In total, 934 individuals belonging to the three studied croaker species (Pseudotolithus elongatus, senegalensis, and typus) were sampled in eight different locations in three countries of West Africa. The three croaker species (210 individuals of P. typus, 174 individuals of P. senegalensis, and 48 individuals of P. elongatus) were sampled in January 2022 in Nigeria (three locations: Ibeno Beach, Ibeno, Akwa Ibom State; Ogulagha, Warri, Delta State; and Makoko, Lagos Island, Lagos State) and Ghana [three locations: Apam Beach, Apam, and Elmina Beach Market, Cape Coast (both in the Central Region); and Albert Bosomtwi Sam Fishing Harbor, Sekondi-Takoradi; Figure 1]. Another 502 individuals of P. senegalensis were obtained between August 2015 and July 2016 in the fish markets of Joal and Mboro (Senegal).

All individuals were analyzed in the laboratory to measure, with high precision, individual parameters: total length (TL, measured to the nearest millimeter) and total weight (W_T, measured to the nearest gram). The macroscopic observation of the gonads was applied to evaluate the sex and sexual maturity stage. Finally, both sagittal left and right otoliths were extracted from the cranial cavity, cleaned with water, and stored dry before being shipped to Ifremer (Boulogne sur mer, France).

All individual data of biological parameters were graphically observed to identify and delete the potential outliers. For each species, to estimate the allometric relationships between total length/total weight, a base-10 logarithm was fitted to data:

where a is the intercept and b is the slope named the growth coefficient (14–16). To test the effect of the explanatory variables {sex [Se], sampling country [Co], sampling location [Lo], and reproduction period [Re, defined from November to March; (17)]} on the relationship between total length and total weight, a completed generalized linear model was applied for each species:

With the interaction between each factor and the length (sex: Log TL:S, geographical area: Log TL:Co, and Log TL:Co:Lo, reproduction period: Log TL:Re, and the physiological state by sex: Log TL:Re:Se). When a factor showed a significant effect (p < 0.05) on the TL–W_T relationship, the analysis was realized for each factor modality.

To understand the difference in the TL−−W_T relationship, the condition factor, Le Cren's (14) index, Kn (i.e., fish showing the high value of Kn are heavy for their length), has been used:

The difference between this observed growth coefficient for each fish species and the theoretical value (b = 3.0) was tested with a one-sample t-test.

As there has been no previous direct growth study for these species, calibrating the tests of the aging method was necessary. Whole and sectioned otoliths were tested to analyze the growth ring patterns. The image processing was carried out using the Icy image analysis system (18). Finally, to limit interpretation bias, each image was independently analyzed by two expert readers, and after this step, the images presenting a difference in age were discussed to increase precision, which, for this study, is defined as the reproducibility of repeated measurements on the otoliths or another calcified structure (19). Alternating translucent and opaque bands were visible in whole otoliths. It was assumed that each annual growth ring consisted of one opaque ring and one translucent band. The age was defined by age group; for example, a fish in age group 0 lived between 1 and 364 days (i.e., between hatching and before 1 year), as recommended by international expert groups (19–21).

No growth validation studies (i.e., mark–recapture of wild individuals, captive rearing of either chemically labeled fish of unknown age or known age, etc.) had not been applied to these species, marginal increment analysis (MIA) was used. This technique assesses the periodicity of increments in calcified structures (22). MIA is a quantitative approach that relies on measuring the size of the increment under formation [named the marginal increment (MI)] as the distance between the last growth ring and the edge of the calcified structures. The relative measure of the MI is given by

where R_O is the otolith radius measured from its focus to the edge, R_n is the distance between the focus and the last growth ring n, and R_n−1 is the distance between the focus and the last-but-one growth ring n−1. If growth rings are formed annually, the MI will thus exhibit an intra-annual periodic pattern that can be observed by plotting the MI against the date of origin, that is, the month at which the specimen was captured. The MI was measured for each specimen using the otolith and plotted against the catch month. A sinusoidal regression of MI against the month of capture m with a period of 12 months was used to test for the annual periodicity of growth ring formation after linearization:

so that

with

Linear regression was used to statistically validate an intra-annual pattern in the growth rings. The classical assumptions of the linear models (normality, homoscedasticity, and absence of trends in the residuals) were verified before to realise the growth models.

To optimize the growth model from the sampled individuals, the length at age (TL_t) was back-calculated from the Fraser–Lee procedure (23) after checking the significant linear relationship between fish and otolith lengths:

where TL_t is the length at age at age t, TL_c is total length at capture, TL_bi is length at the biological intercept, R_t is otolith radius at age t, R_o is otolith radius at capture, and R_bi is otolith radius at the biological intercept.

The growth patterns were described according to four different growth models:

where T_{L1, TLt}, and TL_∞ are, respectively, the length at age 1, the length at age t, and the asymptotic length and K is the rate at which the asymptote is reached, also called the growth coefficient. The best growth model was the bias-corrected form of the Akaike information criterion [AICc; i.e., gave the smallest value of AICc; (27, 28)]. The AICc balances the trade-off between the fit quality and the number of used parameters (29) while accounting for small-sample bias and is defined as

where n is the sample size, k is the total number of parameters of the model, and TL is its likelihood.

Fish growth was estimated using the growth performance index [ϕ; (30)]:

The growth performance index was preferred for growth comparison rather than comparing TL_∞ and K individually. These two parameters of the growth model are correlated (31).

To test the growth variation according to the different effects of the selected species, the sampling country, and sexual dimorphism (difference between the females and males) on the back-calculated length, four mixed-effects models were fitted, with the age and the interaction of age/species, age/country, and age/sex as fixed effects. Random effects were used to account for variability due to each individual, which could present several back-calculated lengths for all ages. Finally, the significance of back-calculated length at 5% was tested by likelihood ratio tests.

Statistical analyses were performed using the following packages in the statistical environment R (40): ggplot2 (32), and car (33).

Measured total length and total weight for all species ranged, respectively, from 16.0 to 58.8 cm and from 20.7 to 1,720.3 g (Table 1). Important differences were observed in weight for the same total length among the three croaker species, with P. senegalensis having the highest weight and P. elongatus having the lowest (Figure 2). Among the three croaker species, all showed a significant correlation (p < 0.05) between total length and total weight (Table 1). The body parameters from the length–weight relationships showed that the initial growth coefficient a varied from 0.005 to 0.162, and the growth coefficient b ranged from 2.142 ± 0.211 (for P. elongatus) to 3.132 ± 0.031 (for P. typus). For P. elongatus and P. senegalensis, the specimens had thin, elongated bodies (i.e., b < 3.0), while P. typus have thicker bodies (i.e., b > 3.0). The geographical and physiological effects on the relationship between total length and weight were also analyzed. For the geographical effect, two geographical scales (i.e., at the country level and, after, the location within the country) were used and for all species, the geographical effect was significant showing the difference among the sampling countries and between the locations within the country too (Table 1). The second effect was the physiological effect with two different levels (i.e., between the males and the females, between the reproduction and the sexual rest period, and, finally, the interaction between these two levels). The TL–W_T relationship showed a significant difference between males and females for P. senegalensis and P. typus (p < 0.05), while for P. elongatus, there was no significant difference. The second level of the physiological effect was tested only for P. senegalensis, which presented the monthly sampling for 1 year. For this species, the TL–W_T relationship changed significantly between the reproduction period and the sexual rest period. Finally, this modification during the year depends on the sex analyzed (Table 1).

The fitness of each fish species according to sex (females vs. males), geographic factor (Senegal, Ghana, and Nigeria and, for the sampling country, the mean value was estimated for each location), and the reproduction period (active vs. sexual rest) was followed from the condition factor (K; Supplementary Table S2). Among the three croaker species, only P. typus showed sexual dimorphism, with the mean condition factor of females being smaller than that for males. For the geographical factor, there were some differences for each species between the sampling countries and/or locations. However, no trend was observed for all species from the eastern part to the western part of the Gulf of Guinea.

The whole sagittal otolith and the sectioned otoliths of the same individuals were tested and the aging data showed that the best otolith preparation method was the sectioned otolith. These preparations were analyzed under transmitted light microscopy to count the growth increments along the longest growth axis. The individuals' ages ranged between 1 and 7 years. Before realizing the first growth analysis, applying the aging validation method is very important. The MI analysis of P. senegalensis allowed a significant intra-annual variation in their growth (p < 0.001; Figure 2, upper panel) to be detected. From November to March, the growth of marginal increments is maximal, and from April to October, it is reduced, revealing an annual rhythm in growth ring formation (Figure 3). For P. elongatus and P. typus, the available data were sampled during 2 or 3 months. Consequently, the MIA was applied only to P. senegalensis.

For the three croaker species (P. elongatus, P. senegalensis, and P. typus), the age range was 1–7 years, except for P. typus, which had only one specimen aged to 4 years (Figure 4). The von Bertalanffy growth model constrained with t₀ = 0 was the best fit for the observed age data from the three species in the sampled countries (Senegal, Ghana, and Nigeria; Table 2). The asymptotic length TL∞ varied from 26.6 cm for P. elongatus along the Nigeria coast to 42.2 cm for P. senegalensis along the Senegal coast (Table 2). The rate at which the asymptotic length was reached, K, ranged from 0.58, for P. senegalensis, to 1.12, for P. typus. Consequently, for these croaker species, the growth performance index (ϕ) varied from 2.73 to 3.24 (Table 2). Comparisons of the growth curves of these three species in West African waters showed that, among these species, growth was very different (Figure 4A). For one species, P. senegalensis, the results of mixed-effects models showed that the growth of one species could be significantly different among the samples from the three countries, whereas for the P. typus specimens from Nigeria and Ghana, the difference in their growth was not significant (Figure 4). Finally, the growth difference between males and females was tested for P. senegalensis from Senegalese waters, and no significant effect of sexual dimorphism on the growth of this species was found (p = 0.251; Figure 4).