Age-based and reproductive biology of the Pacific Longnose Parrotfish Hipposcarus longiceps from Guam

Study area and sampling protocol

The major commercial vendor of reef-associated fisheries was surveyed on the US Pacific Island Territory of Guam (13°28′N, 144°45′E), from Aug 2010 to May 2016, through the NOAA Commercial Fisheries BioSampling Program (CFBP). All research was carried out under permit UOG1202 issued by the University of Guam Institutional Animal Care and Use Committee. Survey methods and sampling protocols are detailed in Sundberg et al. (2015). An average of five specimens of Hipposcarus longiceps was purchased monthly for biological analysis. Specimens were processed immediately after purchasing. For each specimen, general location of capture and fishing method were obtained from the vendor; measurements of fork length (FL, nearest mm) and total body weight (TW, g) were also obtained. Sagittal otoliths and gonad lobes were surgically removed from each specimen for subsequent processing. Otoliths were cleaned in ethanol and stored dry in plastic vials. Gonad lobes were weighed to the nearest 0.001 g (GW). Entire gonads or sections (1 cm thick) of gonad material from the mid-sections of gonad lobes were removed and stored using individually-labelled histological cassettes in a buffered 10% formalin solution or Shandon^® Glyofix^® (Thermo Fisher, Waltham, MA, USA).

Age determination and growth

One sagittal otolith from each specimen’s pair was randomly selected for age analysis. Each otolith was affixed to a glass slide using thermoplastic glue (Crystalbond 509^®; Aremco Products, Inc., Valley College, NY, USA) so that the primordium was focused just inside the slide edge, and the direction of the sulcus ridge was perpendicular to the slide edge. The otolith was ground to the primordium using a 600 grit diamond lap on a grinding wheel along the longitudinal axis to the edge of the slide. The otolith was then removed using a hotplate and reaffixed with the newly flat surface down and ground to produce a thin (∼250 µm) transverse section encompassing the core material. Annuli (alternating translucent and opaque bands) along the face of the transverse sections were counted using transmitted light on a stereo microscope on three separate occasions. Annual ages were assigned when at least two counts were in agreement, which occurred for all but four specimens. Counts of those four specimens differed by only 1 annuli (e.g., 9, 11, 10), so the middle value (e.g., 10) was assigned.

The assumption of annual deposition of increments in transverse sections of H. longiceps otoliths was tested using edge-type analysis. The optical characteristic of the otolith margin was scored as opaque (slower growth, denser material) or translucent (faster growth, less-dense material). Decimal ages (fish ages to a finer resolution than one year) were not derived because the species spawns and presumably recruits throughout the year; hence, there is no appropriate mean birth date for specimens.

Sex-specific and combined patterns of growth were examined using length-at-age data fitted with the von Bertalanffy growth function (VBGF), represented by: ${\displaystyle L_t=L_{\infty }\left[1-e^{-k\left(t-t_0\right)}\right]}$ where L_t is the predicted mean FL (mm) at age t (years), L_∞ is the asymptotic FL, k is the coefficient used to describe the curvature of fish growth towards L_∞, and t₀ is the hypothetical age at which FL is equal to zero, as described by k. The growth model was constrained to a common FL at settlement of 15 mm (following Bellwood & Choat, 1989) to account for early growth of juveniles since specimens below 188 mm FL were not obtained. Male and female growth patterns were compared by plotting 95% bivariate confidence ellipses surrounding paired estimates of L_∞ and k following Kimura (1980).

Mortality

Age composition and instantaneous rates of mortality were explored through age frequency distributions. The sex-specific frequency of individuals sampled from the fishery was plotted by length and age classes. The instantaneous rate of total mortality (Z, yr ⁻¹) was estimated two ways. First, we used a Gaussian likelihood-based linear catch curve with knife-edge selectivity. This method estimated Z as the absolute value of the slope of a line fitted to the natural logarithm of observed frequency of harvested individuals at age t (O_t) for corresponding age classes above t_c, the age at which fish are assumed fully recruited into the fishery (Beverton & Holt, 1957). We conservatively defined t_c as one plus the age at peak observed frequency.

We then used a multinomial likelihood-based catch curve using logistic selectivity by minimizing the multinomial negative log-likelihood λ associated with O_t and the expected proportion (P_t) of fish at or above t_c ${\displaystyle \lambda =-\sum _{t=1}^A{O}_{t}ln\left(P_t\right)}$ ${\displaystyle P_t=\frac{C_t}{\sum _{t=1}^A{C}_t}}$

where A refers to the maximum age assumed in the analysis, arbitrarily set much larger than the observed maximum age, thereby avoiding calculation of a “plus group.” C_t is the expected catch at age t, calculated using the Baranov catch equation: ${\displaystyle C_t=\frac{F_t}{Z_t}{N}_t\left(1-e^{-Z_t}\right)}$ ${\displaystyle Z_t=F_t+M}$ ${\displaystyle F_t=F{V}_t}$ ${\displaystyle V_t=\frac{1}{1+e^{\frac{-ln\left(19\right)\left(t-A_{50}\right)}{\left(A_{95}-A_{50}\right)}}}}$

where F_t is the fishing mortality at age t, F is the “full” instantaneous fishing mortality, and V_t is the estimated gear selectivity at age t following a logistic function with selectivity parameters A₅₀ and A₉₅, the ages at 50% and 95% selectivity, respectively. The equation of Hoenig (1983) was used to estimate natural mortality (M) based on the maximum observed age (t_max), with the assumption that t_max reflects the true life span. Thus, Z = M in the equation: ${\displaystyle Z=e^{\left(1.46-1.01\ast ln\left(t_{max}\right)\right)}.}$

Reproduction

Sampled gonads were histologically processed at the John A. Burns School of Medicine at the University of Hawaii. Fixed sections of gonad tissue were imbedded with paraffin wax, sectioned transversely at 6 µm, and stained on microscope slides with haematoxylin and eosin following standard procedures. Samples were viewed under compound and stereo microscopes to identify features determining sex and maturation stage following criteria detailed in West (1990) and using standardized terminology from Brown-Peterson et al. (2011). Particular attention was given to (1) presence or absence of post-ovulatory follicles (POFs) as confirmation of prior spawning, (2) proof of post-maturational sex change in the form of proliferating spermatogenic material in the presence of degenerative vitellogenic oocytes, and (3) characteristics of an ovarian lumen in male testes signifying secondary (prior sex change from female to male) versus primary (individual originally developed as male) male development.

Length (mm, FL) and age (years) at 50% female maturation and sex change were estimated by fitting a two-parameter logistic curve to the proportion of mature individuals per length and age class (20 mm and 1-year bins). The logistic curve was as follows: P_L = [1 + e^{−ln(19)(L−L₅₀)∕(L₉₅−L₅₀)}]⁻¹, where P_L (P_t for age at maturity) represents the estimated proportion of mature females at a given length (L), and L₅₀ and L₉₅ ( t₅₀ and t₉₅ for age at maturity) represent the FL when 50% and 95% of the population is reproductively mature. The same procedure was carried out to estimate length at sex change, where lengths at 50% and 95% sex change are denoted by L_Δ50 and L_Δ95. Corresponding 95% confidence intervals (CI) for each parameter were derived by bootstrap resampling using 1,000 iterations.

A gonadosomatic index (GSI) was calculated as the ratio of gonad-to-gonad-free body weight for each mature female in an active (not resting or spent) phase. We then examined temporal trends in GSI across calendar months and lunar days by plotting raw GSI values pooled across these timeframes. Occurrence of POFs or hydrated oocytes was also examined across these timeframes to indicate timing and frequency of spawning activity at two temporal scales.

Re-examination of vulnerability to overexploitation

The influence of life-history traits on the incidence and magnitude of vulnerability to overexploitation was examined for Guam parrotfishes in Taylor et al. (2014), using patterns of abundance and mean length across spatial and temporal gradients of fishing pressure. H. longiceps was found to be among the more vulnerable species in the Guam assemblage, but life-history trait values assessed in that study were derived from a surrogate location of Pohnpei, Micronesia. Patterns of life-history variation associated with geomorphological differences between Guam and Pohnpei suggest that Guam H. longiceps would be larger and longer-lived (Taylor, 2014). Hence, we repeated the prior analyses of incidence and magnitude of vulnerability by replacing the Pohnpei H. longiceps data with newly derived Guam data from the present study. Specifically, the traits age at maturity (t₅₀), length at female maturity (L₅₀), length at sex change (L_Δ50), mean maximum length and age (mean of the largest [L_max] and oldest [t_max] quartile of the sampled population), and modeled growth (in mm) from age 1 to age 3 (L₁₋₃) were assessed for their power to predict vulnerability to overexploitation across species (details of analysis in Taylor et al., 2014). We hypothesized that model fits would improve with the use of regional-specific Guam data in lieu of population parameters from Pohnpei.

Age determination and growth

From Dec. 2009 to May 2016, a total of 3,369 individual H. longiceps were surveyed from the Guam commercial fishery. Ninety-seven percent of these were harvested by spearfishing with SCUBA, 1.6% by spearfishing with snorkel, and 1.4% by gill net or unknown method(s). From these, 330 were purchased and sampled for life-history analysis. The smallest and largest specimens sampled were 188 mm (130 g) and 514 mm FL (2,650 g), respectively, and the average length and weight of sampled fish was 366 mm FL (±75 SD) and 1,073 g (±593 SD). The relationship between FL (mm) and TW (g) was described by TW = 1.739 × 10⁻⁵(FL)^3.018 (earlier derivation of this relationship for cm presented in Kamikawa et al., 2015).

Transverse sections of sagittal otoliths revealed overall banding patterns that were highly consistent in deposition among specimens (Fig. 1). These bands were confirmed to be annular by edge-type analysis, whereby opaque zones, representing slow growth, were deposited in the winter months (peaked in December) and were non-existent at the otolith edge from April to July (Fig. 1). Spacing of the first increment was variable among specimens because individuals likely recruited throughout the year (see ‘Reproduction’ section). Because of this, individual ages throughout are reported in years as integers, representing the total number of annular bands.

Sampled female H. longiceps ranged from 188 to 470 mm FL (mean = 345 mm ± 57 SD), and sampled males ranged from 242 to 514 mm FL (mean = 416 mm ± 59 SD; Fig. 2A). Sex-specific length ranges overlapped considerably but frequency distributions demonstrated that males were, on average, much larger than females (two sample t-test, df = 300, t =  − 10.66, P < 0.001). Primary males (based on histological features of testis; represents both IP and TP primary males) spanned the full size range of all males. The modal age for females was two years, and proportional frequencies dropped drastically beyond that age group (mean age = 2.8 years). For males, the modal age ranged from 2–5 years for which there was approximately equal frequencies (mean age = 4.0 years). Males were generally older than females, and the oldest individuals collected for each sex were ten years for males and nine years for females (Fig. 2B). Seven of the aged specimens were considered sex unknown, for which macroscopic designations were undetermined, and no histology was performed.

The overall modal age was two years for all specimens combined. The instantaneous rate of total mortality (Z) based on the linear age-based catch curve with knife-edge selectivity was 0.476 yr ⁻¹ (95% CI [0.375–0.577]), and the Z estimate from the multinomial catch curve was 0.402 yr ⁻¹ (95% CI [0.287–0.518]) (Fig. 2B). Hoenig’s (1983) estimate of M was 0.421 yr ⁻¹, presuming the maximum age of 10 years represents the natural maximum age in the unexploited wild population. This value was greater than the multinomial catch curve estimate of Z, suggesting that estimates of Z and M from the fishery-dependent samples are not robust.

Overall, H. longiceps grew rapidly in the first year, achieving over 300 mm in many specimens with one annuli (Fig. 3). The general asymptotic length was reached in year four, whereas growth profiles separated for the males and females at year three. Males reached a mean asymptotic length of 466 mm FL, whereas females reached 396 mm FL. Sex-specific growth profiles differed considerably, as shown by non-overlapping bivariate confidence ellipses surrounding estimates of L_∞ and K (Fig. 3). The overall growth parameters (sexes combined) for H. longiceps in Guam were L_∞ = 434 mm FL, K = 0.786 yr ⁻¹, and t₀ =  − 0.045 years (Table 1).

Reproduction

A total of 270 specimen gonads were processed histologically for assignment of sex and reproductive stages. These individuals ranged from 195 to 514 mm FL and included 150 females and 120 males. Three of the males were considered to be in the process of sexual transition from female to male and displayed degenerative oocytes in the presence of proliferating spermatogenic material (Fig. 4). Macroscopic field designations of sex were accurate 80% of the time. The length and age at 50% female maturation was found to be 329 mm FL (95% CI [313–345]) and 2.4 years (95% CI [2.2–2.7]), respectively (Figs. 5A, 5B). Length at sex change was determined to be 401 mm FL (95% CI [381–420]; Fig. 5C). Age at sex change was not calculable because sex change is a highly length-based, socially-mediated process, whereby many females do not undergo sex change throughout extended life spans.

We found few insights to spawning periodicity or frequency across calendar months or lunar phases. Mature, active females had GSI values up to 4% of their gonad-free body weight (Fig. 6). We found no annual or lunar trend in GSI values, and we found a high proportion of mature, active females containing post ovulatory follicles and/or hydrated oocytes in every calendar month and across each lunar phase (Fig. 6). The proportional presence of POFs and hydrated oocytes appears to correlate with the maximum observed GSI across both annual and lunar timescales.

Vulnerability analyses

Trait values for Guam H. longiceps were greater than those from Pohnpei for five of the six life-history traits examined in vulnerability models, based on Taylor et al. (2014). As predicted, a majority (ten out of twelve) of models examining the predictive power of life-history traits on vulnerability to overexploitation had an improved fit with the substitution of Guam H. longiceps data (Supplemental Information). Only the predictive power of age at maturity (t₅₀) decreased, although this trait remained the most influential for explaining incidence of response. Overall, the average variance explained increased by 4% (6% without t₅₀) for models predicting incidence of response and by 6.5% (11% without t₅₀) for models predicting magnitude of response (Supplemental Information).