Weak cross-species relationships between body size and trophic level belie powerful size-based trophic structuring in fish communities
Summary
- 1
Body size determines rates of respiration and production, energy requirements, mortality rates, patterns of predation and vulnerability to mortality. Body size distributions are often used to describe structure and energy flux in communities and ecosystems.
- 2
If clear relationships can be established between body size and trophic level in fishes, they may provide a basis for integrating community and ecosystem analyses based on size spectra, food webs and life histories.
- 3
We investigated relationships between the body sizes (weight and length) of north-east Atlantic fishes and their trophic level. The abundance of 15N, as determined by stable isotope analysis, was used as an index of trophic level.
- 4
Cross-species and comparative analyses demonstrated that body size was unrelated or weakly related to trophic level. Thus allometric relationships between body size and trophic level could not be used to predict the trophic structure of fish communities.
- 5
The results of the cross-species analyses contrasted with patterns in the size and trophic structure of entire fish communities. When fish communities were divided into size classes, there were strong positive relationships between size class and trophic level. The slope suggested a mean predator : prey body mass ratio of 80 : 1.
- 6
Our results suggest that body size does not provide a useful surrogate of trophic level for individual species, but that body size is an excellent predictor of trophic level within the community, providing an empirical basis for integrating community analyses based on models of trophic structure and body size distributions.
Introduction
Body size is an obvious and important feature of any animal (Peters 1983; Cohen et al. 1993). It determines potential predators and prey (Peters 1983), rates of production and natural increase (Banse & Moser 1980; Schwinghamer et al. 1986; Brey 1999) and vulnerability to mortality (Adams 1980; Jennings, Reynolds & Mills 1998). Body size distributions have been used to describe structure and energy flux in ecosystems (Sheldon, Prakash & Sutcliffe 1972; Kerr 1974; Platt & Denman 1978; Dickie, Kerr & Boudreau 1987; Boudreau, Dickie & Kerr 1991), and are correlated with species diversity and abundance in communities (Blackburn, Harvey & Pagel 1990; Cotgreave & Harvey 1992, 1994; Siemann, Tilman & Haarstad 1996).
Since body size determines the range of prey sizes a predator can consume (Cohen et al. 1993), larger individuals are expected to feed at higher trophic levels. Studies of size distributions in marine ecosystems are consistent with this hypothesis (Sholtodouglas et al. 1991; Fry & Quinones 1994; France, Chandler & Peters 1998). However, these analyses do not explore relationships among species. If such relationships exist, size could provide an easily measured surrogate of trophic level. Moreover, if changes in the size structure of exploited communities (Pope et al. 1988; Duplisea & Kerr 1995; Rice & Gislason 1996) reflect changes in their trophic level (Pauly et al. 1998; Pauly, Christensen & Walters 2000) then size-based analyses would provide a simple and effective means of assessing the impacts of exploitation on trophic structure.
If there is a relationship between trophic level and body size then this would indicate that there are relationships between trophic level and other life history parameters. This is because maximum body size is correlated negatively with growth, age at maturity, reproductive output and natural mortality (Beverton & Holt 1959; Beverton 1963, 1987; Pauly 1980; Jennings & Beverton 1991, 1992; Charnov 1993; Sadovy 1996; Stamps, Mangel & Phillips 1998), due to trade-offs among life history allocations (Stearns 1976, 1992). Body size is often used as a surrogate for other life history parameters because for many species the detailed population data needed to describe maturation and growth are difficult and costly to gather (Froese & Binohlan 2000).
One impediment to the study of the relationships between body size and trophic level has been the paucity of trophic level estimates for many species. Trophic level has often been estimated from dietary studies (Yang 1982), but massive research effort is needed to obtain dietary data (Greenstreet 1996) and research on fish has tended to focus on a few abundant species of commercial significance (Rice et al. 1991; Christensen 1995). However, nitrogen stable isotope analysis can also be used to estimate trophic level (Fry & Sherr 1984, 1989; Owens 1987) because the abundance of δ15N in the tissues of consumers is typically enriched by 3% relative to their prey (Peterson, Howarth & Garritt 1985; Peterson & Fry 1987; Hobson & Welch 1992; Fry & Quinones 1994). The main benefits of assessing trophic level using stable isotopes are that the abundance of δ15N reflects the composition of the assimilated diet and integrates differences in assimilated diet over time (Hobson & Welch 1992).
The aims of this paper are (i) to describe empirical relationships between trophic level and body size of marine fishes; (ii) to test whether empirical species-based relationships between trophic level and body size can be used to predict the trophic structure of fish communities; and (iii) to describe the trophic structure of entire fish communities based on body size distributions.
Methods
sampling individual species
Individual species of fish were sampled in the Celtic and North Seas with a modified Portuguese High-Headline Trawl and Grande Ouverture Verticale (GOV) demersal trawl, respectively (Knijn et al. 1993; Warnes & Jones 1995). The trawls were fished from the research vessel ‘Cirolana’ and fitted with a cod-end of 20 mm stretched mesh. Tows of 30 min duration were made at a speed of approximately 4 knots. Sixty-one sites in the Celtic Sea were fished during February and March 2000, and 75 sites in the North Sea were fished during August and September 2000 (Fig. 1). In both the Celtic and North Seas, three fish of each species caught were dissected to obtain tissue samples for nitrogen stable isotope analysis. Approximately 2 g of white muscle were dissected from the dorsal musculature of each fish, placed in a vial and immediately frozen to −30 °C. On return to the laboratory the frozen tissue was freeze-dried and ground to a fine powder (particles < 60 µm). This was mixed thoroughly and a 0·7-mg sample was weighed into a tin capsule for stable isotope analysis.
Since trophic level increases with body size within species, fish from different species had to be compared at an equivalent stage of their life history. In the absence of detailed maturity and growth data for all species, we always attempted to sample fish that were between 60% and 80% of the maximum recorded length. Because key events in the life history, such as maturity, occur at a relatively constant proportion of maximum size (Beverton & Holt 1959; Beverton 1963, 1992; Charnov 1993), comparing fishes at a fixed proportion of maximum size is equivalent to comparing them at a specific stage of their life history.
sampling the community
To describe relationships between body size and trophic level for multispecies fish communities, we estimated the 15N content of the entire multispecies catch by log2 body mass class at 15 sites in the northern North Sea during September 2000 (Fig. 1). Catches were made with a GOV trawl, towed for 30 min at approximately 4 knots.
All fish of all species were assigned to log2 size classes. Fish > 512 g total weight were assigned directly to size classes while fish weighing ≤ 512 g were more abundant and had to be subsampled randomly before assignment (fish from all species were combined within size classes). We dissected tissue from 20–25 randomly selected individuals in each size class (or all the fish in the size class if < 20 were caught). The limits on the number of fish processed were imposed by the practicality of dissecting tissue before spoilage had started to occur.
A sample of white muscle tissue was dissected from the dorsal musculature of each fish for stable isotope analysis. Each fish was weighed before the sample was dissected, and the size of the sample retained was set at a fixed percentage of body weight in each size class (Table 1). For fish smaller than 32 g, we cut the largest possible fillet from the dorsal musculature of each fish, and for fish smaller than 4 g we kept all the tissue that remained after removing the head and guts. The tissue from fish in each size class was homogenized in a blender, or crushed and ground manually, to produce a smooth paste. Approximately 4 g of paste were retained and immediately frozen to −30 °C. On return to the laboratory the frozen tissue was freeze-dried and ground to a fine powder (particles < 60 µm). This was thoroughly mixed and a 0·7-mg sample was weighed into a tin capsule for stable isotope analysis.
Size class (g) | % individual body mass sampled | Composition of sample | Number of fish sampled in size class |
---|---|---|---|
> 1–2 | 20% | Whole fish excluding head, tail and guts | 20–25 (all in catch if < 20) |
> 2–4 | 20% | Whole fish excluding head, tail and guts | 20–25 (all in catch if < 20) |
> 4–8 | 20% | White muscle and skin | 20–25 (all in catch if < 20) |
> 8–16 | 20% | White muscle and skin | 20–25 (all in catch if < 20) |
> 16–32 | 10% | White muscle and skin | 20–25 (all in catch if < 20) |
> 32–64 | 5% | White muscle | 20–25 (all in catch if < 20) |
> 64–128 | 5% | White muscle | 20–25 (all in catch if < 20) |
> 128–256 | 2% | White muscle | 20–25 (all in catch if < 20) |
> 256–512 | 1% | White muscle | 20–25 (all in catch if < 20) |
> 512–1024 | 0·5% | White muscle | All in catch |
> 1024–2048 | 0·2% | White muscle | All in catch |
> 2048–4096 | 0·1% | White muscle | All in catch |
> 4096–8192 | 0·05% | White muscle | All in catch |
> 8192–16384 | 0·05% | White muscle | All in catch |
> 16384–32768 | 0·05% | White muscle | All in catch |
stable isotope analysis
The 15N composition of the samples was determined using continuous flow isotope ratio mass spectrometry (CF-IRMS) (Preston & Owens 1983; Preston 1992). Weighed samples of 0·7–1·0 mg ground material were oxidized and the N2 passed to a single inlet dual collector mass spectrometer (Automated Nitrogen Carbon Analysis (ANCA) SL 20–20 system). This was a continuous flow system, so two samples of reference material (an internal standard) were analysed after every five tissue samples in order to calibrate the system and compensate for drift with time (ANCA-SL Dual Isotope v3·4 software).
Ratios of 15N : 14N were expressed relative to N2 in air for nitrogen and calculated as:
The SD for the repeated δ15N measurements made with the reference material was 0·1%.
body size and life histories
The biology and ecology of many fishes in the Celtic and North Seas is poorly known, and detailed growth and maturity data were not always available. For most species, we used maximum recorded body size as a surrogate for other life history parameters (Froese & Binohlan, 2000). Maximum body sizes were taken from Wheeler (1969), Knijn et al. (1993) and unpublished data. Transformations between weight and length were made using conversion factors (Bedford, Woolner & Jones 1986; Coull et al. 1989; CEFAS, unpublished).
Fish communities in the Celtic and North Sea are heavily exploited and the maximum observed size of many species has fallen in recent years. In many cases, this has been associated with reductions in the age and size at maturity and increases in growth rate (Toresen 1990; Rijnsdorp 1993; Millner & Whiting 1996; Rijnsdorp & van Leeuwen 1996). This is the combined effect of genetic and phenotypic responses to fishing mortality (Law & Rowell 1993; Rowell 1993; Stokes & Blythe 1993; Law & Stokes, 2001). In reporting maximum sizes we tried to select those that were indicative of the maximum sizes seen in the fish community at the present time. In many cases, therefore, they are smaller than those reported historically.
data analysis
We used linear regression to explore relationships between maximum body size (as length or weight) and δ15N. In order to confirm that any significant relationships were not confounded by the non-independence of data used in the cross-species analysis, the significance of relationships between body size and trophic level was also assessed using a comparative approach (Felsenstein 1985; Harvey & Pagel 1991; Harvey 1996). Phylogenetically based analyses overcome the fact that related taxa may share traits due to common ancestry and yield data which are statistically independent (Harvey & Pagel 1991). The comparative approach has already been used to assess links between life histories and vulnerability to fishing (Jennings et al. 1998; Reynolds, Jennings & Dulvy 2001) and life histories and recruitment variation (Rickman et al. 2000).
Phylogenetic relationships among species were determined from published taxonomic and molecular evidence (Wheeler 1969; Cohen 1989; Cohen et al. 1990; Nelson 1994). Phylogenetically independent differences (contrasts) in body size within pairs of related species were calculated by subtracting the ln maximum length of the smaller species from the larger one. Preliminary analysis of the data showed that log transformation removed the effects of the relationship between body size and location in the phylogeny (Freckleton 2000). Contrasts in trophic level were then calculated for the same species pair. The significance of the relationship between contrast in maximum length and contrast in trophic level was assessed using binomial tests (Harvey & Pagel 1991).
We analysed the North Sea and Celtic Sea data separately. Contrasts in body size were set to positive since the body size of the smaller species was always subtracted from the larger. If phylogenetic relationships between species could not be resolved, we calculated all (n2 − n)/2 contrasts for each group and presented mean values. The null expectation in this analysis is that the contrasts in trophic level will be randomly distributed around zero when the contrast in body size is set to positive.
Results
species-based analysis
Forty-three species were sampled in the North Sea and 48 species in the Celtic Sea (Tables 2 and 3). The mean length of individuals sampled was 50–77% of maximum length in the North Sea. For five species, only one replicate was obtained in the acceptable size range (Table 2). In the Celtic Sea, mean length of individuals sampled was 54–79% of maximum length, and only one replicate was obtained for the ling Molva molva (Table 3).
Scientific name | Common name | Contrast | Maximum length (mm) | Maximum weight (g) | Sample length (mm) mean ± SD | Mean sample length as % maximum | δ15N% mean ± SD |
---|---|---|---|---|---|---|---|
Scyliorhinus canicula | Lesser spotted dogfish | 880 | 2763 | 647 ± 18 | 74 | 15·0 ± 0·91 | |
Galeorhinus galeus | Tope | 1 | 1670 | 23948 | 1090 ± 164 | 65 | 17·0 ± 0·82 |
Mustelus asterias | Starry smooth hound | 1 | 1700 | 25429 | 1115 ± 15 | 66 | 14·0 ± 0·22 |
Squalus acanthias | Spurdog | 1000 | 4251 | 628 ± 40 | 63 | 12·0 ± 0·57 | |
Raja montagui | Spotted ray | 2 | 790 | 3418 | 626 ± 18 | 79 | 13·7 ± 0·93 |
Leucoraja naevus | Cuckoo ray | 2 | 920 | 5232 | 641 ± 30 | 70 | 13·6 ± 0·44 |
Clupea harengus | Herring | 3 | 315 | 257 | 246 ± 1 | 78 | 13·3 ± 0·23 |
Sardina pilchardus | Pilchard | 3 | 250 | 126 | 190 ± 17 | 76 | 12·6 ± 0·59 |
Sprattus sprattus | Sprat | 3 | 170 | 40 | 126 ± 6 | 74 | 14·3 ± 1·26 |
Engraulis encrasicolus | Anchovy | 200 | 70 | 139 ± 9 | 70 | 14·6 ± 0·95 | |
Argentina sphyraena | Lesser argentine | 270 | 124 | 198 ± 7 | 73 | 12·1 ± 0·22 | |
Lophius piscatorius | Anglerfish | 1060 | 18045 | 653 ± 45 | 62 | 14·3 ± 0·72 | |
Gadiculus argenteus | Silvery pout | 5 | 150 | 29 | 118 ± 7 | 79 | 10·2 ± 0·26 |
Gadus morhua | Cod | 1100 | 14024 | 735 ± 99 | 67 | 15·2 ± 0·77 | |
Melanogrammus aeglefinus | Haddock | 680 | 3157 | 490 ± 30 | 72 | 12·6 ± 0·85 | |
Merlangius merlangus | Whiting | 6 | 420 | 608 | 306 ± 15 | 73 | 17·2 ± 1·24 |
Micromesistius poutassou | Blue whiting | 340 | 315 | 240 ± 10 | 71 | 11·1 ± 0·64 | |
Molva molva | Ling | 4 | 1400 | 16495 | 1107* | 79 | 15·3 |
Phycis blennoides | Forkbeard | 4 | 650 | 3067 | 426 ± 31 | 66 | 13·3 ± 0·50 |
Pollachius virens | Saithe | 6 | 1400 | 19860 | 1054 ± 28 | 75 | 14·4 ± 0·60 |
Trisopterus esmarki | Norway pout | 7 | 230 | 122 | 168 ± 12 | 73 | 13·7 ± 1·03 |
Trisopterus minutus | Poor cod | 7 | 200 | 80 | 146 ± 5 | 73 | 12·9 ± 1·12 |
Merluccius merluccius | Hake | 5 | 1200 | 13046 | 863 ± 120 | 72 | 13·5 ± 0·29 |
Beryx splendens | Beryx | 400 | 319 | 241 ± 2 | 60 | 10·8 ± 0·85 | |
Zeus faber | John dory | 8 | 500 | 2214 | 356 ± 40 | 71 | 14·7 ± 0·74 |
Capros aper | Boarfish | 8 | 160 | 52 | 116 ± 5 | 73 | 10·4 ± 0·20 |
Helicolenus dactylopterus | Bluemouth | 460 | 1750 | 311 ± 17 | 68 | 13·5 ± 0·26 | |
Aspitrigla cuculus | Red gurnard | 9 | 400 | 655 | 258 ± 15 | 65 | 13·2 ± 0·11 |
Eutrigla gurnardus | Grey gurnard | 9 | 460 | 886 | 325 ± 27 | 71 | 12·8 ± 0·23 |
Dicentrarchus labrax | Bass | 10 | 1000 | 11530 | 639 ± 3 | 64 | 15·8 ± 0·42 |
Trachurus trachurus | horse mackerel | 10 | 500 | 1344 | 347 ± 22 | 69 | 13·8 ± 0·41 |
Sponyliosoma canthurus | Black sea bream | 500 | 2068 | 302 ± 22 | 60 | 15·2 ± 0·85 | |
Mullus surmuletus | Red mullet | 11 | 400 | 940 | 280 ± 20 | 70 | 15·3 ± 0·82 |
Echiichthys vipera | lesser weaver | 11 | 160 | 30 | 106 ± 8 | 66 | 15·1 ± 0·21 |
Hyperoplus immaculatus | Corbins sandeel | 11 | 340 | 151 | 267 ± 5 | 79 | 12·2 ± 0·87 |
Callionymus lyra | Dragonet | 11 | 230 | 74 | 158 ± 14 | 69 | 12·7 ± 0·40 |
Scomber scombrus | Mackerel | 11 | 399 | 555 | 289 ± 4 | 72 | 12·7 ± 1·26 |
Lepidorhombus boscii | Four-spot megrim | 12 | 410 | 556 | 284 ± 17 | 69 | 11·9 ± 0·12 |
Lepidorhombus whiffiagonis | Megrim | 12 | 600 | 1970 | 394 ± 38 | 66 | 12·5 ± 0·74 |
Arnoglossus imperialis | Imperial scaldfish | 250 | 93 | 165 ± 8 | 66 | 10·9 ± 0·14 | |
Glyptocephalus cyanoglossus | Witch | 13 | 460 | 736 | 299 ± 24 | 65 | 13·6 ± 0·51 |
Hippoglossoides platessoides | Long rough dab | 250 | 133 | 180 ± 10 | 72 | 14·1 ± 0·16 | |
Limanda limanda | Dab | 310 | 325 | 193 ± 17 | 62 | 14·7 ± 1·66 | |
Microstomus kitt | Lemon sole | 13 | 440 | 1102 | 239 ± 14 | 54 | 12·9 ± 2·00 |
Platichthys flesus | Flounder | 14 | 460 | 1231 | 315 ± 3 | 68 | 13·5 ± 1·17 |
Pleuronectes platessa | Plaice | 14 | 510 | 1454 | 336 ± 13 | 66 | 12·9 ± 0·97 |
Microchirus variegatus | Thick back sole | 15 | 300 | 314 | 202 ± 3 | 67 | 13·5 ± 0·73 |
Solea solea | Sole | 15 | 480 | 1339 | 329 ± 38 | 69 | 14·6 ± 0·71 |
Scientific name | Common name | Contrast | Maximum length (mm) | Maximum weight (g) | Sample length (mm) mean ± SD | Mean sample length as % maximum | δ15N% mean ± SD |
---|---|---|---|---|---|---|---|
Scyliorhinus canicula | Lesser spotted dogfish | 880 | 2763 | 657 ± 49 | 75 | 14·1 ± 0·86 | |
Leucoraja naevus | Cuckoo ray | 1 | 920 | 4220 | 550* | 60 | 12·7 |
Raja radiata | Starry ray | 1 | 660 | 2450 | 433 ± 17 | 66 | 14·3 ± 0·38 |
Clupea harengus | Herring | 2 | 340 | 326 | 262 ± 3 | 77 | 12·1 ± 0·53 |
Sprattus sprattus | Sprat | 2 | 170 | 38 | 127 ± 9 | 75 | 13·3 ± 0·42 |
Argentina sphyraena | Lesser argentine | 270 | 124 | 197 ± 9 | 73 | 13·3 ± 0·16 | |
Lophius piscatorius | Anglerfish | 1060 | 18045 | 672 ± 99 | 63 | 15·3 ± 0·81 | |
Brosme brosme | Torsk | 3 | 890 | 8464 | 550* | 62 | 14·3 |
Gadiculus argenteus | Silvery pout | 4 | 160 | 35 | 120 ± 9 | 75 | 11·3 ± 0·87 |
Gadus morhua | Cod | 1230 | 17650 | 816 ± 94 | 66 | 16·5 ± 1·83 | |
Melanogrammus aeglefinus | Haddock | 720 | 3515 | 434 ± 3 | 60 | 14·7 ± 0·23 | |
Merlangius merlangus | Whiting | 5 | 540 | 1360 | 271 ± 4 | 50 | 14·8 ± 0·28 |
Micromesistius poutassou | Blue whiting | 370 | 410 | 283 ± 12 | 77 | 11·3 ± 0·22 | |
Molva molva | Ling | 3 | 1400 | 16495 | 730 ± 17 | 51 | 13·6 ± 0·58 |
Pollachius virens | Saithe | 5 | 1550 | 23609 | 1070 ± 10 | 76 | 15·2 ± 0·52 |
Enchelyopus cimbrius | Four bearded rockling | 330 | 182 | 222 ± 8 | 67 | 13·3 ± 0·17 | |
Trisopterus esmarki | Norway pout | 6 | 230 | 122 | 166 ± 3 | 72 | 13·6 ± 0·54 |
Trisopterus luscus | Bib | 6 | 420 | 1095 | 227 ± 6 | 54 | 16·8 ± 0·27 |
Trisopterus minutus | Poor cod | 6 | 200 | 80 | 153 ± 6 | 76 | 17·3 ± 0·32 |
Merluccius merluccius | Hake | 4 | 1100 | 10950 | 591 ± 37 | 54 | 14·4 ± 0·71 |
Capros aper | Boarfish | 7 | 140 | 40 | 93 ± 11 | 67 | 11·9 ± 0·39 |
Sebastes viviparus | Norway haddock | 8 | 360 | 876 | 237 ± 38 | 66 | 12·5 ± 0·85 |
Eutrigla gurnardus | Grey gurnard | 7 | 460 | 886 | 325 ± 25 | 71 | 15·3 ± 0·08 |
Myxocephalus scorpius | Bull rout | 8 | 350 | 838 | 250* | 71 | 16·5 |
Agonus cataphractus | Pogge | 8 | 180 | 37 | 132* | 73 | 14·1 |
Cyclopterus lumpus | Lumpsucker | 8 | 500 | 5780 | 371* | 74 | 12·1 |
Trachurus trachurus | Horse mackerel | 9 | 500 | 1344 | 347 ± 13 | 69 | 17·9 ± 0·49 |
Echiichthys vipera | Lesser weaver | 9 | 160 | 35 | 96 ± 9 | 60 | 15·1 ± 0·10 |
Anarhichas lupus | Wolffish | 9 | 1000 | 10392 | 714 ± 144 | 71 | 12·8 ± 0·06 |
Ammodytes marinus | Raitt’s sandeel | 10 | 240 | 48 | 172 ± 22 | 72 | 13·6 ± 0·40 |
Hyperoplus lanceolatus | Greater sandeel | 10 | 350 | 154 | 226 ± 7 | 64 | 14·4 ± 0·51 |
Callionymus lyra | Dragonet | 300 | 148 | 227 ± 16 | 76 | 12·7 ± 0·58 | |
Scomber scombrus | Mackerel | 9 | 399 | 555 | 268 ± 19 | 67 | 12·4 ± 0·56 |
Lepidorhombus whiffiagonis | Megrim | 11 | 610 | 2059 | 357 ± 19 | 58 | 12·1 ± 1·09 |
Phrynorhombus norvegicus | Norwegian topknot | 11 | 140 | 24 | 101 ± 12 | 72 | 13·8 ± 0·30 |
Arnoglossus laterna | Scaldfish | 160 | 38 | 109 ± 7 | 68 | 15·2 ± 0·51 | |
Glyptocephalus cyanoglossus | Witch | 12 | 470 | 696 | 308 ± 39 | 65 | 12·9 ± 0·29 |
Hippoglossoides platessoides | Long rough dab | 250 | 133 | 181 ± 24 | 72 | 13·4 ± 0·28 | |
Limanda limanda | Dab | 13 | 330 | 400 | 190 ± 10 | 58 | 15·5 ± 0·22 |
Microstomus kitt | Lemon sole | 12 | 457 | 1181 | 267 ± 15 | 58 | 13·8 ± 0·79 |
Pleuronectes platessa | Plaice | 13 | 580 | 2157 | 361 ± 7 | 62 | 15·3 ± 0·78 |
Buglossinium luteum | Solenette | 14 | 130 | 22 | 93 ± 6 | 72 | 14·5 ± 0·37 |
Solea solea | Sole | 14 | 460 | 950 | 278 ± 30 | 60 | 16·8 ± 0·11 |
In the North Sea, there was a significant relationship between the mean length of individuals sampled as a proportion of maximum length (Lp) and maximum length (F1,41 = 5·15, P = 0·029). Given that body size and δ15N can be correlated positively within species (e.g. Sholtodouglas et al. 1991), the significant relationship between Lp and maximum length could have affected the results of subsequent analyses. Further exploration of the data showed that the significance of the relationship between Lp and maximum length was due to the small ling that were sampled (Table 2, maximum length 1400 mm, P = 51%). When the data for ling were excluded, the relationship between Lp and maximum length was not significant (F1,40 = 2·38, P = 0·131). In subsequent analyses, the inclusion or exclusion of the North Sea ling did not affect on the significance of our results. Moreover, for both the North Sea and Celtic Sea fishes, the inclusion or exclusion of species for which Lp ≠ 60–80%, or the inclusion or exclusion of those species for which only one replicate was caught, made no difference to the significance of any results that we report. The relationship between Lp and maximum length was not significant for the Celtic Sea (F1,46 = 0·41, P = 0·523).
Mean δ15N ranged from 11·3 to 17·9 in the North Sea and 10·2–17·2 in the Celtic Sea. The δ15N values for the 29 species that occurred in both the North and Celtic Seas (Fig. 2) were significantly but weakly correlated (r = 0·53, d.f. = 27; P < 0·01). The relationship between log2 maximum body mass or log10 maximum length and δ15N (3, 4) was not significant for North Sea fishes (mass: r2 = 0·03; F1,41 = 1·16 P > 0·1 length: r2 = 0·02; F1,41 = 0·86, P > 0·1) but was significant for Celtic Sea fishes (mass: δ15N = 0·249 log2 mass + 11·1, r2 = 0·21; F1,46 = 11·84; P = 0·001 length: δ15N = 2·24 log10 length + 7·55; r2 = 0·17; F1,46 = 9·58; P = 0·003). When the relationships between log2 maximum body mass or log10 maximum length and δ15N were investigated using a phylogenetic comparative approach (Fig. 5), the results were not significant in both the North Sea (binomial test: mass: 6 positive, 8 negative contrasts, P > 0·1; length: 5 positive, 9 negative contrasts P > 0·1) and Celtic Sea (mass: 8 positive, 7 negative contrasts, P > 0·1; length: 9 positive, 6 negative contrasts, P > 0·1).
community-based analysis
At the 15 sites where relationships between δ15N and body mass in the entire community were investigated, there were significant positive relationships between these variables. The pooled data (Fig. 6) showed a significant increase in δ15N with body size in the community as a whole (r2 = 0·93, F = 147·8, d.f. = 1,12; P < 0·0001). The relationship was given by the equation δ15N = 0·335 log2 body mass + 10·4. There was very little variation around the mean δ15N at size among sites, suggesting similar structuring of the fish communities.
Discussion
Cross-species relationships between δ15N and body size were weak or non-significant, in accordance with the results of the comparative analyses. However, the δ15N of the whole community increased consistently and significantly with body size, implying that small species feeding at high trophic levels were relatively scarce. Patterns at the community level were consistent with the expectation that predator–prey relationships lead to powerful size-based trophic structuring.
Body size is used widely to predict production: biomass ratios, natural mortality and respiration rates of animals (Banse & Moser 1980; Schwinghamer et al. 1986; Brey 1999). However, our relationships between body size and δ15N were too weak to provide precise predictions of trophic level. The relationships could be weak (i) because the maximum body size of a species does not constrain its trophic level; (ii) because there are large spatial and temporal differences in the δ15N of the same species that reduce the power of our analyses; or (iii) because δ15N is not a good index of trophic level. We favour (i) and reject (ii) and (iii) for the following reasons.
A large sampling area is needed to catch a range of species of comparable size, while a small sampling area is needed to avoid spatial variance in δ15N among individuals. If the sampling area is reduced, then the number of species that can be included in the analysis will fall. We consider that it is more important to catch fish from narrow and comparable size classes (60–80% of maximum length in this study) than to be overly concerned about spatial variation, since the effects of ontogenetic shifts in δ15N and low replication dramatically reduce the power of our analyses. For example, studies of fishes taken from a single trawl tow (c. 2 nautical miles in length) show a 2% increase in δ15N between the smallest and largest individuals of the same species (S.J., unpublished data), while variation among individuals of the same species and same size is much less (Tables 2 and 3). Moreover, we demonstrated significant correlations between the δ15N of the same species in two widely separated sea areas, suggesting that fishes at the same life history stages are feeding at similar trophic levels.
In our analyses we assumed that δ15N was linearly related to trophic level. We did not attempt to assign actual trophic levels to the fishes, since this requires that the trophic level of a group of animals closer to the base of the food chain is known. Assigning trophic levels to phytoplankton or zooplankton, or species that consume them, is exceedingly difficult, because pure phytoplankton cannot be extracted from plankton samples to make δ15N measurements and because the trophic level of zooplankton can vary by two or more trophic levels due to the effects of the microbial food loop (Azam et al. 1983). We assume a δ15N enrichment of around 3% for each trophic step (e.g. Cabana & Rasmussen 1994; Zanden, Casselman & Rasmussen 1999; Post, Pace & Hairston 2000). However, Adams & Sterner (2000) have commented that the fractionation of 3·4% reported by Minagawa & Wada (1984) was actually the numerical mean of reported values between 1·3 and 5·3%, and this has prompted calls for experimental evaluation and validation (e.g. Gannes et al. 1997; Pinnegar & Polunin 1999; Ponsard & Averbuch 1999; Schoella 1999; Adams & Sterner 2000). We cannot provide such validation for all the complex feeding relationships that apply to the species included in this study, but the choice of value will not affect the conclusions of our analyses unless the increase in δ15N is a non-linear function of trophic level. There is no existing evidence for this.
If there is a linear increase of 3%δ15N per trophic level, our community-based analysis can be used to estimate the mean predator : prey body mass ratio. The slope of the relationship between log2 body mass and δ15N (0·335) suggests that predators in this North Sea fish community have a mean mass approximately 80 times greater than that of their prey (mass ratio = (3/slope)2). This assumes that predator–prey relationships in the community occur predominantly among the species sampled. Clearly, many of the smaller size classes will eat plankton and benthos, but larger ones are predominantly piscivorous.
Attempts to assign trophic levels to small fishes and invertebrates using diet data have been fraught with difficulty, because gut contents can be virtually indistinguishable and diet studies require a massive sampling commitment (Polunin & Pinnegar 2002). If body size could be used as a surrogate for trophic level in marine communities, then size-spectra could be analysed to investigate changes in trophic structure as well as production. It is well known that the size structure of the biota in a marine ecosystem follows regular patterns (Sheldon et al. 1972) and the relationship between body size (as classes) and total biomass by size class can be predicted by theoretical models of energy flow from prey to predators (Kerr 1974; Platt & Denman 1978; Dickie et al. 1987). Size spectra have been used widely to describe aquatic community structure (Kerr 1974; Dickie et al. 1987; Pope et al. 1988; Boudreau et al. 1991; Thiebaux & Dickie 1992; Sprules & Stockwell 1995) and to investigate the effect of perturbations on community structure (Boudreau & Dickie 1992; Duplisea & Kerr 1995; Rice & Gislason 1996). The present study provides empirical evidence, in support of existing theoretical evidence, that size alone is a useful predictor of trophic level in the community. This provides a basis for integrating studies of body size distributions and food webs (e.g. Cohen et al. 1993; Hall & Raffaelli 1993) and for assessing the impacts of fishing on the marine ecosystem (e.g. Hall 1999).
Size-based analyses have shown that fishing leads to predictable changes in the size structure of fish communities (Pope et al. 1988; Boudreau & Dickie 1992; Duplisea & Kerr 1995; Rice & Gislason 1996), while ecosystem models based on trophic relationships suggest that fishing may reduce the trophic level of fish communities (Christensen & Pauly 1992; Walters, Christensen & Pauly 1997; Pauly et al. 1998, 2000). Size-based models of fishing effects are far easier to parameterize than ecosystem models, and our results suggest that size spectra could be reparameterized using relationships between body size and trophic level for the community. This provides a link between the use of size spectra to examine the effects of exploitation on community structure and the use of ecosystem models. The next step is to investigate the stability of relationships between community size structure and trophic level in space and time. If such relationships are stable, then the increases in the slope of size spectra that result from fishing (Pope et al. 1988; Rice & Gislason 1996), will indicate decreases in the trophic level of the community.
Acknowledgements
We are particularly grateful to Christie Stewart for helping with sample collection, to Richard Ayers for accommodating this work on groundfish surveys and to Gillian Taylor for assistance with stable isotope analysis. We would also like to thank the referees for their constructive comments. This research was funded by CEFAS (DP166).