAA Southeast Fisheries Science Center, www.sefsc.noaa.gov/sedar/download/SEDAR6_RW4. pdf?id=DOCUMENT, accessed February 2011.] 2 Kingsley, M. C. S., ed. 2004. The hogfish in Florida: Assessment review and advisory report. Southeast data and assessment review, 15 p. Prepared for the South Atlantic Fishery Management Council, the Gulf of Mexico Fishery Management Council, and the National Marine Fisheries Service. [Available at: http://www.sefsc.noaa.gov/sedar/ download/SEDAR6_SAR2_hogfishall.pdf?id= DOCUMENT, accessed February 2011.] Materials and methods Sampling design Visual observations and hogfish collections were made during scuba dives (to a depth of 69 m) in the central eastern Gulf of Mexico (Fig. 1). To investigate whether increasing depth and distance from shore affected hog- fish distribution and demographics, scuba surveys were allocated to sample a range of depths and were catego- rized as nearshore (<30 m) or offshore (>30 m). Thirty meters was chosen as the dividing point between the nearshore and offshore classification because many rec- reational divers do not exceed this depth on account of the reduced available bottom time and greater hazards associated with diving at deeper depths. Additionally, this 30-m depth corresponds roughly with a distance of 40-50 km from land, beyond which travel becomes more costly in terms of travel time, fuel expense, and risks associated with adverse weather. Sites were also exam- ined by 10 -m depth intervals to identify whether there were finer scale effects of depth on hogfish distribution. Habitat was characterized into one of three major categories according to bottom type and relief: 1) natu- ral habitat of rugose hard bottom with a maximum vertical relief >0.5 m, typically limestone outcroppings 234 Fishery Bulletin 109(2) or ledges; 2) natural habitat of flat hard bottom with low-relief (<0.5 m), typically limestone outcroppings and shallow potholes; or 3) artificial habitat, which was pri- marily shipwrecks but also included other non-natural structures (e.g., bridge pilings, building debris). Other habitats (seagrass, plain sand, or mud bottom) were uncommon and were grouped together. Three to nine research trips were conducted monthly through all seasons: winter ( January-March), spring (April- June), summer (July-September), and fall (Oc- tober-December). Sampling effort was focused on ru- gose hard bottom as recommended by veteran divers with knowledge of hogfish distribution in the study area, and as indicated in published reports regarding hogfish ecology (Davis, 1976; Colin, 1982). Remaining habitats were systematically surveyed less often, mainly to confirm the expectations that hogfish occurred there less frequently or in lower abundance. Attempts were made to visit sites representative of each combination of habitat type and depth category at least quarterly. Research dives Hogfish are in general unwary of divers (Davis, 1976; Colin, 1982) and typically remain in an area when divers are present (senior author, personal observ.) — a characteristic that makes this species a good candidate for visual survey techniques (Jennings et ah, 2001). Underwater observations using scuba were performed to record the presence, density, size distribution, and sex ratio of hogfish. During each dive, a single observer (A. Collins) swam the length of a straight line 50-m transect three con- secutive times. Transects were placed at the observer’s discretion to maximize the length of the transect over the targeted habitat type (typically rugose hard bottom, where transects were laid in a straight line on top of the ledge). The observer waited at least one minute between setting the transect line and beginning the survey. Additionally, the observer waited one minute be- tween the end of one replicate and the beginning of the next. During each replicate, the total number, size, and sex of hogfish observed within 3 m of the line were re- corded (survey band=6x50 m, or 300 m2). The greatest number of fish recorded during a single replicate was used to calculate hogfish density in the transect area. Hogfish are dichromatic and dimorphic (McBride and Johnson, 2007). This attribute typically allowed visual identification of the sex of each fish. Fish were catego- rized as male, female, or, if sex was not obvious, sex un- known. Sex ratio (number of males divided by number of females) was calculated for each transect. The four cases in which a fish was designated as unknown were not included in the calculation of sex ratio. Maximum, minimum, and mean sizes of hogfish observed during each site visit were based on visual survey data (esti- mated FL, cm) as well as on harvested hogfish (mea- sured FL, mm). Hogfish harvested from the survey area were identified during the survey and were measured only once. Horizontal visibility was assessed by the observer during the survey. If visibility was less than 3 m, or if the site was too deep (>45 m) to allow for transect replicates, only data on fish presence were considered in further analyses (i.e., sex ratio and density were not calculated for these dives). The binary relationship between hogfish presence (vs. absence) and habitat, depth, and season were investi- gated by using a general linear mixed model (GLIM- MIX, SAS, vers. 9.1, SAS Inst., Cary, NC), and presence was modeled by using a binary distribution. General linear models (GLM and GLIMMIX) were also used to test for the effects of habitat, depth, and season upon each of the following variables: hogfish density, size, and sex ratio. Density was modeled with a Poisson distribution. Life history Hogfish were typically harvested from dive sites in accordance with fishing regulations; therefore most speared fish were greater than 305 mm FL. However, an effort was made to sample a number of small, suble- gal-size fish during each season of the year. Harvested fish were otherwise randomly chosen throughout the dive. Length (FL, mm) and whole body weight (BW, to the nearest 0.25 kg) were measured for all harvested fish. Gonads were excised immediately after the diver surfaced, were wrapped in plastic, and stored on ice until they could be returned to the laboratory. Within 24 hours, gonads were weighed to the nearest 0.01 g, and a section of tissue approximately 1 cm long was removed from the middle of each gonad and placed in formalin. Histological processing followed the pro- cedures described in McBride and Johnson (2007). Slides were examined (100-200x magnification) at least twice by an individual reader to identify repro- ductive class. Reproductive class was assigned according to the method of McBride and Johnson (2007). Briefly, the most advanced oocyte stage or evidence of previous spawning (i.e., atretic advanced stage oocytes) were used to designate females as immature, mature rest- ing, mature active, or postspawning (classes 1-4, re- spectively). Transitional-stage fish (class 5) were iden- tified by the presence of seminiferous crypts along the boundary of the tunica. Males were classified by the dominant stage of spermatogenesis, the nature of the germinal epithelium, and the connection and size of sperm ducts and were designated as immature, mature inactive, ripening mature, ripe mature, or postspawn- ing (classes 6-10, respectively). Fish were aged by examining sectioned otoliths (sagittae). Age was independently assessed by two individual readers following the methods and cri- teria outlined in McBride and Richardson (2007). Growth was modeled with the von Bertalanffy growth equation: FL = LJ1 - *<>]>), Collins and McBride: Spatially explicit life-history dynamics of a protogynous reef fish 235 where = asymptotic fork length K = the Brody growth coefficient; and t0 = the predicted age at which fish length is equal to zero. Growth was modeled for the entire sample, as well as independently by depth category (nearshore vs. offshore). To test for effects of fish age and depth on fish size, a 2-way analysis of variance (ANOVA) was used to com- pare size at age for age classes common to both depth categories (ages 3-6 yr). Size and age at female matu- rity and sex change were calculated with the logistic curve (binary logit model): PMt = ea + bX/l + ea + bA where PMt is the probability of maturity at a particular age or length class; a and b = constants; and X is either length or age. Size or age at 50% maturity = |a/b|. Model structure and fitting followed Allison (1999). Size and age at first maturity (i.e., class 1 vs. classes 2-4) and at sex change (i.e, classes 1-4 vs. classes 5-10) were modeled for each depth category, as well as for the aggregate sample. Additional otoliths and gonads were collected oppor- tunistically through spearfishing tournaments, trawl research cruises (Fisheries-Independent Monitoring Program of the Fish and Wildlife Research Institute), and independent diver donations. Fish were used for life history analyses only if the location and depth at capture within the central eastern Gulf of Mexico could be verified. Results Research dives Hogfish presence was significantly related to habitat and depth. Fish were recorded most often and in high- est densities nearshore over rugose hard bottom (Fig. 2A). Hogfish were present during 74% of all surveys (318/431) and were observed during all months of the year throughout the sampled depths and major habitat types (Tables 1 and 2). Hogfish density was greater near- shore (range 0-25; mean=5.4) than offshore (range 0-15; mean=1.3) during all seasons, and highest densities were recorded during summer (Fig. 2B). No significant relationship between presence and season was detected, nor was there a significant interaction between habitat and depth or season and depth (Table 2). Hogfish observed during research dives nearshore were half the size of those offshore (nearshore mean=24 cm FL [range: 6—56 cm, n = 1352]; offshore mean=51 cm FL [range: 18-77 cm, n=296]). Nearshore hogfish were larger in summer than in winter (P= 0.0029), and offshore hogfish were larger in spring than in fall 9 A $ ^ CD 7 - 6 - 5 - 4 - 3 - 2 - 1 • • offshore o nearshore i O o CO B CD E o CD CD Artificial 3 - 2 - Flat Habitat type Rugose 800 700 E 600 ■ E £ 500 ' CT> c ® 400 - Ad ° 300 - c 03 4; 200 - c f 100 - 0 1 winter * f spring summer i i fall Figure 2 Geometric mean density of hogfish (Lachnolaimus maximus) recorded during visual transects (50x6 m bands with replication) by (A) habitat type and ( B) season. (C) Mean fork length for hogfish observed over all seasons during all research dives. Depth cat- egories were classified as nearshore (<30 m depth; open circles) or offshore ( >30 m depth; filled circles), and error bars represent 95% confidence limits. (P=0.0141), but otherwise, no significant relationship was detected between fish size and season (Fig. 2C). Although density decreased with depth (PcO.OOOl; Fig. 3A), FL exhibited a positive relationship with depth (PcO.OOOl; Fig. 3B). Males were larger than fe- males within each depth category (P<0.0001), but both sexes were larger offshore than nearshore (PcO.OOOl; Fig. 3B). Within depth categories, further analysis by 236 Fishery Bulletin 109(2) Table 1 Number of dives, visual transects, and hogfish (Lachnolaimus maximus) sampled (August 2005-August 2007) at nearshore (<30 m) and offshore (>30 m) sites. Visual transects (research dives where replicates could be completed and visibility was >3 m) are indicated by habitat type as artificial (A), flat hard bottom (F), rugose hard bottom (R). The number of transects (No. of tran- sects) during which at least one hogfish was observed (present) and the total number of transects performed (total) are listed for each month. Only seven dives were performed over other habitat (O); therefore this category was excluded from further analyses. Survey samples were harvested during research dives. Additional fish (included in the number of total fish sampled) were col- lected during spearfishing tournaments, trawl cruises or through private donations, n = number of fish sampled. No. of transects (present/total) No. of dives ti Total n Month (near/offshore) Total A F R O ( survey) (near/offshore) Jan 38 (25/13) 29 2/2 3/5 19/20 0/2 46 46 (25/21) Feb 32 (16/16) 23 0/1 0/1 10/18 0/3 28 28 (11/17) Mar 56(55/1) 47 7/11 2/5 28/31 0 31 36 (32/4) Apr 63 (30/33) 32 0/2 2/5 22/25 0 65 110(52/57) May 34 (30/4) 26 0/2 1/3 21/21 0 26 75 (34/41) Jun 37 (23/14) 22 2/3 0/1 17/18 0 25 33 (6/14) Jul 27 (10/17) 14 2/3 0/1 10/10 0 25 29 (5/24) Aug 21 (8/13) 10 0 1/2 7/8 0 21 115 (14/63) Sep 20 (16/4) 18 0 1/2 14/16 0 16 38(13/25) Oct 31 (16/15) 17 2/4 3/5 8/8 0 44 80 (47/33) Nov 31 (23/8) 22 3/9 0/3 8/8 0/2 24 27 (11/16) Dec 41 (12/29) 23 3/7 1/2 11/14 0 35 36 (14/22) Total 431(264/167) 283 21/44 14/35 175/197 0/7 386 653 (264/337) Table 2 Relationship of hogfish ( Lachnolaimus maximus ) presence and density to habitat type, depth zone, and season (main effects), as well as the interaction effects between habitat type and depth zone. Hogfish were considered present if at least one individual was observed. Surveys where hogfish were present and the total survey number are indicated in parentheses (no. of surveys pres- ent/no. of surveys performed). Hogfish presence and density were significantly related to habitat and depth, and they were most common and abundant on shallow, rugose habitat. There were no significant seasonal effects on hogfish presence or density, or interactions between depth and habitat or season. LSM indicates least squares means. Hogfish presence Hogfish density P>F F LSM P>\t\ P>F F LSM P>|/| Habitat <0.0001* 32.38 <0.0001* 13.40 Artificial (23/55) 0.3943 0.1797 0.9641 0.9003 Flat (16/43) 0.3248 0.0606 0.9847 0.9682 Rugose (278/324) 0.8734 <0.0001 3.5074 <0.0001 Depth zone <0.0001* 8.7 <0.0001* 18.46 Deep (112/166) 0.4284 0.3376 0.7591 0.3607 Shallow (205/256) 0.6904 <0.0001 2.9395 <0.0001 Season 0.6439 0.56 0.2998 1.23 Fall (66/101) 0.5285 0.671 1.5843 0.0125 Spring (106/133) 0.5741 0.2902 1.2634 0.2131 Summer (53/68) 0.6387 0.1101 1.7192 0.0084 Winter (92/120) 0.5111 0.8787 1.4467 0.0424 Depth zonexhabitat 0.4968 0.7 0.3469 1.06 Depth zonexseason 0.1488 1.79 0.0659 2.44 Collins and McBride: Spatially explicit life-history dynamics of a protogynous reef fish 237 o o CO <® offshore (transect n= 54) O nearshore (transect n- 229) * f • Male (n=92) B v Transitional (n=61) O Female (n= 342) • l i - 5 v * _ 5 O £ I i s ^ 5 ° - <10 10-19 20-29 30-39 Depth (m) Figure 3 Mean observed densities and fork lengths (FL) for all sex phases of hogfish ( Lachnolaimus maximus) as observed over increasing depths (by 10-m intervals). (A) Densities recorded during visual transects. Calculations were not performed for sites deeper than 45 m because of limited survey time, and are designated by *. (B) Mean fork lengths for each sex phase, as determined by histological examination. The dotted line separates nearshore and offshore data. Error bars indicate standard error of the mean. 10-m intervals did not reveal significant differ- ences for density or size distribution (Fig. 3, A and B). Hogfish aggregations varied in number and sex ratio. Females were most common and were re- corded during 206 out of 283 transects (mean n = 6), whereas males were recorded during only 103 out of 283 transects (mean n= 1.5). As many as 25 individuals were recorded during a single tran- sect. The maximum number of females observed during a transect was 23, and the maximum number of males observed was 4. Occasionally, more than four males were noted at a site beyond the boundaries of the transect, but typically, if males were observed, it was more common to see only one or two during the survey. When both sexes were present (n = 94 transects), the larg- est fish observed were always males. Sex ratio (males:females) ranged from 0.0 to 1.0 (Fig. 4), with a mean of 0.14 (overall), 0.14 (nearshore), and 0.20 (offshore). Sex ratio showed no relationship to depth (P= 0.90) or season (P=0.99). Visual surveys were completed between Novem- ber 2005 and June 2007, when bottom tempera- ture, dissolved oxygen, and salinity were mea- sured within the following ranges: 15.7-31.2°C, 6. 0-9. 6 mg/L, and 29-36 PSU, respectively. Life history Life history analyses supported visual survey observations, with hogfish size and age positively related to depth. Ages were assigned to 622/653 fish (95%), and ranged from 0 to 19 years old. Collection depth data were available for 92% of all harvested hogfish (601/653). Hogfish collected nearshore (71=264) ranged from 102 to 492 mm FL and from 0 to 8 years old; those from offshore (ti = 337) ranged from 319 to 774 mm FL and from 2 to 19 years old (Fig. 5). Fish at a common age were larger offshore, indicating that faster grow- ing fish occur in deeper water (Fig. 6). Reproductive classes were assigned to 472 aged individuals. As expected for a protogynous hermaphrodite, the majority of hogfish were female (classes 1-4; n = 342). The remaining individuals were classified as transitional or immature males (class 5 or 6, respectively; ti = 61) or mature males (classes 7-10; 72 = 92). Size and age at 50% maturity for females were 151.6 mm FL and 0.9 years. It is assumed that females completed maturation nearshore because immature females were not observed at depths >22 m. Females were smaller and younger nearshore (means: 246 mm FL, 2.3 yr; ti=159) than offshore (means: 479 mm FL, 6.7 yr; 72 = 161) (PcO.OOOl). Sex change oc- curred across a wide range of sizes (197-727 mm FL) and ages (1-11 yr) and was observed both nearshore and offshore. Median size and age at sex change were significantly less nearshore (327 mm FL; 2.8 yr; 72 = 15) than offshore (592 mm FL; 7.8 yr; 72 = 18) (PcO.OOOl) (Fig. 7). The smallest transitional fish collected off- shore was 449 mm FL. All fish >685 mm FL or older than 10.5 years were in the process of sex change or were already male. Discussion We identified distinct cross-shelf patterns in the presence and density of hogfish; both were greater nearshore. Hogfish were distributed widely, but not randomly. Across all depths sampled, their presence and density were greatest over complex, natural hard bottom habitats. In the Florida Keys, hogfish actively select habitat, preferring sandy rubble and gorgonian microhabitats (Munoz et al., 2010). 238 Fishery Bulletin 109(2) 100 80 60 offshore nearshore 40 - 1 0.0 n I I 0.4 0.6 Sex ratio (M:F) 08 1.0 Figure 4 Frequency distribution of hogfish ( Lachnolaimus maximus) sex ratio nearshore (<30 m depth; n= 229 transects) and offshore (30-45 m depth; n = 54 transects) recorded during visual tran- sects. A value of 0.0 indicates that no males were observed; a value of 1.0 indicates that only males were observed. Sex ratios were not calculated for sites >45 m due to limited survey time. The spatially explicit demographic patterns evident within this study were not detected in previous research in the eastern Gulf of Mexico, probably because the data were analyzed in ag- gregate from collections over a broad geographic area (McBride and Richardson, 2007; McBride et al. 2008). These new results reveal distinct demographic structure across the shelf. Near- shore, hogfish occurred in higher densities and were younger, smaller, and slower growing than those offshore. Moreover, fish changed sex at a smaller size and younger age nearshore — per- haps as a response to social cues that maintain harem structure and increase spawning success. Given these facts, the potential would be great for fishing-induced genetic shifts, except for the existence of larger, faster growing fish offshore. Potential mechanisms are evaluated in the follow- ing sections to synthesize these ecological findings and elucidate the resilience of these reef fishes to fishing and environmental factors. Cross-shelf dynamics Spatial variation in demographic parameters is not unusual for widely distributed reef fishes (Gust, 2004; DeVries, 2006; Allman, 2007; Lombardi- Carlson et al., 2008). It is likely that the underlying cross-shelf gradients of density and life history param- eters observed for hogfish reflect their bipartite life cycle. Hogfish are broadcast pair-spawners whose larvae are planktonic for 30-45 d (Colin, 1982) before settling in shallow inshore habitat such as seagrass beds (Roessler, 1965; Victor, 1986; Lindeman et al., 2000). Along the west coast of Florida, juvenile hogfish use as nursery areas Tampa Bay, Charlotte Harbor, and the shallow inshore waters off Tarpon Springs and the Big Bend region (McMichael, unpubl. data3). Ontogenetic migration offshore is suspected but is difficult to verify without tagging studies. Our research provides strong support for this hypothesis. Immature females were not collected from depths >22 m, and the youngest fish collected offshore (>30 m) was 2 years old, indicating that it takes at least two years to mi- grate from inshore settlement areas to offshore habitat. Although many reef fish have limited home ranges af- ter settlement (e.g., Williams et al., 1994), ontogenetic habitat shifts to deeper water are not uncommon (e.g., surgeonfish [Acanthurus chirurgus] and parrotfish [Sco- rns spp.], Nagelkerken et al., 2000; gag grouper [Myc- teroperca microlepis], Brule et al., 2003; gray snapper [Lutjanus griseus], Faunce and Serafy, 2007). It is likely that nearshore and offshore differences in maximum fish size and age were also, at least partially, related to the persistent, severe red tide ( Karenia bre- vis) that occurred off the west coast of Florida during 3 McMichael, Robert. 2011. Unpubl. data. Fish and Wildlife Research Institute, Fisheries-independent monitoring group, 100 8th Avenue SE, Saint Petersburg, Florida 33701. 2004-05, the year before this study began. Nearshore benthic communities in the study area suffered sig- nificant mortality during and following that red tide (Landsberg et al., 2009), when widespread fish kills and dead or reduced benthic fauna were reported in waters <30 m deep off Tampa Bay (Hu et al., 2006; Gannon et al., 2009). During the last red tide outbreak of similar severity (in 1971), hogfish died or were displaced from many reefs in 13-30 m (Smith, 1975) but recolonized the af- fected areas within 4-10 months (Smith, 1979). Smith did not report length data, and therefore it was not possible to identify whether the source of recovery was new recruits or transient fish from unaffected reefs. Our findings regarding nearshore demographics may partially reflect the recovery of the population in that area after a major (but uncommon) toxic event. Resiliency to localized environmental perturbations such as red tides is likely related to a species’ distribu- tion over a wide geographical range. The existence of large individuals in deep water offshore should provide a reservoir of spawning individuals to help replenish inshore areas (e.g., Simberloff, 1974). Although there were no reef-specific demographic data for the study area before the 2005 red tide, local divers recalled that the hogfish in shallow water were larger and more abundant before the toxic event. Additionally, greater numbers of relatively larger hogfish have been observed in shallow waters during research dives performed since the completion of this study (senior author, un- publ. data). The pronounced size and age truncation observed nearshore is also likely related to greater fishing mor- Collins and McBride: Spatially explicit life-history dynamics of a protogynous reef fish 239 tality associated with increased accessibility of fish to fishing vessels. Hogfish feed on slow mov- ing, benthic invertebrates (Randall and Warmke, 1967) and are less vulnerable to hook-and-line fishing methods than most other reef species in the region. Consequently, they are harvested pri- marily by spearfishing. Most recreational diving is done at depths <130 ft (40 m; PADI, 1999); at greater depths a diver’s bottom time is limited and restricted to divers with higher skill levels. Additionally, because deep sites are farther from shore, fuel expense and travel time are greater. Together, these factors potentially contribute to decreased fishing-induced mortality of hogfish offshore. Tupper and Rudd (2002) noted a similar pattern in the Caribbean, where larger hogfish were present in deeper and unfished areas. This pattern has also been observed for other species in the Gulf of Mexico. Gray triggerfish ( Balistes capriscus ) exhibit decreasing mortality with in- creasing distance from shore (Ingram, 2001), and vermillion snapper ( Rhomboplites aurorubens ) display a spatial size dichotomy that has been related to higher exploitation rates within waters closer to shore (Allman, 2007). Notably different patterns of sex change were observed for nearshore and offshore regions. In aggregate, sex change occurred over a broad range of ages and sizes (1-11 years and 197-727 mm FL), indicating that it is likely to be under social control (e.g., removal of the dominant male initiates sex change in a large female). The size advantage model predicts that sexual transition will occur at an earlier age in populations expe- riencing higher mortality (Warner, 1988), and it has often been observed that the continued removal of males results in reduced size at sex change and that size and age at the onset of sex change are lower in areas of greater fishing pres- sure (Warner 1975; Hawkins and Roberts, 2003; Hamilton et al., 2007). The smaller size and younger age of hogfish at sex change indicates shorter life spans and greater mortality in nearshore waters. In this study, median size at sex change nearshore (327 mm FL) just exceeded the legal minimum size (305 mm FL). These data indicate that many nearshore fe- males are changing sex within one year after reaching legal size, since hogfish take about one year to complete sex change (McBride and Johnson, 2007). The probabil- ity of moving offshore may be related to an individual’s growth rate because hogfish of the same age were larger offshore than nearshore. These faster-growing fish may have had greater energy reserves (perhaps by delaying sex change), allowing successful migration offshore. Alternately, resource (e.g., food, habitat) availability or another environmental factor may have allowed for faster growth within deeper habitat. The higher den- sities observed nearshore may result in an increased competition for food; however, a qualitative assessment of stomach fullness (stomach weight divided by total 0 2 4 6 8 10 12 14 16 18 20 22 Age (years) Figure 5 Hogfish (Lachnolaimus maximus ) fork length (FL) at age for (A) nearshore (<30 m), and (B) offshore (>30 m) collections. Gonad histology determined sexual classification as female (classes 1-4), transitional or immature male (classes 5-6), or mature male (classes 7—10). Estimated von Bertalanffy growth parameters: L^= 380.5 mm, K= 0.5614 and t0=-0.1619 (near- shore) and Lm= 896 mm, 7f=0.0940 and t0=-1.9752 (offshore). body weight) did not show any relationship with depth. A more quantitative assessment of prey availability and prey quality should be performed to address this question. It is possible that differences in life history traits could reflect genetically distinct populations. Although this scenario was considered unlikely (because of the absence of immature hogfish offshore), DNA samples were collected from a subsample of individuals from both depth ranges (n = 82; authors of this article, un- publ. data). Preliminary genetic analysis of microsatel- lite loci provided no evidence of separate stocks in our sampling area (Seyoum, unpubl. data4). The level of analysis available at this time cannot completely ex- clude the possibility, but it seems unlikely. 4 Seyoum, Seifu. 2011. Unpubl. data. Fish and Wildlife Research Institute, 100 8th Avenue SE, Saint Petersburg, Florida 33701. 240 Fishery Bulletin 109(2) Spawning harems and management Mature hogfish form harems; isolated males are some- times observed but females tend to occur in pairs or groups (Davis, 1976; Colin, 1982; this study). Previous 550 r 500 - 450 - 350 • offshore o nearshore Age Figure 6 Mean fork length (mm) at age of hogfish ( Lachnolaimus maxi- mus) collected from nearshore (<30 m) and offshore (>30 m) depths for four age classes commonly collected from both depth categories. Error bars represent 95% confidence limits. Figure 7 Maturity schedule for male hogfish (Lachnolaimus maximus) by fork length (top) and age (bottom). Nearshore ( < 3 0 m) hogfish are indicated by hollow squares and offshore (>30 m) hogfish by filled squares. Lines indicate the predicted curve. reports of hogfish sex ratios (0.1 M:F in Puerto Rico [Colin, 1982] and 0.2 M:F in Cuba [Claro et ah, 1989]) coincided with the modal range that we observed (0.1- 0.4 M:F). The variability in sex ratios reported herein was not related to season; therefore we conclude that harems are probably maintained throughout the year. Colin (1982) and Munoz et al. (2010) reported high site fidelity and restricted home ranges for hogfish, at least during their spawning season (primarily winter-spring). The wide range of sizes observed for transi- tional hogfish indicates the mechanism is un- der social (rather than genetic) control. Warner (1984) showed that female wrasse change sex at smaller sizes when densities are high because a single small male could not monopolize mating, increasing female incentive to change sex. How- ever, large male size or low density discouraged competition, and sexual transition by females was postponed. Smaller sizes and higher densities of hogfish observed nearshore would indicate that social mechanisms were likely responsible for the cross shelf patterns of size and age at sex change for this protogynous fish. Spawning success was much higher in a protect- ed area of the Florida Keys than in an adjacent area open to fishing, even though the frequency of contact between sexes was the same in both areas (Munoz et al., 2010). Munoz et al. proposed that lower rates of mortality will create a familiar social order, facilitating courtship and increasing spawning rates. Higher levels of mortality in near- shore waters may thus potentially disrupt harem structure and reduce reproductive output in more heavily fished areas. Conclusions Although there is evidence of fishing effects in nearshore waters, the continued escapement of fast-growing fish to deeper waters reduces con- cerns about fishery-induced evolution of life his- tory traits that could occur if fast growers were being harvested at such a rate that they could no longer spawn successfully. The maximum size and age of hogfish reported herein are similar to those reported for Cuban waters, where there is a relatively “unfished population” (Claro et al., 1989), and to those measured previously within the cur- rent study area (1995-2001; McBride and Richard- son, 2007). The technical and logistic limits that prevent most spearfishing in offshore waters and the behavioral peculiarities that make hogfish less vulnerable to hook-and-line fisheries appear to sup- port a de facto refuge for some of the faster growing and largest hogfish. Offshore females spawn for longer periods and produce larger batches of eggs than do nearshore females (authors of this article, unpubl. data), and therefore the persistent escape- Collins and McBride: Spatially explicit life-history dynamics of a protogynous reef fish 241 ment to offshore waters may contribute notably to the reproductive success of hogfish in the eastern Gulf of Mexico (Johannes, 1998; Birkeland and Dayton, 2005). Still, because the conspicuous nature and inquisitive behavior of hogfish make them very vulnerable to fish- ing, routine monitoring of fishing effort or fishing power by the spearfishing sector is warranted, as is periodic monitoring of spatially explicit densities, harem struc- ture, and life history traits of this species. Acknowledgments This work could not have been completed without the dedication of the members of the Saint Peters- burg Underwater Club (SPUC). D. O’Hern acted as the SPUC spokesman throughout this research. We would like to thank captains B. Bateman, L. Borden, W. Butts, J. DeLaCruz, C. Gardinal, C. Grauer, T. Grogan, B. Hardman, J. Hermes, S. Hooker, I. Lathrop, S. Lucas, M. Joswig, D. O’Hern, D. Palmer, H. Scar- boro, C. Schnur, and R. Taylor. E. Leone, M. Murphy, and M. Greenwood provided statistical guidance. A. Richardson and K. McWhorter provided laboratory assistance. FWRI’s Fisheries-Independent Monitoring Program provided trawl samples. D. DeVries served as the NOAA/NMFS partner and provided useful com- ments throughout the project. G. Shepherd, J. Colvo- coresses, M. Murphy and three anonymous reviewers provided comments that improved this manuscript. The majority of the work described herein was funded by grant NA05NMF4540040 to the Florida Fish and Wildlife Conservation Commission from the National Oceanic and Atmospheric Administration’s Cooperative Research Program. Literature cited Adams, S., B. D. Mapston, G. R. Russ, and C. R. Davies. 2000. Geographic variation in the sex ratio, sex spe- cific size, and age structure of Plectropomus leopardus (Serranidae) between reefs open and closed to fishing on the Great Barrier Reef. Can. J. Fish. Aquat. Sci. 57:1448-1458. Allison, P. D. 1999. Logistic regression using the SAS System, 288 p. SAS Institute, Inc., Cary, NC. Allman, R.J. 2007. Small-scale spatial variation in the population structure of vermilion snapper ( Rhomboplites aurorubens ) from the northeast Gulf of Mexico. Fish. Res. 88:88—99. Alonzo, S. H., and M. Mangel. 2005. Sex-change rules, stock dynamics, and the perfor- mance of spawning-per-recruit measures in protogynous stocks. Fish. Bull. 103:229-245. Bannerot, S., W. W. Fox, and J. E. Powers. 1987. Reproductive strategies and the management of snappers and groupers. In Tropical snappers and grou- pers: biology and fisheries management ( J. J. Polovina and S. Ralston, eds.), p. 561-603. Westview Press, Boulder, CO. Birkeland, C., and P. K. Dayton. 2005. The importance in fishery management of leaving the big ones. Trends Ecol. Evol. 20:356-358. Brooks, E. N., J. E. Powers, and E. Cortes. 2010. Analytical reference points for age-structured models: application to data-poor fisheries. ICES J. Mar. Sci. 67:165-175. Brule, T., C. Deniel, T. Colas-Marrufo, and M. Sanchez-Crespo. 2003. Reproductive biology of gag in the southern Gulf of Mexico. J. Fish Biol. 63:1505-1520. Claro, R., A. Garcia-Cagide, and R. Fernandez de Alaiza. 1989. Caraeteristicas biologicas del pez perro, Lachno- laimus maximus (Walbaum), en el golfo de Batabano, Cuba. Rev. Investig. Mar. 10:239-252. [In Spanish ] Colin, P. L. 1982. Spawning and larval development of the hogfish, Lachnolaimus maximus (Pisces: Labridae). Fish. Bull. 80:853-862. Davis, J. C. 1976. Biology of the hogfish, Lachnolaimus maximus (Wal- baum), in the Florida Keys. M.S. thesis, 86 p. Univ. Miami, Coral Gables, FL. DeVries, D. A. 2006. The life history, reproductive ecology, and demog- raphy of the red porgy, Pagrus pagrus, in the north- eastern Gulf of Mexico. Ph.D. diss., 160 p. Florida State Univ., Tallahassee, FL. Faunce, C. H., and J. E. Serafy. 2007. Nearshore habitat use by gray snapper ( Lutja - nus griseus) and bluestriped grunt ( Haemulon sciurus ): environmental gradients and ontogenetic shifts. Bull. Mar. Sci. 80:473-495. Gannon, D. P., E. J. Berens McCabe, S. A. Camilleri, J. G. Gannon, M. K. Brueggen, A. A. Barleycorn, V. I. Palubok, G. J. Kirkpatrick, and R. S. Wells. 2009. Effects of Karenia brevis harmful algal blooms on nearshore fish communities in southwest Florida. Mar. Ecol. Prog. Ser. 378:171-186. Gust, N. 2004. Variation in the population biology of protogynous coral reef fishes over tens of kilometres. Can. J. Fish. Aquat. Sci. 61:205-218. Hamilton, S. L., J. E. Caselle, J. D. Standish, D. M. Schroeder, M. S. Love, J. A. Rosales-Casian, and O. Sosa-Nishizaki. 2007. Size-selective harvesting alters life histories of a temperate sex-changing fish. Ecol. Appl. 17:2268-2280. Harris, P. J., and J. C. McGovern. 1997. Changes in the life history of red porgy, Pagrus pagrus , from the southeastern United States, 1972— 1994. Fish. Bull. 95:732-747. Hawkins, J. P., and C. M. Roberts. 2003. Effects of fishing on sex-changing Caribbean par- rotfishes. Biol. Conserv. 115:213-226. Heppell, S. S., S. A. Heppell, F. C. Coleman, and C. C. Koenig. 2006. Models to compare management options for a pro- togynous fish. Ecol. Appl. 16:238—249. Hu, C., F. E. Muller-Karger, and P. W. Swarzenski. 2006. Hurricanes, submarine groundwater discharge, and Florida’s red tides. Geophys. Res. Lett. 33 (LI 1601 ): 1—5. Ingram, G. W., Jr. 2001. Stock structure of gray triggerfish Balistes capriscus on multiple spatial scales in the Gulf of Mexico. Ph.D. diss., 228 p. LTniv. South Alabama, Mobile, AL Jennings, S. D., M. J. Kaiser, and J. D. Reynolds. 2001. Marine fisheries ecology, 417 p. Blackwell Sci- ence Ltd., Malden, MA. 242 Fishery Bulletin 109(2) Johannes, R. E. 1998. The case for data-less marine resource manage- ment: examples from tropical nearshore finfisher- ies. Trends Ecol. Evol. 13:243-246. Landsberg, J. H., L. J. Flewelling, and J. Naar. 2009. Karenia brevis red tides, brevetoxins in the food web, and impacts on natural resources: Decadal advance- ments. Harmful Algae 8:598-607. Lindeman, K. C., R. Pugliese, G. T. Waugh, and J. S. Ault. 2000. Developmental patterns within a multispecies reef fishery: management applications for essential fish habi- tats and protected areas. Bull. Mar. Sci. 66:929-956. Lombardi-Carlson, L., G. Fitzhugh, C. Palmer, C. Gardner, R. Farsky, and M. Ortiz. 2008. Regional size, age and growth differences of red grouper (Epinephelus morio ) along the west coast of Florida. Fish. Res. 91:239-251. McBride, R. S., and M. R. Johnson. 2007. Sexual development and reproductive seasonal- ity of hogfish (Labridae: Lachnolaimus maximus ), an hermaphroditic reef fish. J. Fish. Biol. 71:1270-1292. McBride, R. S., and A. K. Richardson. 2007. Evidence of size-selective fishing mortality from an age and growth study of hogfish (Labridae: Lach- nolaimus maximus ), a hermaphroditic reef fish. Bull. Mar. Sci. 80:401-417. McBride, R. S., P. E. Thurman, and L. H. Bullock. 2008. Regional variations of hogfish ( Lachnolaimus maxi- mus) life history: Consequences for spawning biomass and egg production models. J. Northw. Atl. Fish. Sci. 41:1-12. Munoz, R. C., M. L. Burton, K. J. Brennan, and R. O. Parker. 2010. Reproduction, habitat utilization, and movements of hogfish ( Lachnolaimus maximus) in the Florida Keys, U.S.A.: comparisons from fished versus unfished habi- tats. Bull. Mar. Sci. 86:93-116. Nagelkerken, I., M. Dorenbosch, W. C. E. P. Verberk, E. Cocheret de la Moriniere, and G. van der Velde. 2000. Importance of shallow-water biotopes of a Carib- bean bay for juvenile coral reef fishes: patterns in biotope association, community structure and spatial distribu- tion. Mar. Ecol. Prog. Ser. 202:175-192. PADI (Professional Association of Diving Instructors). 1999. The PADI divemaster manual (D. Richardson , eel.), 200 p. International PADI, Inc., Santa Margarita, CA. Randall, J. A., and G. L. Warmke. 1967. The food habits of the hogfish (Lachnolaimus maxi- mus), a labrid fish from the western Atlantic. Caribb. J. Sci. 7:141-144. Roessler, M. 1965. An analysis of the variability of fish populations taken by otter trawl in Biscayne Bay, Florida. Trans. Am. Fish. Soc. 94:311-318. Simberloff, D. S. 1974. Equilibrium theory of island biogeography and ecology. Annu. Rev. Ecol. Syst. 5:161-182. Sluka, R. D., and K. M. Sullivan. 1998. The influence of spear fishing on species composi- tion and size of groupers on patch reefs in the upper Florida Keys. Fish. Bull. 96:388-392. Smith, G. B. 1975. The 1971 red tide and its impact on certain reef communities in the mid-eastern Gulf of Mexico. Envi- ron. Lett. 9:141-152. 1979. Relationship of eastern Gulf of Mexico reef-fish communities to the species equilibrium theory of insular biogeography. J. Biogeogr. 6:49-61. Tupper, M., and M. A. Rudd. 2002. Species specific impacts of a small marine reserve on reef fish production and fishing productivity in the Turks and Caicos Islands. Environ. Conserv. 29:484- 492. Victor, B. C. 1986. Duration of the planktonic larval stage of one hundred species of Pacific and Atlantic wrasses (family Labridae). Mar. Biol. 90:317-326. Warner, R. R. 1975. The reproductive biology of the protogynous her- maphrodite Pimelometopon pulchrum (Pisces: Labridae). Fish. Bull. 73:262-283. 1984. Deferred reproduction as a response to sexual selection in a coral reef fish: a test of the life historical consequences. Evolution 38:148-162. 1988. Sex change and the size-advantage model. Trends Ecol. Evol. 3:133-136. Williams, D., McB., S. English, and M. J. Milicich. 1994. Annual recruitment surveys of coral reef fishes are good indicators of patterns of settlement. Bull. Mar. Sci. 54: 314-331. 243 Errata Fishery Bulletin 109:20-33 Friess, Claudia, and George R. Sedberry Age, growth, and spawning season of red bream ( Beryx decadactylus) off the southeastern United States Page 26, Table 1. Please note that in the column labeled A14C (%c) (±error) that some values were shown incorrectly as negative. The following table shows the correct values for that column. Table 1 Summary of radiocarbon (A14C ) results from red bream ( Beryx decadactylus) otoliths collected off the southeast coast of the United States. NOSAMS accession no.= identification number assigned by the Woods Hole National Ocean Sciences Accelerator Mass Spectrometry Facility. NOSAMS accession no. Collection year Birth year Sample weight (mg) Reading 1 age (yr) Reading 2 age (yr) Reading 3 (joint) age (yr) AUC (%o) (±error) OS-67042 2007 1945 81.1 62 62 62 -61.5 (3.2) OS-66866 2007 1951 40.6 58 59 56 -54.1 (4.0) OS-66870 2006 1952 52 54 62 54 -62.1 (3.6) OS-67041 2006 1958 56.5 48 50 48 -52.3(2.9) OS-66869 2007 1959 48.4 54 55 48 -67.7 (3.6) OS-68036 2004 1959 74 38 43 45 -42.6 (3.3) OS-68037 2005 1963 56.9 41 47 42 -67.8 (2.8) OS-66868 2005 1964 86.6 44 50 41 -65.7 (3.2) OS-68041 2004 1966 85.1 37 38 38 -18.7(3.1) OS-68142 2006 1966 152.3 40 43 40 25.3 (3.7) OS-68035 2003 1967 75.3 24 32 36 14.4 (3.2) OS-66867 2005 1969 42.7 33 38 36 48.6 (3.6) OS-68038 2006 1970 99 38 42 36 -34.3 (3.2) OS-66998 2004 1974 39.6 23 29 30 93.4 (4.1) OS-68034 2005 1982 75.5 18 27 23 67.0(3.6) OS-67038 2004 1989 35.6 11 11 15 89.8 (4.1) OS-68040 2003 1989 77.2 14 19 14 85.7 (5.1) OS-67040 2005 1991 70.6 10 11 14 82.6 (3.4) 244 Fishery Bulletin 109(2) 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, 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 key words 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 Scientific 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. Equations and mathematical symbols Set equa- tions from a standard mathematical program (Math- Type) or tool (Equation Editor in MS Word). LaTex is acceptable for more advanced computations. For mathe- matical symbols in the general text (a, %2, n, ±, etc.), use the symbols provided by the MS Word program and italicize all variables. Do not use the photo mode when creating these symbols in the general text. Literature cited comprises published works and those accepted for publication in peer-reviewed litera- ture (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). Abbreviations of serials should conform to abbreviations given in Cambridge Scientific Abstracts (http://www.csa. com/ids70/serials_source_list.php?ab=biolclast-set-c). 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 significance 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. 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 245 name and number of the figure. Avoid placing labels vertically (except for 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. • 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. • 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). 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. Flowever, 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 Submit manuscript online at http://mc.manuscriptcentral. com/fisherybulletin. Commerce Department personnel should submit papers under a completed NOAA Form 25-700. For further details on electronic submission, please contact the Scientific Editorial Office directly at julie.scheurer@noaa.gov 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 format. Figures should be sent as PDF files, Windows metafiles, tiff files, or EPS files. Send a copy of figures in the original 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 Julie Scheurer, Associate Editor. 245 name and number of the figure. Avoid placing labels vertically (except for 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. • 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. • 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). 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 Submit manuscript online at http://mc.manuscriptcentral. com/fisherybulletin. Commerce Department personnel should submit papers under a completed NOAA Form 25-700. For further details on electronic submission, please contact the Scientific Editorial Office directly at julie.scheurer@noaa.gov 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 format. Figures should be sent as PDF files, Windows metafiles, tiff files, or EPS files. Send a copy of figures in the original 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 Julie Scheurer, Associate Editor. Fishery Bulletin Subscription form Superintendent of Documents Publications Order Form *5178 1 1 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. IT’S EASY! Please Choose Method of Payment: I | Check Payable to the Superintendent of Documents ] GPO Deposit Account 1 1 | 1 | | | ] — Q I VISA or MasterCard Account To fax your orders (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