Evaluating Summer Flounder Spatial Sex-Segregation in a Southern New England Estuary

Marine fish species can exhibit sex-specific differences in their biological traits. Not accounting for these characteristics in the stock assessment or management of a species can lead to misunderstanding its population dynamics and result in ineffective regulatory strategies. Summer Flounder Paralichthys dentatus, a flatfish that supports significant commercial and recreational fisheries along the northeastern U.S. shelf, expresses variation in several traits between the sexes, including growth and habitat preference. To further understand these patterns, 1,302 Summer Flounder were collected and sexed in 2016 and 2017 from fisheries-independent surveys conducted in Rhode Island state waters. Female flounder were more prevalent in shallow waters (≤15 m) through all months, but males had a greater presence in deeper waters (>15 m) from May through September. The probability of a collected flounder being female was evaluated with generalized linear models and covariates representing depth, temperature, month, year, and TL. Summer Flounder were more likely to be female at larger sizes, in shallower waters, and late in the season. When compared with landings data in the recreational fishery over the sampling period, the results suggest Subject editor: Kenneth Rose, University of Maryland Center for Environmental Science, Cambridge *Corresponding author: joseph_langan@uri.edu Received August 1, 2018; accepted December 30, 2018 This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. Marine and Coastal Fisheries: Dynamics, Management, and Ecosystem Science 11:76–85, 2019 © 2019 The Authors. Marine and Coastal Fisheries published by Wiley Periodicals, Inc. on behalf of American Fisheries Society. ISSN: 1942-5120 online DOI: 10.1002/mcf2.10065

that nearly all flounder harvested in the sector were female. This work provides further evidence for and characterization of Summer Flounder sex-segregation and highlights, for management purposes, the importance of considering fine-scale spatial dynamics in addition to broader distribution patterns. The fitted model represents an effective first step toward understanding the implications of an aggregated fishing effort for disproportionate removals of male or female flounder and for exploring resulting consequences for regional spawning stock biomass and stock resiliency.
Incorporating sex-specific information in fisheries stock assessments can be critical in understanding the structure and resiliency of fish populations. Research on several fish species has indicated that failing to account for sexual dimorphism in fish growth and natural mortality (Swain 1997;Swain and Morin 1997;Wang et al. 2007;Gerritsen et al. 2010;Su et al. 2011;Loher et al. 2016), differing fishery selectivity (Gerritsen et al. 2010;Myers et al. 2014;Loher et al. 2016;Gonçalves et al. 2017), and/or spatiotemporal sex-segregation (Swain 1997;Gerritsen et al. 2010;Loher and Hobden 2012;Okamura et al. 2014;Gonçalves et al. 2017;Haugen et al. 2017) may bias biological reference points and compromise the success of management strategies. Nonetheless, relatively few stock assessments incorporate sex-specific life history characteristics due to the intensive data required (Okamura et al. 2014;Terceiro 2018) and uncertainty in whether local patterns represent stock-wide characteristics. Although some of these data are available from fisheries-independent surveys or landings (Wang et al. 2007;Su et al. 2013;Morson et al. 2015;Loher et al. 2016;Haugen et al. 2017;Terceiro 2018), there is little information characterizing spatial variability or describing the sex-composition of fishery discards, further complicating incorporation of sex-specific information into stock assessments.
Summer Flounder Paralichthys dentatus is one of the most commercially and recreationally important finfish species on the northeastern U.S. Continental Shelf (Terceiro 2018). Like many flatfishes (Swain and Morin 1996;Clark 2004;Gerritsen et al. 2010;Loher and Hobden 2012;Loher et al. 2016), this species exhibits sexual dimorphism; females grow larger and faster than males, and sexual differentiation begins between 6 and 12 cm in length or 4.5 to 8 months in age (King et al. 2001). Natural mortality may also differ between male and female Summer Flounder, in that females may have a lower mortality rate (Maunder and Wong 2011); however, this attribute is still debated (Terceiro 2018). Morson et al. (2012) identified a significant relationship between the sex ratio of recreational landings and the port at which Summer Flounder were collected, indicating that this species exhibits spatial sex-segregation. Such spatial differences between the sexes suggest that there is a risk of disproportionately removing female spawning stock biomass if commercial and recreational fishing is localized, which could alter the population's reproductive capacity (Su et al. 2011(Su et al. , 2013Loher et al. 2016;Alonso-Fern andez et al. 2017;Haugen et al. 2017). A validating example of this concern is seen in the New Jersey recreational Summer Flounder fishery, where the harvest was found to be predominantly female (Morson et al. 2012(Morson et al. , 2017. Therefore, it is evident that spatial sex-based population dynamics represent an important consideration in the successful management of this species (Terceiro 2018).
While sexual dimorphism is well documented for Summer Flounder (King et al. 2001;Terceiro 2018), the data needed to construct sex-specific stock assessment models remain scarce. For example, it is unknown whether male Summer Flounder migrate inshore later than females. Such a pattern would shift the sex ratio of the inshore population throughout the season (NEFSC 2013). Although the stock assessment uses trawl surveys covering the latitudinal and bathymetric range of Summer Flounder (NEFSC 2013;Politis et al. 2014;Terceiro 2016), these surveys lack the fine spatiotemporal resolution required to develop a clear understanding of sex-specific distribution patterns in the shallow inshore areas targeted by recreational anglers, who are responsible for a major component of the annual harvest (Terceiro 2016(Terceiro , 2018, and how the sex composition changes throughout the fishing season (NEFSC 2013). To begin addressing these research needs, this work aims to use specimens from two nearshore trawl surveys in Rhode Island waters with the spatiotemporal coverage necessary to evaluate coastal patterns of sex-specific distribution in Summer Flounder. Conducted in the center of southern New England, an area of significant commercial and recreational Summer Flounder harvest (NEFSC 2013;Terceiro 2018), the results of this investigation will help elucidate important spatiotemporal considerations for successful regional fishery management.

METHODS
Sample collection.-Summer Flounder were collected opportunistically in Rhode Island state waters throughout 2016 and 2017 from fish trawl surveys conducted by the University of Rhode Island Graduate School of Oceanography (URI GSO) and the Rhode Island Department of Environmental Management (DEM; Figure 1). Using an otter trawl net with an effective opening of 6.5 m and towed at 2 knots (3.7 km/h) for 30 min, the weekly URI SUMMER FLOUNDER SPATIAL SEX-SEGREGATION GSO survey is conducted at two stations in Narragansett Bay: Fox Island, a midbay habitat with a mean depth of approximately 7 m, and Whale Rock, an outer-bay habitat with a mean depth of approximately 22.9 m (Collie et al. 2008). The DEM survey is conducted at 13 set locations in Narragansett Bay monthly and is augmented with additional fixed and randomly chosen stations during the spring (from approximately mid-April to mid-May) and fall (from approximately mid-September to early October) in the Bay, Rhode Island Sound, and Block Island Sound. The DEM survey uses an otter trawl net with a 6.7-m wingspan and tow durations of 20 min traveling at 2.5 knots (4.6 km/h). Accompanying information during the tows, including date, location, depth, and bottom temperature, is recorded for both surveys. A total of 1,328 flounder were collected across the URI GSO and DEM surveys for this analysis. Summer Flounder were measured (TL, cm) and dissected for sexual identification. Dissections occurred within 4 h of capture when collected from the URI GSO survey, and individuals obtained from the DEM survey were frozen and processed at a later date. Previous research has indicated that freezing samples has negligible effect on the TL measurements of the specimen (Morson et al. 2012). Sex was determined by visual inspection of the gonads as in Morson et al. (2015). Small individuals with underdeveloped gonads that were unable to be sexed were excluded from analyses. Because females were generally easier to distinguish at TLs less than 30 cm, the sex ratio of this small portion of the overall sample may be biased by such exclusions.
Statistical analyses.-The probability of a flounder being male or female was estimated by logistic regression in R (R Core Team 2017). The regression predicted sex as a function of the spatiotemporal capture information (depth, month, year), bottom temperature, and the TL of each individual. Model variants using all combinations of variables considered were compared through minimization of the Akaike information criterion (AIC; Akaike 1973) and maximization of the area under the receiving operator characteristic curve (AUC), a metric describing the statistical power of a model as a function of the type-I error rate of the decision threshold. The classification accuracy, calibration, and bias of the selected model variant were then evaluated via cross-validation analyses (see the Supplement available in the online version of this article).

Nearshore Sex Ratio Description
Of the total 1,328 Summer Flounder collected from the URI GSO (n = 1,113) and DEM (n = 189) fish trawl surveys, 1,302 could be successfully sexed (Table 1). Despite similar sampling effort between years, approximately twice as many Summer Flounder were collected in 2017 (n = 889) than in 2016 (n = 456). In the URI GSO survey, the median catch per tow during the study period increased from six individuals in 2016 to 11 in 2017. Further, the proportion of specimens in this survey that were male significantly increased from 46.6% to 54.9% between the sampled years (v 2 test: P = 0.0145).
Male Summer Flounder were found to occupy a smaller TL distribution than did females ( Figure 2; Kolmogorov-Smirnov [K-S] test: P < 0.001). However, male Summer Flounder captured in 2017 spanned a wider range of sizes than those in 2016 ( Figure 2; K-S test: P < 0.001). In contrast, female Summer Flounder exhibited approximately the same TL distribution in both years (K-S test: P > 0.05). While both sexes were prevalent at small sizes, females made up the majority of captured fish larger than the 45.7-cm (93.0%) and 48.3-cm (97.7%) minimum length limits implemented in the Rhode Island recreational fishery in 2016 and 2017, respectively (Jason McNamee, Rhode Island Department of Environmental Management, personal communication) (Figure 2). Specifically, 21.4% and 13.2% of captured females exceeded the 45.7-cm and 48.3-cm minimum length limits, respectively, while 1.6% and 0.3% of captured males were large enough for recreational harvest.
The TL distribution and sex ratio (v 2 test: P < 0.001) of captured Summer Flounder shifted by month (Figures 3, 4). Individuals captured in May (median TL = 38.0 cm) tended to be small in size before larger fish arrived in the sampling area in June (median TL = 41.0 cm; K-S test: P < 0.001). The TL distribution then remained similar throughout the summer months (K-S test: P > 0.05) before again sequentially shifting toward smaller sizes between August (median TL = 40.6 cm) and September (median TL = 39.0 cm; K-S test: P < 0.001) and again between September and October (median TL = 38.0 cm; K-S test: P = 0.004).
Sex-segregation of Summer Flounder was evident across the sampled water depths. Females were the more prevalent sex in shallow waters (≤15 m) across all months, and across all depths in October ( Figure 4). Meanwhile, male Summer Flounder were predominant in the samples from deeper  SUMMER FLOUNDER SPATIAL SEX-SEGREGATION waters (>15 m) and across all depths combined, from May through September. In collections from sites no deeper than 15 m, the ratio of females to males ranged from approximately 1.25:1 to 6:1 throughout the sampling season (Figure 4). At sizes greater than 30 cm, females tended to be more prevalent in shallow waters (≤15 m) and males more prevalent in deeper waters (>15 m; v 2 test: P < 0.05). However, the overall sex ratio was male dominated across all depths at sizes less than 40 cm TL (Figure 4).

Predicting Male-Female Probability
Upon reviewing the tested variants, the selected logistic model to explain patterns in the observed sex ratio included depth, month, and TL. While the inclusion of the year in which a Summer Flounder was captured resulted in a statistically significantly improved fit compared with the selected model variant (likelihood ratio test: P = 0.0004), the reduction of the AIC was small compared with that when other covariates were included (Table 2), and the improvement in classification accuracy was statistically insignificant (DeLong's test: P > 0.05). The effect of year in the model was attributed to the increased abundance of males, particularly those greater than 40 cm in length (Figure 2), in 2017 and did not have a significant interaction with any of the other included covariates. Given that the inclusion of year did not

80
significantly improve classification accuracy and would inhibit out-of-sample use of the model, this variable was excluded in favor of the selected variant representing the average pattern of sex-segregation between the 2 years in which sampling occurred.
As indicated by the coefficients of the selected regression equation, larger fish and those in shallower water were estimated to have a higher probability of being female (Table 3). Specifically, sites no deeper than 15 m had a 1:1 predicted sex ratio at a TL of 36.5 cm, while predictions at sites deeper than 15 m did not reach an even ratio until 43.2 cm ( Figure 5). Further, there was greater predicted probability of a flounder being male from May through September and being female in October (Table 3). Monte Carlo cross-validation indicated that the estimated out-of-sample prediction error rate was minimized at a value of 22.79%, with a decision threshold of a 0.61 probability of being female (Supplement). When applying the 0.61 threshold to the entire data set's predictions, 77.65% of the samples were correctly predicted as male or female (Table 4), and an AUC value of 0.845 was achieved (Table 2). Finally, additional cross-validation procedures indicated that the probabilities that flounder were female predicted by the model were well calibrated and that the unbalanced sampling design between the URI GSO and DEM surveys did not introduce bias into the model fit (Supplement).
Based on the logistic regression predictions, the probability of a collected Summer Flounder being female was greater at smaller sizes in October than in other months at both shallow and deeper depths (Figure 5). At a depth of 7 m (average depth of the Fox Island site), for example, a 50% probability of a captured flounder being female was reached at TLs ranging from 36 to 40 cm between May and September, compared with 21 cm in October. Similarly, at a depth of 22.9 m (average depth of the Whale   (Figure 5). In October, model predictions indicated that Summer Flounder collected at 7 and 22.9 m depths at TLs greater than or equal to 40 and 35 cm, respectively, were expected to be all female. In the other sampled months, flounder between 50 and 54 cm TL caught at 7 m and those between 55 and 59 cm TL caught at 22.9 m were predicted to all be female ( Figure 5).

DISCUSSION
The size difference in sampled male and female Summer Flounder agrees with the growth dimorphism reported in previous work (King et al. 2001;Morson et al. 2012Morson et al. , 2015Morson et al. , 2017. Females dominated the captured fish larger than 45 cm (Figure 2), and the sex ratio of observed flounder legal for recreational harvest was more extreme than those reported by the fishery-dependent sampling efforts of Morson et al. (2012Morson et al. ( , 2015Morson et al. ( , 2017 and the 2013 Benchmark Stock Assessment (NEFSC 2013). This discrepancy is perhaps due to the greater emphasis on shallower depths in this study. Given that previous results suggest females make up a larger proportion of the recreational catch at a given depth in the mid-Atlantic region (Morson et al. 2017), it is possible that the skewed sex ratios observed here may be less extreme than in other inshore areas to the south. Further, the scarcity of large males implies that these older individuals may stay farther offshore, as has been hypothesized in past work (Morson et al. 2015(Morson et al. , 2017. As a result, the habitat preferences of such individuals cannot be characterized by the findings presented here. In agreement with Morson et al. (2017), females were found to prefer shallower habitats at all sizes greater than 30 cm and to dominate the sex ratio of these areas. The consistency of these patterns throughout the sampling season, in addition to recreational fishing data collected in Rhode Island waters over the study period (Table 5) and coupled with the logistic model predictions, suggests that inshore fishing efforts in this region will be targeting a female-dominated population segment. Commercial fishing effort focused in deeper coastal habitats, operating under a 35.6-cm minimum length limit, is expected to have an increased catchability of male Summer Flounder.
Suggested causes for spatiotemporal sex-segregation in flatfishes have included depth-specific preferences (Swain 1997;Swain and Morin 1997;Gerritsen et al. 2010), temperature selection (Swain 1997;Swain and Morin 1997),    (Swain and Morin 1996;Gerritsen et al. 2010). This analysis revealed that the addition of bottom water temperature did not improve the logistic model fit when depth was already included (Table 2). Recorded temperature measurements during sampling efforts ranged between 8°C and 24°C, yet the sex ratio of captured Summer Flounder did not appear to track this pattern throughout the season (Figure 4). While shallower sampled sites were indeed warmer than deeper areas throughout the summer months, the sex ratios in each location were not found to follow monthly changes in temperature. Therefore, as observed in American Plaice Hippoglossoides platessoides by Swain and Morin (1997), documented patterns of sex-specific habitat selection in Narragansett Bay Summer Flounder could not be characterized by temperature alone, but rather were best explained by water depth. Similar to findings reported in New Jersey by Morson et al. (2012), the sex ratio of sampled Summer Flounder varied significantly by month of the year (Figure 4). Controlling for the effects of the TL of the catch and sampling depth, the fitted logistic regression indicated that significantly more females were present in May before the sex ratio shifted toward males during the summer months ( Figure 5). As the offshore migration began in August (Packer et al. 1999), the observed sex ratio shifted back toward females in September and October. This pattern suggests that male Summer Flounder migrate into inshore areas later and move offshore considerably earlier than females. Counter to what would be expected in a more female-dominated regime, the Summer Flounder captured in May tended to be smaller than in the following 3 months (Figure 3). Further, the samples had a higher proportion of smaller fish in September and October, despite the increased share of female fish. These patterns were somewhat in contrast to reported recreational catch data from Rhode Island, where the TL frequencies of captured flounder were similar from May to August before increasing at the end of the season (Table 5). While further study is required to understand these trends, we hypothesize that smaller, immature females that were not actively spawning migrated to inshore sampling areas earlier in the spring and remained there longer into the fall. Additionally, the emergence of flounder less than 25 cm in TL in July and growing in prevalence through the fall (Figure 3) suggests that young-of-the-year individuals (NEFSC 2013) migrating from shallow nursery habitats (Taylor et al. 2016) appeared in the survey samples and also remained present after larger flounder had migrated offshore. If these hypotheses are correct, the observed increased size of recreational landings late in the season could be due to anglers targeting larger female fish as they migrate offshore.
While Morson et al. (2012) found a statistically significant difference in the monthly trend in the sex ratio among the years of sample collection, no such pattern was found here. However, the overall sex ratio did shift between years of sample collection. While incorporating a year effect did not improve the classification ability of the model, this does indicate that there may be year-to-year variability in model accuracy. Further, data from the URI GSO fish trawl survey suggest that the timing of annual Summer Flounder ingress and egress have each shifted significantly in recent decades (J. S. Collie, unpublished data), matching results found for other flatfishes (van Walraven et al. 2017). Therefore, it may be necessary to investigate and monitor such interannual and long-term patterns if the spatiotemporal sex-specific distribution of this species is to be successfully considered in management of the fishery.
The strong diagnostics of the fitted model (Supplement) and use of few covariates highlights the prospective application of this tool in predicting spatial sex ratio patterns in Summer Flounder in southern New England. Month of sampling, depth, and the size of fish are data frequently collected as part of other fisheries-independent surveys (Politis et al. 2014) and may even be available from recreational anglers or party and charter vessels (Morson et al. 2015). Such predictions could provide further insight into the finer spatial characteristics of the species' sex-specific habitat preferences. However, while the reported relationship between the selected covariates and the probability of a given fish being female represents dynamics in Rhode Island waters and is similar to past results (Morson et al. 2015), it is unclear how the model fit corresponds to other regions. Further, while the model is built with a large sample size, many of the specimens come from depths typical of the coastal zone. Thus, predictions at greater depths farther offshore will have greater uncertainty.
The interaction of patterns of sex-segregation and fishing activities in Rhode Island waters could result in the disproportionate removal of Summer Flounder by size and sex, thereby affecting the regional spawning stock biomass or the stock's reproductive potential. While limited in spatial scope, the patterns characterized by fisheryindependent sampling in this work resemble and further elucidate some of those described in the sex ratio of Summer Flounder landings previously (Morson et al. 2012(Morson et al. , 2015. As discussed by Morson et al. (2017), the implementation of high minimum length limits may only compound the population impacts of harvest. However, the finding that sex-specific habitat preferences in Summer Flounder are both consistent and predictable suggests that with additional sampling in other regions, management strategies incorporating more targeted spatial considerations in fishing effort are possible. The predictive model proposed here offers a first step toward this goal, but should be further validated and expanded upon by future work. In doing so, the incorporation of spatiotemporal sex-structure into the management of the coastwide Summer Flounder fishery may be successfully achieved.