Age and growth of Pomatomus saltatrix in the south-western Pacific Ocean (eastern Australia), with a global comparison

Keywords: age validation, Australian, Bayesian growth models, bluefish, life history, mortality, recreational fishing, Schnute growth curve, tailor, whole otolith.

Pomatomus saltatrix (Linnaeus) has many common names throughout its global distribution, including bluefish, tailor, elf, shad, tassergal and anchova. It is a mobile pelagic mesopredatory fish found throughout subtropical and temperate latitudes (Fig. 1). There are eight major populations, located in the western Mediterranean Sea, eastern Mediterranean–Black Sea, the north-western Atlantic, eastern Atlantic, south-western Atlantic, western Indian, eastern Indian and south-western Pacific oceans (Juanes et al. 1996). Their circumglobal distribution has been shaped by paleoclimate, water temperatures and water depths, and has resulted in extremely limited genetic mixing among populations (Goodbred and Graves 1996; Miralles et al. 2014).

Fig. 1.  Global distribution of Pomatomus saltatrix (in black) overlaid on the average annual sea-surface temperature (°C). These population ranges (except south-western Pacific) are based on the following reported ranges: eastern Indian Ocean (Smith et al. 2013), north-western Atlantic Ocean (Robillard et al. 2009), western Indian Ocean (Govender 1999; Mann 2000), eastern and western Mediterranean Sea (Sabatés et al. 2012), south-western Indian Ocean (Haimovici and Krug 1996) and the eastern Atlantic population (Champagnat 1983). Samples in the current study were from the population on the eastern coast of Australia (south-western Pacific Ocean).

As a traditional food source and artisanal fishery species (Silvano and Begossi 2005), P. saltatrix is targeted around the world primarily by recreational fishers for sport and food, although commercial operations continue to harvest significant quantities (14 731 tonnes in 2015; Food and Agriculture Organization of the United Nations 2017). P. saltatrix is typically found around surf beaches, headlands, bays and estuaries (Pollock 1984; Zeller et al. 1996; Potts et al. 2016; Schilling et al. 2017). During the 20th century, many populations of P. saltatrix were noted to be in decline, most likely because of overexploitation (Northeast Fisheries Science Center 1997; Maggs et al. 2012; Nieto et al. 2015; Lovett et al. 2020). This prompted the introduction of regulations and management plans, including seasonal closures, size limits and bag limits, and commercial quotas, which resulted in most stocks showing signs of stabilisation or improvement (Maggs et al. 2012; Northeast Fisheries Science Center 2015; Lovett et al. 2020).

Across populations, P. saltatrix shows similarities in migration (Shepherd et al. 2006; Brodie et al. 2018), spawning patterns (Sabatés and Martin 1993; Schilling et al. 2019), and early life history or recruitment (Juanes et al. 1996; Schilling et al. 2018, 2020). Although their life history is broadly similar globally, P. saltatrix may have different growth and mortality rates across different populations (Juanes et al. 1996). These differences are evident in the maximum ages observed in each population. For example, in the north-western Atlantic, fish aged 12 years or older occur regularly (Robillard et al. 2009); however, the other populations rarely contain fish older than 6–8 years (van der Elst 1976; Ceyhan et al. 2007; Smith et al. 2013; Leigh et al. 2017).

Given the potential for such large variability in life-history parameters among populations, it is necessary to estimate population-specific parameters. The eastern Indian Ocean, south-western Pacific Ocean and eastern Atlantic Ocean P. saltatrix populations are the only stocks without peer-reviewed studies on their growth and mortality rates. The two Australian stocks (eastern Indian Ocean and south-western Pacific Ocean) are genetically distinct (Nurthen et al. 1992), and stock assessments have been conducted (Smith et al. 2013; Leigh et al. 2017; Lovett et al. 2020).

A key aspect of stock assessment is age estimation, which defines the important parameters of growth and mortality, the two most influential demographic characteristics controlling the productivity of fish populations (Campana and Thorrold 2001). If these parameters are not accurately understood, fished populations may be at risk of becoming overfished. In Australia, P. saltatrix age composition, growth and mortality are underpinned by ages estimated from counts in annuli in whole otoliths (Leigh et al. 2017). However, the use of whole otoliths has not been fully accepted as an ageing method for P. saltatrix in Australia, so some uncertainty still surrounds the growth, mortality and age structure of Australian P. saltatrix populations. Hoyle et al. (2000) and Brown et al. (2003) validated the annual periodicity of growth checks in whole otoliths in eastern Australia by using fluorescent marking methods (Brown et al. 2003) and marginal increment analysis (Hoyle et al. 2000; although the results were conclusive only for otoliths with one translucent ring), but comparisons of whole and sectioned otoliths were limited. This is despite sectioned otoliths being the preferred method of age estimation in the north-western Atlantic population, where annuli have been validated with high precision (Robillard et al. 2009).

In eastern Australia, P. saltatrix has historically supported important commercial and recreational fisheries (Lovett et al. 2020). Total commercial catch taken from eastern Australia was ~120 tonnes in 2020, with recreational harvest being estimated to be of a similar magnitude (Murphy et al. 2020; Teixeira et al. 2021). Owing to the importance of P. saltatrix to commercial and recreational fisheries in eastern Australia, together with past concerns for the stock, it is important to understand its key life-history parameters to accurately assess and sustainably manage the south-western Pacific (eastern Australia) population. Fishery-dependent monitoring of the length and age structure of the recreational and commercial harvest is routinely undertaken by state fisheries agencies. To complement this, the aims of this study were to (1) validate the use of whole otoliths as an appropriate ageing tool, (2) estimate growth and mortality rates, and (3) compare these key life-history parameters with those in other global populations of P. saltatrix, so as to identify key differences and potential drivers of differences in life history.

To model growth using size-at-age data, otoliths were collected from 4049 P. saltatrix individuals in eastern Australia (25–37°S) that were sourced from both the commercial and recreational fishing sectors in conjunction with targeted sampling of juvenile fish (Table 1). In New South Wales (NSW), fish from the commercial sector were sampled from ports of landing (the Sydney Fish Market and regional fishermen’s co-operatives). NSW recreationally harvested fish were sampled through surveys of beach fishers and donations of biological material from a recreational citizen science program (NSW Department of Primary Industries Research Angler Program https://www.dpi.nsw.gov.au/fishing/recreational/resources/fish-tagging/researchangler). Where possible, all fish were measured for fork length (FL) and total length (TL) to the nearest 0.1 cm, and weight to the nearest gram, and sex was determined by macroscopic examination of the gonads (Schilling et al. 2019). Otoliths were removed by dissection in the laboratory, cleaned in fresh water and stored dry.

Table 1.  Collection information for the two otolith datasets used to estimate the age of Pomatomus saltatrix in eastern Australia.

In Queensland (Qld), fish were sampled between the NSW–Qld border (28.1°S) and Baffle Creek (24.5°S) through the Fisheries Queensland Fishery Monitoring Program (Fisheries Queensland 2013). Commercial catches, to be representative of the Qld harvest, were routinely sampled year-round from fishers’ residences and seafood processors. Recreational catches, to be representative of the Qld harvest, were routinely sampled year-round using a variety of sampling strategies, including boat-ramp surveys, roving beach surveys, competition sampling and the Keen Angler Program. Whole fish or fish frames were measured for FL to the nearest 1 cm, and, where possible, sex was recorded. Otoliths were removed from a subset of fish (with an annual calendar year cap of 10 otoliths per each 1-cm-FL length class), cleaned in fresh water and stored dry.

This research includes computations using the computational cluster Katana supported by Research Technology Services at UNSW Sydney (https://doi.org/10.26190/669x-a286). This research was approved by the NSW Department of Primary Industries Fisheries Ethics Committee (ACEC# 14/14) and fish were collected under NSW DPI Scientific Collection Permit (P03/0086(F)-8.1).

To validate the use of whole otoliths as an ageing method for P. saltatrix, increment counts from whole otoliths were compared to increment counts from sections of the same otoliths in a blind reading (randomised order), by first reading whole otoliths then sectioning the same otoliths. Increments were defined as opaque rings on the otoliths. Age estimates from sectioned otoliths have been fully validated and show high precision in the north-western Atlantic P. saltatrix stock (Robillard et al. 2009). Using the NSW Department of Primary Industries (DPI) Fisheries database, five fish were randomly selected from each 5-cm-FL size bin (16 bins; 0–80 cm). All available otoliths were used for size bins with fewer than five otoliths (total n = 71). The left otolith was used for consistency, except when it was broken. Whole otoliths were read by immersing the otolith in water and illuminating them by using reflected light.

Interpretation of otolith increment macrostructure was undertaken according to the fish ageing protocols for the Fisheries Queensland Fishery Monitoring Program Ageing protocols (Fisheries Queensland 2017, 2018), with some minor modifications such as immersion in water instead of oil, a variation to the classification of edge type and a variation in the reader training and qualification procedure. Briefly, the methods used here involved counting the number of annual increments between the primordium and the otolith edge, starting with the first annual increment approximately forming near the anterior edge of the antirostrum (Brown et al. 2003), and an edge classification assigned as ‘thin’, ‘medium’ or ‘wide’. A ‘thin’ edge was defined as <20% width of the previous translucent zone or increment not fully visible around the whole edge of the otolith; a ‘medium’ edge was defined as 20–70% width of the previous translucent zone; and a ‘wide’ edge was defined as >70% width of the previous translucent zone. Images were taken using a dissecting microscope (Leica M80, Switzerland) with HD camera at 7.5× magnification. For each otolith, two whole otolith readings were performed without knowledge of fish length or sample data.

To prepare a sectioned sample, an otolith was embedded in resin and sectioned transversely across the primordium to 30–35-μm thickness by using a low-speed saw. The section was polished to remove saw marks, and covered with mounting medium (Safety Mount #4, Fronine, Australia) and a coverslip. Sectioned otoliths were read using the procedure described by Robillard et al. (2009); however, reflected light was used instead of transmitted light and the sections were not burnt. For each sectioned otolith, two counts were made without knowledge of fish length or sample data. An example of an otolith read using both methods is included in the Supplementary material (Supplementary Fig. S1).

The two whole otolith reads of each otolith and two sectioned otolith reads of each otolith were averaged for use in the comparison. Within-method and among-method comparisons were conducted using age-bias plots and average percentage error (APE). As recommended by McBride (2015), bias was assessed using a combination of a maximally pooled method (McNemar 1947), a diagonally pooled method (Evans and Hoenig 1998) and an unpooled method (Bowker 1948).

APE was defined as:

where x_ij is the ith age for the jth fish, is the mean age for the jth fish, R is the number of times that each fish was aged (assumed to be the same for all fish, here R = 2), and n is the number of aged fish in the sample (Beamish and Fournier 1981), here n = 71. Bias and precision calculations were conducted using the ‘FSA’ package (ver. 0.8.20, D. H. Ogle, see https://cran.r-project.org/package=FSA) in R (ver. 3.4.3, R Foundation for Statistical Computing, Vienna, Austria, see https://www.R-project.org/).

Age estimates were collated for two otolith collections taken from NSW and Qld (Table 1), the two eastern Australian states where P. saltatrix is most abundant. All otoliths were read using whole otoliths, following the Fishery Monitoring Fish Ageing Protocols (Fisheries Queensland 2017, 2018), except for three otoliths that were broken but had intact cores, and so were read from sectioned otoliths. For the Qld collection, otoliths are read annually, immediately after a reader passes a competency test on a reference collection. Readers must exceed a set level of precision and bias for increment count and edge type (Fisheries Queensland 2017, 2018). To pass the competency test, a reader must not show unacceptable bias or precision. Unacceptable bias for this species occurs when the calculated bias is ≥1 increment for all groups of increment counts or when the calculated bias shows a divergent trend across increment counts. Unacceptable precision occurs when the index of average percentage error (IAPE) for increment counts is ≥8%. Readers must also obtain an average agreement on edge classifications (New 50%, Intermediate 50%, Wide 70%). After reading the annual sample, a reader must immediately re-read 200 otoliths and pass the same prescribed level of precision and bias for increment count and edge type before the data are added to the database. For the NSW collection, otolith reading was conducted in batches over 2 months by a single qualified reader (having recently passed the Qld competency test described above). After every batch of otolith readings, 200 random otoliths were re-read and tested for precision and bias to ensure consistent interpretation before the data were entered into the database.

Annual formation of increments has been validated for sectioned P. saltatrix otoliths in the north-western Atlantic by using marginal increment analysis (MIA) up to Age 8 (MIA; Robillard et al. 2009), and previously in eastern Australia by using otolith fluorescence marking with a tag–recapture program (Brown et al. 2003) and MIA (Hoyle et al. 2000). The previous MIA from this region showed only clear results for otoliths with a single translucent ring. To further validate annual increment formation in eastern Australia in this study, MIA and assessment of the otolith edge were conducted on the whole otoliths in the NSW dataset. Measurements of the distance from the most recently formed increment to the otolith edge and between the most recently formed increment and the previous increment in whole otoliths, were taken to calculate the marginal increment ratio (MIR). The otolith edge was also classified as being translucent or opaque. For otoliths that displayed translucent edges, the width of the translucent edge was recorded. The translucent-edge widths were defined relative to the width of the previous translucent increment as narrow (5–30%), medium (30–60%) or wide (>60%). MIA and edge-type analysis were restricted to fish with one or two visible increments, to ensure sufficient sample sizes each month. All otoliths were examined using the camera and microscope setup described above. The MIA classification rules were slightly different from the edge-classification rules used in the ageing analysis (to conform to local Fisheries standards); however, as the sole purpose of the MIA was to confirm the annual formation of otolith increments, this did not affect the ageing process.

To provide a better estimate of juvenile growth rates in eastern Australia, a subset of 61 otoliths from Age-0 fish were hand-polished with fine lapping paper until daily increments could be counted. These fish ranged in size from 3.9 to 19.5 cm FL. Daily increments in young-of-the-year P. saltatrix have previously been validated (Nyman and Conover 1988), but it has been difficult to identify daily growth rings in older fish (Brown et al. 2003).

The ageing method (Fisheries Queensland 2017) advises that to assign an age class (number of birthdays a fish is assumed to have), otoliths with a ‘wide’ edge observed during the months of increment formation have one increment added to their increment count to account for the delay in the increment formation in that individual. This adjustment corrects the age of fish in which the otolith increment should have formed in the last month, but is not yet visible on the very edge of the otolith (Fisheries Queensland 2017, 2018). To assign each fish to an age class, we followed the adjustment procedure in Fisheries Queensland (2017), whereby, according to the month of capture and type of otolith edge, the increment count was adjusted to generate an age class. If the edge type was ‘new’ and the capture month was August, then the age class was equal to the increment count – 1. If the edge type was wide and the capture month was September–December, the age class was equal to the increment count + 1. In all other scenarios, the age class was equal to the increment count. These adjustments account for the fact that the marginal opaque zones (new increments) do not become visible in all fish at the same time and that all cohorts of tailor move from one age class to the next on the nominal birthdate of 1 September. A biological age (months) for each fish was calculated using the age class of each fish and the following formulae:

where A_m is biological age (months), C_m is capture month, and B_m is biological birth month (9 = September in our circumstances).

Before calculating growth rates, the biological age in months or daily increments was converted to decimal years. Fish of Age-class 0 (years) that did not have their age in days estimated were excluded from growth model analyses (n = 659).

Because of the heteroscedastic error structure of the data, owing to the Age-0 fish being aged on a scale (days) different from the annuli-aged fish (years), multiple growth models were fitted to the size-at-age data within a Bayesian framework, which allowed the variance parameter for the Age-0 (daily aged) and adult (annually aged) fish to vary. This process ensured more accurate parameter and error estimations than a shared variance. Eight growth models were fitted, including von Bertalanffy, linear, logistic, power, Schnute Variation 1, Schnute Variation 2 (equivalent to Gompertz), Schnute Variation 3 and Schnute Variation 4 (Supplementary Table S1). Model selection was undertaken on the basis of the-leave-one-out information criterion and the difference in expected log pointwise predictive density from the best model. The Bayesian model-fitting process involved four chains using Markov-chain Monte Carlo sampling with a 25 000 iteration burn-in (warm up) and 25 000 iteration sampling period with a thinning parameter of 5. We used semi-informative priors to assist in model convergence (Supplementary Table S2). Model convergence was assessed by visually inspecting the chains to ensure mixing and checking the values were ≤1.01. This was undertaken using the ‘brm’ function in the ‘brms’ R package (ver. 2.16.3, see https://github.com/paul-buerkner/brms; Bürkner 2017), with models fit using a High Performance Computing system (Katana, University of New South Wales, see https://doi.org/10.26190/669x-a286).

Juvenile growth rates were calculated using the same model selection process as described above, but using only otoliths that were daily aged. We present the best-fitting model in addition to the mean growth rate from a linear model. Using this fitted juvenile growth curve, the predicted birth dates of all small (5–20 cm FL) P. saltatrix individuals collected as part of this study were estimated. It is worth nothing that the juvenile collection may not be representative of the population.

Using the best-performing overall growth model, we refit sex-specific growth curves. The daily aged juveniles (unknown sex) were included in both male and female models to help anchor the growth curves. Differences between sexes were assessed by comparing the 95% credible intervals for the parameters in the growth model.

The fishery for P. saltatrix in eastern Australia comprises both recreational and commercial harvests in multiple management jurisdictions. To estimate instantaneous total mortality (Z), an overall catch composition was compiled by combining weighted length frequencies representative of annual commercial and recreational catches from both NSW and Qld (Litherland et al. 2016).

To generate representative length frequencies for the Qld commercial P. saltatrix fishery, gill-net, tunnel-net and ocean beach-net catches were sampled year-round as described above. Commercial catch records were used to post-weight the length samples by using spatial and temporal stratification and are considered representative of the harvest (Fisheries Queensland 2013).

To generate representative length frequencies for the Qld recreational P. saltatrix fishery, line catches were sampled year-round as described above. Recreational catch estimates from the state-wide recreational fishing survey (Webley et al. 2015) were used to post-weight the length samples by using spatial stratification and are considered representative of the harvest (Fisheries Queensland 2013). Sampling was undertaken to be representative of the harvest. All fish were measured for FL to the nearest 1 cm, and the percentage of catch generated for each 1-cm-FL bin (Table 2). The length frequency was supplied as the percentage of catch per 2-cm-TL-class bin length. These TL bin percentages were back-converted to FL by using our FL and TL relationship (Eqn 3) and the proportion method of Booth and Isted (1997). Briefly, this method involves all possible TL measurements between 0 and 90 cm (to nearest 0.1 cm) for P. saltatrix being converted into FL by using Eqn 3. The proportion of these TL measurements in each FL centimetre size class was noted and used to construct a TL–FL key in the form of a matrix, which was then multiplied by the number of measurements within each TL size class, generating the predicted FL length frequency data.

Table 2.  Sample sizes of the datasets used to generate length frequency distributions for Pomatomus saltatrix in eastern Australia.

To generate a NSW commercial harvest length frequency, samples of NSW commercial fishing catches were measured to the nearest 1 cm rounded down (FL) in various locations along the coast (Table 2). The state-wide commercial-catch records and gradings were used to scale up length–frequency samples on the basis of latitude, sample grade and month, similar to the Qld data (Stewart et al. 2018). The recreational length frequencies for NSW were determined by measuring fish caught by recreational fishers (Table 2). These measured fish came from a combination of boat ramp surveys, walking beach surveys and fish donated by recreational fishers through the NSW DPI Research Angler Program.

Age compositions were generated using a combined age–length key that was created using all the aged fish from both NSW and Qld otolith collections (NSW and Qld) and the combined representative harvest (weighted by relative harvest estimates) length frequency (Supplementary Table S3). All length frequencies were binned to 2-cm-FL bins to correspond to the coarsest bins in any dataset (the Qld datasets). Ages were assigned to fish using the method described in Isermann and Knight (2005). The length composition of the population was assumed to be captured by the length composition of the harvest above a certain size (see Supplementary Table S4) and was used to estimate Z by using the Chapman–Robson method (Chapman and Robson 1960). The first age group included was 1 year older than the age of peak abundance (see Supplementary Table S5), as recommended by Smith et al. (2012). This accounts for the truncated size distributions resulting from the minimum legal lengths. The Chapman–Robson estimator presented by Hoenig et al. (1983) is:

where is the mean age of fish in the sample of fish that are greater than or equal to age T_C, T_C is the age of full recruitment, and N is the sample size of fish greater than or equal to age T_C. The first term transforms the Chapman–Robson estimate of survival into an estimate of Z, and the second term reduces the bias that is induced by the transformation. Age compositions and Z were calculated using the ‘FSA’ package (ver. 0.8.20, D. H. Ogle) in R (ver. 3.4.3, R Foundation for Statistical Computing).

Comparing the life-history parameters of the eight major global populations of P. saltatrix required a compilation of the most recent estimates of growth and mortality from the five populations from the eastern Atlantic, north-western Atlantic, south-western Atlantic, western Indian and eastern Indian oceans. The two populations (eastern and western) of the Mediterranean Sea showed the same life history and so were considered together (Ceyhan et al. 2007; Sabatés et al. 2012). The collected data from the literature were combined with the estimates in this study for the south-western Pacific Ocean population.

In assembling the data, peer-reviewed literature was used when available; however, there are no estimates of total or fishing mortality available for the eastern Atlantic population. The growth model provided for the eastern Atlantic population (Champagnat 1983) did not match the size–at-age data or the fitted curve; therefore, parameters for this equation were recalculated using the size-at-age data in Champagnat (1983). A von Bertalanffy growth curve was fitted using the ‘FSA’ package (ver. 0.8.20, D. H. Ogle) in R (ver. 3.4.3, R Foundation for Statistical Computing).

All reported growth models for global P. saltatrix populations took the form of the von Bertalanffy growth function (VBGF); however, the parameters of the VBGF were not compared because some of these growth models had large |t₀| values (>1). This suggests poor modelling of juvenile growth and therefore a questionable k growth parameter (Supplementary Table S6). This is because if the t₀ value is not accurate, the slope of the growth curve (the rate of growth) is likely to be biased (either too shallow or too steep). Thus, comparisons of size-at-age and growth rates were made among populations. Comparisons of growth rates among populations therefore consisted of (1) comparing Sizes-at-age 1–5 years, and (2) comparing growth in the year preceding each of these ages. These ages were chosen because the growth equations for each population fit the data reasonably well across this age range. Size-at-age was calculated directly from the reported growth equations. All populations were constrained to a biologically relevant size 0 cm FL at Age 0 to overcome the poor fits to juvenile growth in the reported studies by using the VBGF to provide an estimate of growth rate in the first year.

A single large P. saltatrix individual (90 cm FL, 103 cm TL, 7.8 kg, female) was donated to the NSW DPI Research Angler Program in May 2018, after the sampling period of this study had finished. This fish was captured during November 2017 in the atypical environment of an intermittently closed and open lagoon (ICOLL) on the NSW south coast (St George’s Basin; 35.12°S, 150.60°E). This fish was by far the largest individual recorded in the current study and was estimated to be double the maximum biological age for the other 4049 aged fish (14 years; Supplementary Fig. S2) and larger than all 31 778 fish measured as part of the length–frequency sampling. The exceptional size and longevity of this individual may be due to its atypical collection environment where it may have been trapped. As an outlier, this individual was not included in the statistical comparisons of whole and sectioned otoliths or growth models.

Using the sampled fish from the NSW portion of the south-western Pacific population, the relationship between TL (cm) and FL (cm) was described by (r² = 0.997, P < 0.001):

and the relationship between FL (cm) and weight (W; g) was described by (r² = 0.998, P < 0.001):

No significant bias was observed between the mean whole increment counts and mean sectioned otolith increment counts for up to six increments (Fig. 2, Table 3). The differences in mean increment counts between both methods were all ≤1 increment and spread evenly across predicted ages up to Age 6 (Fig. 2a) and sizes up to 80 cm (Fig. 2b). Five otoliths were available in all length classes except for the following: 60.1–65 cm FL (four otoliths), 65.1–70 cm FL (three otoliths), 70.1–75 cm FL (three otoliths) and 75.1–80 cm FL (three otoliths).

Fig. 2.  Comparison of increment counts on Pomatomus saltatrix otoliths read using both whole and sectioned methods, showing (a) comparison of mean whole otolith increment counts and mean sectioned otolith increment counts, and (b) the difference between sectioned and whole increment count fork lengths (cm). The solid blue line represents a LOESS smoother and the grey band represents the 95% confidence region in both plots. The dashed lines represent the relationship if there was no difference. Note a jitter has been applied to all points to prevent overlap (n = 81). Increment counts for each method were calculated as the mean of the two independent reads.

Table 3.  Results of the precision and bias comparisons within and between otolith preparation methods using 71 otoliths from the NSW dataset.

The variable appearance of the whole otoliths was reflected by a medium precision when re-reading the otoliths (APE = 4.51; Table 3). Nonetheless, there was agreement in increment counts for 92.6% of otoliths. For the otoliths with inconsistent counts, there was a maximum difference of one between the first and second readings. The APE for sectioned otoliths was also acceptable, although greater than for whole otoliths (APE = 6.678; Table 3).

For the exceptionally large P. saltatrix, eight increments were observed in the whole otolith, but when sectioned, there were 13 increments clearly visible (Fig. S2). The otolith was also considerably more opaque than were otoliths from all other sampled fish and was of an irregular shape compared with otoliths from most large P. saltatrix individuals.

Using whole otoliths, both MIA and assessment of otolith edge-type analysis throughout the year confirmed that one annulus was formed per year, which is consistent with other global populations. For the south-western Pacific population, this annulus was shown to form between September and January (Fig. 3).

Fig. 3.  Validation of annual increment formation for Pomatomus saltatrix in eastern Australia, showing (a) the marginal increment ratio for fish with one visible increment, (b) the edge type of the otolith for otoliths with one and two visible increments, (c) the marginal increment ratio for fish with two visible increments, and (d) the thickness of otolith translucent edges for otoliths with one and two visible increments. These data use only the NSW dataset (Table 1) and the numbers above the error bars show the sample size for each group.

The best-fitting growth model was the Schnute growth model Variation 1 (Table S1). This model is expressed as:

where a = −0.15 (−0.24 to −0.08), b = 2.56 (2.35–2.87), Age 1 (T₁) = 1, Age 2 (T₂) = 4, size at Age 1 (y₁) = 24.38 (23.99–24.89) cm FL, size at Age 2 (y₂) = 47.36 (47.00–47.71) cm FL (Fig. 4; estimates are the median and 95% credible intervals for the parameter estimates). Unlike the VBGF, this model has no asymptotic length. The sex-specific growth models (Schnute Variation 1) showed that female fish grow faster than male fish, although this difference is minor (female being 1.2 cm larger, on average, at Age 1 and 2.2 cm larger at Age 4, Table 4, Supplementary Fig. S3).

Fig. 4.  Size at age (biological age) for Pomatomus saltatrix in the south-western Pacific population (eastern Australia; n = 3390). The solid blue line shows the fitted Schnute growth curve (Variation 1), with the dark grey ribbon showing the 95% credible interval and the light grey ribbon showing the 95% prediction interval. Note the semi-transparent points that show the density of data.

Table 4.  Estimated parameters for the Schnute growth curve (Variation 1) for adult Pomatomus saltatrix from the south-western Pacific (Eqn 5).

The VBGF did not accurately describe the growth of smaller or large fish (Fig. 4). We report the fit because the VBGF is the standard growth curve in all fish growth literature; however, we do not recommend the use of these values. There was large variance in the parameter estimates; the best-fitting parameters (with 95% CI) were L_∞ = 141.7 cm FL (116.7, 169.8), k = 0.07 year⁻¹ (0.05, 0.09) and t₀ = −2.02 (−2.24, −1.79).

Using only the data from fish that had their daily ages estimated (sizes: 3.9–19.5 cm; ages: 32–236 days, n = 61), juvenile P. saltatrix grew ~0.09 cm day⁻¹ (Bayesian linear model; 95% CI 0.08–0.09, size at Age 0 = 1.88 mm (95% CI: 0.97–2.77) r² = 0.87). The best-fitting growth curve between FL (cm) and age (d; days) is given by the logistic equation (n = 61, P < 0.01):

where L_∞ = 20.50 (95% CI: 20.31–20.69) cm FL, k = 0.02 (95% CI: 0.02–0.02) and t₀ = 94.31 (95% CI: 0.02–0.02; Fig. 5) days. Using this equation, the estimated birth dates for 659 juvenile fish suggested that some spawning may be occurring in all months except March and June (Supplementary Fig. S4).

Fig. 5.  Size at age for juvenile Pomatomus saltatrix, which were aged using daily increments with a logistic growth curve (n = 61). The blue line shows the fitted growth curve. The dark grey ribbon shows the 95% credible interval of the growth curve and the light grey ribbon shows the 95% prediction interval around the growth curve.

In total, 95% of the harvest in eastern Australia (south-western Pacific Ocean population) was between 28 and 46 cm FL, which represents fish aged 1–3 years old, with the majority of the harvest being 2 and 3 years of age (Fig. 6, Tables S5, S6). Because of the truncation of sizes owing to the larger minimum legal length (MLL) in Qld (35 cm TL) than in NSW (30 cm TL), the Qld harvest had proportionally more 3-year-old and fewer 1-year-old fish than did NSW. Very few large old fish were collected from eastern Australia in the present study, with just 35 individuals sampled being >60 cm FL and 11 fish being >5 years of age (Fig. 4, 6). Excluding the outlier described above, the largest (and oldest) P. saltatrix individual sampled was 80.7 cm FL and 7 years old (Fig. 4, 6). Using Ages 3–7, total mortality (Z) was estimated to be 1.62 (95% CI 1.43–1.82).

Fig. 6.  Representative length composition of harvested Pomatomus saltatrix in eastern Australia. Each bar represents a 2-cm length class (n = 31 778).

Except for the P. saltatrix population in the north-western Atlantic, which showed a low estimate of Z, all other global P. saltatrix populations had moderate to high estimates of Z (0.90–1.62; Table 5). This pattern is reflected in the maximum observed age in each population, with the north-western Atlantic population containing the oldest fish. Growth was similar between all populations up to approximately Age 3. After this age, the north-western Atlantic and eastern Atlantic populations showed larger sizes at each subsequent age (Fig. 7, 8). All populations, except the south-western Atlantic and eastern Atlantic, showed that fast growth reached a plateau after Age 2 (Fig. 7). Juvenile growth (<1 year) was poorly described by the VBGF with |t₀| values being >1 in all populations except for the eastern Indian, eastern Atlantic and north-western Atlantic ocean populations (Fig. 7, Table S6).

Table 5.  Summary of published life history parameters for global populations of Pomatomus saltatrix.

Fig. 7.  Comparison of the growth curves of different populations of Pomatomus saltatrix around the world (Table 5). All populations except the south-western Pacific Ocean (eastern Australia; this study) are taken from the literature as follows: eastern Atlantic (Champagnat 1983), south-western Atlantic Ocean (Brazil; Haimovici and Krug 1996), Mediterranean Sea (Ceyhan et al. 2007), western Indian Ocean (South Africa; Govender 1999), north-western Atlantic Ocean (Robillard et al. 2009), eastern Indian Ocean (western Australia; Smith et al. 2013) and are restricted to the sizes modelled in the original research.

Fig. 8.  Visual representation of the growth rates of each population up to Age 5. Growth in previous year was determined by using the fitted growth curve for each population in Fig. 7. All populations were assumed to be 0 cm FL at Age 0 to overcome the poorly fit juvenile growth in the published growth curves.

The south-western Pacific Ocean (eastern Australian) P. saltatrix population is characterised by fast, non-asymptotic growth up to the normal maximum reported age (7 years). Modelled growth of the species likely fails to reach an asymptotic length because of high mortality rates (i.e. low numbers of old, large fish). Growth and mortality in P. saltatrix shows two main patterns among global populations. All populations have fast growth and high mortality rates, except for the north-western and eastern Atlantic ocean populations. The north-western Atlantic population has fast growth but considerably lower mortality rates, which is reflected by higher proportions of large, old fish not found elsewhere. The general similarity in life histories over such a broad geographical range indicate that P. saltatrix represents a key example of the pelagic mesopredator niche, but one that may be strongly influenced by fishing pressure as demonstrated by the potential for high mortality rates.

Our study has validated the use of whole otoliths for accurate age estimation of P. saltatrix for fish routinely found in catches in eastern Australia, with sections being better for atypical exceptionally old P. saltatrix. Increments counted in whole otoliths showed close agreement with those counted in sectioned otoliths up to estimated Age 6. Whole otoliths are used for ageing P. saltatrix in the Mediterranean, eastern Indian, western Indian and the south-western Pacific populations (Govender 1999; Ceyhan et al. 2007; Leigh et al. 2017), which are regions where large, old individuals are scarce.

Marginal increment and edge-type analyses both showed one annulus forming per year. This agrees with previous validation using sectioned otoliths reported from the north-western Atlantic (Robillard et al. 2009) and south-western Pacific (Hoyle et al. 2000). Our study found that formation of increments occurred earlier in the year (c. November) than in a previous marginal increment analysis (December–January), which is likely to be due to the previous use of sectioned otoliths (Hoyle et al. 2000).

In the current study, we conclude that whole otoliths are a valid method of age determination in the south-western Pacific up to Age 6. Our conclusion is based on validating only the first two increments, although there was high agreement with sectioned otolith counts, which have been validated up to 8 years in the north-western Atlantic (Robillard et al. 2009). The current validation agrees with the previous validation work of the first increment and increment periodicity in the south-western Pacific (Hoyle et al. 2000; Brown et al. 2003), and confirms the high agreement (>95%) between increment counts in sectioned and whole otoliths up to Age 4 in the north-western Atlantic (Sipe and Chittenden 2002). Despite the close agreement between whole and sectioned otoliths up to six increments, there was a large difference in increment counts between whole and sectioned otoliths in the largest fish. This was due to the difficulty in interpreting whole otoliths of large fish, particularly increments near the otolith edge. Therefore whole otoliths can be used as an alternative to sectioned otoliths for routine ageing of P. saltatrix in eastern Australia; however, the otoliths from fish larger than 70 cm FL should be sectioned to ensure unbiased age assignment.

The daily growth analysis showed a fast juvenile growth rate of ~0.09 cm day⁻¹, which is within the reported juvenile growth rate in the north-western Atlantic (0.09–0.21 cm day⁻¹; Juanes et al. 1996), faster than juvenile growth in the eastern Indian Ocean based upon tagging (0.02–0.04 cm day⁻¹; Young et al. 1999), and aligns with previous estimates for the western Pacific (0.075–0.13 cm day⁻¹) based upon length frequencies (NSW SPCC 1981). The differences among these populations could potentially be due to the relatively low productivity of the eastern Indian Ocean, in comparison to the high productivity of the north-western Atlantic and western Pacific (Longhurst et al. 1995; Reygondeau et al. 2013). Using our estimated growth rates, we calculated birth dates that span most of the year. Although we saw an obvious peak in births between August and December in our samples, this may be due to biased sampling for juveniles conducted mainly between February and May in NSW. This extended spawning period aligns with previous research, which showed that spawning was not restricted to August–December as originally thought, and confirmed that reproduction and recruitment may occur over a prolonged time period (Schilling et al. 2019).

The model selection process used in our current study found that the classic VBGF was not well suited to describing growth for the south-western Pacific Ocean (eastern Australian) population. This is because it did not accurately describe juvenile growth and our data did not show a clear asymptotic length. We found that a model without an asymptotic length (Schnute Variation 1) was more accurate. Although few studies of growth in fish suggest models with no asymptotic lengths (e.g. Stewart et al. 2004), this case is consistent with the high mortality rates estimated for the south-western Pacific population, whereby few fish survive long enough to provide size-at-age data from large old fish. Despite the overall better fit provided by the Schnute Variation 1 growth model, this model still underestimated growth of larger, older classes of fish. Because biological theory suggests that there should be an asymptotic length, we caution about extrapolating our given growth rates beyond the ages presented in this study. For example, our growth rate if extrapolated would estimate a 14-year-old fish to be 112 cm FL; however, our large 14-year-old outlier was only 90 cm FL. By including the outlier fish in the growth models, this may have slightly changed the shape of the curve, but this fish may be the outcome of highly atypical growth conditions not experienced by the rest of the population. We found that female P. saltatrix individuals grew slightly faster than did males, which is consistent with the fact that females mature at slightly larger sizes than males (Schilling et al. 2019). Sex-specific differences in growth rates have not been observed in the north-western Atlantic population (Robillard et al. 2009), but have been observed in the Mediterranean (Ceyhan et al. 2007).

All previous studies of P. saltatrix growth have provided estimates of the VBGF parameters; however, these model parameters are likely to be biased by poorly modelled juvenile growth (Fig. 7, Table S6), and therefore should not be compared. Our comparison of growth rates shows that all populations except the south-western Atlantic share an initial fast growth rate (20–30 cm in the first year) before slowing in the second year. The different growth rate observed in the south-western Atlantic population (Haimovici and Krug 1996) may be due to ageing errors, with the first annual increment potentially being misidentified because of the use of scales instead of otoliths to derive age estimates. If this were the case, then the growth patterns for this population would be likely to match the pattern observed in most other populations. The north-western Atlantic population has rapid initial growth that is maintained longer than in the other populations, potentially owing to delayed investment in reproduction, because this population has a larger size at maturity than do the other populations (van der Elst 1976; Champagnat 1983; Robillard et al. 2008; Schilling et al. 2019).

Differences in growth rates among and within populations of other species have previously been linked to temperature, with faster growth occurring in higher-temperature environments because of increased metabolic rate (Morrongiello and Thresher 2015; Nicolle et al. 2022). Given that all global P. saltatrix populations experience similar average annual water temperature (Goodbred and Graves 1996; Fig. 1), it is considered unlikely that variations in environmental temperature are driving the difference in growth between the north-western Atlantic population and the other populations of P. saltatrix. However, migrations are not considered when simply looking at average annual sea-surface temperature over the whole species distribution, and all populations of P. saltatrix migrate. For example, the south-western Pacific P. saltatrix population is heavily influenced by sea-surface temperature, with abundance peaking at 21.5°C (Brodie et al. 2018). Population-specific studies may be required to fully understand the importance of environmental temperature for driving basic life-history parameters, such as growth, in P. saltatrix populations.

It is also possible that growth rates are not a response to temperature, but rather regional productivity, although there are still mismatches in productivity and growth rate. The north-western Atlantic and eastern Atlantic populations, which have the highest growth rates, are not the regions with the highest regional productivity, whereas the south-western Atlantic population also occurs in an area with high regional productivity and does not have a higher growth rate (Antoine et al. 1996; Behrenfeld et al. 2006; Stock et al. 2017). Despite the variations in temperature and productivity, growth rates are remarkably consistent among all the populations. This suggests that neither productivity nor temperature are the primary driver of growth rate differences among different populations of P. saltatrix.

We propose the most likely cause of these differing growth rates is fishery-induced evolution, with the populations with higher levels of Z (either from M or F) investing in reproduction earlier and therefore slowing growth (Kuparinen and Merilä 2007; Quince et al. 2008a, 2008b). This is highlighted by the north-western Atlantic and eastern Atlantic populations having the lowest mortality and the fastest growth rates. Indeed a reduction of fishing pressure has been linked to a larger reproductive size in the south-western Pacific population (Schilling et al. 2019).

Although we did collect data from a single, exceptionally large individual, this fish was an outlier from the population because of its size (far exceeding the length of 31 778 fish sampled) and age (14 years estimated by sectioned otolith, which made it the oldest of the 4049 fish sampled). The exceptional size and longevity of this individual was attributed to its atypical collection environment of an intermittently closed estuary system on the NSW southern coast, despite the typical habitat for adult P. saltatrix being coastal marine waters. Despite the unusual nature of this fish, it does provide evidence that P. saltatrix in the south-western Pacific, under some circumstances, have potential longevity similar to that observed in the north-western Atlantic population (Robillard et al. 2009). It is worth noting that the most recent stock assessment of the population did not include this individual because it was an atypical individual not representative of the population.

Estimates of total mortality (Z) are similar among global populations (0.90–1.62), with the exception of the north-western Atlantic population, which has by far the lowest Z estimate of 0.34. The south-western Pacific (eastern Australian) population has the highest Z, which is likely to be driven by continued heavy recreational fishing pressure on annual spawning aggregations in the north of their distribution and the combination of variable recruitment and a population consisting of very few age classes (Pollock 1984; Zeller et al. 1996; Leigh et al. 2017). It should be noted that catch-curve analyses of Z assume constant vulnerability, constant recruitment and an unbiased sample; however, these cannot be assumed for this population, particularly when recruitment is thought to be variable (Leigh et al. 2017; Lovett et al. 2020). Although the current study combined multiple fishery sectors, it had only five age classes (Ages 3–7) with which to estimate Z. This means that the estimate of Z in this study should be interpreted cautiously because its heavy reliance on fishery-dependent samples means that it may not be representative of the population overall.

All global populations of P. saltatrix show similar spawning migration patterns, and are also subject to various management strategies that attempt to limit fishing pressure (Maggs et al. 2012). The Mediterranean population appears to be the only population (except the eastern Atlantic) that is not subject to strict management. Although the Mediterranean population does not have the highest estimate of Z, it is assessed to be overexploited, with most of the harvest consisting of small fish because of the lack of a minimum size limit (Ceyhan et al. 2007; Cengiz et al. 2013).

Differences in the life-history parameters of growth and mortality are potentially driving differences in other life-history parameters, including reproductive age. In the populations of P. saltatrix that have sustained fast growth and greatest longevity, length at 50% female maturity (L₅₀) is 45.1 cm FL in the north-western Atlantic population (Robillard et al. 2008) and 38.0 cm FL in the eastern Atlantic population (Champagnat 1983). In the other populations, L₅₀ ranges from 25 to 31.5 cm FL (van der Elst 1976; Bade 1977; Smith et al. 2013; Schilling et al. 2019) and is reflected in the slowing of growth rates by Age 2. This could be a strong evolutionary response signal and may reflect the lower Z in the north-western Atlantic population (and potentially the eastern Atlantic; Swain et al. 2007). With higher survival, these populations may delay the start of their reproductive development to increase investment in somatic growth and, therefore, be larger at first spawning, with potentially higher fecundity in the first spawning season (Schilling et al. 2019).

P. saltatrix in the south-western Pacific (eastern Australia) shows fast growth and high mortality, which is consistent with the global pattern of growth and mortality for most P. saltatrix populations. This species demonstrates a persistent life-history strategy as a successful globally distributed mesopredator, across multiple isolated genetic populations, with differences potentially driven by variation in mortality.

Supplementary material is available online.

The data and code used in these analyses are available at https://github.com/HaydenSchilling/TailorGrowthMortality.

The authors declare that they have no conflicts of interest.

This work was supported by an Australian Research Council Linkage Project (LP150100923) and the NSW Recreational Fishing Trust. The funders had no involvement in the design, data collection or analysis of this project.