Matthiessen et al., Macroramphosus CM 2002/M: 07 1

Ices CM 2002 / Oceanography and Ecology of Seamounts - Indications of Unique Ecosystems (Session M) Not to be cited without prior reference to the author.

Seamounts, hotspots for high speciation rates in bentho-pelagic fishes? A case study on Macroramphosus spp. (Syngnathidae) from Great Meteor Seamount

BIRTE MATTHIESSEN, HEINO O. FOCK, HEIN VON WESTERNHAGEN

Alfred-Wegener-Institut für Polar- und Meeresforschung, Am Handelshafen 12, D-27570 Bremerhaven, Germany, Phone: 49-0471-4831 1384, [email protected]

Snipefishes of the genus Macroramphosus (Syngnathidae) are the most abundant demersal fish species on the Great Meteor Seamount (GMR, subtropical NE Atlantic, 30° N, 28,5° W). High morphological variability within the genus at GMR questions the presence of only one species of Macroramphosus. Thus, 202 individuals of Macroramphosus spp. were investigated with respect to diet composition and morphology. A deep-bodied benthos feeding type (b-types) whose diet consisted of foraminiferans, pteropods, decapods and polychaetes, could be distinguished from a slender planktivorous type (p-types, n=140) mainly feeding on ostracods, copepods, pteropods and foraminiferans. Few specimens showed no specialized feeding mode (p/b-types). Both feeding types showed significantly morphologically differences after applying bi- and multivariate statistical analysis considering the variables body depth, length of second dorsal spine, diameter of orbit and standard length. Based on a randomisation test, the analysis of dietary overlaps revealed significant differences between feeding types. This result reinforces the hypothesis of two competing species of snipefishes at GMR.

Notwithstanding similar observations on snipefishes from shelf habitats from the Pacific, our results suggest that isolated seamounts may represent hotspots of speciation where special selection pressure favours the evolution of two ecological divergent species within the genus Macroramphosus.

Keywords: Macroramphosus, Great Meteor Seamount, habitat partitioning, sympatric species, null-model, randomisation, competition Matthiessen et al., Macroramphosus CM 2002/M: 07 2

Introduction In all oceans, snipefishes of the genus Macroramphosus Lacepède 1803 (Centriscidae,

Macroramphosinae)1 abundantly inhabit shelf regions and seamounts, principally between 20 and 40° of latitude in both hemispheres (e.g. Parin et al., 1997; Wilson and Kaufmann, 1987).

Two vernacular names have been established for Macroramphosus spp. to account for a world-wide observed variability, i.e. longspine snipefish and slender snipefish. At present, the world-wide record of thirteen species can be traced back to these two different types of species, i.e. the longspine snipefish M. scolopax (L. 1758, originally described as Balistes scolopax) and the slender snipefish M. gracilis (Lowe 1839, originally described as Centriscus gracilis) (Eschmeyer, 1998). Though present in all oceans, the taxonomic status for M. gracilis is still unclear, since it is both treated as a genuine species and as a juvenile form of M. scolopax (see Froese and Pauly, 2001;

Mohr 1937). A review of the status of Macroramposus in the NE Atlantic by Ehrich and

John (1973) initially distinguished two distinct species in the North Atlantic. However, the findings of intermediate specimens then forced Ehrich (1976) to revise the status of

M. gracilis so that it became a juvenile stage of M. scolopax. Consequently, Ehrich

(1976) subsumed all his specimens of the genus under M. scolopax. Clarke (1984) investigated Macroramphosus spp. from near the edge of the continental shelf of SE

Australia. Due to two distinct dietary types which also differed significantly in morphological characters, he postulated the existence of two separate species of

Macroramphosus in SE Australia. In turn, he also discussed whether this could resemble the variability in other parts of the world's oceans as well. A further analysis of recent material from GMR applying methodologies of both Ehrich (1976) and Clarke

(1984) revealed a strong dependence of morphology on feeding type and thus supported Matthiessen et al., Macroramphosus CM 2002/M: 07 3

Clarke's (1984) hypothesis, indicating that two sympatric species of Macroramphosus were also present in the NE-Atlantic (Matthiessen et al., submitted).

Selection pressure by means of competition promotes divergence in adaptive radiation

(Schluter, 1994). Thus, new exploitative capabilities are attained (Schluter, 1993). The amount of overlap in resource utilization is assumed to be proportional to the intensity of competition between two species (Schoener, 1974a). In turn, morphologically similar species are likely to be stronger competitors than morphologically different species

(Juliano and Lawton, 1990; Ricklefs et al., 1981; Warren and Lawton, 1987). Sympatric competing members of a guild often vary in decisive morphological features by a ratio known as Hutchinson's rule in order to partition diet and habitat (Schoener, 1974a).

Hence, the exploitation of new resources and the associated interspecific differences in morphology are both signatures of adaptive radiation (Futuyama, 1990). This rationale paves the way for a synoptic investigation strategy considering both the exploitation of resources, i.e. diet, as well as conjoint differences in morphology and diet composition.

Significant results of conjoint morpho-dietary differences were already analysed and are summarised by Matthiessen et al. (submitted). Further, competition theory warrants significant differences in diet utilisation between the observed morpho-dietary types in order to evidence sufficient specific separation. This is investigated in this paper by means of randomisation procedure.

To analyse the observed pattern of morphologically distinct diet-types (benthic and planktonic feeders) with regard to resource partitioning and limiting competition it is useful to compare the observed data with an assumed pattern-free randomised pseudo- community. The central prediction of ecological and evolutionary models of displacement is that species should overlap less in resource use than in the absence of competition (Schoener, 1974a; Schoener, 1974b). There is wide support for the theory

1 Nomenclature follows Eschmeyer (1998). Matthiessen et al., Macroramphosus CM 2002/M: 07 4 that similar species reduce interspecific competition by subdividing their use of available resources (Pianka, 1974; Schoener, 1974a; Winemiller and Pianka, 1990).

By making use of morphological and dietary analysis within the genus

Macroramphosus we are trying to elucidate the existence of two sympatric species of snipefishes on Great Meteor Seamount.

Materials & Methods

Sampling

For the study Macroramphosus from the Great Meteor Seamount (subtropical NE

Atlantic, 30°N 28.5°W) obtained during the M42/3-cruise of the RV Meteor in

September 1998 (Pfannkuche et al., 2000) was analysed. Sampling was conducted with a GOV-bottom trawl with 450 meshes opening width and 32 m wing span (Uiblein,

1999); detailed description in Fock et al. (2002b). Stretched codend mesh size was about

15 mm. Sampling depth varied between 295 to 349 m. On Great Meteor Seamount

Macroramphosus spp. and Capros aper were the numerically dominant fish species

(Uiblein, 1999). For Macroramphosus spp. a total of 82700 fish were caught. For the extensive analyses of morphology and diet, 202 fishes were randomly drawn from deep- frozen sub-samples with the intension to cover a diurnal cycle (Table 1).

Morphometrics

In order to generate data which were comparable to earlier investigations (Clarke, 1984;

Ehrich, 1976) the same morphometric characters were measured (Fig. 1).

. Matthiessen et al., Macroramphosus CM 2002/M: 07 5

Analysis of diet and dietary overlap

Due to lack of a well defined stomach, the whole alimentary tract was examined for dietary analysis. Following Clarke (1984) the alimentary tract was separated into an anterior, mid and posterior section and the rectum. Prey items were identified to major taxa by means of a dissection microscope. Items with fleshy tissue were counted as dietary items, items without any flesh were counted as remains, e.g. single echinoderm spines, single polychaete setae, calcareous deposits of unknown origin. With regard to the main diet, specimens were classified as either plankton feeders (p-type), benthic feeders (b-type) or mixed feeders (p/b-type) after Clarke (1984). Length-frequency distributions were applied to check whether comparable size ranges of the fishes were obtained within different categories compared to the total population on the Great

Meteor Seamount. Distributions were based on total length measurements, since total length was measured onboard during the 1998-cruise.

For each prey item an `Index of Relative Importance` (IRI) was calculated after Hyslop

(1980), based on numbers (N), frequencies (F) and weight (W):

100 Ai IRIi [%] = n ∑ Ai i

with Ai = % Fi + % N i + %Wi , and

Fi (%) = Percentage of guts with food item i of the total number of analysed guts Ni (%)

= Percentage of food item i of the total number of food items

Wi (%) = Percentage of ashfree dryweight (AFDW) of food item i of the total AFDW n = Number of different food items

Matthiessen et al., Macroramphosus CM 2002/M: 07 6

Empty alimentary tracts were included in the calculation.

Statistical analysis of dietary overlap

The analysis of dietary overlap (Gotelli and Graves, 1996; Lawlor, 1980; Winemiller and Pianka, 1990) was carried out for a sub-sample of 80 specimens for which full taxonomic resolution of dietary items was available (Table 1). Pianka´s index of overlap

(Ojk) was deployed (Lawlor, 1980; Pianka, 1973; Winemiller and Pianka, 1990):

p p = ∑ ij ik O jk 2 2 ∑∑pij pik

were pij and pik represent the proportion of consumed prey item i by consumer types j and k. An overlap-matrix was obtained comprising all combinations of p- and b-type specimens (Table 2).

To generate a pattern-free feeding matrix, a modification of Lawlor’s Randomisation

Algorithm 3 (RA 3) was deployed (Lawlor, 1980). Whereas RA 3 randomises numbers of prey items within one predator species to retain niche breadth and specialization, we permuted the number of prey items within each prey category for all predator specimens

(Table 3). Accordingly, we retained specific abundance profiles for each prey item (e.g. overall number of consumed copepods, pteropods, etc.). Thus, the test procedure investigates deviation from randomness of utilization within a given prey ensemble, and each predator specimen (b- or p-type) is allowed to feed on both planktonic and benthic prey in varying numbers (niche breadth not retained, but prey characteristics). In order to indicate its relatedness to RA3, we call this randomisation procedure ortho-RA3, Matthiessen et al., Macroramphosus CM 2002/M: 07 7 since its permutation (within rows) is perpendicular to the direction of permutation in

RA3 (within columns), if comparisons are carried out between columns (see Table 3).

Overlaps were calculated using ECOSIM software (Gotelli and Entsminger, 2002).

Permutations were repeated 105 times (see Winemiller, 1990). Overlaps from permutations are named simulated values. Direct counts of prey items instead of averages were deployed to avoid bias in overlap estimates (Ricklefs and Lau, 1980).

The index was calculated with the nine most important major prey items (see IRI, Table

5) assumed to be equally frequent (Gotelli and Graves, 1996; Haefner, 1988; Lawlor,

1980).

Three hypotheses (H1, H2, H 3, for details on respective sub-divisions of the overlap matrix see Table 2) with respect to an idealized pattern-free community (H0) were tested.

H1: The observed overlap between all specimens is smaller than for simulated matrices;

H2: The observed overlap for p-type specimens is bigger than for simulated matrices;

H3: The observed overlap for b-type specimens is bigger than for simulated matrices.

Statistical analysis of differences between simulated and observed overlap matrices were carried out by means of nonparametric Man-Whithney U-test (Lawlor, 1980; Case,

1983a). Results were considered significant at the α = 0.05 level.

Morpho-dietary relationships

Relationships between diet and morphology were analysed with respect to classifications in diet-types and with respect to morphological measurements.

Analysis of covariance (ANCOVA) was carried out to analyse differences in morphological characters between different feeding types (Backhaus et al., 1990;

Underwood, 1997). Multivariate analysis of morphological characters based on the Matthiessen et al., Macroramphosus CM 2002/M: 07 8 classification of feeding types was carried out by means of discriminant analysis

(Backhaus et al., 1990). Prior to ANCOVA, data were tested for normality and homogeneity of variances. The Bartlett’s test on homogeneity of variances was applied since this test takes account of heterogeneous sample sizes (Zar, 1996). Multivariate methods appear to be robust even if the homogeneity criterion is violated (Fock, 2000).

Results were considered significant at the α = 0.05 for ANCOVA and at the α = 0.01 level for discriminant function analysis.

Results

Diet

Two major feeding types were identified, based on dietary composition. Following

Clarke (1984), the first feeding type was classified as b-type, predominantly feeding on pelagic foraminiferans and pteropods (Table 4). In terms of importance these two prey categories accounted for 48.4 % of IRI scores (Table 5). Although being of planktonic origin, these organisms were always found accompanied by sedimented particles such as calcareous deposits, grains of sand and remains of echinoderms. Further important prey items were polychaetes, decapods as well as already settled decapod larvae, indicated by the metamorphosis of the thoracopods. Together, these groups of prey organisms accounted for 83.3 % of IRI scores. Individuals feeding on decapod larvae were found to exclusively feed on these items.

The second feeding type was classifed as p-type after Clarke (1984). No bottom material was ingested. Main food items were ostracods and calanoid copepods, accounting for 49.6 % of IRI scores for this type. Still, pteropods and foraminiferans constituted a major portion of the diet, contributing 29.7 % of IRI scores. Matthiessen et al., Macroramphosus CM 2002/M: 07 9

Eight specimens belonged to a mixed feeding type (p/b-type, Table 4) mainly utilising foraminiferans and pteropods. Decapods, calanoid copepods and ostracods appeared to be secondary items. Whereas copepods indicated prey of planktonic origin, occasionally sediment particles also indicated a benthic feeding mode.

Length distribution

The uni-modal distribution of the total population of Macroramphosus spp. had a peak at 130 mm and a range of 100 to 170 mm (Fig. 2). The same length range was covered by both major feeding types, i.e. b-type and p-type specimens (Fig. 3). Whereas the p- type specimens fairly well represented the uni-modal distribution of the total population with a maximum at 130 mm, two modes appeared for the b-type specimens (Fig. 3 B).

The major mode was at 130 mm, accompanied by a minor mode at 110 mm. However, the number of b-type specimens was much smaller than for the p-type, so that the minor mode probably was an artefact. From the entire group of 202 specimens, 140 were classified as p-type, 50 as b-type and 12 as p/b-type. For the analysis of dietary overlap, representative subgroups were chosen, i.e. 106 to 152 mm length for the b-type and 112 to 166 mm length for the p-type specimens. Both sexes were equally represented.

Morpho-dietary relationships

Morphological variability between feeding types described by a set of scatterplots and analysed by means of ANCOVA (Fig. 4) revealed significant differences in 4 cases. In relation to standard length, b-type specimens had significantly longer dorsal spines (Fig.

4 A; F = 235,04; p < 0.05), bigger eyes (Fig. 4 B; F = 115,04; p < 0.05), longer pre- dorsal lengths (Fig. 4 C; F = 173,2, p < 0.05) and deeper bodies (Fig. 4 D; F = 290.65; p

< 0.05) than p-type snipefish. P/b-type specimens ranged between both groups. All Matthiessen et al., Macroramphosus CM 2002/M: 07 10 variables except body depth (character 7 in Fig. 1) possessed homogeneously distributed variances between categories. However, the strongest separation of groups was obtained using body depth. The p-type specimens showed smaller variability within this character than b-type specimens. This was an important observation with respect to the commonly accepted distinction between slender and lLongspine snipefishes. The significant differences between the b- and p-type specimens were preserved, when combinations, i.e. ratios of variables, were deployed (Fig. 4 E, F).

Consequently, the results of multivariate discriminant analysis showed a clear separation with respect to feeding type. The resulting discriminant function 1 yielded an eigenvalue of 2.11 and contributed more than 99 % to the overall variance (Table 6).

Thus, function 2 could be neglected. Based on four morphological characters, characters and feeding types could be grouped in the following way: (1) b-type specimens, body depth and length of second dorsal spine were correlated, (2) p-type specimens, orbit diameter and standard length were correlated. This takes into account, that on average p-type specimens appeared to be slightly larger than b-type specimens (Fig. 3) and that

'slenderness' was the decisive character in separating the types.

The re-classification (Table 7) showed, that a high proportion of p- and b-type specimens were correctly classified. However, none of the p/b-type specimens were correctly classified. Splitting this group up into either p- or b-type specimens indicated that the recognition of a transient type was probably a misperception.

Analysis of dietary overlap

The comparison between the simulated and the observed distributions of dietary overlap-values for the three hypotheses (H1, H2, H3) is displayed in Figure 5. The pattern-free simulated distributions had a median overlap value of O = 0.25. Though Matthiessen et al., Macroramphosus CM 2002/M: 07 11 considerable differences in line with the underlying hypotheses were observed in all three cases (Fig. 5 A – C), only H1 and H2 were accepted (Table 8).

Discussion

We showed that within the population of Macroramphosus spp. at GMR two clearly defined types of specimens can be distinguished, differing in diet composition and morphology. This was confirmed by multivariate analysis discovering clear morphological differences with respect to feeding types. In the multivariate analysis, only one discriminant function was necessary to separate the categories, indicating a very strong separation of already distinct groups. Furthermore, the analysis of dietary overlaps yielded results in line with the underlying hypothesis of the presence of two distinct groups.

The literature says that Macroramphosus gracilis-like forms are known from pelagic and bottom catches, feeding predominantly or exclusively on plankton (Clarke, 1984;

Ehrich, 1973). The Macroramphosus scolopax-like forms are described to be associated with bottom habitats and feed on both planktonic and benthic prey (Ehrich, 1973; Mohr,

1937). Clarke (1984) showed apparent differences in depth distributions among planktivorous and benthivorous specimens. Depth dependent distribution could not be proven for the Great Meteor Seamount, since only samples from the seamount plateau were available.

Mainly daytime feeding was observed in SE Australia as well as on Great Meteor

Seamount in the NE Atlantic (Clarke, 1984; Matthiessen et al., submitted). This feeding mode was explained by interception feeding with the sound scattering layer, hypothesising that maintenance of bentho-pelagic fish populations is provided through feeding on the advected plankton components in the vicinity of seamounts and shallow Matthiessen et al., Macroramphosus CM 2002/M: 07 12 topography (Fock et al., 2002a; Genin et al., 1988; Isaacs and Schwartzlose, 1965;

Rogers, 1994). Being trapped above the plateau, this plankton is vulnerable to predation by visual predators. The entrapment and feeding on diel migrators is expected to be most significant on seamounts with a top at depths between 100 and 300 m (Pearcy et al., 1977). Correspondingly, Macroramphosus spp. started feeding early in the morning when trapped organisms of the sound scattering layer were likely to be available.

Evidence for two sympatric species

Non-ambiguous evidence was presented for differences in morphology and diet composition between two groups termed p- and b-type specimens. It was statistically proven that no true transient stage in terms of a p/b-type exists. On the basis of the observed length distribution of Macroramphosus spp., with the gracilis-type being smaller than the scolopax-type, Ehrich (1976) established the juvenile stage-hypothesis, suggesting that gracilis-type specimens were juvenile scolopax-type specimens.

However, opposing to the juvenile stage-hypothesis we showed that under both classification schemes after Ehrich (1976) and after Clarke (1984) the p- and the gracilis-type specimens were larger than the corresponding b- and scolopax-type specimens (Matthiessen et al., submitted). Our data support Clarke's (1984) conclusion, that at least two sympatric species exist. Differential larval morphologies further confirm the existence of at least two species of Macroramphosus (Kuranaga and Sasaki,

2000).

Sympatric species show different strategies for habitat and resource partitioning

(Schoener, 1974a). It is assumed, that ecological differences in diet composition and habitat use of sympatric species favour the evolution of corresponding morphological differences (Schluter, 1993). Based on experimental work with different species of Matthiessen et al., Macroramphosus CM 2002/M: 07 13 sticklebacks, Schluter (1994) showed that resource competition actually promotes morphological diversification. Comparable to Macroramphosus spp., sympatric sticklebacks differed in body depth and resource use between benthos and open water species (Schluter, 1993). An idea about the minimum necessary difference for sympatric species is given by Hutchinson's rule (Hutchinson, 1959; Schoener, 1974a). The rule says that in decisive morphological characters a minimum difference of 1.2 to 1.9 must be obtained to maintain a sufficiently functional difference, i.e. ecological separation in the utilization of a resource category, to permit coexistence under a given set of resource and consumer densities (Gladfelter and Johnson, 1983). Actually, for

Macroramphosus spp. the length dependent difference in body depth between b- and p- type is 1.38 in the given size range (Fig. 6). The difference in body depth for

Macroramphosus spp. is likely to be an important adaptation. Body depth and body shape are responsible for swimming capabilities. The slender pelagic p-type probably is a better swimmer and well adapted to prey in open waters whereas the deep-bodied benthic feeding b-type optimises manoeuvrability near the bottom to detect and hunt benthic prey. Planktivorous pelagic fishes adapted to cruising have a compressed fusiform body whereas slow swimming epibenthic fishes that pick or suck their prey off the substrate are of gibbose form (Motta et al., 1995). Capture success in open water might be enhanced by the elongate body shape of pelagic living individuals by providing reduced drag (Schluter, 1993). Though this is only an indirect proof of the two species-hypothesis, frequency-dependent selection could provide a mechanism for adaptive morphological peak-shifts and has been implicated in the process of speciation

(Schluter, 1994).

Studies on chaetotontid fish indicate that species with similar diets reach higher abundances when they show spatial separation on the reef (Bouchon-Navaro, 1986) and in a similar way coexisting anemone-fishes with the same host anemone had different Matthiessen et al., Macroramphosus CM 2002/M: 07 14 distribution patterns on the reef (Elliott and Mariscal, 2001). Clarke (1984) reported apparent differences in depth distribution between p- and b-type specimens of

Macroramphosus spp. in south-east Australia. A potential resource partitioning by means of partitioning of feeding habitat was postulated from differences in the diets of three estuarine sunfishes (Lepomis spp.) (VanderKooy et al., 2000). In the same way as displayed by Macroramphosus spp. at GMR, long-snouted morphologically similar species of the genus Forcipiger from same habitats showed strong dietary separation

(Hobson, 1974). In contrast to the omnivorous F. flavissimus, F. longirostris had a restricted diet consisting of small caridean decapods (Hobson 1974).

Further indirect prove for the existence of two sympatric species of Macroramphosus spp. is based on the theory that similar species reduce interspecific competition by subdividing their use of available resources (Pianka, 1974; Schoener, 1974a; Schoener,

1982; Winemiller and Pianka, 1990).

If competition influences resource utilization at the community level, niche overlap should be significantly smaller than in an randomised pattern-free matrix (Schoener,

1974a; Schoener, 1974b).

For isolated habitats like seamounts speciation is claimed to be a highly relevant process

(de Forges et al., 2000; Rogers, 1994; Wilson and Kaufmann, 1987) and should be taken into consideration in the case of the suggested speciation process of

Macroramphosus spp.

Acknowledgements

The authors are thankful to all persons that gave support to the progress of this work, especially Angelika Brandt and Stefanie Bröhl. We are also thankful to Marcus Engel for helpful comments and improvement of the English language. Matthiessen et al., Macroramphosus CM 2002/M: 07 15

References

Backhaus, Erichson, Plinke and Weiber (1990) Diskriminanzanalyse. In: Backhaus, Erichson, Plinke and Weiber (ed) Multivariate Analysemethoden. Eine anwendungsorientierte Einführung. Springer-Verlag, Berlin Heidelberg, pp 161- 220 Bouchon-Navaro Y (1986) Partitioning of food and space resources by chaetodontid fishes on coral reefs. J. Exp. Mar. Biol. Ecol. 103: 21-40 Case T (1983a) Niche overlap and the assembly of island lizards communities. Oikos 41: 427-433 Clarke TA (1984) Diet and morphological variations in snipefishes, presently recognized as Macroramphosus scolopax, from Southeast Australia: Evidence for two sexually dimorphic species. Copeia 3: 595-608 de Forges BR, Koslow JA and Poore GCB (2000) Diversity and endemism of the benthic seamount fauna in the southwest Pacific. Nature 405: 944-947 Ehrich S (1976) Zur Taxonomie, Ökologie und Wachstum von Macroramphosus scolopax (Linnaeus, 1758) (Pisces, Syngnathiformes) aus dem subtropischen Nordostatlantik. Ber. dt. wiss. Kommn. Meeresforsch. 24: 251-266 Ehrich S, John, H.-C. (1973) Zur Biologie und Ökologie der Schnepfenfische (Gattung Macrorhamphosus) vor Nordwestafrika und Überlegungen zum Altersaufbau der adulten Bestände der Großen Meteorbank. "Meteor" Forsch.-Ergebnisse 14:(Reihe D): 87-98 Elliott JK and Mariscal RN (2001) Coexistence of nine anemonefish species: differential host and habitat utilization, size and recruitment. Marine Biology 138: 23-36 Eschmeyer W, 1998. Catalogue of fishes. California Academy of Science, San Francisco. Fock H (2000) Analysis of community structure. In: Harris RP, Wiebe P, Lenz J, Skjodal HR and Huntley M (ed) ICES Zooplankton Methodology Manual. Academic Press, San Diego, pp 154-173 Fock H, Matthiessen B, Zidowitz H and v. Westernhagen H (2002a) Diel and habitat dependent resource utilisation of deep-sea fishes at the Great Meteor seamount (subtropical NE Atlantic): niche overlap and supportfor the sound scattering layer-interception hypothesis. Mar. Ecol. Progr. Ser., in press Matthiessen et al., Macroramphosus CM 2002/M: 07 16

Fock H, Uiblein F, Köster F and von Westernhagen H (2002b) Biodiversty and species- environment relationships of the demersal fish assemblage at the Great Meteor Seamount (subtropical NE Atlantic), sampled by different trawls. Marine Biology published online: 12 April 2002 Froese R and Pauly D (2001). Fishbase 99. http://www.fishbase.org/home.htm. Futuyama DJ (1990) Konkurrenz und adaptive Radiation. In: Futuyama DJ (ed) Evolutionsbiologie. Birkhäuser Verlag, Basel Boston Berlin, pp 37-39 Genin A, Haury L and Greenblatt P (1988) Interactions of migrating zooplankton with shallow topography: predation by rockfishes and intensification of patchiness. Deep Sea Research 35:(2): 151-175 Gladfelter WB and Johnson WS (1983) Feeding niche separation in a guild of tropical reef fishes (Holocentridae). Ecology 64(3): 552-563 Gotelli NJ and Entsminger G (2002) EcoSim: Null models software for ecology. Version 7. Acquired Intelligence inc. & Kesey-Bear, Burlington, VT 05465. Gotelli NJ and Graves G (1996) Null models in ecology. Smithsonian Institution. Haefner J (1988) On niche shifts in greater Antillean Anolis communities: effects of niche metric and biological resolution on null modell tests. Oecologia 77: 107- 117 Hobson ES (1974) Feeding relationships of teleostean fishes on coral reefs in Kona Hawaii. Fishery Bulletin 72(4): 915-1031 Hutchinson (1959) Homage to Santa Rosalia, or why are there so many kinds of animals? The American Naturalist 93: 145-159 Hyslop EJ (1980) Stomach content analysis - a review of methods and their application. Journal of Fish Biology 17: 411-429 Isaacs JD and Schwartzlose RA (1965) Migrant sound scatterers: Interactions with the seafloor. Science 150: 1810-1813 Juliano S and Lawton J (1990) The relationship between competition and morphology patterns among co-occurring dytiscid beetles. Journal of Animal Ecology 59: 403-419 Kuranaga I and Sasaki K (2000) Larval developement in a snipefish (Macroramphosus scolopax) from Japan with notes on eastern Pacific and Mediterranean Macroramphosus larvae (Gasterosteiformes, Macroramphosidae). Ichthyological Research 47(1): 101-106 Matthiessen et al., Macroramphosus CM 2002/M: 07 17

Lawlor L (1980) Structure and stability in natural and randomly constructed competitive communities. American Naturalist 116: 394-408 Matthiessen B, Fock HO and von Westernhagen H (submitted) Evidence for two sympatric species of snipefishes Macroramphosus spp. (Syngnathiformes, Centriscidae) on Great Meteor Seamount. Helgoland Marine Research Mohr E (1937) Revision der Centriscidae (Acanthopterygii Centrisciformes). Dana Rept. 13 Motta JP, Clifton KB, Hernandez P and Eggold BT (1995) Ecomorphological correlates in ten species of subtropical sea grass fishes: diet and microhabitat utilization. Environmental Biology of Fishes 44: 37-60 Parin NV, Mironov AN and Nesis KN (1997) Biology of the Nazca and Sala y Gomez Submarine Ridges, an outpost of the Indo-West Pacific fauna in the Eastern Pacific Ocean: composition and distribution of the fauna, its communities and history. Advances in Marine Biology 32: 147-228 Pearcy WG, Krygier EE, Mesecar R and Ramsey F (1977) Vertical distribution and migration of oceanic micronekton off Oregon. Deep-Sea Research 24: 223-245 Pfannkuche O, Müller TJ, Nellen W and Wefer G (2000) Ostatlantik 1998, Cruise No. 42, 16 June-26 October. METEOR-Berichte 00-1: 259 pp. Pianka E (1973) The structure of lizard communities. Annual Review of Ecology and Systematics 4: 53-74 Pianka E (1974) Niche overlap and diffuse competition. Proc. Nat. Ac. Sc. USA 71: 2141-2145 Ricklefs R, Cochran D and Pianka E (1981) A morphological analysis of the structure of communities of lizards in desert habits. Ecology 62: 1474-1483 Ricklefs R and Lau M (1980) Bias and dispersion of overlap indices: results of some Monte Carlo Simulations. Ecology 61: 1019-1024 Rogers AD (1994) The Biology of Seamounts. Advances in Marine Biology 30: 305- 340 Schluter D (1993) Adaptive radiation in sticklebacks: size, shape, and habitat use efficiency. Ecology 74:(3): 699-709 Schluter D (1994) Experimental evidence that competition promotes divergence in adaptive radiation. Science 266: 798-800 Schoener TW (1974a) Resource partitioning in ecological communities. Science 185: 27-37 Matthiessen et al., Macroramphosus CM 2002/M: 07 18

Schoener TW (1974b) Some methods for calculating competition coefficients from resource-utilization spectra. The American Naturalist 108: 332-340 Schoener TW (1982) The controversy over interspecific competition. American Scientist 70: 586-595 Uiblein F, Geldmacher A, Köster F, Nellen W, Kraus G (1999) Species composition and depth distribution of fish species collected in the area of the Great Meteor Seamount, eastern central Atlantic, during cruise M 42/3, with seventeen new records. Informes Tecnicos del Instituto Canario de Ciencias Marinas 5: 1-77 Underwood AJ (1997) Analysis involving relationships among variables. In: Underwood AJ (ed) Experiments in ecology. Their logical design and interpretation using analysis of variance. University Press, Cambridge, pp 444- 471 VanderKooy KE, Rakocinsky CF and Heard RW (2000) Trophic relationships of three sunfishes (Lepomis spp.) in an estuarine bayou. Estuaries 23(5): 621-632 Warren P and Lawton J (1987) Invertebrate predator-prey body size relationships: an explanation for upper triangular food webs and patterns in food web structure? Oecologia 74: 231-235 Wilson RR and Kaufmann RS (1987) Seamount biota and biogeography. In: Keating BH, Fryer P, Batiza R and Boehlert GW (ed) Seamounts, Islands and Atolls. American Geophysical Union, Washington, pp 319-334 Winemiller K and Pianka E (1990) Organisation in natural assemblages of desert lizards and tropical fishes. Ecological Monographs 60:(1): 27-55. Zar H (1996) Biostatistical Analysis. Prentice Hall, Inc., New Yersey.

Matthiessen et al., Macroramphosus CM 2002/M: 07 19

Figures

Fig. 1: Morphometric measurements for Macroramphosus spp. Character #10 oblique due to variable position of anus in relation to insertion of dorsal spine. 1 - Total length

(TL), 2 - Standard length (SL), 3 - Snout length (SnL), 4 - Orbit diameter (O), 5 – Inter- orbital length (IOL), 6 - Smallest depth of tail (DT), 7 - Body depth (BD), 8 - Pre-dorsal length (PdL), 9 – Pre-anal length (PaL), 10 - Distance from anus to insertion of second dorsal spine (AP), 11 - Length of second dorsal spine (DS)

Matthiessen et al., Macroramphosus CM 2002/M: 07 20

1000

800 ) n (

h 600 s fi f o r e b 400 m Nu

200

0 80 100 120 140 160 180 Total length (mm)

Fig 2: Frequency distribution of the total catch of Macroramphosus (n = 2769). Length classes of total length indicated. Matthiessen et al., Macroramphosus CM 2002/M: 07 21

A p-Type B b-Type 40 16

30 12 ) n ( h is f f

o 20 8 r Numbe 10 4

0 0 80 100 120 140 160 180 80 100 120 140 160 180

Total length (mm)

Fig. 3: Frequency distribution of the sub-samples of p- and b-Type snipefish (p-Type: n=140; b-Type: n=50). Length classes of total length indicated. Matthiessen et al., Macroramphosus CM 2002/M: 07 22

50 A 13 B 120 C ) m

m 110 ( 40 )

12 ) m ne m i m p m

( 100 s l ( h t r a 30 11 e s et eng 90 l dor l . am a c

20 s e 10 r di t o s 80 d bi of r e h O t Pr g 10 9 70 Len 0 8 60 80 100 120 140 160 80 100 120 140 160 80 100 120 140 160 h r t 45 e 5 D ng 0.5 E F et e l

l am a s

40 di t bi dor

) 0.4 4 e or r m

p 35 / e / n e i h (m n p t i s p p 30 0.3 3 s al s al de r s y r d o do

o 25 . d c . B e c 0.2 2 e s f 20 s h o of h t ngt

15 ng 0.1 1 Le Le 80 100 120 140 160 80 100 120 140 160 80 100 120 140 160 Standard length (mm) Standard length (mm) Standard length (mm)

Fig. 4: Bivariate scatterplots of morphological characters versus standard length for

different feeding types. A) Length of second dorsal spine (DS), B) orbit diameter (O),

C) pre-dorsal length (PdL) and D) body depth (BD) vs. standard length (SL) of

Macroramphosus. E) displayed length of second dorsal spine to pre-dorsal length and F)

ratio of length of sec. dorsal spine to orbit diameter vs. standard length of

Macroramphosus. Open circles: b-types (n = 50); solid circles: p/b-Types (n = 12);

crosses: p-Types (n = 140). Explanations see text.

Matthiessen et al., Macroramphosus CM 2002/M: 07 23

80 A 60 B 60 C ) 60 y (% 40 40 uenc

eq 40 r f e v

ti 20 20 la 20 Re

0 0 0

0 .1 .2 3 4 .5 .6 7 .8 .9 0 1 2 .3 .4 5 6 7 .8 9 0 .1 .2 3 4 5 6 7 8 9 0 0 0. 0. 0 0 0. 0 0 0. 0. 0 0 0. 0. 0. 0 0. 0 0 0. 0. 0. 0. 0. 0. 0. Dietary overlap Dietary overlap Dietary overlap

Fig 5 The white bars show the simulation of dietary overlaps (N = 105). The grey bars

display (A) the observed overlap values between p- and b-type individuals (N = 1216),

(B) the observed overlap values only between p-type specimens (N = 986), and (C) the

observed overlap values only between b-types (N = 335). The triangles above show the

median values, respectively.

Matthiessen et al., Macroramphosus CM 2002/M: 07 24

45 Average ratio of b-Type to p-Type body depth = 1,38

40

35 ) m 30 m ( h pt e

d 25 dy Bo 20

15

10

80 90 100 110 120 130 140 150 160 Standard length (mm)

Fig. 6: Regressions of body depth vs. standard length for p-type (crosses; R2 = 0.63) and b-type (squares; R2 = 0.25) specimens resembling part of Fig. 6 D. Within the given size range the average ratio between the body depth of b- and p-type specimens is

1.38±0.15. The ratio was calculated from the regression graphs for either feeding type. Matthiessen et al., Macroramphosus CM 2002/M: 07 25

Tables

Table 1: Specimens of Macroramphosus spp. selected for the analysis of diet and morphology from Great Meteor Seamount in 1998. a, aa indicate numbers of fishes processed within different steps of the analysis: a)

Specimens analysed for feeding types and morphology, aa) Subsample taken from (a) for detailed analysis of food items and dietary overlaps. * = Dusk, 'Position' is mid position of trawl, 'Depth' is mean depth.

Station Time of catch Type of analysis Number of Position specimens Depth 488 00.40-00.55 a) 36 30.05°N aa) 21 28.54°W 291.5 m 495 05.23-05.39 a) 16 30.09°N aa) - 28.44°W 327 m 469 10.15-11.00 a) 19 30.07°N aa) - 28.54°W 298 m 470 12.51-13.37 a) 33 30.08°N aa) 19 28.56°W 299.5 m 482 15.43-16.01 a) 40 29.91°N aa) 20 28.40°W 321 m 483 18.15-18.31* a) 19 30.08°N aa) - 28.45°W 326 m 487 22.21-22.38 a) 39 29.92°N aa) 20 28.38°W 320 m Total 202 80

Matthiessen et al., Macroramphosus CM 2002/M: 07 26

Table 2: Subdivisions of the overlap matrix with regard to hypotheses H1 to H3. p1 ... pn, b1 ... bm indicate p- and b-type specimens, o... indicate the respective overlap values.

Values for the test of H1 are op1b1 to opnbm (=Obspb), for H2 are above the diagonal op1p1 to opnpn (=Obspp), for H3 above the diagonal ob1b1 to obmbm (=Obsbb).

p1 L pn b1 L bm

p1 op1p1 L op1pn op1b1 L op1bm M O M M M

pn 0 opnpn opnb1 L opnbm

b1 ob1b1 L ob1bm M 0 O M

bn 0 obmbm

Matthiessen et al., Macroramphosus CM 2002/M: 07 27

Table 3: Demonstration of randomisations: (A) Observed and simulated matrix for RA

3 after Lawlor (1980), and (B) observed and simulated matrix for the presently used randomisation algorithm (ortho-RA 3). p1,p2,b1,b2 display exemplary individuals of p- and b-type snipefish and prey 1 to 4 describe exemplary prey categories.

A Number of prey items A Number of prey items Observed Prey Prey Prey Prey Simulated Prey Prey Prey Prey matrix 1 2 3 4 matrix 1 2 3 4 p1 0 0 1 1 p1 1 0 1 0 p2 0 0 1 1 p2 1 0 0 1 p3 1 1 1 1 p3 1 1 1 1

Feeding type p4 1 0 1 0 Feeding type p4 0 0 1 1

B Feeding type B Feeding type Observed p1 p2 p3 p4 Simulated p1 p2 p3 p4 matrix matrix Prey 1 0 0 1 1 Prey 1 1 1 0 0 Prey 2 0 0 1 0 Prey 2 0 1 0 0 Prey 3 1 1 1 1 Prey 3 1 1 1 1

Number of of Number prey items Prey 4 1 1 1 0 of Number prey items Prey 4 1 1 0 1

Table 4: Numbers (n) and frequency of occurrence (F) for different food items for p-, b-

, und p/b-types of Macroramphosus spp.. Feeding types defined after Clarke (1984). Matthiessen et al., Macroramphosus CM 2002/M: 07 28

Feeding type b p p/b Number of examined 26 46 8 fishes (n=) Prey item n F n F n F Foraminifera 18103 26 243 28 1222 7 Pteropoda 8239 22 241 38 1164 6 Gastropoda 5561 20 13 6 76 7 Ostracoda 4 3 971 45 75 4 Calanoida 18 11 854 46 59 4 Polychaeta 148 12 4 3 - - Decapod larvae 51 8 1 1 - - Decapoda 34 10 2 1 13 4 Amphipoda 21 4 4 1 10 1 Mysidacea 9 4 8 6 11 3 Cirripedia 5 2 - - - - Harpacticoidea 2 2 - - 1 1 Isopoda 2 2 - - 1 1 Bivalvia 2 2 - - - 1 Euphausiacea - - - - 2 - Tanaidacea 2 2 - - - - Crustacea unidentified 46 15 103 33 23 7 Cephalopoda - - - - 1 1 Echinodermata 2 2 - - - - Fishes, unidentified pieces - - 1 1 - - Bryozoa - 13 - 1 Unidentified items 25 9 48 21 5 3 Remains Echinoderm remains 219 11 2 2 26 2 Sponge remains 25 8 - - 1 1 Calcareous deposits - 19 2 2 Sand grains - 17 1 3 Tube fragments - 13 2 3 Polychaete remains - - - 1 Total 32517 2496 2691 Matthiessen et al., Macroramphosus CM 2002/M: 07 29

Table 5: Ash free dry weight (AFDW) per item (mg Ind-1) and calculated `Index of

Relative Importance´ (IRI ) for b-, p- and p/b-type Macroramphosus spp. according to abundance and frequency data given in Table 2.

Feeding type b p p/b Taxon AFDW (mg Ind-1) IRI IRI IRI Foraminifera 0.040 28.3 12.8 25.0 Pteropoda 0.141 20.1 16.9 22.4 Polychaeta 1.026 9.9 3.6 1.5 Ostracoda 0.304 2.5 25.2 10.4 Calanoida 0.049 7.8 24.4 9.9 Euphausiacea 0.909 1.3 1.3 3.7 Decapoda 3.868 12.4 5.8 15.1 Decapod larvae 5.176 12.9 7.7 7.6 Amphipoda 1.362 4.7 2.3 4.4

Matthiessen et al., Macroramphosus CM 2002/M: 07 30

Table 6: Eigenvalues and standardised canonical discriminant function coefficients for morphological characters and canonical discriminant scores for associated feeding types. Eigenvalues indicate the variance explained by the respective discriminant function.

Function 1 Function 2 Statistics Eigenvalue 2.11 0.010 Cumulative eigenvalues 0.995 1.000 Coefficients for morphological characters Standard length 0.918 0.444 Orbit diameter 0.341 0.726 Body depth -0.914 -1.632 Length of second dorsal spine -0.629 1.095 Scores for feeding types p-type (group means) 0.94 -0.02 b-type (group means) -2.38 -0.52 Matthiessen et al., Macroramphosus CM 2002/M: 07 31

Table 7: Re-Classification matrix after discriminant analysis.

Rows refer to observed classification, columns to predicted classification, respectively.

Specimens classified as p-type b-type p/b-type Observed feeding % correctly classified types p-type 96.8 135 5 0 b-type 86.00 7 43 0 p/b-type 0.00 9 3 0 total 151 51 0

Matthiessen et al., Macroramphosus CM 2002/M: 07 32

Table 8 Results of Man-Whithney U-test for comparisons of simulation (Sim) versus observed overlap values (Obs). For H1: Obspb , for H2 :Obspp for H3: Obsbb . U: U-Value;

Z: transformed Z-Value; m: median-value.

Sim vs. Obspb Sim vs.Obspp Sim vs. Obsbb

U 7794.5 5539.5 9482.0

Z -15.766 -15.07 -7.148

p 0.00 0.00 0.00

m Obspb 0.00 < Sim 0.25 Obspp 0.87 > Sim 0.25 Obsbb 0.06 < Sim 0.25

H1 to 3 H1 accepted H2 accepted H3 rejected