Life history traits vary between geographically distinct populations in a protogynous hermaphrodite

Abstract

Sequential hermaphroditic species, such as blue cod (Parapercis colias), are particularly vulnerable to selective harvesting as it can directly influence the sex ratio, timing of sex change, and consequentially recruitment success. We analyzed the population structure, at which males dominate the populations, and modeled fecundity of blue cod from the Marlborough Sounds in the north of New Zealand's South Island, and compared these results with previously collected data from a blue cod population in Fiordland on the southwest coast of the South Island. In addition to the geographical difference, there were differences in the time the two populations were sampled, and the fisheries management regime between regions. Within the Marlborough Sounds population, the length-frequency distribution was skewed toward high numbers of smaller individuals with an average length of 290 ± 59 mm. In Fiordland, the average length of blue cod was 380 ± 78 mm. Larger proportions of males within each length class were observed in the Marlborough Sounds, whereas females were the dominant sex in the Fiordland population. The estimated length at which males dominate the population was significantly smaller for the Marlborough Sounds population (182 mm) than for blue cod from the Fiordland population (419 mm). Geographical and temporal differences in sampling likely played important roles in the observed life history differences, particularly in the observed differences in growth. However, when compared to length and age data from national surveys, the life history traits of both sampled populations have remained relatively stable over the past 15 years, indicating that differences in population management and fishing pressure are the most likely explanation for the observed differences in length frequency and length-at-sex-change. For effective and sustainable management of species with sequential hermaphroditism, size-sensitive management, particularly protecting larger individuals, is required to avoid declines in length structure and recruitment limitations.

INTRODUCTION

Growth, mortality, and recruitment drive variation in abundance and biomass of marine populations (Hamilton et al., 2011; Stawitz & Essington, 2019). Among distinct geographic regions, populations may encounter large differences in environmental conditions as well as in fishing mortality and management, resulting in distinct demographic structures of regional populations. Accordingly, resolving the impacts of fisheries on demographics of marine populations has been the subject and aim of a large number of studies during the past century (Berkeley et al., 2004; Beverton & Holt, 1957; Hixon et al., 2014; Hjort, 1914). Fishing can result in the selective removal of older, larger individuals from a population, and can, thereby, truncate a population toward younger and smaller individuals, as well as reduce the spatial heterogeneity of exploited populations (e.g., Jack & Wing, 2010). Consequently, age truncation can make populations less resilient against environmental and anthropogenic stressors. In contrast, heterogeneous age/size and spatial structure increase the probability that fish populations can withstand variability in environmental conditions and localized fishing pressure (Hsieh et al., 2010). Therefore, to understand fluctuations in abundance, and to implement effective management strategies, it is important to understand the mechanisms by which the demographics of marine fish populations respond to both spatial and temporal patterns in exploitation.

Many marine populations are distributed as spatially structured population networks across their distribution range, experiencing differences in prey availability and environmental conditions. These differences can lead to local populations exhibiting natural spatial variation in demographic parameters, such as growth or the timing of reproduction (Berkeley et al., 2004; Hsieh et al., 2010). For example, latitudinal changes in temperature and productivity can affect growth and reproductive output (Wakefield et al., 2013). The resulting spatial heterogeneity and connectedness of population networks can increase the stability and resilience of a species by buffering recruitment fluctuations (Thorson et al., 2018).

In some overharvested populations, compensatory responses leading to declines in the length- and age-at-maturity have been observed (Trippel, 1995). In these cases, reduced intraspecific competition for food at low density can result in increased somatic growth and gonadal development, resulting in younger ages and smaller lengths at maturity (Rose et al., 2001). In contrast, the reduction of population size structure and density can also negatively impact reproduction, as a result of suppressed recruitment, slow recovery rates, and local extinctions (Liermann & Hilborn, 2001). In small populations, the reproductive potential may be reduced when adults are unable to locate mates during the reproductive season, or if delays in locating a suitable mate delay spawning events to a time inadequate for egg and larval development (Rowe & Hutchings, 2003).

Consequentially, reproductive potential can also be reduced due to altered or skewed sex ratios. Sequential hermaphroditic species are particularly vulnerable to size-selective harvesting methods, which are often indirectly sex-selective. Sex-selective harvesting increases the bias toward one sex and can directly reduce fertilization success and resulting reproductive rates (Alonzo & Mangel, 2004; Armsworth, 2001; Hamilton et al., 2007). Sequential hermaphroditic species may also be subject to indirect effects of sex-selective harvesting if the reduction in the size of terminal males results in earlier sex change in females. In protogynous species, sex change is often controlled by social mechanisms, such as the presence of a dominant male, or the size of the dominant male relative to females in its social group (Ross, 1990; Shapiro, 1980). Consequentially, if large individuals of the terminal sex are present or the sex ratio is near-optimal, sex change among small individuals will be delayed (Matthias et al., 2019). If the number of terminal sex individuals is reduced, or the sex ratio skewed severely from its optimum, sex change may occur at smaller lengths and younger ages (Caselle et al., 2011; Hamilton et al., 2007). A shift in the timing of sex change toward smaller lengths and younger ages due to the selective harvest of larger individuals has been reported for some exploited protogynous hermaphroditic species (Caselle et al., 2011; Hamilton et al., 2007; Muñoz & Warner, 2003). The observed effect has important implications for understanding how fisheries mortality affects fecundity in protogynous fish populations.

The present study focused on two populations of blue cod (Parapercis colias, Forster 1801) from the Marlborough Sounds (northernmost region of the South Island) and Fiordland (southwestern coast of the South Island), New Zealand. Endemic to New Zealand's temperate reefs, blue cod are one of the most commonly landed commercial and recreational species (Carbines, 2004). Blue cod are protogynous hermaphrodites, with individuals maturing as male or female initially and some females changing into males later in life (Brandt et al., 2017; Carbines, 2004). They are thought to live in a harem structure where a large male shares his territory with three to five smaller females (Mutch, 1983). Adult individuals are relatively sedentary with limited adult dispersal among rocky reefs (Beer et al., 2011; Beer & Wing, 2013; Rodgers & Wing, 2008). Due to the sedentary lifestyle, a short larval pelagic phase, and high territoriality, a high level of localized population structures has been observed among individual rocky reefs (Carbines & McKenzie, 2001; Mace & Johnston, 1983). The discrete spatial structure and protogynous reproductive strategy may underpin particular vulnerabilities of blue cod to overexploitation.

The Marlborough Sounds, a series of drowned river valleys, are located at the northernmost coast of the South Island of New Zealand (Figure 1). The area has been particularly popular for recreational fishing, which accounted for approximately 75 tonnes of blue cod landed in 2015/2016, the second-largest recreational blue cod fishery in the country (Hartill et al., 2017). The commercial blue cod fishery, confined to the outer Marlborough Sounds and Cook Strait, reported landings of 50–70 tonnes per annum over the last decade (Fisheries New Zealand, 2019). Potting surveys to monitor the blue cod populations have been carried out since 1995. By 2001, results from the surveys indicated that the abundance of blue cod in the Marlborough Sounds had halved and continued to decline until 2007 (Cole et al., 2000; Davidson, 2001; Fisheries New Zealand, 2019). In 2008, a 2-year fishing closure for blue cod was implemented in the inner Marlborough Sounds, which was followed by an increase in abundance during subsequent surveys. Following the closure, the minimum landing size (MLS) was increased from 300 to 330 mm, and the daily bag limit was lowered from three to two fish (Fisheries New Zealand, 2019). As a result of the fishing closure and changes in the MLS, the estimated recreational blue cod harvest decreased from ~148 tonnes in 2008 to ~75 tonnes in 2017 (Davey et al., 2008; Hartill et al., 2017).

FIGURE 1

Map of (a) New Zealand, and the collection sites, indicated by circles, in (b) Fiordland (green) (from Wing et al., 2012) and the (c) Marlborough Sounds (purple).

In contrast, the Fiordland Marine Area (Te Moana o Atawhenua) is a region of natural and cultural significance to New Zealand. Two no-take areas were established in 1993, with eight further no-take areas to follow in 2005, bringing the total to 10 no-take areas covering 10,421 ha. Additionally, the inner regions of 15 fjords were closed to commercial fishing (11,364 ha). Jack and Wing (2013) reported that the abundance of blue cod had increased within some of the no-take marine reserves 5 years after their establishment and that the marine reserve network had provided insurance against regional population decline by supporting mature, fecund populations.

In the present study, the population structure and modeled reproductive output of blue cod from the Marlborough Sounds were investigated and then compared with previously collected data of the blue cod population in Fiordland. The specific aims of the present study were to (1) compare the length-frequency distributions of blue cod from Fiordland and the Marlborough Sounds, (2) investigate differences in growth between the regions, and (3) compare the sex ratio, length at which males dominate population, and fecundity between the blue cod populations in Fiordland and the Marlborough Sounds. The approach offered a unique regional comparison of demographics between populations subject to contrasting fishing efforts with important implications for understanding responses of protogynous hermaphroditic species to exploitation.

METHODS

Fish collection and measurements

From the Marlborough Sounds and Tasman Bay (hereafter collectively called Marlborough Sounds), blue cod (n = 294) were collected during two research cruises on the RV Polaris II in February and November 2018 (Figure 1). Blue cod were sampled using modified commercial cod pots with fine gage mesh (20 and 10 mm), with squid as bait. The total length (TL ± 1 mm) of each sampled blue cod was recorded to estimate the length-frequency distribution. The majority of sampled blue cod from the Marlborough Sounds were released alive. A subsample of 5 fish/1 cm length class was retained for age and growth, sex ratio, and fecundity analyses. Marlborough Sound blue cod in each subsample were humanely euthanized using the Iki-method (Close et al., 1997) under University of Otago ethics protocol AUP-18-193. Blue cod from the Marlborough Sounds were compared with blue cod from Fiordland, collected by Wing et al. (2012) during surveys in 2000, 2002, 2003, and 2004. Blue cod were sampled using modified commercial cod pots with fine gage mesh (20 and 10 mm) and hook and line fishing with a range of hook sizes. Length, age, and sex of blue cod collected from Fiordland were recorded by Wing et al. (2012). All blue cod were sexed macroscopically, and female gonads were retained for fecundity analysis. The sex ratio was calculated as the ratio of females to males, and as the proportion of males and females for each region.

Age and growth

The sagittal otoliths were removed by cranial dissection to determine the age (years) of blue cod. Otoliths were rinsed in deionized water and transferred to sterile Eppendorf tubes. One of each pair of otoliths was embedded in K36 epoxy resin (Epoxy Kit, Nuplex Industries Inc., Auckland, New Zealand) and transverse sections (~1.5 mm thickness) were cut through the primordium using a Buehler Isomet low-speed diamond-bladed saw. The sections were mounted on glass slides using crystalbond 509 (Amerco Products Inc., NY, USA). The otoliths were then ground using wet-dry sandpaper (grades P600 and P800) until the yearly growth increments were clearly visible. The slides were polished using ultrafine sandpaper (grade P1500) along with alumina silicate polishing powder. Ages were estimated from photomicrographs of sectioned otoliths under transmitted light. Only opaque zones (winter growth) bordered by translucent zones (summer growth) on both sides were counted. Image processing software (ImageJ) was used to improve the contrast and clarity of images and to allow for a more accurate reading of the annual growth increments.

Growth models were constructed using the von Bertalanffy growth model (von Bertalanffy, 1934):

$$$$ {\mathrm{TL}}_t={\mathrm{TL}}_{\infty}\left(1-{e}^{-k\left[t-{t}_0\right]}\right), $$$$ (1)

where TL_t is the total length at time t (in years), TL_∞ is the asymptotic length (in millimeters, i.e., the length an individual would reach if it would grow to an infinite age), k is the growth constant expressing the rate at which length approaches the asymptote, and t₀ is the theoretical age from settlement of an individual at zero size (this can be negative for species with large larvae). The von Bertalanffy growth models and confidence intervals (CIs) were constructed using the FSA package (Ogle et al., 2021) in R 4.1.1 (R Core Team, 2021). The von Bertalanffy growth models were compared using the Kimura likelihood ratio test (Kimura, 1980) using the growthlrt function in the fishmethods package in R (Nelson, 2017).

Equation 1 was rearranged to estimate the age at the MLS (330 mm):

$$$$ {t}_{330}=\frac{\ln \left(1-\left[{\mathrm{TL}}_{\mathrm{MLS}}/{\mathrm{TL}}_{\infty}\right]\right)}{-k}, $$$$ (2)

where t₃₃₀ is the age at MLS and TL_MLS is the length at time (330 mm).

Mortality

Mortality estimates were generated using otolith-based age data. The otolith-based age-class data were used in an age-based catch curve (Chapman & Robson, 1960) to estimate instantaneous mortality (Z) and total annual survival rate (S) with 95% CIs in the FSA package (Ogle et al., 2021) in R 4.1.1 (R Core Team, 2021). Blue cod younger than the peak age in each population were excluded from calculations. Following Ricker (1975), estimates of the annual mortality rates (A) were then calculated as:

$$$$ A=1-{e}^{-Z}. $$$$ (3)

Fecundity

The wet weight (WW) of each female was taken before the ovaries were carefully removed, without rupturing the ovarian membrane. The ovaries were fixed in 2% seawater-buffered formalin, which has been shown to preserve oocytes with minimal changes to size and appearance (Lowerre-Barbieri & Barbieri, 1993). Fecundity was estimated using the gravimetric method. Four subsamples of 0.03–0.05 g were taken from the left ovarian lobe. Subsamples were weighed to the nearest 0.001 g, transferred into Eppendorf tubes containing distilled water and shaken gently to release all oocytes from the ovarian lumen. Subsamples were analyzed using a Bogorov counting chamber under a dissecting microscope. The number of hydrated oocytes, recognizable by their size, transparency, and large yolk sac, was recorded. Batch fecundity (F_B) was calculated as follows:

$$$$ {F}_{\mathrm{B}}={F}_{\mathrm{S}}\times \frac{W_{\mathrm{O}}}{W_{\mathrm{S}}}, $$$$ (4)

where F_S is the number of hydrated oocytes in the subsample, W_O is the ovary weight, and W_S is the subsample weight. The mean batch fecundity and SD of subsamples 1 and 2 were calculated and used to determine the coefficient of variation (CV). If the CV exceeded 5%, further subsample pairs were analyzed. Additional pairs of subsamples were taken until the CV was <5%. The relationship between F_B and TL for the whole population was determined as follows:

$$$$ {F}_{\mathrm{B}}=a\times {\mathrm{TL}}^b. $$$$ (5)

The length-specific relative batch fecundity (F_TL) was calculated as follows:

$$$$ {F}_{\mathrm{TL}}=\frac{F_{\mathrm{B}}}{\mathrm{TL}}. $$$$ (6)

The cumulative batch fecundity for each population was modeled according to the observed length-frequency distribution and the relationship between TL and fecundity. Only female blue cod were included in the analysis. The relationship between F_B and TL (retrieved from Equation 5) was used to model the individual batch fecundity (F_ind):

$$$$ {F}_{\mathrm{ind}}=8\times {10}^{-7}\times {\mathrm{TL}}^{3.9827}. $$$$ (7)

The cumulative batch fecundity (F_C) was then calculated as the sum of the individual modeled batch fecundity for each population:

$$$$ {F}_{\mathrm{C}}=\sum \limits_{i=1}^n{F}_{\mathrm{ind}}, $$$$ (8)

where n is the number of individuals sampled for each population. The cumulative fecundity was then calculated for an equal population size of 200 individuals (F_CPop):

$$$$ {F}_{\mathrm{C}\mathrm{Pop}}={F}_{\mathrm{C}}\times \frac{200}{n}. $$$$ (9)

As regions may differ in egg production based on spatial differences in the sex ratio, the F_C was then adjusted for the sex ratio:

$$$$ {F}_{\mathrm{adj}}={F}_{\mathrm{C}}\times \mathrm{ratio}\ \mathrm{F}:\mathrm{M}, $$$$ (10)

where F_adj is the cumulative batch fecundity adjusted for the sex ratio F:M (females:males) that was observed in each region: Marlborough Sounds = 0.529 and Fiordland = 1.356.

Statistical analysis

The average fish lengths were compared using t tests between the two regions. The length-frequency distributions of whole blue cod populations, female, and male blue cod were compared using the Kolmogorov–Smirnov two-sample test within the FSA package in R (Ogle et al., 2021; R Core Team, 2021). A X² contingency table was used to test for differences in the sex ratio between the populations from the Marlborough Sounds and Fiordland. The relationships between batch fecundity/length-specific relative fecundity and TL and WW were analyzed using linear regressions with log-transformation when necessary. The length at which males dominated the population L_50♂ (length at 50% male) and the proportion of females per length class between the two regions were modeled using logistic regression. Kruskal–Wallis tests were used to test for differences in the modeled cumulative fecundity for an equal population size (F_CPop) and modeled cumulative fecundity adjusted for the sex ratio (F_adj) between the two populations in R. Statistical analyses were conducted in JMP Pro 11 (SAS Institute Inc., Cary, NC, USA) if not stated otherwise.

RESULTS

Length-frequency distribution

The average length of the blue cod population in Fiordland was significantly larger than that of the blue cod population in the Marlborough Sounds (t = 15.22, df = 534, p < 0.0001) (Table 1). Similarly, female and male blue cod in Fiordland had significantly larger average lengths than female and male blue cod in the Marlborough Sounds (females: t = 11.66, df = 211, p < 0.0001; males: t = 11.51, df = 246, p < 0.0001) (Table 1). The Kolmogorov–Smirnoff two-sample test also demonstrated significantly different length-frequency distributions between whole blue cod populations from Fiordland and the Marlborough Sounds (D = 0.486, p < 0.0001), as well as female (D = 0.592, p < 0.0001) and male (D = 0.512, p < 0.0001) blue cod from the two regions (Figure 2). The length-frequency distributions of male and female blue cod differed significantly within Fiordland (D = 0.203, p < 0.05) and within the Marlborough Sounds (D = 0.289, p < 0.001) (Figure 2).

TABLE 1. Overview of the total length, minimum and maximum for whole blue cod populations, females (♀), males (♂), and juveniles.

Region/sex	Total length (mm)			n
	Average ± SE	Minimum	Maximum
Fiordland whole population	378 ± 4.4	185	627	315
Marlborough Sounds whole population	279 ± 4.1	111	469	257
Fiordland ♀	364 ± 5.4	185	545	169
Fiordland ♂	400 ± 7.2	249	627	135
Fiordland juvenile	336 ± 20.0	215	415	11
Marlborough Sounds ♀	272 ± 5.8	144	409	84
Marlborough Sounds ♂	301 ± 5.0	179	469	143
Marlborough Sounds juvenile	185 ± 9.1	111	267	27

• Note: n is sample size.

FIGURE 2

Length-frequency distributions of (a) whole blue cod populations from Fiordland (green bars, n = 319) and the Marlborough Sounds (purple bars, n = 294), (b) male (n = 140, blue bars) and female (n = 179, red bars) blue cod from Fiordland, and (c) male (n = 195, blue bars) and female (n = 99, red bars) blue cod from the Marlborough Sounds. Lines represent the fitted normal distribution. Note the scaled x-axis and that some bars overlap, represented in dark purple.

Growth

The von Bertalanffy growth model indicated that blue cod populations pooled across sexes from Fiordland grew at a faster rate and to larger sizes than the blue cod population from the Marlborough Sounds (Figure 3, Table 2). Blue cod sampled from Fiordland reached the MLS t₃₃₀ at 6.3 years old, while the Marlborough Sounds population reached t₃₃₀ at 9.1 years old. The likelihood ratio test found significant differences in growth between the populations from the two areas (X² = 25.85, p < 0.0001) (Table 3).

FIGURE 3

Raw length-at-age data and von Bertalanffy growth models for (a) whole blue cod populations from Fiordland and the Marlborough Sounds, (b) male and female blue cod from Fiordland, and (c) male and female blue cod from the Marlborough Sounds. Sample size and von Bertalanffy growth parameters are given in Table 2.

TABLE 2. The von Bertalanffy growth model parameter estimates (L_∞, k, and t₀) for blue cod whole populations, males and females from Fiordland and the Marlborough Sounds.

Region/sex	L∞ (mm)		k		t0		t330	n
	Estimate	CI	Estimate	CI	Estimate	CI
Fiordland whole population	487.5	460.1, 525.9	0.18	0.13, 0.24	−0.12	−1.82, 1.08	6.3	326
Marlborough Sounds whole population	401.9	377.5, 439.0	0.19	0.15, 0.25	−0.92	−1.58, −0.38	9.1	284
Fiordland ♀	471.2	440.3, 527.3	0.15	0.09, 0.21	−1.17	−3.83, 0.47	8.0	179
Fiordland ♂	615.4	522.7, 840.5	0.11	0.05, 0.19	−1.12	−4.75, 0.99	6.9	147
Marlborough Sounds ♀	366.6	333.3, 422.1	0.20	0.12, 0.29	−1.18	−2.41, −0.44	11.5	111
Marlborough Sounds ♂	420.4	391.4, 462.6	0.21	0.15, 0.27	−0.68	−1.43, −0.07	7.3	173

• Note: Upper and lower 95% CIs for L_∞, k, and t₀, the age (t₃₃₀) at minimum landing size (330 mm) and the sample size (n) of whole blue cod populations, and male (♂) and females (♀) are given. Juvenile individuals were used in male and female calculations.

TABLE 3. Likelihood ratio tests (Kimura, 1980) for growth parameters estimated (L infinity [Linf], K, and t0) for blue cod from Fiordland (Fiord.) and the Marlborough Sounds (Marlb.) and between male (♂) and female (♀) blue cod.

Scenario	Test	Hypothesis	X^2	p
Fiordland versus Marlborough Sounds (whole population)	H0 versus H1	Linf(Fiord.) = Linf(Marlb.)	6.58	0.010
	H0 versus H2	K(Fiord.) = K(Marlb.)	0.47	0.493
	H0 versus H3	t0(Fiord.) = t0(Marlb.)	0.49	0.484
	H0 versus H4	Linf(Fiord.) = Linf(Marlb.), K(Fiord.) = K(Marlb.), t0(Fiord.) = t0(Marlb.)	25.85	0.000
Fiordland ♀ versus Fiordland ♂	H0 versus H1	Linf(♀) = Linf(♂)	6.16	0.013
	H0 versus H2	K(♀) = K(♂)	0.00	1.000
	H0 versus H3	t0(♀) = t0(♂)	0.21	0.647
	H0 versus H4	Linf(♀) = Linf(♂), K(♀) = K(♂), t0(♀) = t0(♂)	35.86	0.000
Marlborough Sounds ♀ versus Marlborough Sounds ♂	H0 versus H1	Linf(♀) = Linf(♂)	1.92	0.166
	H0 versus H2	K(♀) = K(♂)	0.00	1.000
	H0 versus H3	t0(♀) = t0(♂)	0.14	1.708
	H0 versus H4	Linf(♀) = Linf(♂), K(♀) = K(♂), t0(♀) = t0(♂)	12.32	0.006
Fiordland ♀ versus Marlborough Sounds ♀	H0 versus H1	Linf(Fiord.♀) = Linf(Marlb.♀)	5.26	0.022
	H0 versus H2	K(Fiord.♀) = K(Marlb.♀)	0.88	0.348
	H0 versus H3	t0(Fiord.♀) = t0(Marlb.♀)	0.05	0.823
	H0 versus H4	Linf(Fiord.♀) = Linf(Marlb.♀), K(Fiord.♀) = K(Marlb.♀), t0(Fiord.♀) = t0(Marlb.♀)	23.69	0.000
Fiordland ♂ Marlborough Sounds ♂	H0 versus H1	Linf(Fiord.♂) = Linf(Marlb.♂)	9.36	0.002
	H0 versus H2	K(Fiord.♂) = K(Marlb.♂)	1.55	0.213
	H0 versus H3	t0(Fiord.♂) = t0(Marlb.♂)	0.28	0.597
	H0 versus H4	Linf(Fiord.♂) = Linf(Marlb.♂), K(Fiord.♂) = K(Marlb.♂), t0(Fiord.♂) = t0(Marlb.♂)	39.22	0.000

• Note: Significant results (p < 0.05) appear in boldface.

In both sampled populations, the von Bertalanffy growth models indicated that male blue cod reached larger sizes than female blue cod in Fiordland and the Marlborough Sounds (Figure 3, Table 2). In Fiordland, male and female blue cod reached t₃₃₀ at 6.9 years old and 8.0 years old, respectively. Male blue cod from the Marlborough Sounds reached t₃₃₀ at 7.3 years old, while female blue cod in the Marlborough Sounds reached t₃₃₀ at 11.5 years old. The likelihood ratio test (Kimura, 1980) found significant differences in growth between sexes in Fiordland (X² = 35.86, p < 0.0001) and the Marlborough Sounds (X² = 12.32, p < 0.01) (Table 3). Further, the likelihood ratio test detected significant differences between females (X² = 23.69, p < 0.0001) and males (X² = 39.22, p < 0.0001) from Fiordland and the Marlborough Sounds (Table 3).

Sex ratio and length at which males dominate population

The sex ratio (females:males) was significantly different between Fiordland (1.28F:1M) and the Marlborough Sounds (0.51F:1M) (X² = 31.401, p < 0.0001), with a larger proportion of females observed in the Fiordland population. Logistic regression indicated that the length at which males began to dominate the populations (L_50♂) was larger for the blue cod population in Fiordland (419 mm) than for the blue cod population in the Marlborough Sounds (182 mm) (Figure 4). The proportion of females per length class was also significantly higher in the blue cod population from Fiordland compared to the proportion of females per length class in the Marlborough Sounds (logistic regression main test: X² = 20.76, p < 0.0001, area: X² = 52.67, p < 0.0001, TL, X² = 26.34, p < 0.0001) (Figure 4).

FIGURE 4

Logistic regression model results showing the probability of females per total length (in millimeters) for blue cod populations in Fiordland $$$ \left(\mathrm{logistic}\ \mathrm{regression}\ \mathrm{equation}:1/\left(1+{e}^{\left(-2.78+\left(-0.007\right)\times \mathrm{TL}\ \right)}\right)\right) $$$ and the Marlborough Sounds $$$ \left(\mathrm{logistic}\ \mathrm{regression}\ \mathrm{equation},:,1,/,\left(1+{e}^{\left(-1.21+\left(-0.007\right)\times \mathrm{TL}\right)}\right)\right) $$$. Circles indicate the length at which 50% of the population are male (♂), arrows show the 95% CIs.

Mortality

The Chapman–Robson estimates for annual mortality (A) was lower for the Fiordland population than the Marlborough Sounds populations. Similarly, survivorship (S) was higher for blue cod in Fiordland than those in the Marlborough Sounds (Table 4). When separated by sex and region, A was lowest (17.8% year⁻¹) while S was highest (82.1% year⁻¹) among females from Fiordland. Females from the Marlborough Sounds had the highest A (35.3% year⁻¹) and lowest S (64.4% year⁻¹) (Table 4). A and S were relatively similar between males from Fiordland and the Marlborough Sounds (Table 4).

TABLE 4. Chapman–Robson estimates of instantaneous mortality (Z) and survivorship (S; %) (mean ± SE) with upper and lower 95% CIs of age-frequency data for blue cod sampled from Fiordland and the Marlborough Sounds.

Region	N	Min. age	Max. age	Z		S		A
				Estimate	CI	Estimate	CI
Fiordland whole population	306	3	26	0.254 ± 0.016	0.222, 0.286	77.5 ± 1.4	74.7, 80.3	22.4
Fiordland ♀	170	3	24	0.196 ± 0.019	0.159, 0.234	82.1 ± 1.3	79.6, 84.7	17.8
Fiordland ♂	136	4	26	0.340 ± 0.041	0.259, 0.422	71.1 ± 2.6	65.9, 76.2	28.8
Marlborough Sounds whole population	230	1	16	0.323 ± 0.029	0.266, 0.381	72.3 ± 1.7	68.9, 75.6	27.6
Marlborough Sounds ♀	84	1	16	0.435 ± 0.044	0.348, 0.553	64.4 ± 4.1	56.3, 72.5	35.3
Marlborough Sounds ♂	146	2	15	0.331 ± 0.027	0.277, 0.385	71.7 ± 2.2	67.5, 76.0	28.2

• Note: Annual mortality (A; %) was calculated using the estimated Z values (Equation 3). Sample size (N) and maximum (max.) and minimum (min.) sampled ages are given.

Fecundity

Twenty gravid female blue cod were sampled from the Marlborough Sounds in 2018. The TL of spawning females ranged from 221 to 334 mm. Batch fecundity (F_B), the number of hydrated oocytes (HO), ranged from 975 to 11,268 HO. Length-specific fecundity (F_TL) ranged from 3.61 to 37.56 HO/mm TL. Batch fecundity and length-specific relative fecundity both significantly increased with TL and WW (Table 5).

TABLE 5. Maternal total length (TL, in millimeters), and wet weight (WW, in grams) effects on batch fecundity (F_B) and length-specific relative fecundity (F_TL, in larvae per millimeter of TL).

Dependent variable	Trait	Equation	r^2	F	p	df
Batch fecundity	TL	log(FB) = −13.985 + 3.983 × log(TL)	0.33	8.90	0.008	1.18
	WW	log(FB) = 0.932 + 1.289 × log(WW)	0.33	8.89	0.008	1.18
Length-specific relative fecundity	TL	log(FTL) = −13.984 + 2.985 × log(TL)	0.22	4.98	0.0384	1.18
	WW	log(FTL) = −2.813 + 0.965 × log(WW)	0.22	4.98	0.0384	1.18

• Note: Significant results appear in boldface.

The cumulative fecundity for an equal population size of 200 individuals (F_CPop) was calculated for the Fiordland and Marlborough Sounds populations. The maximum F_CPop based on length-frequency distribution was 3.11 × 10⁶ for the blue cod population in Fiordland and 9.73 × 10⁵ for blue cod in the Marlborough Sounds (Figure 5). The F_CPop was significantly larger in Fiordland than in the Marlborough Sounds (Kruskal–Wallis: X² = 42.46, p < 0.0001). When F_CPop was adjusted for the sex ratio (F_adj, Equation 9) (Marlborough Sounds = 0.529 F:M, Fiordland = 1.356 F:M), F_adj increased for the blue cod population in Fiordland (maximum 3.77 × 10⁶, average ± SE 1.18 × 10⁶ ± 7.75 × 10⁴) but decreased for the blue cod population in the Marlborough Sounds to a maximum of 2.55 × 10⁵ (average ± SE 8.13 × 10⁴ ± 7.10 × 10³) (Figure 5). A Kruskal–Wallis test showed a significant difference in F_adj between Fiordland and the Marlborough Sounds (Kruskal–Wallis: X² = 122.06, p < 0.0001).

FIGURE 5

Modeled cumulative fecundity for an equal population size (F_CPop, solid lines) and cumulative fecundity adjusted for the sex ratio (F_adj, dashed lines) per individual total length (in millimeters) of blue cod populations from Fiordland and the Marlborough Sounds.

DISCUSSION

In the present study, population structure and reproductive attributes were compared between blue cod (P. colias) populations from the Marlborough Sounds and Fiordland. The Marlborough Sounds, located at the northern extreme of New Zealand's South Island, are characterized by a high number of anthropogenic activities including both recreational and commercial fishing (Davey et al., 2008; Urlich & Handley, 2020). In contrast, the relatively pristine Fiordland region is located in the southwest corner of the South Island, with a network of no-take marine reserves and fishing closures as part of the Fiordland Marine Management Act 2005 (Jack & Wing, 2013; Wing & Jack, 2013, 2014). The two populations differed in their length-frequency distributions, growth, mortality, sex ratios, length at which males dominate population, and modeled reproductive outputs.

We observed a smaller range of lengths, a lower abundance of individuals within larger length classes, and slower somatic growth within the blue cod population from the Marlborough Sounds, compared to blue cod from Fiordland. In Fiordland, individuals were abundant within larger length classes, and growth was significantly faster. To reach the MLS of 330 mm, female blue cod in the Marlborough Sounds needed approximately 11.5 years, compared to 8 years for female blue cod in Fiordland, highlighting the especially suppressed individual growth of female blue cod in the Marlborough Sounds. There are multiple, likely confounding, explanations for the observed differences between the two sampled populations: geographical, temporal, some methodological differences, and differences in local management.

The two sampled regions are located more than 500 km apart on opposite ends of New Zealand's South Island. There are important environmental and oceanographic differences between these two regions that may have contributed to the observed differences in population parameters as latitudinal changes in temperature and productivity can affect life history traits in protogynous hermaphrodites (Wakefield et al., 2013). Previous studies have indicated that the size at maturity of blue cod increases with increasing latitude (Fisheries New Zealand, 2019). Blue cod in the south of New Zealand mature later and at larger sizes than blue cod in the north of New Zealand, likely exhibiting a counter gradient relationship between length-at-maturity and water temperature (Wakefield et al., 2017). Latitudinal changes may also affect local productivity and the abundance and quality of prey between the two regions (Udy, Wing, O'Connell-Milne, Durante, et al., 2019). Differences in diet may directly affect growth rates in blue cod (Beer & Wing, 2013; Jiang & Carbines, 2002), and the related life history traits, survivorship, and reproduction. Previous studies on blue cod have indicated that growth rates, size structure, and mortality can vary geographically in response to differences in diet (Beentjes & Carbines, 2005, 2012; Beer & Wing, 2013).

In the present study, blue cod from the Marlborough Sounds stayed smaller than blue cod from Fiordland. However, counterintuitively, the Marlborough Sounds populations (lower latitude) also had slower growth than blue cod from Fiordland (higher latitude), likely indicating that differences in the nutritional quality of prey between the two regions played an important role (Udy, Wing, O'Connell-Milne, Durante, et al., 2019; Udy, Wing, O'Connell-Milne, Kolodzey, et al., 2019). For example, the observed decline in kelp forest food webs in the Marlborough Sounds habitat may have led to suppressed growth, smaller sized fish, and lower condition among blue cod from the region. The lack of old-growth age structures among the population in the Marlborough Sounds may have severe consequences, such as decreased reproductive success and increased recruitment variability (Hixon et al., 2014; Secor, 2000), which can be detrimental to a population recovering from overexploitation (Murawski et al., 2001; Rouyer et al., 2011).

The two populations from Fiordland and the Marlborough Sounds have been sampled 14–18 years apart, between 2000 and 2004 and in 2018, respectively. The temporal difference in sample collection needs to be acknowledged when interpreting the observed differences in life history between the two populations. While both regions are generally characterized by similar wave-exposed rocky reef habitats at the entrances of the fiords, there have been substantial differences in the protection and use of the surrounding land (Shears & Babcock, 2007; Udy, Wing, O'Connell-Milne, Kolodzey, et al., 2019). For example, the abundance of kelp forests, important habitat for blue cod, within the Marlborough Sounds has substantially decreased in recent decades, likely due to an increase in sea surface temperature and marine heatwaves (Hay, 1990), as well as an increase in fine sediment into the Sounds due to land clearing and deforestation (Handley & McLean, 2016). The loss of important kelp habitat has been shown to affect fish condition and growth, as well as indirectly reproduction and survival (Beer & Wing, 2013; Udy, Wing, O'Connell-Milne, Durante, et al., 2019; Udy, Wing, O'Connell-Milne, Kolodzey, et al., 2019), and likely attributed to the observed slow growth and small body length of blue cod in the region.

The differences in the observed length classes in Fiordland and the Marlborough Sounds could have been somewhat influenced by differences in the sampling methodologies used. Blue cod from the Marlborough Sounds were sampled using modified commercial cod pots with a smaller mesh size to retain small and large individuals to fully represent the occurring lengths, while blue cod from Fiordland were sampled using a combination of hand lines with artificial lures with a range of hook sizes and modified commercial cod pots (Wing et al., 2012; Wing, unpublished data). These specified sampling techniques could have had an effect on the observed length-frequency distributions and growth models by over- or underrepresenting small and large individuals, but not on the sex ratio calculations.

The geographic, temporal, and methodological sampling differences between the two regions are accompanied by differences in spatial management of blue cod populations in the Marlborough Sounds and Fiordland. In Fiordland, a network of marine reserves was established, the inner waters of Fiordland were closed to commercial fishing, and recreational fishing was heavily restricted as part of the Fiordland Marine Management Act 2005 (Fisheries New Zealand, 2020b; Wing & Jack, 2014). In the Marlborough Sounds and Tasman Bay, blue cod have undergone various management strategies as a response to regional population decline, including temporal and spatial closure, size and slot limits. Today, the Marlborough Sounds blue cod are protected by four small individual marine reserves and a fishing closure from September to December (Fisheries New Zealand, 2020a). Due to their sedentary lifestyle, blue cod are thought to respond extremely well to spatial management, particularly no-take marine reserves. Studies have demonstrated that in newly established marine reserves, the average length and abundance of blue cod increase within a relatively short time frame (<10 years) compared to non-reserve areas (Cole et al., 2000; Jack & Wing, 2013; Pande et al., 2008).

The latest available data on blue cod from Fiordland are from 2014 as part of the Dusky Sound blue cod survey in the southwest corner of New Zealand's Fiordland National Park (Beentjes & Page, 2016). Beentjes and Page (2016) observed a sex ratio of 1.26:1 females to male, which was similar to the observed sex ratio of 1.28:1 females to male observed for the Fiordland population in the present study. In Dusky Sound, males were on average larger than females with the largest individuals being males (Beentjes & Page, 2016), which was similar to findings in the present study for blue cod from multiple fjords in Fiordland (Table 1).

In 2004, the Marlborough Sounds blue cod survey resulted in 3645 sampled blue cod, ranging from 130 to 500 mm fork length (Blackwell, 2006). Analysis of the sex ratio indicated that males were the dominant sex with a ratio of 0.7:1 females to males, which was higher than the observed 0.5:1 females to males in the present study. In 2004, the larger length classes were male-dominated, while female size classes ranged from 140 to 350 mm fork length (Blackwell, 2006). These results are similar to the results of the present study, where males dominated the larger length classes and occupied a larger range of lengths than females in the Marlborough Sounds (Table 1). Importantly, large blue cod have not always been absent from the Marlborough Sound and Tasman Bay region (Leach et al., 1999). Using otoliths and bone measurements from historical catches dating back to the 19th and 15th centuries, the length of blue cod was estimated for sites on Mana Island in Cook Strait near the Marlborough Sounds (Leach et al., 1997). The comparison of the size-frequency distributions of the historic and modem catches suggest that blue cod were much larger in the pre-European and early European periods than they are today (Leach et al., 1999).

The Marlborough Sounds area has been a favored spot for many recreational activities, including fishing. Blue cod in these areas have undergone various management interventions within the last decades, including a fishing closure, size and slot limits, bag limits, and seasonal closures. The observed truncated length distributions toward large proportions of smaller individuals and sex ratios skewed toward males indicated that, despite the efforts to manage and stabilize the populations, fishing mortality in the region likely remained high with the potential for negative fishery-induced effects on the sex ratio and egg production (e.g., Collins & McBride, 2011; Hamilton et al., 2007).

Protogynous hermaphrodites, such as blue cod, are particularly vulnerable to overexploitation due to the display of sexual dimorphism—adult males are usually larger than adult females. Heavily fished populations of blue cod around New Zealand often display sex ratios skewed toward males with females dominating the smallest size classes (Beentjes & Carbines, 2005, 2009). In the present study, the Marlborough Sounds population had a sex ratio of 0.5 female blue cod per male, with an average of 26.6% females in each length class. In contrast, the population in Fiordland, which is under spatial management with large no-take marine reserves and commercial exclusion zone, had an overall sex ratio of 1.28:1 females to male, with an average proportion of 62.9% females in each length class, and only the largest length classes being dominated by males.

The mechanisms controlling sex change in blue cod are poorly understood. It is suspected that the presence of large males suppresses sex change in females. When large males are removed, higher numbers of females change into males, explaining the male-dominated sex ratios in fished populations of blue cod (Beentjes & Carbines, 2005). The observed inhibitory effect of large males on sex change indicated that sex change in blue cod is likely under social control.

If sex change was under social control in blue cod, the removal of large males from the population, and therefore the inhibition of sex change in smaller females, may cause females to change sex at a higher rate and smaller lengths. The result likely explains the observed large proportion of smaller males in the Marlborough Sounds compared to males in Fiordland in the present study. We estimated the length at which 50% of the population was male at 182 mm for blue cod observed in the Marlborough Sounds, compared to 419 mm for blue cod within Fiordland. The occurrence of sex change at different lengths between the Marlborough Sounds and Fiordland populations provided evidence that sex change in blue cod is under social control, rather than occurring at a critical length. The pattern is consistent with observations of earlier onset of sex change within heavily fished populations for other protogynous hermaphrodites, such as the Caribbean parrotfishes (Sparisoma viride and Scarus vetula), hogfish (Lachnolaimus maximus), and the California sheephead (Semicossyphus pulcher) (Collins & McBride, 2011; Hamilton et al., 2007; Hawkins & Roberts, 2003).

As a result of very few and small females observed in the Marlborough Sounds, the modeled cumulative fecundity (F_CPop) was three times higher for the population in Fiordland, due to the abundance of large females. When F_CPop was adjusted for the sex ratio of females to males in each region, the theoretical reproductive output (F_adj) of the blue cod population in Fiordland was 14.8 times higher than the estimate for the population in the Marlborough Sounds. The observed disparity can be attributed largely to the significantly larger proportion of large female blue cod in Fiordland. The lack of large individuals in the Marlborough Sounds may have negative effects on the fecundity and subsequently on recruitment.

The results of the present study demonstrate the sensitivity of the demographic structure of blue cod populations to fishing pressure, providing an important case study for how protogynous fish respond to exploitation. While the observed differences between blue cod populations from the Marlborough Sounds and Fiordland were likely affected by geographical and temporal differences, recent survey data of both populations indicated that they remained relatively stable during the last 10 years. The results of the present study highlight the need for management efforts that recognize the particular sensitivity of protogynous hermaphroditic fish to modification of sex ratios and subsequent recruitment overfishing, providing an important case study demonstrating how species-specific understanding of reproductive biology can help improve the sustainability of fisheries.

AUTHOR CONTRIBUTIONS

Stina Kolodzey and Stephen R. Wing designed the study and carried out fieldwork and sampling. Stina Kolodzey conducted the analysis and drafted the initial manuscript. Stephen R. Wing provided published data and revised the manuscript.

ACKNOWLEDGMENTS

The authors thank Leonardo Durante, Rebecca McMullin, Nichola Salmond, Anna Katrin Stroh, Bill Dickson, and Evan Kenton for their assistance during fieldwork. Technical support was provided by the Department of Marine Science and the Portobello Marine Laboratory. Stina Kolodzey was supported by a University of Otago doctoral scholarship. Funding support was provided by the National Science Challenge: Sustainable Seas (4.1.1 Ecosystem Connectivity) and from the University of Otago's Research Committee to Stephen R. Wing.

CONFLICT OF INTEREST

The authors have no conflict of interest to declare.

Open Research

DATA AVAILABILITY STATEMENT

Marlborough Sound blue cod population data (Kolodzey, 2022) are available from Figshare: https://doi.org/10.6084/m9.figshare.18671567.v1.