Sampling occurred along the northeastern coast of Brazil between the states of Rio Grande do Norte (5°S) and Alagoas (9°S) (Fig. 1). This region contains high biodiversity and several areas considered a priority for management and conservation (CBD 2014, Eduardo et al. 2018). The continental shelf is 40 km wide on average and almost completely covered by carbonate sediments (Vital et al. 2010), and the average depth varies between 40 and 80 m (Vital et al. 2010). This area holds many MPAs (e.g. APA Costa dos Corais, APA Guadalupe, APA Santa Cruz and APA Barra de Mamanguape) (Ferreira and Maida 2007, Prates et al 2007).

The samples were collected during the Acoustic Along Brazilian Coast (ABRACOS) project, carried out during two scientific expeditions (Bertrand 2015, 2017). The first expedition was carried out between August 2015 and September 2015 and the second between April 2017 and May 2017, both aboard the R/V Antea. There were 37 sampling stations along the transect of the continental shelf (Fig. 1). Trawls were performed at depths of between 10 and 60 m for 5 minutes with a bottom trawl net (body mesh =40 mm; cod-end mesh =25 mm; mouth aperture =280 m²), with diurnal and nocturnal tows. The geometry of the net was evaluated through the SCANMAR system to give an estimate of its height, depth and aperture width. Temperature, salinity and oxygen profiles were collected for each haul using a Sea-Bird 911 CTD.

Three major types of habitat were identified and classified: (i) sand; (ii) algae; (iii) and sand with rocks, coralline formations and sponges (SWCR) (Eduardo et al. 2018). The 40 km average width of the continental shelf was used as the basis for classifying shelf position, according to the distance of the sampling point from the shore: inner-shelf (<20 km from the shoreline) and outer-shelf (>20 km from the shoreline). Also, the latitudinal gradient was stratified in four levels according to the sampling point position: A (<6°S), B (6°S to 7°S), C (7°S to 8°S) and D (>8°S); these levels were in accordance with Eduardo et al. (2018) for the same area. Onboard, each fish sample was identified, counted, weighed on a motion-compensating scale and preserved in a 4% formalin solution or frozen until further processing. In the laboratory, morphometric data were obtained for each specimen. Stomachs were removed through a ventral longitudinal incision and immediately fixed in 10% formalin solution for 48 h, before being stored in a 70% alcohol solution. The contents of each stomach were sorted, counted, weighed (g) and identified to the lowest possible taxonomic level.

The length-weight relationship (LWR) was determined for grouped and separated sexes by using the log-transformed morphometric data collected in the laboratory. The relationship was expressed using the equation TW= aSLb, where TW is the total weight in g, SL is the standard length in cm, a the intercept (constant), and b the regression slope (allometry coefficient). The allometry coefficients were further tested for significant deviations from the expected b=3 (Student t-test). The condition factor (K) was determined according to the method proposed by Ricker (1975) and adapted by Albieri and Araújo (2010), and was expressed using the equation

where b is the allometry coefficient of the LWR. The allometry coefficient and the condition factor were both tested for differences between habitats, shelf positions, latitude levels and sexes. The standard length was used because of the degradation of various individuals’ tails. However, for purposes of comparison with other published studies that used fork length, the following equation is recommended: FL=0.136+1.062SL.

Gonadal development was assessed by macroscopic evaluation and classified in four stages: A, immature; B, maturing; C, mature; and D, spawned or resting. Length at first maturity (L₅₀) followed the King (2007) classification and was determined only for pooled sexes because of the low the number of sexed juveniles, which prevented the determination of sex and thus separation of the sexes. Sex ratio was determined as the total number of males/total number of females (Vazzoler 1996) and was statistically tested for significant deviations from the expected 1:1 ratio with a chi-squared test (α=0.05).

The relative index of biomass, defined as capture per unit of effort (CPUE), was determined on the basis of the number of fish captured per trawled area (kg km^–2). The area was determined by multiplying the tow distance by the mouth opening area, estimated by the SCANMAR system. Additionally, possible influence of seasonal variables (e.g. temperature, salinity and dissolved oxygen) on the surveys was tested to assess its effects on the CPUEs but, although significant, the results did not show major differences between campaigns. Distribution patterns were evaluated through analysis of the length-frequency distribution and CPUE. To test significant differences in the mean standard length between shelf positions, habitats and latitude levels, the Kruskal-Wallis test was performed after the basic assumptions of normality, homoscedasticity and homogeneity were proven inadequate for a parametric test.

Feeding intensity was classified by observing the degree of fullness as follows: I, empty; II, partially empty; III, partially full; and IV, full (Viana et al. 2010). The Zavala-Camin (1996) fullness index (FI) was used as an indication of feeding intensity. FI differences between habitats, shelf position and latitude levels were tested using the Kruskal-Wallis test.

The contribution of each prey taxon to the diet composition was assessed by following three relative metrics of prey quantity: frequency of occurrence (%FO), numerical abundance (%N) and weight percentage (%W) (Hyslop 1980, Bowen 1996). All items collected in the stomach content analysis were considered for subsequent analyses. For a solid assessment of prey importance, two indexes were calculated: the index of relative importance (IRI) (Pinkas et al. 1971) and the alimentary index (IAi) (Kawakami and Vazzoler 1980), adapted by Oliveira et al. (2004). The IAi is based on frequency of occurrence and on the weight of each item. In contrast, the IRI combines the frequency of occurrence weight and numerical occurrence of each item. We chose to use these two indexes to have a more robust assessment of the diet.

Niche breadth was estimated using Levin’s standardized index (Bi). This index varies from 0 (species that feed on only one item) to 1 (species that feed on the same proportion of all evaluated items) (Levins 1968).

The feeding strategy was characterized using a modified Costello diagram (Amundsen et al. 1996), which is a graphic representation of prey items illustrating the degree of diet variability of a predator by plotting the prey-specific importance (Pi) of each prey taxon against the frequency of occurrence (Fo) in a 2D diagram.

To assess the degree of similarity in the diet, the multivariate technique of multidimensional scaling (nMDS) based on the Bray-Curtis similarity matrix was performed, with stomachs being considered as the sampling units. Relative weight of prey (%W) was square root transformed to reduce any dominance effect of individual prey items. Differences between habitats, shelf position and latitude levels were tested through ANOSIM (Clarke 1993). The items that discriminated the groups were also evaluated using a SIMPER routine (Clarke 1993). The R software (R Core Team 2019) and the vegan package (Oksanen et al. 2019) were used for the multivariate analyses.

Up to ten individuals from each sampled station were selected for isotopic analysis. Fish had white muscular tissue extracted and cleaned with distilled water to remove exogenous material such as scales and bones. The samples were dried in an oven at 60°C for 48 h and ground into a fine powder with a mortar and pestle. Each sample was analysed for carbon and nitrogen isotope ratios through a mass spectrometer (Thermo Delta V+) coupled with an element analyser (Thermo Flash 2000, interface Thermo ConFio IV) in the Platform Spectrometry Ocean (PSO; IUEM, France). Results of stable isotope analysis for carbon (δ¹³C) and nitrogen (δ¹⁵N) are derived from the relation of the isotopic value from the sample and a known standard, δ¹³C or δ¹⁵N=[(R_sample/R_standard) – 1]×10³, in which R corresponds to the ratio between ¹³C:¹²C or ¹⁵N:¹⁴N. The mean values of δ¹³C and δ¹⁵N were tested for differences among habitats, shelf positions and along the latitude gradient. Also, the ontogenetic variation of δ¹³C and δ¹⁵N isotopes was assessed across the continental shelf.

The trophic position of the spotted goatfish across habitats and shelf positions and along the latitudinal gradient was assessed using the formula proposed by Post (2002); TDF is the trophic discrimination factor and was considered to be 2.54‰ (Vanderklift and Ponsard 2003). As there was no long-living filter feeder to be the baseline, the mean δ¹⁵N of zooplankton was used as a reference. Zooplankton community sampling was conducted at each station by horizontal haul with a bongo net, and the samples were organized into Eppendorf micro tubes and freeze-dried for at least 24 hours. To obtain the correct values from δ¹³C, part of the samples was separated to remove the carbonates (CaCO₃). Part of the sample was acidified according to Fry (1988) by adding hydrochloric acid (HCl). After the procedure, the samples were redried at 60°C for 24 hours and homogenized, and subsamples were reanalysed.

To give a more robust representation of the trophic level, the trophic position based on the stomach content analysis was determined to compare with the results of the stable isotopes analysis, based on the methodology proposed by Adams et al. (1983).

Of the 505 individuals captured and identified aboard, 291 were analysed in the laboratory. Standard length (SL) ranged from 4.4 to 23.8 cm (15.05±4.01[mean±SD]), while total weight (TW) varied from 1.04 to 315.01 g (91.38±54.86). From these specimens, 154 gonads were in good enough condition to be retrieved, 28% being males, 34% females, and 38% juveniles whose gonads could not be sexed. In general, growth was classified as positively allometric (b=3.20; p<0.05), and the length at first maturity was estimated to be 13.7 cm SL (CI=12.8-14.4) for the pooled sexes (Fig. 2). Significant differences in the allometric parameter were found between sexes, with females classified as negatively allometric (b=2.75; p<0.05) and males as isometric (b=3.03; p>0.05). When we compared the b values within spatial categories, such as shelf positions and habitats, no significant differences were found. However, the b values varied among latitude levels, being isometric in the A (<6°S) and B (6 to 7°S) strata and positively allometric in the C (7 to 8°S) and D (>8°S) strata.

Overall, females and males were equally abundant (1:1), with no significant differences in the sex ratio (p>0.05). The most representative size class was 16-18 cm, comprising 29% of the individuals. Females were most frequent at lower sizes, with their most representative size class being 16-18 cm; males, mostly represented in the 18-20 cm size class, were most frequent at larger sizes (Fig. 3). Most individuals with sex identified were adults (90.6% adult males and 90.3% adult females). Differences in size frequency distribution were found between SWCR and the other habitats, with a high presence of juveniles in SWCR. Despite this higher prevalence of juveniles, the mean SL in SWCR was higher than that in other habitats due to the bimodal distribution characterized by a high frequency of juveniles and larger individuals (SL>17 cm) and a large range of lengths. Mean SL and size distribution were higher in outer-shelf than inner-shelf areas (p<0.05) (Fig. 4A).

No significant differences in mean SL were found among the latitude levels, however, at stratum D a high frequency of juveniles was found. The CPUE analysis (Fig. 4B) revealed higher abundance of individuals at the southern sampling points (below 8°S). There were also some abundance peaks off the coast of Pernambuco and Paraíba, where fisheries for spotted goatfish are present, and off the northern Rio Grande do Norte coast.

Significant differences in the condition factor (K) were identified among habitats, with SWCR showing lower values than the other habitats (p<0.05) (Fig. 5A). Significant differences in K were also identified among latitude strata (Fig. 5C), with the northernmost areas (<6°S) presenting higher values than other areas (p<0.05), while the southernmost stratum exhibited the lowest K values (p<0.05) (Fig. 4C). No statistical differences in K were found between shelf positions and sexes.

A high proportion of P. maculatus had food contents in their stomach (degrees II, III and IV) (63%), mainly those found in algae (88%) and SWCR (85%) habitats, whereas the proportion of those inhabiting sand with food in their stomach was considerably lower (56%). Also, 65% of those fish with no stomach content (degree I) were captured at night-time. Of those with food content in their stomach, only 20% were collected during night tows, and 85% of these night tows were performed earlier than 9 pm. There was no significant difference (p>0.05) in FI between habitats and shelf positions (Fig. 4D), but differences were found between latitude stratum D and the other levels (p<0.05) (Fig. 4F).

The diet of P. maculatus was composed mainly of crustaceans, teleosts and Polychaeta. In general, crustaceans dominated over the other food categories, especially the decapods, which were the most abundant. The most frequent items were unidentified shrimp, Polychaeta and teleosts (%FO=34, 20 and 17, respectively). Among decapods, shrimps (%FO=45) dominated, followed by Brachyura (%FO=13) (Table 1). Unidentified shrimp were the most abundant in number (%N=31), followed by Caridea shrimp (%N=18), Polychaeta and teleosts (%N=10 each), whereas teleosts had the highest biomass of the food groups (%W=28), followed by shrimps and Polychaeta (%W=23 and 7, respectively) (Table 1). In general, crustaceans were the most important food group found in P. maculatus, followed by teleosts and Polychaeta (Table 2).

The largest niche breadth was observed for outer-shelf over inner-shelf habitats (Bi=0.44 and 0.39 respectively) and, among habitat types, algae showed a higher value than sand and SWCR (Bi=0.50, 0.37 and 0.36 respectively).

According to the modified Costello diagram (Fig. 6), no specific species was preferentially targeted by P. maculatus. Instead, considering the vertical axis, most prey are on the left side of the diagram, with a low frequency of occurrence (Fo). Considering prey importance, the diet of P. maculatus consisted mostly of rare prey taxa present in small proportions (low Fo/Pi). Other items were also not often seen but had a high contribution in weight when present, such as isopods, stomatopods and teleosts. Based on the position of the items relative to the diagonal axis from top-left to bottom-right, the niche breadth is mostly due to a high between-phenotype component. The spotted goatfish seemed to be a generalist predator, feeding on many rare prey taxa. On the other hand, the population had some individuals with some degree of specialization that fed substantially on infrequent prey (high Pi and low Fo).

The arrangement obtained by the multidimensional analysis (nMDS) showed a high overlap of the diet between habitats (Fig. 7A), with very weak, although significant, differences (ANOSIM: R=0.04; p<0.05). Also, weak significant differences in diet were detected between shelf positions (Fig. 7A) (ANOSIM: R=0.02; p<0.05) and between latitude strata (Fig. 7B) (ANOSIM: R=0.06; p<0.05). The SIMPER routine indicated that mainly shrimps, stomatopods and teleosts contributed to the similarity within habitats, shelf positions and latitude levels (Table 3).

A total of 101 P. maculatus muscle samples were analysed for stable isotope composition (δ¹³C and δ¹⁵N). The SL of the individuals selected varied between 6.8 and 23.8 cm 15.91±3.16 (mean±SD). The values for δ¹³C ranged from –18.15 to 14.07‰, and δ¹⁵N varied from 6.9 to 11.4‰. Significant differences were found in δ¹⁵N along the latitudinal gradient. Kruskal-Wallis tests revealed differences in δ¹⁵N only between strata D and C (p<0.05). Significant differences were also found in δ¹³C along the latitude gradient, with stratum A showing lower values than the other strata (p<0.05). There were no significant differences in δ¹³C and δ¹⁵N among habitats and shelf positions (p<0.05), and these values were not correlated with size (SL) (R²=0.15). As values of δ¹³C and δ¹⁵N varied greatly, any model approach was designed to fit the data. The δ¹³C and δ¹⁵N values of the isotopic baseline (zooplankton) were also evaluated and compared between habitat, shelf position and latitude gradient, and no statistical significances were detected (Fig. 8). Also, no ontogenetic or cross-shelf distribution pattern of δ¹³C and δ¹⁵N isotopes was identified (Fig. 9).

Overall, the trophic level of the spotted goatfish was between the third and fourth position, classifying it as a carnivore predator. The effects of habitat, shelf position and latitude gradient were virtually non-existent. The results of the trophic position analysis based on the diet (TP_SC) were very close to the results provided by the isotopic analysis (TP_SI) (Table 4).

Munro (1976) affirms that there are no migratory movements of P. maculatus, which means they do not move far from the areas in which they settle. In our study, juveniles were found well distributed along the continental shelf, including in areas near the shelf break, but a slight difference in size distribution was noted, mainly due to the fact that larger individuals seemed to be more related to water that is deeper and further from coastal areas. Fishing pressure is concentrated near the coast (Lessa et al. 2004) and might play a role in the aforementioned pattern by removing larger individuals from the coastal areas.

The CPUE patterns indicated that P. maculatus is more abundant in the area below 8°S, but that it also had relatively higher abundance in Northern Pernambuco and in Southern Paraíba, where the largest fishery fleet that targets the spotted goatfish is located (Lessa et al. 2004, Ribeiro 2004). The analysis of the size distribution indicated that most juveniles were found in SWCR habitats, mainly in the southern areas. The higher CPUE values may be due to the presence of a great extension of coral reefs in this region (Costa et al. 2007, Leão et al. 2016) and the fact that most of this area is included in two MPAs (APA Costa dos Corais and APA Guadalupe) (Floeter et al. 2007, Steiner et al. 2015, Eduardo et al. 2018). Furthermore, upwelling processes have been reported in these areas, in which nutrients are transferred from deep to shallow layers, and this is linked to a higher abundance of prey for fish assemblages (MMA 2006). Some studies have reported that fishing activity affects marine ecosystems by modifying habitats and depleting fish populations, consequently altering their structures and affecting fish abundance (Garcia et al. 2003, Worm et al. 2006, Kenchington 2003). Therefore, MPAs are critical to the conservation and sustainable use of marine resources because they protect reproductive, nursery and feeding areas.

According to Lessa et al. (2004), length-frequency distribution does not change over the year and recruitment occurs mainly between March and September. The modal class (16-18 cm SL) was slightly lower than that reported by Lessa et al. (2004) (18-20 cm SL). These factors could be explained by several factors, such as fishing pressure over time and spatial-temporal contrasts, although both are unlikely causes. Instead, we assumed that the main cause is differences in gear selectivity, because our specimens were caught using bottom trawling, whereas Lessa et al. (2004) used trap fishing samples from commercial catches.

Adult individuals of P. maculatus are usually present in deeper areas with a muddy or sandy substratum available for foraging activity, while juveniles occupy seagrass bed areas (Munro 1976). Krajewski et al. (2006) also observed a high frequency of P. maculatus in areas of mixed substratum composed of algae, sand and coral reef, but they were also present foraging over substrata other than sand. In this study, P. maculatus displayed a similar pattern, occupying all types of habitat.

The condition factor (K) showed different values between habitats and latitude levels, with individuals from SWCR and stratum D showing smaller values. However, these differences seem to be more related to the fact that individuals caught in this area came mostly from the first ABRACOS campaign, which occurred in August-September 2015. Fish from the second campaign (April-May of 2017) were caught during the reproductive season of the spotted goatfish (Munro 1983, Santana et al. 2006), which can explain the higher K values for specimens with mature gonads caught during this campaign than during the first campaign.

The obtained length at first maturity (L₅₀) differed from that of Santana et al. (2006), based on Munro (1976), which estimated the age corresponding to the L₅₀ through the inverted Von Bertalanffy growth. Munro (1976) suggested P. maculatus reach maturity size at a minimum of 16 cm SL, a value somewhat higher than what we found. This discrepancy could be a result of differences between sampling, since our samples covered a broad range of lengths in different areas. Also, the reduced L₅₀ might be a reflection of fishing pressure. The official statistics of Brazil state that there was a marked increase in the landings of P. maculatus, from 18 t in 1994 to 296 t in 2007 (IBAMA 2008), with a peak of 626 t in 2000 (IBAMA 2002). Intense fishing pressure may affect growth and maturity of fish, with those species prone to high fishing mortality rates maturing early (Olsen et al. 2004, Green 2008, Tsikliras and Stergiou 2014). Since 2007, there have been no official reports and no studies about stock assessment, so we do not know the exact current state of the stocks. Nevertheless, Lessa (2004) warned that these fish populations have already been impacted by the removal of immature specimens, compromising the state of these stocks.

Males were significantly larger than females and were more frequent in larger size classes. Sexual dimorphism has been reported previously, with males being larger and having "higher arched backs and more angular facial profiles”, while females are more fusiform with a more hydrodynamic shape (Caldwell 1962, Munro 1976). Our LWR results differed from those obtained by Campos and Oliveira (2001), but they only had specimens above L₅₀, lacking smaller individuals. Also, their study area was restricted to regions explored by the artisanal fleet of North Pernambuco. Similar results were found by Munro (1983), with females showing negative allometric growth, confirming Caldwell’s (1962) observations that males and females of P. maculatus have sexual dimorphism.

Feeding activity of P. maculatus was classified as diurnal, since most individuals caught nocturnally had low or no food contents in their guts, which confirms previous results (Bohlke and Chaplin 1968, Munro 1976, 1983). The feeding intensity described by the FI can be influenced by the availability and type of prey, body length and season (Shanti Prabha and Manjulatha 2010, Mondal and Mitra 2016, Perelman et al. 2017). In general, the FI did not vary between habitats and shelf positions, but was higher in the southern areas, where the marine protected area APA Costa dos Corais is located. An MPA may provide high diversity of habitat and abundance of prey (Bonaldo et al. 2017, Novaczek et al. 2017), which may result in higher feeding success.

The presence of several epibenthic organisms in the diet of P. maculatus, such as decapods, polychaetes and amphipods, suggests that it is a generalist predator specialized in zoobenthos. Krajewski et al. (2006) described the feeding behaviour of P. maculatus as skimming the sea bottom with its barbels before shovelling and turning over the substratum to seek its prey. As described by several authors (Gosline 1984, Krajewski et al. 2006, Sazima et al. 2010), the spotted goatfish plays an important role in the benthic ecosystem by disturbing the substratum, changing the bottom topography and the distribution of benthic organisms and then attracting other carnivorous fishes that prey on the small organisms flushed out during the substratum disturbance.

The highest spectrum of prey was seen within the SWCR habitats and the lowest in algal turfs. Though there was a higher spectrum of prey in SWCR habitats, the niche breadth was smaller than that in the other habitats, suggesting a possible preference for a specific prey in this habitat or a possible low abundance of prey in the algae habitat, forcing the individuals in this substratum to widen the range of its prey. The diet of P. maculatus was not significantly different between habitats, shelf positions or latitudes. Though the waters of the study area usually have a high diversity and abundance of prey (CBD 2014, Eduardo et al. 2018), P. maculatus feeds mostly on crustaceans (shrimps, Brachyura, Stomatopoda) and, more rarely but still importantly, on Polychaeta and teleosts. No clear δ¹³C and δ¹⁵N patterns between either habitats or positions along the shelf were found for P. maculatus. Small differences between latitude strata were observed, although these were biologically irrelevant (less than 1‰). While other studies reported a possible positive relationship between δ¹⁵N and growth (Kolasinski et al. 2009), no relationship was found between δ¹⁵N and body size, possibly due to the size-specialized bottom‐foraging feeding mode of P. maculatus. The trophic positions estimated with stable isotope analysis showed that P. maculatus is ranked between the third and fourth trophic level. These results are endorsed by previous studies based on stomach content analysis for P. maculatus (Randall 1967, Sierra et al. 1994, Gómez-Canchong et al. 2004) and for other species of the family Mullidae (Hobson 1974, Golani 1994, Kulbicki et al. 2005). A similar estimation of the mean trophic position was obtained from our stomach content analysis, suggesting stability in the feeding habits of P. maculatus during its lifetime. Indeed, this species seems to be capable of maintaining a constant diet as it is a very specialized feeder with sophisticated chemo-sensorial barbels that allow it to search and retrieve its prey in the substratum (Krajewski et al. 2006).

The results of the present work provide information about the distribution of the spotted goatfish and its trophic relations among different habitats and locations. Dietary analysis is a particularly important tool for comprehending aspects of biology and ecology, especially when allied with spatial marine planning aimed at sustainable management of fishing stocks and the development of conservation measures (La Mesa et al. 2007, Kitsos et al. 2008). In fact, our results endorse Eduardo et al. (2018), who identified several potential areas for implementation of MPAs, contributing to future management and conservation measures for this species.