Abstract

Sweeting, C. J., Badalamenti, F., D'Anna, G., Pipitone, C., and Polunin, N. V. C. 2009. Steeper biomass spectra of demersal fish communities after trawler exclusion in Sicily. – ICES Journal of Marine Science, 66: 195–202.

The effects of trawling on Mediterranean demersal fish communities were assessed by comparing them with the normalized biomass spectra among three gulfs on the north Sicilian coast, one of which had been subject to 15 years of trawler exclusion. Comparisons were conducted across seasons and among depth strata. Biomass-spectra slopes were significantly steeper in the gulf that was closed to trawling than in the unprotected gulfs. This was attributed to the exclusion of trawlers, which lacked catch-size selectivity, and to the continued fishing by artisanal gears that are more size selective. The biomass of all size classes was higher in the protected gulf than in unprotected areas, and increases were greatest in smaller size classes. Community mean individual fish mass was similar among all areas with a wider range of body masses present in the trawl exclusion area, compensating for the greater abundance of small fish. Size-spectra slopes were generally shallower and midpoint heights lower with increasing depth and were greater in autumn than spring; the effect of these seasonal and depth factors was as great as that of protection. Depth patterns are explainable by bathymetric trends in within- and among-species fish size. Seasonal differences were attributed to variation in spawning behaviour and fishery recruitment, with seasonal differences being greater in unprotected locations as a result of recruitment overfishing.

Introduction

In marine ecosystems, the exploitation of fishery resources often has consequences beyond the reduction of fish biomass (Jennings and Blanchard, 2004), giving rise to community changes in fish body size and diversity (Bianchi et al., 2000; Greenstreet and Hall, 1996), and damage to the physical habitat (Jennings and Kaiser, 1998). Fishing can change the mean fish community trophic level (Pauly et al., 1998) and the foodweb source of production (Kaiser and Spencer, 1994). The ecosystem approach to fishery management can account for fishing impacts on different ecosystem components and recognizes that species are not exploited in isolation from each other (Jennings, 2004).

The impacts of fishing in the Mediterranean are wide ranging, the result of a very high fishing intensity employing a great variety of fishing gears and practices, and exploiting a great diversity of species and habitats (Tudela, 2004). Catches and earnings are dominated by bottom-trawl fleets, which utilize small mesh sizes (40 mm in Italy; Machias et al., 2004). This results in a multispecies fishery characterized by a large proportion of juvenile fish in catches, high bycatch and discard rates, and frequent violation of minimum landing size (Machias et al., 2004; Tudela, 2004).

Finding robust ecosystem metrics to test ecosystem effects of fishing has proved difficult (Rice, 2000; Rochet and Trenkel, 2003). Of the measures available, size-based indicators such as size spectra seem promising. The abundance or biomass of individuals decreases log-linearly with size, the result of energy use and energy transfer efficiency among organisms of different size. Body size is also a significant determinant of catchability (Bianchi et al., 2000) and influences a species’ ability to recover from fishing through links to life history and population dynamics (Charnov, 1993).

Fishing can directly alter the size composition of exploited communities by triggering disproportionate mortality on larger individuals or species. Indirectly, fishing may increase abundance of small prey species through predation release (Shin et al., 2005). Therefore, as fishing mortality increases and the relative abundance of large individuals decreases, the mean size of individuals in the community may decrease, resulting in a steeper decrease in biomass–size spectra and changes in species composition (Figure 1).

Figure 1.

How the size spectra from (a) an unexploited community, (b) alters as a result of size-selective fishing, and (c) potential prey release. The dashed lines in (b) and (c) represent the unexploited size spectra, and the circles define the midpoint height relative to the unexploited size spectra.

Figure 1.

How the size spectra from (a) an unexploited community, (b) alters as a result of size-selective fishing, and (c) potential prey release. The dashed lines in (b) and (c) represent the unexploited size spectra, and the circles define the midpoint height relative to the unexploited size spectra.

Fishing closures may mitigate many of the direct and indirect effects of exploitation (Collie et al., 2000; Pipitone et al., 2000), and closed areas have proved useful in elucidating ecosystem effects of, and recovery from, fishing (Murawski et al., 2000; Badalamenti et al., 2002). However, most protected areas are too small to assess the impacts of commercial trawl fisheries on the size composition of associated communities at appropriate scales. Such work is essential to understanding both what impact trawling has on the ecosystem and whether or not areas closed to trawling represent a viable management tool for ecosystem restoration or fishery management.

This study examines the effects of trawling and subsequent closure on the biomass–size spectra of demersal marine species by comparing the Gulfs of Castellammare (GCAST), Termini Imerese (GTERM), and Sant'Agata (GSANT; North Sicily; Figure 2). The Gulf of Castellammare has been closed to trawling since 1990, and the total fish biomass there has increased eightfold (Pipitone et al., 2000). The assumptions that trawler exclusion resulted in an increased size-spectrum midpoint height (a measure of community-wide abundance) and a shallower size spectrum were tested, and effects examined across depths and seasons for consistency of response.

Figure 2.

Charts for each of the three gulfs, showing depth strata (10–50, 51–100, and 101–200 m), the associated elemental statistical units (ESUs) for subsampling, and the no-trawl area in the Gulf of Castellammare. Areas in which commercial fishing is allowed, but that are not conducive to scientific trawling, are shown in grey.

Figure 2.

Charts for each of the three gulfs, showing depth strata (10–50, 51–100, and 101–200 m), the associated elemental statistical units (ESUs) for subsampling, and the no-trawl area in the Gulf of Castellammare. Areas in which commercial fishing is allowed, but that are not conducive to scientific trawling, are shown in grey.

Material and methods

GTERM and GSANT are next to each other in the middle of the north coast of Sicily. GCAST, which is located 200 km farther west (Figure 2), has been subject to a 200 km2 ban on commercial trawling since 1990 (Pipitone et al., 2000). Fishing within the trawl exclusion area is restricted to artisanal vessels employing mainly size-selective static gears such as trammelnets and set gillnets. Artisanal fishing extends over the whole continental shelf of GCAST. GTERM and GSANT have large trawling fleets (Cavaliere et al., 1988; Greco et al., 1993), as well as artisanal fisheries. The only regulation imposed on GTERM and GSANT is a ban on trawling in water shallower than 50 m, which applies to all Italian coastlines. Both GTERM and GSANT are overfished (Greco, 1994).

All three areas are characterized by large, central, sandy shores with cliffs at both outer edges. The seabed is largely soft sediment, except immediately under the surrounding cliffs where there are boulder slope habitats. Sampling was done only over soft sediment habitats.

The demersal community was sampled using a benthic otter trawl, with headline and groundrope lengths of 31.3 and 41 m, respectively, and a stretched mesh codend of 35 mm deployed from the commercial fishing vessel FV “Giaguaro”. Trawls were undertaken at 2.5 knots and lasted 30 min in GCAST and 60 min elsewhere.

Each gulf (location) was divided into three depth strata (10–50 m (A), 51–100 m (B), 101–200 m (C)), each of which was divided into boxes of approximately 2.25 square nautical miles, a number of which were randomly sampled during each survey in each stratum. Surveys were carried out over four consecutive seasons in GCAST starting in summer 2004; surveys were carried out in autumn 2004 and spring 2005 only in GTERM and GSANT.

Fish catch was sorted on board. Where the catch of one or more species was excessively large, a subsample was taken after sorting. Samples were returned to the laboratory daily for subsequent measurement. Total length and weight of all fish were measured to the nearest 5 mm and 1 g. Catch was standardized to a 60 min trawl, then all individuals were allocated to log2 body mass classes (grammes) from −2 (<0.5 g) to 11 (>2048 g), regardless of taxonomy.

Size-spectra analyses are only applicable to data in the size range over which the gear effectively samples the community (Jennings and Dulvy, 2005). Selectivity was determined based on theoretical expectations that normalized fish biomass and body weight class are negatively log-linearly related in marine ecosystems. This expectation was not met below size class 4 (8–16 g) and was direct evidence of inefficient gear sampling for fish <8 g, which were excluded from the analysis.

The slopes of size spectra were calculated from the linear regression of log2 normalized biomass [total biomass (g) per weight class divided by weight span (g) of the weight class] against the midpoint of each log2 weight class. Normalized biomass spectra were centred on the predominant median weight class (weight class 6; 16–32 g) to calculate midpoint height, rather than use the regression intercept. This was to remove the influence of correlation between slope and intercept (Daan et al., 2003). These spectra were compared among gulfs and within GCAST among all four seasons.

Abundance size spectra were also constructed using log10 fish abundance instead of normalized biomass on the y-axis. Abundance size spectra exhibited patterns very similar to those based on normalized biomass with the same interpretation and, therefore, are not presented here.

A permutational analysis of variance (PERANOVA; Anderson, 2001) was used to test the differences in slopes and midpoint heights of the size spectra. The PERANOVA approach was selected to overcome difficulties in testing differences in the first design, where the number of protected and unprotected gulfs was different (one protected, GCAST, vs. two unprotected, GSANT and GTERM). Heteroscedascity was checked using Cochran's test and, when possible, removed using the appropriate transformations (see tables for details). Two analytical models were used to assess (i) the effects of the trawl exclusion in different seasons and strata and (ii) the seasonal variability within GCAST. The first design included three factors, treated as fixed and orthogonal: gulf (Lo—three levels), season (Se—two levels), and stratum (St—three levels), with experimental trawls treated as replicates (n = 6). Variance was partitioned for gulf (Lo) by contrasting protected vs. unprotected gulfs and among unprotected gulfs because the a priori hypothesis was that there were differences between the three gulfs mediated by the trawling ban. The second design included two fixed and orthogonal factors: season (Se—four levels) and stratum (St—three levels), with experimental trawls treated as replicates (n = 6).

Similarity matrices were created using PERMANOVA+ for PRIMER v. 6 (Anderson et al., 2008), choosing Euclidean distances as a distance measure. For all tests, 9999 permutations were allowed under the reduced model. For both designs, significance was set at α = 0.05. A posteriori pairwise tests were also performed within the PERANOVA test (Anderson, 2001; Anderson et al., 2008).

Results

The slopes and midpoints of the normalized biomass spectra displayed significant season × stratum interactions and differences among the gulfs (Table 1). The slopes and midpoints were similar between the two unprotected sites, GSANT and GTERM, with the slope significantly shallower, the midpoint height significantly lower than the slope, and midpoint height found in the gulf closed to trawling (Tables 1 and 2).

Table 1.

Analysis of the effects of trawl exclusion on size spectra variables: (top panel) PERANOVA for analysis of protected (P) vs. unprotected (UP) locations; (bottom panel) pairwise comparisons of main effects and interaction terms.

Source d.f. Slope
 
Midpoint
 
MS Pseudo-F p (perm) MS Pseudo-F p (perm) 
Season (Se) 0.9477 14.403 0.0003 0.0023 0.1269 0.7219 
Stratum (St) 1.5900 24.164 0.0001 0.0813 4.5608 0.0123 
Location (Lo) 1.1323 17.208 0.0001 2.4064 134.9500 0.0001 
 P vs. UP 2.2419 33.809 0.0001 4.8114 281.4600 0.0001 
 UP 0.0227 0.27836 0.5881 0.0014 0.0776 0.7842 
Se × St 0.2060 3.1303 0.0456 0.1762 9.8826 0.0002 
Se × Lo 0.0361 0.54862 0.5777 0.0082 0.4620 0.6384 
 Se × P vs. UP 0.0151 0.22835 0.6364 0.0002 0.0105 0.9153 
 Se × UP 0.0571 0.70118 0.4091 0.0163 0.8799 0.3608 
St × Lo 0.0494 0.75068 0.5569 0.0088 0.4961 0.7316 
 St × P vs. UP 0.0069 0.10371 0.9013 0.0087 0.5100 0.5981 
 St × UP 0.0919 1.1295 0.3327 0.0090 0.4844 0.6196 
Se × St × Lo 0.0517 0.78531 0.5309 0.0045 0.2545 0.9031 
 Se × St × P vs. UP 0.0131 0.19712 0.8208 0.0088 0.5146 0.5993 
 Se × St × UP 0.0903 1.1094 0.3384 0.0003 0.0151 0.9836 
Residuals 90 0.0658   0.0178   
Total 107       
Cochran's test  C = 0.2022 (n.s.) C = 0.2022 (n.s.) 
Transformation  None   Ln(X + Constant), Constant = 3 

 
Pairwise comparisons:       
Location (Lo) GCAST < GSANT = GTERM GSANT = GTERM < GCAST 
Se × St interactions:       
Season (Se)  A B C A B C 
 n.s. Au < Sp n.s. Au < Sp n.s. Au > Sp 
Stratum (St)  Au Sp  Au Sp  
 A = B < C A < B = C  A < B = C A = B > C  
Source d.f. Slope
 
Midpoint
 
MS Pseudo-F p (perm) MS Pseudo-F p (perm) 
Season (Se) 0.9477 14.403 0.0003 0.0023 0.1269 0.7219 
Stratum (St) 1.5900 24.164 0.0001 0.0813 4.5608 0.0123 
Location (Lo) 1.1323 17.208 0.0001 2.4064 134.9500 0.0001 
 P vs. UP 2.2419 33.809 0.0001 4.8114 281.4600 0.0001 
 UP 0.0227 0.27836 0.5881 0.0014 0.0776 0.7842 
Se × St 0.2060 3.1303 0.0456 0.1762 9.8826 0.0002 
Se × Lo 0.0361 0.54862 0.5777 0.0082 0.4620 0.6384 
 Se × P vs. UP 0.0151 0.22835 0.6364 0.0002 0.0105 0.9153 
 Se × UP 0.0571 0.70118 0.4091 0.0163 0.8799 0.3608 
St × Lo 0.0494 0.75068 0.5569 0.0088 0.4961 0.7316 
 St × P vs. UP 0.0069 0.10371 0.9013 0.0087 0.5100 0.5981 
 St × UP 0.0919 1.1295 0.3327 0.0090 0.4844 0.6196 
Se × St × Lo 0.0517 0.78531 0.5309 0.0045 0.2545 0.9031 
 Se × St × P vs. UP 0.0131 0.19712 0.8208 0.0088 0.5146 0.5993 
 Se × St × UP 0.0903 1.1094 0.3384 0.0003 0.0151 0.9836 
Residuals 90 0.0658   0.0178   
Total 107       
Cochran's test  C = 0.2022 (n.s.) C = 0.2022 (n.s.) 
Transformation  None   Ln(X + Constant), Constant = 3 

 
Pairwise comparisons:       
Location (Lo) GCAST < GSANT = GTERM GSANT = GTERM < GCAST 
Se × St interactions:       
Season (Se)  A B C A B C 
 n.s. Au < Sp n.s. Au < Sp n.s. Au > Sp 
Stratum (St)  Au Sp  Au Sp  
 A = B < C A < B = C  A < B = C A = B > C  

Significant values of p are emboldened. GCAST, Gulf of Castellammare; GSANT, Gulf of Sant'Agata; GTERM, Gulf of Termini Imerese; A, 10–50 m; B, 51–100 m; C, 101–200 m; Au, autumn; Sp, spring; n.s., not significant.

Table 2.

Slope and midpoint height of normalized biomass–size spectra for the fish communities sampled.

Slope
 
Midpoint height
 
Term Mean 1 s.d. Term Mean 1 s.d. 
GCAST −1.383 0.260 GCAST 6.225 1.091 
GSANT −1.059 0.400 GSANT 2.874 0.757 
GTERM −1.095 0.291 GTERM 2.951 0.924 
A*Au −1.423 0.318 A*Au 3.404 2.132 
A*Sp −1.348 0.349 A* Sp 4.291 1.828 
B*Au −1.366 0.252 B*Au 4.342 1.757 
B*Sp −1.006 0.296 B*Sp 4.340 1.558 
C*Au −1.029 0.302 C*Au 4.272 1.769 
C*Sp −0.902 0.211 C*Sp 3.452 1.803 
Slope
 
Midpoint height
 
Term Mean 1 s.d. Term Mean 1 s.d. 
GCAST −1.383 0.260 GCAST 6.225 1.091 
GSANT −1.059 0.400 GSANT 2.874 0.757 
GTERM −1.095 0.291 GTERM 2.951 0.924 
A*Au −1.423 0.318 A*Au 3.404 2.132 
A*Sp −1.348 0.349 A* Sp 4.291 1.828 
B*Au −1.366 0.252 B*Au 4.342 1.757 
B*Sp −1.006 0.296 B*Sp 4.340 1.558 
C*Au −1.029 0.302 C*Au 4.272 1.769 
C*Sp −0.902 0.211 C*Sp 3.452 1.803 

GCAST = Gulf of Castellammare, Gulf of Sant'Agata = GSANT, and GTERM = Gulf of Termini Imerese; A = 10–50 m, B = 51–100 m, and C = 101–200 m; Au = autumn, Sp = spring.

Seasonal differences in slope were observed in depth stratum B (50–100 m depth) only, with slopes being steeper in autumn than in spring (Tables 1 and 2). For both strata A and C, no seasonal effects on slope were found. In autumn, strata A and B had similar slopes, and both were steeper than stratum C; in contrast, in spring, strata B and C were comparable, and both had shallower slopes than stratum A (Tables 1 and 2).

Midpoint height did not vary among seasons in stratum B, but in the shallower stratum A, the midpoint heights were lower in autumn than in spring, whereas they were lower in spring than in autumn in the deeper stratum C. In autumn, the midpoint heights were lowest in stratum A, whereas strata B and C were comparable. In contrast, in spring, the midpoint heights were comparable between strata A and B and significantly lower in stratum C (Tables 1 and 2).

Despite differences in the slope of size spectra, mean mass of the fish assemblage was comparable among locations (Table 3), except for significantly greater mean fish mass in GCAST in spring. The largest individuals collected in both seasons were obtained from GCAST (Table 3), and more than 70% of individuals >500 g were caught in GCAST.

Table 3.

Differences in fish mass among locations represented by mean weight and the weight at the 95%-ile of the biomass spectra of all fish >8 g at each location.

Season Parameter Fish mass (g)
 
  GCAST GSANT GTERM 
Autumn 95%-ile 211 215 225 
Mean 61 66 62 
Spring 95%-ile 363 227 332 
Mean 89 63 70 
Season Parameter Fish mass (g)
 
  GCAST GSANT GTERM 
Autumn 95%-ile 211 215 225 
Mean 61 66 62 
Spring 95%-ile 363 227 332 
Mean 89 63 70 

Within GCAST, slope was independent of season but differed among strata (Table 4). Statistically, slopes in strata A and B were comparable and both steeper than observed in stratum C, so that slope became shallower with depth (Tables 4 and 5).

Table 4.

Analysis of the effects of season on size-spectra variables within GCAST: (top panel) PERANOVA table of the seasonal analysis, and (bottom panel) pairwise comparisons of main effects and interaction terms.

Source d.f. Slope
 
Midpoint
 
  MS Pseudo-F p (perm) MS Pseudo-F p (perm) 
Season (Se) 0.0494 1.3866 0.2435 0.3158 0.3513 0.7984 
Stratum (St) 0.9346 26.2190 0.0001 3.8684 4.3041 0.0156 
Se × St 0.0130 0.3654 0.9007 2.4319 2.7059 0.0212 
Residuals 60 0.0356   0.8988   
Total 71       
Cochran's test  C = 0.148 (n.s.) C = 0.419 (p < 0.01) 
Transformation  None   Unsuitable   

 
Pairwise comparisons:       
Season (Se)  n.s.      
Stratum (St)  A = B < C      
Se × St interactions:        
Season     A, B, C   
     n.s.   
Stratum     Su Au, Wi, Sp  
     A = C < B n.s.  
Source d.f. Slope
 
Midpoint
 
  MS Pseudo-F p (perm) MS Pseudo-F p (perm) 
Season (Se) 0.0494 1.3866 0.2435 0.3158 0.3513 0.7984 
Stratum (St) 0.9346 26.2190 0.0001 3.8684 4.3041 0.0156 
Se × St 0.0130 0.3654 0.9007 2.4319 2.7059 0.0212 
Residuals 60 0.0356   0.8988   
Total 71       
Cochran's test  C = 0.148 (n.s.) C = 0.419 (p < 0.01) 
Transformation  None   Unsuitable   

 
Pairwise comparisons:       
Season (Se)  n.s.      
Stratum (St)  A = B < C      
Se × St interactions:        
Season     A, B, C   
     n.s.   
Stratum     Su Au, Wi, Sp  
     A = C < B n.s.  

Significant values of p are emboldened. A, 10–50 m; B, 51–100 m; and C, 101–200 m; Su, summer; Au, autumn; Wi, winter; Sp, spring; n.s., not significant.

Table 5.

Slope and midpoint height of normalized biomass–size spectra among seasons within the Gulf of Castellammare.

Slope
 
Midpoint height
 
Term Mean 1s.d. Term Mean 1s.d. 
−1.523 0.231 A*Su 5.618 0.865 
−1.457 0.154 A*Au 5.498 2.126 
−1.153 0.160 A*Wi 7.222 1.519 
   A*Sp 6.491 1.219 
   B*Su 7.236 0.373 
   B*Au 6.695 0.422 
   B*Wi 6.596 0.654 
   B*Sp 6.747 0.269 
   C*Su 5.992 0.379 
   C*Au 6.587 0.557 
   C*Wi 5.836 0.444 
   C*Sp 5.832 0.506 
Slope
 
Midpoint height
 
Term Mean 1s.d. Term Mean 1s.d. 
−1.523 0.231 A*Su 5.618 0.865 
−1.457 0.154 A*Au 5.498 2.126 
−1.153 0.160 A*Wi 7.222 1.519 
   A*Sp 6.491 1.219 
   B*Su 7.236 0.373 
   B*Au 6.695 0.422 
   B*Wi 6.596 0.654 
   B*Sp 6.747 0.269 
   C*Su 5.992 0.379 
   C*Au 6.587 0.557 
   C*Wi 5.836 0.444 
   C*Sp 5.832 0.506 

A, 10–50 m; B, 51–100 m; C, 101–200 m; Su, summer; Au, autumn; Wi, winter; Sp, spring.

For midpoint height, the season × stratum interaction was significant (Table 4), but was driven by a significantly greater midpoint height in stratum B than the other two strata in summer only (Table 4). In all other seasons, midpoint height was similar among strata within season and among seasons within stratum, suggesting only minor effects of season and depth on midpoint height (Table 4).

Discussion

Size spectra are believed to be robust ecosystem metrics for examining the impact of fishing, where fishing causes mortality on larger individuals or species (Rice, 2000; Rochet and Trenkel, 2003). The resultant disproportionate removal of large individuals and theoretical increase in small individuals through prey release (Shin et al., 2005) commonly results in steeper size spectra and lower midpoint height with increased fishing pressure (Rice and Gislason, 1996; Dulvy et al., 2004; Graham et al., 2005). It was hypothesized that establishment of the trawl exclusion area in GCAST would reverse these trends, resulting in shallower size-spectra slopes and higher midpoint heights.

Response of the size spectra

Observations of midpoint height were consistent with previous size-spectra studies, where increasing fishing pressure resulted in lower midpoint heights (Bianchi et al., 2000; Dulvy et al., 2004), and protection from fishing reverses the decline as recovery occurs (McClanahan and Graham, 2005). Given that the area under the size spectrum (a function of slope and midpoint height) is a measure of total community abundance, the results also agree with an eightfold increase in total cpue in GCAST (Pipitone et al., 2000). Increased fish abundance and biomass have also been observed for other Mediterranean MPAs (Guidetti and Sala, 2007), although increases varied among species and were based largely on rocky reef or seagrass habitats rather than open, soft sediments.

The response of the size-spectra slopes was not consistent with expectations from previous studies on fishing effects (Bianchi et al., 2000; Dulvy et al., 2004). The unprotected gulfs exhibited shallower slopes than the slopes calculated for GCAST, where trawling was excluded. This effect was unexpected. For size spectra to be a suitable indicator of the impacts of fishing, size-spectra responses should be predictable in relation to fishing (Macpherson and Gordoa, 1996; Bianchi et al., 2000).

Expectations that increasing fishing pressure results in steeper size spectra stem from fisheries that are more size selective, e.g. 120 mm mesh applied in North Sea trawl fisheries (Rice and Gislason, 1996). In contrast, the codend mesh size employed by most commercial fisheries in Italy is 40 mm (Machias et al., 2004), which exploits species over the entire size spectrum that was studied (>8 g). Under such conditions, three mechanisms link patterns of fishing to size-spectra response. Perhaps, (i) trawler exclusion did not uniformly reduce fishing mortality across all size classes. Legitimate artisanal gears were allowed in the protected area, which are more size selective (Pipitone et al., 2000). Therefore, although total fishing mortality was reduced by trawler exclusion, this effect was greater for small size classes. Alternatively, similar effects may be induced where fishing with small-mesh nets in unprotected areas results in either (ii) recruitment-overfishing (limited production of small fish) or (iii) heavy fishery mortality on small size classes. All three mechanisms link fishing directly to the size-spectra response and account for the observed patterns in response to protection from trawling. The similarity among unprotected locations suggests that differences were a direct effect of fishing.

Although biomass spectra slopes were steeper in GCAST than elsewhere, the size range of fish caught was commonly greater and, consequently, the mean across-community fish size was similar or slightly greater in protected than in unprotected locations. Positive MPA effects on fish size have been noted in other Mediterranean MPAs, e.g. Bell (1983), although these effects are by no means ubiquitous and have been rarely tested at community levels.

Alternative drivers of slope

Although the above mechanisms are believed to be the most probable causes of observed differences in slope with protection, other possibilities should be considered. These include the correlation of body size with other biological characteristics. Smaller species generally have faster life histories (Jennings et al., 1999) and shorter population doubling times, allowing smaller size classes to increase faster. Qualitatively, the species that increased the most (e.g. Lepidotrigla cavillone; Pipitone et al., 2000) did have a short population doubling time. Such an interpretation supports the assertions that MPAs may take many years to develop mature community structures (McClanahan and Graham, 2005), despite rapid initial change (Halpern and Warner, 2002).

Body size also correlates positively with mobility (Schmidt-Nielsen, 1984). Individuals that are more mobile have greater potential for movement outside the trawl exclusion area and may therefore experience greater fishing mortality. It is unlikely that this explains the pattern here, because the narrow continental shelf of GCAST presents a physiological and physical boundary, reducing movement potential and therefore the magnitude of this effect.

Finally, greater survival of large, fecund individuals in the initial years following MPA establishment can result in high recruitment and disproportionately large numbers of small individuals (Agardy, 1994). Within the GCAST trawl exclusion area, a high recruitment of some species after protection has been observed, e.g. Mullus barbatus, a species that accounts for 20–30% of total biomass. However, after >15 years of protection, the size spectrum should have settled to a new equilibrium (Bianchi et al., 2000). Based on size spectra alone, it is not possible to determine which of the above mechanisms drive patterns in slope, although processes operating through fishing mortality are most parsimonious.

Background variation

The effect of fishing on size spectra should be considered against other sources of variation. The slope of size spectra decreased with depth, and these changes were of a magnitude similar to that of protection. Few studies have specifically examined depth effects on size spectra, despite expectations that the relative proportions of small to large fish will vary with depth. Larger individuals within species are frequently observed in deeper water, reflecting a migratory ontogenetic movement to cooler water (Macpherson and Duarte, 1991), and many species utilize complex shallow inshore habitats as nursery areas (Franco et al., 2006; Planes et al., 2000). Some studies control for this depth effect (Bianchi et al., 2000; Stobberup et al., 2005), whereas others have been conducted over relatively narrow depth ranges (<50 m) and are unlikely to be confounded by depth. However, given that the main or interaction effects of depth and season on slope are of a magnitude similar to those of protection in our results, the importance of accounting for depth in size-spectra analyses is evident.

The patterns in size-spectra slope with respect to depth varied seasonally. Fish populations are rarely static, but move inshore or offshore for feeding, spawning, or as a result of seasonal environmental conditions. Such movements should be reflected in the size spectra. However, when strata were pooled, seasonal effects were still evident, suggesting that seasonal effects cannot be explained by within-gulf movement alone. Seasonal effects appear greater in the unprotected locations, being largely absent from analyses of GCAST across four seasons. Size-spectra analyses are increasingly being promoted as useful metrics to assess the impacts of fishing at the community level (Rice, 2000; Rochet and Trenkel, 2003; Jennings and Dulvy, 2005), and they demonstrate predictable changes in response to fishing for temperate demersal fisheries (Gislason and Rice, 1998). This study's observations that slope does not always decrease with declines in fishing mortality highlights the connection between fishing practices and the response of size spectra, but cautions against generalized assumptions that increasing slope represents greater fishing pressure. This study also suggests a need to control size spectra for depth and season. However, more research into seasonal distribution of species by size class would be needed to examine this issue fully.

Acknowledgements

The authors sincerely thank Simon Jennings for initial discussion in establishing the project, M. Coppola and G. Di Stefano for assistance with fish sampling, and N. A. J. Graham for comments on the manuscript. This study was co-funded by the Italian Ministry of University and Research.

References

Agardy
M. T.
Advances in marine conservation: the role of marine protected areas
Trends in Ecology and Evolution
 , 
1994
, vol. 
9
 (pg. 
267
-
270
)
Anderson
M. J.
A new method for non-parametric multivariate analysis of variance
Austral Ecology
 , 
2001
, vol. 
26
 (pg. 
32
-
46
)
Anderson
M. J.
Clarke
K. R.
Groley
R. N.
PERMANOVA+ for PRIMER: Guide to Software and Statistical Methods
2008
Plymouth, UK
Primer-E
Badalamenti
F.
D'Anna
G.
Pinnegar
J. K.
Polunin
N. V. C.
Size-related trophodynamic changes in three target fish species recovering from intensive trawling
Marine Biology
 , 
2002
, vol. 
141
 (pg. 
561
-
570
)
Bell
J. D.
Effects of depth and marine reserve fishing restrictions on the structure of a rocky reef fish assemblage in the northwestern Mediterranean Sea
Journal of Applied Ecology
 , 
1983
, vol. 
20
 (pg. 
357
-
369
)
Bianchi
G.
Gislason
H.
Graham
K.
Hill
L.
Jin
X.
Koranteng
K.
Manickchand-Heileman
S.
, et al.  . 
Impact of fishing on size composition and diversity of demersal fish communities
ICES Journal of Marine Science
 , 
2000
, vol. 
57
 (pg. 
558
-
571
)
Cavaliere
A.
Greco
S.
Perdichizzi
F.
Azzaro
F.
Arena
G.
Valutazione dell risorse demersali: primi risultati di due campagne di pesca (anno 1985) condotte nell'area compresa fra Capo Calava e Capo Gallo (Sicilia)
Atti Seminari Pesca e Aquacoltura. Pubbl. Min. Mar. Mercantile–C. N. R.
 , 
1988
, vol. 
III
 (pg. 
1337
-
1367
)
Charnov
E. L.
Life History Invariants
1993
Oxford, UK
Oxford University Press
Collie
J. S.
Hall
S. J.
Kaiser
M. J.
Poiner
I. R.
A quantitative analysis of fishing impacts on shelf-sea benthos
Journal of Animal Ecology
 , 
2000
, vol. 
69
 (pg. 
785
-
798
)
Daan
N.
Gislason
H.
Pope
J.
Rice
J.
Changes in the North Sea fish community: evidence of indirect effects of fishing?
2003
ICES Document CM 2003/N: 10
Dulvy
N. K.
Polunin
N. V. C.
Mill
A. C.
Graham
N. A. J.
Size structural change in lightly exploited coral reef fish communities: evidence for weak indirect effects
Canadian Journal of Fisheries and Aquatic Sciences
 , 
2004
, vol. 
61
 (pg. 
466
-
475
)
Franco
A.
Franzoi
P.
Malavasi
S.
Riccato
F.
Torricelli
P.
Mainardi
D.
Use of shallow water habitats by fish assemblages in a Mediterranean coastal lagoon
Estuarine and Coastal Shelf Science
 , 
2006
, vol. 
66
 (pg. 
67
-
83
)
Gislason
H.
Rice
J.
Modelling the response of size and diversity spectra of fish assemblages to changes in exploitation
ICES Journal of Marine Science
 , 
1998
, vol. 
55
 (pg. 
362
-
370
)
Graham
N. A. J.
Dulvy
N. K.
Jennings
S.
Polunin
N. V. C.
Size-spectra as indicators of the effects of fishing on coral reef fish assemblages
Coral Reefs
 , 
2005
, vol. 
24
 (pg. 
118
-
124
)
Greco
S.
Considerazioni sullo stato di sfruttamento delle risorse demersali (Capo Suvero-San Vito Lo Capo)
Biologia Marina Mediterranea
 , 
1994
, vol. 
1
 (pg. 
61
-
66
)
Greco
S.
Cavaliere
A.
Perdichizzi
F.
Cammaroto
S.
Sintesti dell'attivita svolta dall'unita operativa “T6”: Terreno Meridionale (Da Capo Suvero a Capo San Vito). Atti del seminario nazionale delle unita operative italiane svoltosi presso l'Instituto di Tecnologia della Pesca e del Pescato di Mazara del Vallo. “La valutazione delle risorse demersali dei mari italiani”
1993
(pg. 
139
-
141
NTR-ITPP, Special Publication 2
Greenstreet
S. P. R.
Hall
S. J.
Fishing and the ground-fish assemblage structure in the north-western North Sea: an analysis of long-term and spatial trends
Journal of Animal Ecology
 , 
1996
, vol. 
65
 (pg. 
577
-
598
)
Guidetti
P.
Sala
E.
Community-wide effects of marine reserves in the Mediterranean Sea
Marine Ecology Progress Series
 , 
2007
, vol. 
335
 (pg. 
43
-
56
)
Halpern
B. S.
Warner
R. R.
Marine reserves have rapid and lasting effects
Ecology Letters
 , 
2002
, vol. 
5
 (pg. 
361
-
366
)
Jennings
S.
The ecosystem approach to fishery management: a significant step towards sustainable use of the marine environment?
Marine Ecology Progress Series
 , 
2004
, vol. 
274
 (pg. 
279
-
282
)
Jennings
S.
Blanchard
J. L.
Fish abundance with no fishing: predictions based on macroecological theory
Journal of Animal Ecology
 , 
2004
, vol. 
73
 (pg. 
632
-
642
)
Jennings
S.
Dulvy
N. K.
Reference points and reference directions for size-based indicators of community structure
ICES Journal of Marine Science
 , 
2005
, vol. 
62
 (pg. 
397
-
404
)
Jennings
S.
Greenstreet
S. P. R.
Reynolds
J. D.
Structural change in an exploited fish community: a consequence of differential fishing effects on species with contrasting life histories
Journal of Animal Ecology
 , 
1999
, vol. 
68
 (pg. 
617
-
627
)
Jennings
S.
Kaiser
M. J.
The effects of fishing on marine ecosystems
Advances in Marine Biology
 , 
1998
, vol. 
34
 (pg. 
203
-
314
)
Kaiser
M. J.
Spencer
B. E.
Fish scavenging behaviour in recently trawled areas
Marine Ecology Progress Series
 , 
1994
, vol. 
112
 (pg. 
41
-
49
)
Machias
A.
Maiorano
P.
Vassilopoulou
V.
Papaconstantinou
C.
Tursi
A.
Tsimenides
N.
Sizes of discarded commercial species in the eastern central Mediterranean Sea
Fisheries Research
 , 
2004
, vol. 
66
 (pg. 
213
-
222
)
Macpherson
E.
Duarte
C. M.
Bathymetric trends in demersal fish size: is there a general relationship?
Marine Ecology Progress Series
 , 
1991
, vol. 
71
 (pg. 
103
-
112
)
Macpherson
E.
Gordoa
A.
Biomass spectra in benthic fish assemblages in the Benguela system
Marine Ecology Progress Series
 , 
1996
, vol. 
138
 (pg. 
27
-
32
)
McClanahan
T. R.
Graham
N. A. J.
Recovery trajectories of coral reef fish assemblages within Kenyan marine protected areas
Marine Ecology Progress Series
 , 
2005
, vol. 
294
 (pg. 
241
-
248
)
Murawski
S. A.
Brown
R.
Lai
H. L.
Rago
P. J.
Hendrickson
L.
Large-scale closed areas as a fishery-management tool in temperate marine systems: the Georges Bank experience
Bulletin of Marine Science
 , 
2000
, vol. 
66
 (pg. 
775
-
798
)
Pauly
D.
Christensen
V.
Dalsgaard
J.
Froese
R.
Torres
F.
Fishing down marine food webs
Science
 , 
1998
, vol. 
279
 (pg. 
860
-
863
)
Pipitone
C.
Badalamenti
F.
D'Anna
G.
Patti
B.
Fish biomass increase after a four-year trawl ban in the Gulf of Castellammare (NW Sicily, Mediterranean Sea)
Fisheries Research
 , 
2000
, vol. 
48
 (pg. 
23
-
30
)
Planes
S.
Galzin
R.
Rubies
A. G.
Goni
R.
Harmelin
J. G.
Le Direach
L.
Lenfant
P.
, et al.  . 
Effects of marine protected areas on recruitment processes with special reference to Mediterranean littoral ecosystems
Environmental Conservation
 , 
2000
, vol. 
27
 (pg. 
126
-
143
)
Rice
J.
Gislason
H.
Patterns of change in the size spectra of numbers and diversity of the North Sea fish assemblage, as reflected in surveys and models
ICES Journal of Marine Science
 , 
1996
, vol. 
53
 (pg. 
1214
-
1225
)
Rice
J. C.
Evaluating fishery impacts using metrics of community structure
ICES Journal of Marine Science
 , 
2000
, vol. 
57
 (pg. 
682
-
688
)
Rochet
M. J.
Trenkel
V. M.
Which community indicators can measure the impact of fishing? A review and proposals
Canadian Journal of Fisheries and Aquatic Sciences
 , 
2003
, vol. 
60
 (pg. 
86
-
99
)
Schmidt-Nielsen
K.
Scaling: Why Is Animal Size so Important?
1984
New York
Cambridge University Press
pg. 
241
 
Shin
Y-J.
Rochet
M-J.
Jennings
S.
Field
J. G.
Gislason
H.
Using size-based indicators to evaluate the ecosystem effects of fishing
ICES Journal of Marine Science
 , 
2005
, vol. 
62
 (pg. 
384
-
396
)
Stobberup
K. A.
Inejih
C. A. O.
Traore
S.
Monteiro
C.
Amorim
P.
Erzini
K.
Analysis of size spectra off northwest Africa: a useful indicator in tropical areas?
ICES Journal of Marine Science
 , 
2005
, vol. 
62
 (pg. 
424
-
429
)
Tudela
S.
Ecosystem effects of fishing in the Mediterranean: an analysis of the major threats of fishing gear and practices to biodiversity and marine habitats. Studies and Reviews
2004
Rome
General Fisheries Commission for the Mediterranean, FAO
pg. 
44