Specimens were collected during the deep-sea fishery exploratory surveys of the Fishery Oceanographic Research Vessel (FORV) Sagar Sampada (71.5 m LOA: 2285 hp) (Cruise no 349) in Andaman and Nicobar waters in April 2016 using a High-Speed Demersal Trawl -crustacean version (HSDT-CV) at a towing speed of 2.5 to 3.5 knots. Eight stations were surveyed (one operation at each station) along the continental margins of the Andaman and Nicobar Islands (7.29-13.76°N and 92.14-93.11°E) at depths ranging from 300 to 650 m (Fig. 1). The locations were scanned using a SIMRAD EK60 echo sounder before trawling operations and stations were selected on the basis of the suitability of the grounds for trawling. The fishing operations were carried out from 6 am to 6 pm depending upon the weather conditions.

The lophiiforms were identified following standard identification keys (Alcock 1891, 1894, Rajeeshkumar et al. 2016, Ho et al. 2016a, b). Only non-damaged adult fishes were selected for meristic and morphological measurements and to extract the otoliths. The catch per unit effort (CPUE) and the spatial distribution of each species along with their geographical positions are given in the Figure 2.

Otoliths were collected and washed with distilled water to remove exogenous matter, dried and kept in plastic vials for further analysis. Otoliths from the right side of each fish were oriented with the inner side (sulcus acusticus) uppermost on a slide in order to digitize their form using a microscope (S8APO Camera, Leica DFP-425). Otolith length (OL, mm), height (OH, mm), area (OA, mm²) and perimeter (OP, mm) were measured using ImageJ with magnification depending on otolith size. Otolith weight (OW, mg) was obtained using an electronic balance (Metler Toledo, ML 503) (see descriptive values in Appendix 1). The morphological characteristics of each species were described following Tuset et al. (2008).

Sixteen morphological variables were measured on each specimen using a Vernier calliper (0.1 mm precision): total length (TL), standard length (SL), eye diameter (ED), mouth opening (MO), head depth (HD), eye height (EH), pectoral fin base (PFB), pectoral fin insertion, pectoral fin length (PFL), pectoral fin surface (PFS), caudal peduncle depth (CPD), caudal fin surface (CFS), caudal fin depth (CFD), body depth (BD), body length, body width (BW), mouth height (MH) and mouth width (MW) (Fig. 3). From these measurements, the following 11 ecomorphological attributes correlated with foraging, manoeuvrability and locomotion were selected for detailed studies. The formulas for estimating the functional traits are given in italicized letters.

– Oral gape surface (Osf)=(MW×MH)/(BW×BD), which indicates the nature/size of the prey that can be captured. A large oral gape allows feeding on a wide size range including large prey (Karpouzi and Stergiou 2003).

– Oral gape shape (Osh)=MH/MW, which defines the method for capturing food items. A greater width allows species to capture highly mobile prey and have a more aggressive behaviour (Karpouzi and Stergiou 2003, Wainwright et al. 2007).

– Oral gape position (Ops)=MO/HD, which shows the feeding position in the water column. The position of the oral gape influences the retention of prey during ingestion (Kumar et al. 2017a, Villéger et al. 2017).

– Eye size (Edst)=ED/HD, which defines the prey detection efficiency. It also influences the feeding rhythms (nocturnal vs diurnal) and predator avoidance and indicates the availability of light in the microhabitat (Boyle and Horn 2006, Bellwood et al. 2014).

– Eye position (Eps)=EH/HD, which displays the vertical position in the water column. High values indicate dorsally located eyes (Watson and Balon 1984, Ribeiro et al. 2016).

– Body transversal shape (Bsh)=BD/BW, which indicates the vertical position of the fish in the water column as well as hydrodynamic efficiency (Villéger et al. 2017).

– Caudal peduncle throttling (Cpt)=CFD/CPD, which shows the caudal propulsion efficiency through the reduction of drag (Webb 1984, Zhao et al. 2014).

– Fin surface ratio (Fsr)=(2×PFS)/CFS, which indicates the type of propulsion between caudal and pectoral fins. Higher values denote a swimming driven by pectoral fins, whereas lower values correspond to a greater caudal fin propulsion (Mouillot et al. 2013, Zhao et al. 2014).

– Fin surface to body size ratio (Fsb)= ((2×PFS) + CFS)/(π/4×BW×BD), which indicates the acceleration and/or manoeuvring competence. Higher values indicate prolonged sustained swimming speed and fitness, which positively influence endurance, acceleration and manoeuvring capacities (Zhao et al. 2014, Kumar et al. 2017a).

– Aspect ratio of the pectoral fin (ArPF)= PFL²/PFS, an indicator of swimming ability, which helps sustained swimming. Longer pectoral fins favour sustained swimming speed (Watson and Balon 1984, Casatti and Castro 2006).

– Aspect ratio of the caudal fin (ArCF)=CFD²/CFS, which indicates the caudal fin use for propulsion and/or direction. A higher ratio produces the maximum thrust (Webb 1984, Bridge et al. 2016).

To estimate the functional traits, the morphological data were standardized to remove the allometric effect using the total weight (Mouillot et al. 2005, Kumar et al. 2017a). The allometric relationship between morphological data (X) and body mass (M) is X=aM^b, where ‘b’ varies with species. The effect of body mass was eliminated by using the residuals of the common within-group slopes of linear regressions for each component of body mass.

The Kolmogorov-Smirnov and Levene tests were used to check normality of the data distributions and variance homogeneity, respectively. The intraspecific variability was analysed considering the fish size-otolith measurement relationships as a tool in the feeding ecology to estimate fish size and biomass (Kumar et al. 2017b, c) and the otolith relative size as a resemblance to fish habitat and depth distribution (Lombarte and Cruz 2007). For the first analysis, the relationships between otolith morphometric variables (OL, OH, OA, OP, OW) were described using the allometric power equation (Y = aX^b) (Huxley 1924). Measurements were converted into logarithmic values (log10) to identify and exclude possible outliers in the data (Froese et al. 2011). Regression parameters a and b were estimated by the least square regression method, where b represents the constant of differential growth rate (Froese 2006). An analysis of covariance (ANCOVA) was performed to compare the regression slopes between species, treating the species as the main factor and fish size (SL) as a covariate. Specific difference was analysed using a post-hoc Tukey-HSD test. In the second analyses, the otolith measurements were standardized for each species by removing the effect of allometry (Lleonart et al. 2000). Different relative sizes were estimated for each otolith morphometric variable using the following criteria (Lombarte and Cruz 2007): ORi= (otolith variable)i SL^b, with b=1 for OL, OH and OP variables, b=2 for OA and b=3 for OW. An ANOVA was conducted for each variable on the relative size to test differences in the averages among species. A post-hoc test (Dunn’s test) was performed to elucidate the pairwise comparison of relative otolith sizes (Pohlert 2014). All statistical analyses were performed in PAST (PAlaeontologicalSTatistics, version 3.26) (Hammer et al. 2001).

To order species in the functional space, a principal component analysis (PCA) based on the correlation matrix of the functional traits was performed. The choice of which principal components to interpret was based on a broken-stick model, which constructs a null distribution of eigenvalues and compares it with observed ones (Collar and Wainwright 2006, Villéger et al. 2011). Our hypothesis of significant difference among the species and Bonferroni’s correction for post-hoc pairwise multiple comparisons were tested using multivariate analysis of variance (MANOVA) (Marcus 1993, Layman et al. 2005, Marrama and Kriwet 2017).

The degree of functional niche overlap among species was performed using a non-parametric kernel density function (NO_K) (Mouillot et al. 2005, Mason et al. 2008, Geange et al. 2011):

NO_K (i, j, t) is the niche overlap between species i and j for the trait t, T is the number of functional traits and wt is the weighting parameter, which is calculated as:

r_tl is the Pearson correlation coefficient between traits t and l over all five species selected for the study. To understand the niche differences between the anglerfishes, permutation tests were performed to assess whether the observed niche overlap was significantly low based on the potential distribution of niche overlap values (Mouillot et al. 2005, Mason et al. 2008, Geange et al. 2011). Pseudo-values were calculated through randomly permuting species types in the corresponding data set for more than 1000 runs followed by computing the distribution of the average niche overlap for the null model to create the statistical null distributions. A Bonferroni adjustment of type I (Quinn and Keough 2002) was performed for the multiple comparisons. Density functions available in R (R Development Core Team 2017) were used to calculate niche overlap and for the subsequent null model tests. We followed the source code provided by Geange et al. (2011) for the above analysis in the R environment.

All species shared otolith features such as dorsal lobes and the lightly marked sulcus acusticus, with a well-defined crista inferior (Fig. 4). The otoliths of Chaunacidae (C. apus and C. multilepis) are characterized by a sulcus acusticus with undifferentiated ostium and cauda referred to as archaesulcoid. Indeed, they maintain an oval shape throughout growth with a smoothed and deep convex ventral margin. The dorsal margin has a variable number of lobes depending on species. In general, C. apus have more lobes (5 to 7) that are less angled than in C. multilepis. In Ogcocephalidae species, otoliths show a stronger differentiation in shape: H. coccinea has a semi-circular pattern (in the largest specimens), with a high number of deep lobes (6 to 10), some irregularities on the dorsal margin, a smooth, convex ventral margin, a rounded to angled anterior margin, and an angled end at the posterior margin for the largest specimens, providing an oblong shape. In contrast, M. lutea has an oval shape, with a sinuous to lightly lobed dorsal margin (3 to 6 lobes) and a smooth, shallow, convex ventral margin, and the anterior margin is oblique, lacking a rostrum. In both species the sulcus acusticus is archaesulcoid, mesial and ascendant, with an oval ostium (poorly defined) and a cauda smaller than the ostium. In particular, the sulcus acusticus of M. lutea is placed in an inframedian position. Finally, the otolith of L. lugubris (Lophiidae) is characterized by a semi-circular to oblong shape (in the largest specimens), with a sinusoidal ventral margin and a deeply lobed (6 to 9) irregular dorsal margin, a blunt anterior margin and an undefined rostrum and pointed end of the posterior margin of the largest specimens. The sulcus acusticus is a homosulcoid type, with oval ostium and cauda.

All otolith morphometric variables showed a statistically significant relationship with fish length for all species (Table 1, Appendix 2). However, otolith length and weight were the best variables correlated with fish size (r² ranges from 0.740 to 0.936 for OL, and between 0.708 and 0.959 for OW). The other variables showed a high intraspecific variation, and even attained very low values in the otolith height (r²=0.287) for M. lutea and the otolith perimeter (r²=0.243) for H. coccinea (Table 1). The ANCOVA exhibited no differences between species in the slopes of relationships SL-OH (F= 0.879, df=4, p=0.482) and SL-OA (F=2.158, p=0.085), but it indicated interspecific variability for the SL-OL (F=4.764, df=4, p=0.002), SL-OP (F=2.705, df=4, p=0.039) and SL-OW (F=6.787, df=4, p<0.001) relationships (Appendix 3). In particular, the slope (b) for the SL-OW relationship was higher in M. lutea than in C. apus-H. coccinea, and higher in L. lugubris than in H. coccinea. In fact, H. coccinea and L. lugubris also varied for the SL-OP and SL-OL relationships, and L. lugubris also showed differences with C. apus for the latter.

The ANOVA tests revealed significant differences for all relative variables (OA_R, F=166.2, df=4, p<0.05; OL_R, F=120.1, df=4, p<0.05; OH_R, F=309.8, df=4, p<0.05; OP_R, F=46.6, df=4, p<0.05; OW_R, F=124.1, df=4, p<0.05). Pairwise comparison using Dunn’s test (Bonferroni, p<0.05) indicated significant inter-species differences, except between Chaunax spp. (Fig. 5, Table 2). The highest interspecific variability was obtained for OH_R and the lowest for OP_R. No particular clustering was noted between species. In general, L. lugubris showed the lowest values for all relative sizes and C. apus the highest ones.

The first five PCA axes explained 97.9% of the total variance and the first three explained 93.8%. The PC1 axis alone contributed 63.7% of the total variance and was mainly correlated with Fsb (r= 0.868) (Appendix 4). The positive values represented species with a more dorso-ventrally flattened body and higher swimming capabilities (M. lutea, H. coccinea and L. lugubris) versus species with higher body depth and lesser swimming abilities (C. multilepis and C. apus) (Fig. 6). The PC2 axis (19.1% of variance) was related to propulsion and acceleration capabilities (ArCF, r=0.893), showing a similar pattern in all five species. The PC3 axis (10.9% of variance) was mainly related to swimming performance (Arcf, r=–0.834). The remaining PC scores (4 to 11) cumulatively explained 6.2% of the variance and were related to locomotion traits (Appendix 4). MANOVA confirmed the occurrence of significant differences among these deep-sea anglerfishes (Wilk’s Lambda=00.0023, F_44,258.3=22.88, p<0.001). The pairwise comparisons among species using sequential Bonferroni correction indicated significance differences among all species (p<0.001) (Appendix 5).

The functional traits Ops, Edst and Eps showed the highest interspecific differences, whereas Osf, Cpt and Fsr showed the lowest (Table 3, Fig. 7). The overall niche overlap ranged between 0.32 for C. apus-M. lutea and 0.65 for H. coccinea-L. lugubris. The species with highest niche partitioning was M. lutea due to its differentiation in the variables such as Osf, Ops, Edst and Eps. The analysis revealed significant differences between species, with M. lutea having a more differentiated functional niche, and both species of Chaunax showed more resemblance between them (Table 4). In any case, the findings indicated that functional niches did not overlap among the common five anglerfishes from the Indian Ocean.