Comparative demography of commercially-harvested snappers and an emperor from American Samoa

Study area and sampling protocol

Commercial reef-associated fisheries were surveyed on the island of Tutuila, American Samoa (14.3 °S, 170.7 °W), from March 2011 to September 2015, through the NOAA Commercial Fisheries Biosampling Program (CFBP; Sundberg et al., 2015; IACUC Permit number 13-1696). Surveys were conducted from 5:30 to 8:00 am at roadside temporary vendors (typically individual or groups of fishers selling their catch from the previous night) or at the newly-established centralized fish market in Pago Pago. During sampling times, all landed fish were measured (nearest 0.1 cm fork length (FL)) and recorded by species. Subsamples of measured fish were purchased for life-history analysis for the following species: the yellow-lip emperor Lethrinus xanthochilus, the humpback red snapper Lutjanus gibbus, and the yellow-lined snapper Lutjanus rufolineatus. These three species are common commercial and subsistence targets in the American Samoa fishery, representing the 9th, 12th, and 66th most common species by mass, respectively, in the nearshore catch out of nearly 300 harvested species (Pacific Islands Fisheries Science Center, 2018). Purchased fish were selected non-randomly to encompass all length classes targeted by the fishery. For each purchased specimen, samplers measured length (nearest 0.1 cm FL) and total body mass (g). Sagittal otoliths and gonad lobes were surgically extracted from each specimen for age and reproductive assessment. Otoliths were cleaned with ethanol and stored dry in individually labeled vials. Gonads were weighed to the nearest 0.001 g and macroscopically designated by sex. Entire gonad lobes or cross sections (3 mm thick) of gonad material from the mid-sections of gonad lobes were removed and stored using individually-labeled histological cassettes in a 10% buffered formalin solution.

Age determination and growth

For the three species, one sagittal otolith from each specimen was selected at random and affixed to a glass slide using thermoplastic glue (Crystalbond 509^®), such that the primordium was focused just inside the edge of the slide, and the sulcul ridge was perpendicular to the slide edge. The otolith was ground to the slide edge using a 600-grit diamond lapping wheel with continuous water flow along the longitudinal axis until flush with the edge of the slide. The otolith was then removed with heat (∼200 °C) and reaffixed with the newly flat surface down and ground to produce a thin (∼250 μm) transverse section encompassing the core material. Annuli, represented by alternating translucent and opaque bands, were counted along a consistent axis on the face of the sections to derive an estimate of age in years. Using reflected light (L. xanthochilus) and transmitted light (L. gibbus and L. rufolineatus) on a stereomicroscope, age readings for all specimens were conducted on three separate occasions by the primary author and final age (number of annuli as a proxy for age in years) was assigned when agreement in counts occurred. If three counts differed by one presumed annulus (e.g., 13, 12, 14), the middle age was assigned (e.g., 13). Daily growth increments were enumerated for one small (10.1 cm) specimen of L. xanthochilus with preparation and reading protocols following Taylor & Choat (2014). Daily growth increment profiles of recently recruited L. xanthochilus were first summarized in Wilson & McCormick (1999).

The assumption that annuli are deposited on a yearly cycle was tested using edge-type analysis for each species (Manickchand-Heileman & Phillip, 2000). The outermost otolith margin was scored as within either an opaque or translucent zone across all specimens, and proportions of opaque zone deposition were plotted across the calendar year for each species. Plots were presented alongside annual patterns of sea surface temperature, using remotely sensed data (one day resolution, Pathfinder data base available through NOAA Coastwatch, Jan 2006 through May 2011).

Sex-specific patterns of growth were modeled by fitting the von Bertalanffy growth function (VBGF) to length-at-age data using least squares estimation. The VBGF is represented by: $L_t=L_{\infty }\left[1-{\text{e}}^{-K\left(t-t_0\right)}\right]$ where L_t represents the predicted mean FL (cm) at age t (years), L_∞ is the mean asymptotic FL, K is the coefficient used to describe the curvature of fish growth toward L_∞, and t₀ is the hypothetical age at which FL is equal to zero, as described by K. As the sampling protocol was fisheries-dependent, very few specimens were below the 15–20 cm range. Hence, to approximate early growth trajectories, VBGF models were constrained to an established or inferred (from similar taxa) length at settlement (i.e., FL at age 0). For L. xanthochilus, this was 3.0 cm (Nakamura et al., 2010); for L. gibbus and L. rufolineatus, this was 3.5 cm (Mori, 1984; Nanami & Yamada, 2009).

Mortality

Total mortality (Z) was estimated using a multinomial catch curve fitted to the age classes at or above the assumed age at full recruitment to the fishery (t_rec). t_rec is defined as one plus the peak frequency age (Dunn, Francis & Doonan, 2002). For fish at or above t_rec, the per recruit survival of fish (S_t) to integer age (t) was calculated as: $S_t={\text{e}}^{-Z\left(t-t_{\text{rec}}\right)}$

${\widehat{P}}_t$, the expected proportion of fully-recruited fish at age t was calculated as: ${\widehat{P}}_t=\frac{S_t}{{\displaystyle \sum _{t=t_{\text{rec}}}^{t_{\text{max}}}{S}_t}}$ where t_max refers to the maximum observed age. The catch curve was fitted by maximizing the multinomial log-likelihood (λ) associated with the observed and expected proportions at age: $\lambda ={\displaystyle \sum _{t=t_{\text{rec}}}^{t_{\text{max}}}}{f}_{t}ln\left({\widehat{P}}_t\right)$ where f_t refers to the observed frequency at age t.

The value for the natural mortality (assumed constant across age) in this analysis was calculated using the Hoenig (1983) equation: $Z={\text{e}}^{1.46-1.01*\text{log}\left(t_{\text{max}}\right)}$, whereby Z is equal to M through this derivation if the true maximum age was derived. For empirical derivations of M in this study, we cautiously assumed this to be true.

Uncertainty in the estimate of total mortality was calculated using a length-stratified bootstrap procedure. The proportion in each length bin (5 cm bin width) from the full size distribution from the fishery was calculated and used as weights in the bootstrap resampling of the length-age samples. For example, if the sample size of the length-age data was 100 and the proportion of individuals in the first length bin in the full size distribution was 0.05, five data points from the first length bin in the length-age data were sampled with replacement. With the exception of a small individual in the L. xanthochilus dataset (10.1 cm, 0.35 years old (estimated based on daily growth increments)), the length ranges of the length-age data were similar to those from the full length distribution of the fishery. Each dataset was bootstrapped 10,000 times. The median of the bootstrap distribution was reported and the 2.5th and 97.5th percentiles of the bootstrap distribution of total mortality were reported as the 95% percentile confidence interval. The sensitivity of the size of the length bins on the total mortality estimate was evaluated using length bins of 2, 5, 10, and 20 cm, and also across individual years to examine potential effects of compounding multi-year data. The total mortality estimate using a conventional bootstrap was also evaluated as a part of the sensitivity analysis. This length-stratified bootstrap method was used because our sampling program was not designed to derive a random sample representative of the fishery, whereas the full market survey was designed for this purpose. Additionally, all otoliths and gonads that were originally sampled were not retained for final processing, which further precluded us from considering the sample to be representative.

Reproduction

Fixed sections of gonadal tissue were histologically processed at the John A. Burns School of Medicine at the University of Hawaii. Sections were imbedded in paraffin wax, sectioned transversely at 6 μm, and stained on microscope slides with haematoxylin and eosin. Slides were viewed under dissecting and compound microscopes with transmitted light to determine sex and level of reproductive development following criteria and terminology from Brown-Peterson et al. (2011). The female maturation schedule by length and age was modeled for each species. Specimens were scored by their maturity status (immature, 0: representing “Immature” and “Developing” phases from Brown-Peterson et al. (2011); mature, 1: representing “Spawning capable,” “Regressing,” and “Regenerating”), and maturity-at-length/age data were fitted with a two-parameter logistic curve as follows: $P_L={\left\{1+{\text{e}}^{-\text{ln}\left(19\right)\left(L-L_{50}\right)/\left(L_{95}-L_{50}\right)}\right\}}^{-1}$ where P_L (P_t for age at maturity) is the estimated proportion of mature females at a given length (L), and L₅₀ and L₉₅ (t₅₀ and t₉₅ for age) are the FL at 50 and 95% maturity, respectively. Corresponding 95% confidence limits (CLs) for the maturity schedules were derived by bootstrap resampling with replacement through 1,000 iterations.

Although not a primary objective of this study, we further examined aspects of the reproductive developmental ontogeny of L. xanthochilus based on histological features of specimens and length- and age-based patterns of sex ratio. Specifically, we searched for evidence of pre- or post-maturational female-to-male sex change, as has been commonly identified in species of Lethrinus, following criteria established in Sadovy & Shapiro (1987) and Sadovy de Mitcheson & Liu (2008).

Age determination and growth

From March 2011 to September 2015, a total of 2,559, 2,788, and 847 commercial specimens were measured of L. xanthochilus, L. gibbus, and L. rufolineatus, respectively. Of these, 372 L. xanthochilus, 481 L. gibbus, and 217 L. rufolineatus were dissected for life-history analysis, but otoliths and/or gonads were retained for only 244, 311, and 149 individuals, respectively. Length-frequency distributions from the commercial harvest are displayed in Fig. 1. All species showed unimodal distributions with modal FL bins of 38.5, 27.5, and 21.5 cm, respectively. However, L. gibbus displayed a considerable second “hump” in the size distribution at 36.5 cm. Ages were estimated from sagittal otoliths from a total of 236 specimens for both L. xanthochilus and L. gibbus and from 134 specimens for L. rufolineatus (Table 1). Although no species were sampled through all calendar months, annual patterns of edge deposition demonstrated an annual periodicity in the formation of opaque zones, confirming that increments are indeed annual (Fig. 2; Supplemental Information). Opaque zones were fully deposited in August and September during peak low sea surface temperatures in American Samoa. All three species deposited clearly defined annuli that were highly characteristic of otolith patterns previously identified for snappers or emperors (Figs. 3A–3C; Marriott & Mapstone, 2006; Grandcourt et al., 2010b). Otolith weight was a strong predictor for age in all species, whereby relationships were best explained using a standard quadratic equation (Fig. 4). The relationship between otolith weight and age did not differ between males and females for L. xanthochilus or L. rufolineatus, but it differed substantially between the sexes for L. gibbus (Fig. 4B). This pattern was retained in the length-at-age derived growth profiles, whereby patterns of growth were nearly identical between the sexes for L. xanthochilus and L. rufolineatus. Growth profiles diverged between males and females for L. gibbus, with males reaching a much larger asymptote (difference of ∼10 cm; Fig. 5; values in Table 1). Overall VBGF parameter values L_∞ and K for the combined sexes were as follows: L. xanthochilus, L_∞ = 40.2 cm, K = 0.64 year⁻¹; L. gibbus, L_∞ = 32.9 cm, K = 0.46 year⁻¹; L. rufolineatus, L_∞ = 22.9 cm, K = 0.82 year⁻¹. Sex-specific VBGF values and associated CLs are presented in Table 1. Sex-specific maximum observed ages for each species were 12 and 19 years for female and male L. xanthochilus, respectively, 27 and 19 years for female and male L. gibbus, and eight and 12 years for female and male L. rufolineatus.

Mortality

Median total mortality estimates were 0.35 (95% percentile CI [0.29–0.41]), 0.22 (95% percentile CI [0.19–0.26]), and 0.54 (95% percentile CI [0.44–0.75]) for L. xanthochilus, L. gibbus, and L. rufolineatus based on data pooled across surveyed years (Fig. 6). The estimate for total mortality and the distribution of total mortality bootstrapped estimates were sensitive to neither the size of the length bin, nor to the type of bootstrap procedure or timeframe (annual versus pooled) of survey data (Supplemental Information). Estimates from individual years were consistent and slightly higher than for pooled years for L. xanthochilus alone. The distributions of bootstrapped total mortality estimates were approximately symmetric for the L. xanthochilus and L. gibbus datasets, while the distribution of bootstrapped total mortality estimates for L. rufolineatus had a slight positive skew (Fig. 6C). Estimated natural mortality rates were 0.22, 0.15, and 0.35 year⁻¹ for L. xanthochilus, L. gibbus, and L. rufolineatus, suggesting fishing mortality rates were approximately half the presumed natural mortality across species.

Reproduction

We confirmed sexual identity and characterized female maturation profiles using histological sections of gonads from 161 L. xanthochilus (87 females), 157 L. gibbus (93 females), and 100 L. rufolineatus (20 females). Our fisheries-dependent sampling program covered the length and age ranges over which female maturation occurred for L. xanthochilus and L. gibbus. Length at maturation (but not age) was modelled for L. rufolineatus based on only three immature individuals, and therefore, the maturation profile and L₅₀ estimate should be interpreted cautiously. The median length at female maturity (L₅₀) was estimated at 30.0 cm (27.3–31.8 cm 95% C.L.) for L. xanthochilus, 24.9 cm (23.8–25.8 cm 95% C.L.) for L. gibbus, and 16.4 cm (14.6–18.1 cm 95% C.L.) for L. rufolineatus (Figs. 7A–7C). Age at female maturity was only modelled for L. xanthochilus and L. gibbus. Median age at maturity values were 2.1 years (1.8–2.3 years 95% C.L.) for L. xanthochilus and 3.2 years (2.9–3.9 years 95% C.L.) for L. gibbus (Figs. 7D–7E).

Further examination of male L. xanthochilus found that testes were consistently characterized by a well-developed ovarian lumen, and the vast majority of testes (>80%) contained internal parasites. We found one specimen (39 cm FL) that contained an atretic vitellogenic oocyte in the presence of proliferating male material, indicating post-maturational sex change (functional protogyny). This specimen also had large gonad walls and displayed tight rounded folds of the epithelium (Fig. 8). Testes of other specimens were found to be in similar condition—presumed to be in a transitional mode—but did not contain identifiable atretic oocytes. Further, we found at least two potentially bisexual individuals (both aged two years) under or near the size at first female maturation (24.0 and 29.4 cm FL; Fig. 9), indicating primary development into males. In both cases, primary oocytes were sparse (compared with other immature females), and stromal tissue proliferated with tenuous evidence of sperm crypt development. Length-based and age-based relative proportions of females and males across length and age classes provided further evidence of both pre- and post-maturational sex change (Supplemental Information). It is clear that sex ratios change considerably across size and age classes and that some females are retained in the largest length classes, but more robust investigations are required to determine developmental ontogeny for this species.