SEARCH

SEARCH BY CITATION

Keywords:

  • exploitation;
  • marine fish populations;
  • predator–prey relationships;
  • thresholds;
  • trophic cascades;
  • trophic structure

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion and conclusions
  7. Acknowledgements
  8. References
  9. Appendices

Correlations between time series of the abundance of predator and prey fish species in heavily exploited western North Atlantic marine fisheries vary temporally but are generally positive in southern, warmer waters and negative in northern, colder ones. The correlations provide an index of trophic structure and dynamics. We construct a framework to quantify critical thresholds between states in which the predator–prey correlations are positive or negative. We do so by developing a quantitative model of the distribution of the correlations between predator (15 species) and prey (8 species) functional groups based on the annual predator depletion rates and bottom temperatures (or alternatively species richness). The model accounts for 58% of the variance of the correlations with a root mean square error of 0.3. This index of trophic structure indicates that warmer, species-rich, southern fish populations resist transformation from positive to negative predator–prey correlations at exploitation rates that can be double those in the colder, relatively species-poor, northern areas. The model can be used to set limits for exploitation rates that preserve the functional relationships between predator–prey groups in emerging fisheries, and to assess the potential for and measures required to achieve recovery of degraded fish communities.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion and conclusions
  7. Acknowledgements
  8. References
  9. Appendices

The dramatic reduction/elimination of top predators from exploited marine ecosystems has led repeatedly to population explosions of mesopredators and planktivores. This loss has resulted in profound ecosystem transformations (Jackson et al., 2001; Levin et al., 2006; Oguz and Gilbert, 2007). Analogous changes have also occurred in terrestrial ecosystems (Schmitz et al., 2000; Berger et al., 2001; Bowyer et al., 2005; Souléet al., 2005; Terborgh et al., 2007). A recent example from the Northwest Atlantic involved cascading changes in four trophic levels and nitrate concentrations following the collapse of cod and other demersal fish species (Frank et al., 2005). Frank et al. (2005) found negative correlations between time series of adjacent trophic levels and indirect effects, i.e., positive correlations between time series separated by a trophic level (e.g., level 4 and level 2). This and other examples (Myers and Worm, 2003; Hjermann et al., 2004; Bascompte et al., 2005; Horwood et al., 2006) have led to a call for a more integrated, ecosystem approach to fisheries management (Quinn et al., 1993; Pikitch et al., 2004; Huggett, 2005). Central to this approach is the concept of ‘ecosystem over-fishing’ (Murawski, 2000; Sissenwine and Murawski, 2004; Steele and Collie, 2005), a common attribute of which is a fisheries-induced imbalance between predator and prey abundances, generally at the highest trophic levels. A fundamental challenge in moving to an ecosystem-based approach is the development of the ability to define quantitatively, and ultimately predict, how the state of these ecosystems is modified by directed fishing and, increasingly, by environmental change.

A central question for those committed to the ecosystem-based approach to fishery management is the extent to which apex predators can be depleted before the resulting instability leads to large scale ecosystem transformation. A clear example of such transformations is found in the Northwest Atlantic where the collapse of cod and other top predators was followed by up to 1000-fold increases in the biomass of forage fishes following their release from predation (Bundy, 2005; Bundy and Fanning, 2005; Savenkoff et al., 2007a,b). Apex predator collapse was followed by cascading effects at all lower trophic levels (Frank et al., 2006). Some idealized models predict specific relationships between predator and prey communities. For example, Lotka–Volterra theory predicts that predator–prey populations should behave cyclically and, by extension, that top predator biomass should eventually recover if exploitation is curtailed or eliminated (May et al., 1979). Smooth, asymptotic convergence to steady state populations occur even in the case of heavy exploitation (May et al., 1979). Cyclical variability and recovery of top predator biomass to prior levels has frequently occurred in terrestrial systems (Yom-Tov et al., 2007) but in the marine environment, recovery from collapse is less certain (Hutchings and Reynolds, 2004), and in many formerly cod-dominated fish communities in the Northwest Atlantic, the expected recovery of these top predators has not occurred in spite of reduction or cessation of exploitation for more than a decade (Shelton et al., 2006; Brander, 2007). In these fisheries, the now highly abundant (due to lack of predation) forage species appear capable of consuming and/or out-competing the eggs and larvae of their former predators, thereby preventing their recovery. In effect, the once-dominant apex predators move to a lower level in the trophic hierarchy and become prey, while their former prey assume the role of top predators. This role reversal limits the supply of new recruits to the populations of once-dominant predators and inhibits stock recovery (May et al., 1979; Swain and Sinclair, 2000; Shin and Cury, 2004). Role reversals occur less frequently in terrestrial ecosystems in which top predators typically occupy the same trophic level throughout life (Frank and Leggett, 1994).

Equally detrimental to the capacity of these disturbed fish populations to recover from such perturbations is the observed parallel decline in individual growth rate (a product both of selective fishing and environmental change) of several apex species (Choi et al., 2004; Brander, 2007). Size is strongly correlated with a broad suite of life history traits that influence population productivity and stability (Roff, 1992; Olsen et al., 2004; Swain et al., 2007) and the strength of ecological interactions such as predation (Post et al., 2008). Reductions in body size thus have a negative effect on reproductive success and/or predation efficiency, thereby increasing the susceptibility of top predators to collapse and the fish community to major trophic changes.

We investigate eastern North American continental shelf fisheries that have been subjected to annual biomass depletion rates for cod in excess of 50% and for which fishing is the largest contributor to mortality and has been identified as a prime cause of fisheries collapse in parts of the area (Hutchings, 1996; Frank et al., 2005). Species richness and ocean temperature have also been identified as important to the trophic balance in this region (Frank et al., 2006). Worm and Myers (2003) and Frank et al. (2005, 2006, 2007) have shown that correlations between predator and prey time series are a measure of the trophic balance within the fish community and possibly within the ecosystem. Our approach is an empirical investigation of the relationships between predator–prey time series, referenced to the total annual depletion of stocks and bottom temperature/species richness.

Our objectives are:

  • to develop a quantitative, first-order model relating patterns in the governing dynamics, represented by predator–prey functional group correlations, of fish populations in the Northwest Atlantic to local variation in exploitation, ocean temperature and species richness, variables that are readily measurable, and for which a rich database exists;
  • to use the model to characterize the susceptibility of fish communities to transitions in their governing dynamics in response to exploitation and environmental change;
  • to test the model using independent data from other regions.

Methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion and conclusions
  7. Acknowledgements
  8. References
  9. Appendices

Assessment of trophic structure

Trophic structure is a major characteristic of ecosystems and has been defined as the partitioning of biomass among trophic levels (Leibold et al., 1997). In this study, we focus on the top two trophic levels, namely, the top predatory fish and their prey. Following Frank et al. (2006), time series of abundance data for functional groups of predators and their prey (see below) were subjected to correlation analysis to characterize the pattern of trophic interactions, i.e., the balance or imbalance between predators and prey, as they relate to fish community structure. For fish communities in which the abundance and ecological role of top predators remains intact, positive correlations between predator and prey abundance are expected. Conversely, where top predator abundance has been greatly reduced and their ecological role destabilized (e.g., through a focussed fishing effort), negative correlations between predator and prey should be evident. Such analyses of time series data can provide a meaningful diagnostic of dynamic changes in trophic structure in continental shelf fish communities (Worm and Myers, 2003; Richardson and Schoeman, 2004; Frank et al., 2006). We note that it is possible to obtain a positive correlation in situations where both predator and prey functional groups are subjected to excessive exploitation. This is not the case in the Northwest Atlantic areas we examine.

We used the East Coast of North America Strategic Assessment Project (Mahon et al., 1998) database to develop time series of abundance of predator and prey functional groups. These data originate from annual scientific trawl surveys in nine areas in the Northwest Atlantic (Fig. 1) from 1970 to 1994, and form one of the most comprehensive, fishery-independent data series in the world. All species data were expressed as numbers per tow. The nine areas are large, 140 000 km2 on average with a range of 54 000 km2 (area 7)–392 000 km2 (area 1).

image

Figure 1.  Map showing locations of Northwest Atlantic fish communities (1–9) for which correlations between functional group predator–prey time series were calculated and a multiple regression model was developed with annual depletion rate and bottom temperature as independent variables. Also identified are nine of 10 fish communities examined in an independent test of the regression. Flemish Cap (FC), West Greenland (WG), Iceland (I), Faeroe Shelf (FS), Irish Sea (IrS), North Sea (NS), Iberian Seas (IbS), Barents Sea (BS), and the Baltic Sea (BlS). The NE Pacific area is not shown. The background ocean temperature is the annual average value (0–200 m) for the North Atlantic Ocean with selected isotherms (°C) shown, and is from the World Ocean Atlas 2005 (http://www.nodc.noaa.gov/OC5/WOA05/pr_woa05.html; accessed 27 February 2009). The southward (northward) extent of colder (warmer) water on the western (eastern) side of the ocean reflects the influence of the Labrador (North Atlantic) Current.

Download figure to PowerPoint

In this region the dominant predator species, and hence those comprising the predator group, are Atlantic cod Gadus morhua, haddock Melanogrammus aeglefinus, pollock Pollachius virens, longfin hake Urophycis chesteri, silver hake Merluccius bilinearis, white hake Urophycis tenuis, red hake Urophycis chuss, redfish Sebastes spp., thorny skate Amblyraja radiata, spiny dogfish Squalus acanthias, Greenland halibut Reinhardtius hippoglossoides, American plaice Hippoglossoides platessoides, winter flounder Pseudopleuronectes americanus, witch flounder Glyptocephalus cynoglossus, and yellowtail flounder Limanda ferruginea. All 15 of these species have been commercially exploited, often in mixed fisheries that are prosecuted throughout the continental shelf in depths of <200 m.

The prey species group consists of eight species: arctic cod Boreogadus saida, Atlantic argentine Argentina silus, Atlantic mackerel Scomber scombrus, butterfish Peprilus triacanthus, alewife Alosa pseudoharengus, herring Clupea harengus, capelin Mallotus villosus, and northern sand lance Ammodytes dubius. Commercial exploitation of these species (herring being the exception) has been relatively low or non-existent relative to that experienced by most of the groundfish species listed above.

Annual depletion of top predators

Annual fishing mortality is the most informative measure of fishing pressure; however, the only mortality estimate available from stock assessments across the nine areas is for Atlantic cod. We therefore used estimates for cod to represent annual mortality of the predator functional group. Sinclair (1996; is Table 4) and NEFSC (2005) decomposed the estimates of total mortality, derived from Virtual Population Analyses, into fisheries and natural components by assuming a constant, instantaneous natural mortality rate (M) of 0.2 yr−1. We recombined their results to derive an annual depletion of cod, expressed as the annual proportion of biomass removal (see Appendix A).

There are compelling reasons why these mortality estimates are broadly representative for the functional group: cod’s dominant position in the food chain throughout the nine areas, its high spatio-temporal overlap with most of the other species, particularly haddock and pollock (Gabriel, 1992; Halliday et al., 1992), the non-selective manner in which cod and all the other top predator species are captured in the fishery, and the similarity of the cod life cycle to that of haddock and pollock, the two other dominant groundfish, and flatfish (Appendix B). The life cycles of other species, notably the hakes and especially the elasmobranch spiny dogfish, differ and, as a consequence, comparable high annual mortality rates for cod and spiny dogfish would be expected to have considerably different consequences. Thus, our overall measure of total mortality is an imperfect one. Refinement would require aging information on all species within the predator group; this information is not available. In our modelling of trophic structure, we used annual cod mortality as a primary variable, given the similar life cycles of the dominant predators (cod, haddock and pollock). We acknowledge that further refinements may be required within the predator functional group.

Treatment of data

The 25-yr time series of annual depletion rates and derived predator–prey correlation coefficients were analysed for each region in sliding 15-yr data blocks. This was done partly to account for temporal variability of the trophic structure and the annual depletion rates within fish communities (Frank et al., 2006). It has the added benefit of expanding the range of the correlations and the annual depletion rates. This ‘sliding window’ approach and the significance levels of the correlations are described in Frank et al. (2006), see their (Fig. 6). They accounted for autocorrelation (Chelton, 1983) in the time series and found that 10 of 93 correlations were significant at the 0.05 level.

image

Figure 6.  Comparison between functional group predator–prey and single species correlations for the nine modelled fish communities (Fig. 4). All single species comparisons were made between smoothed abundances of cod (predator) and either sandlance (areas 1, 6, 7, 8, 9) or capelin (areas 2, 3, 4, 5) for the period 1970–1994.

Download figure to PowerPoint

Temperature was treated differently, with long-term averages used instead of annual values because there are generally insufficient data to determine the annual average for each area every year. The long-term mean bottom temperatures for the nine areas were obtained from the Canadian Atlantic Atlases (http://www.mar.dfo-mpo.gc.ca/science/ocean/tsdata.html; accessed 26 February 2009). Moreover, because the time series were examined in 15-yr segments, we analyzed 1970–1994 bottom temperature records for four areas which were sampled frequently: area 1 (long-term hydrographic site, 175 m depth at 47.55°N, 52.57°W off St. John’s, NL, near the southern limit of the area and sampled on average 11.3 months yr−1), area 6 (July scientific trawl survey hydrographic stations covering entire area every yr), area 8 (July scientific trawl survey) and area 9 (50 m depth over Georges Bank, sampled on average 6.5 months yr−1). At these four sites, the standard deviations of bottom temperature were 0.05, 0.11, 0.10 and 0.08°C, respectively. We conclude that the within-area variance of bottom temperature is very small relative to the ∼8°C range recorded across the nine areas.

Temperature versus species richness

Frank et al. (2006) suggested that species-rich fish communities are more stable and are able to resist or recover from disturbances more readily than are low diversity ones. They also argued that temperature-dependent physiology and population dynamics could be important factors contributing to community stability and, consistent with this expectation, found that temperature offered comparable or greater power in accounting for trophic control than did taxonomic diversity. In the nine fish communities forming the basis of their analysis, water temperature was highly correlated with species richness (= 0.90; Frank et al., 2006; Table 1, Fig. 7a). They concluded that water temperature and species richness are interchangeable variables.

Table 1.   Geographic areas used for independent test of the threshold detection model.
Location Predator/prey identityPredator/prey correlationAverage annual depletion/predatorTPeriod
  1. Time series for fish species biomass and mortality estimates were obtained from the ICES website at http://www.ices.dk.committe/acom/comwork/report/asp/advice.asp and from the NAFO website at http://www.nafo.int/publications/frames/science.html. Temperature (T) estimates were average annual bottom or 200-m temperatures (whichever was shallower) from optimally estimated values from the World Ocean Atlas 2005 of the National Oceanographic Data Center (http://www.nodc.noaa.gov/). Predators and prey are designated as follows: 1. Cod. 2. Haddock. 3. Pollock, saithe. 4. Atlantic wolffish. 5. Spotted wolffish. 6. American Plaice. 7. Plaice. 8. Sole 9. Greenland halibut. 10. Starry skate 11. Redfish. 12. Whiting. 13. Hake. 14. Megrim. 15. Herring. 16. Shrimp. 17. Norway pout. 18. Sandlance, sandeel. 19. Capelin 20. Sardine. 21. Sprat. 22. Phytoplankton. 23. Grenadier. 24. Echinoderms.

  2. Iceland: ICES database.

  3. Faeroe Shelf: Steingrund and Gaard (2005), updated from ICES database.

  4. North, Barents, Iberian, Irish and Baltic Seas: ICES website.

  5. NE Pacific: Bailey et al. (2006).

  6. Assumed instantaneous natural mortality of 0.2 yr−1.

  7. Temperature at 4000 m.

Flemish Cap1, 4, 5, 6, 9, 10, 11/16+0.290.42/1, 113.811988–2006
W Greenland1, 4, 5, 6, 9, 10, 11/16+0.800.54/14.351992–2006
Iceland1, 2, 3/15, 19−0.550.51/1, 2, 34.571981–2007
Faeroe Shelf1, 2/22+0.690.44/1, 26.671990–2006
North Sea1, 2, 7, 8, 12/15 1, 2, 3, 7, 8, 12/15, 17, 18−0.18 +0.380.58/1, 2, 7, 8, 12 0.57/1, 2, 3, 7, 8, 127.931963–2007 1983–2007
Barents Sea1, 2, 3, 9/15, 19−0.190.51/1, 2, 3, 92.301973–2006
Iberian Seas8, 13, 14/20+0.370.49/8, 13, 1412.71986–2006
Irish Sea1, 7, 8/15+0.630.57/1, 7, 89.91970–2007
Baltic Sea1/15, 21−0.270.67/14.11977–2007
NE Pacific23/24+0.730.181.51999–2004
image

Figure 7.  (a) Time series of observed predator–prey correlations (vertical bars) and average annual depletions (line, solid dots) for the North Sea based on 15-yr data blocks. (b) Time series of observed predator–prey correlations (black dots) and bottom temperature (grey dots) based on 15-yr data blocks. (c) Scatter plot of data in panel (B). (d) Scatter plot of observed and predicted correlation based on 15-yr data blocks.

Download figure to PowerPoint

Based on the Jack1 estimate of species richness from Frank et al. (2006) and given the temperature range (∼1–9°C) observed for the nine areas, species richness is related to bottom temperature by the equation 50 + 10*Annual bottom temperature. Similar correlations have been reported for other marine (e.g., Rohde, 1992; Roy et al., 1998; Macpherson, 2002) and terrestrial (e.g., Turner et al., 1987, 1988) ecosystems. Given this strong correlation (r2 = 0.81, = 0.001) and the greater availability of water temperature data for the nine regions we studied, and for fish communities over a much broader area, we used temperature as the primary ecological variable (and, in effect, as a surrogate for species richness). Because temperature and species richness are important in structuring ecosystems, in locations where a high correlation between these variables is not found, both should be treated as independent variables.

Model development

A multiple regression model was developed and consists of the correlation between predator–prey functional groups as the dependent variable and annual depletion rates and bottom temperatures as the independent variables:

  • image(1)

where we define the correlation ‘r’ as the trophic structure index. Both independent variables have been shown to be relevant to fish community restructuring (Frank et al., 2006, 2007). Several modelling approaches were assessed including fitting the correlations with linear and quadratic functions of bottom temperature and annual depletion rates, and generalized linear modeling. In practice, all fits accounted for equivalent variances and exhibited virtually identical residual root mean square differences between the observations and the fit. Consequently, we adopted the simpler linear fit.

These predator–prey correlations are generally subject to large uncertainty because the data series are strongly autocorrelated (Frank et al., 2006) and are the greatest source of model error. To evaluate the multivariate linear regression, we ran 1000 simulations based on the correlations and their 0.05 confidence levels (determined from the effective number of degrees of freedom of the time series in the Supplementary On-line Material; Frank et al., 2006).

We note that the predator–prey correlations could vary at smaller spatial scales within the nine large areas we investigated. However, the random stratified sampling of the fisheries surveys, the source of our data, should provide adequate spatial representation and reliable estimates of predator and prey abundances for the areas we consider. Nonetheless, because the sizes of our nine areas vary by a factor of 7, we checked the relationship between area size and predator–prey correlations and species richness. The area/predator–prey correlation was weak (r2 = 0.17), with the magnitude of r determined primarily by area 1. The area/species richness relationship was weaker (r2 = 0.09); however, positive relationships are expected at smaller geographic scales, e.g., banks (Frank and Shackell, 2001).

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion and conclusions
  7. Acknowledgements
  8. References
  9. Appendices

Fish community structure, as indexed by the functional group predator–prey correlations (Fig. 6, Frank et al., 2006), differed among areas and generally changed with time (Fig. 2). Top predators in all areas except 4 and 5 experienced high annual depletion levels (∼50%), which increased with time, on average by 16 ± 11% (SD), range 4–34%, for areas 1–3 and 6–9 from least-squares fits to annual depletion rates (Fig. 2). In area 4, the annual depletion rate decreased slightly with time, whereas in area 5 the rate was at a minimum for the 1976–1990 block, and then rose slightly. The trophic structure index was initially strongly positive in the southernmost areas 8 and 9, indicating the presence of largely unaltered predator–prey dynamics; these areas are characterized by higher average bottom water temperatures (Fig. 1), or equivalently high species richness, and high annual depletion rates. The magnitude of the positive correlation in these areas decreased with time. Area 7 experienced the highest annual depletion rate (∼70% annual biomass removal), yet unexpectedly evolved towards a positive trophic structure index. Frank et al. (2006) hypothesized that this variability may have resulted from an intensive, unregulated fishery, presumably with higher annual mortality (no quantitative estimates are available) than shown for the later period (Fig. 2). This gave rise to negative correlations followed by a period of recovery due to stricter management measures associated with extended fishery jurisdiction to 200 nautical miles in 1977. Geographically intermediate areas 5 and 6 were characterized by transitions in trophic structure, for area 6 potentially in response to increased annual depletion rates, and lower average bottom temperatures (or equivalently lower species richness) relative to southern areas. The northernmost areas, characterized by cold bottom water temperatures (low species richness) and high annual exploitation rates, had consistently negative predator–prey correlations (areas 1 and 2) and a sharp transition in trophic structure (area 3) indicative of perturbed predator–prey dynamics and altered trophic states. Overall, the magnitude of the negative correlations tended to increase with time in these northernmost areas.

image

Figure 2.  Temporal variability (based on sliding, 15-yr segments) of the sign and magnitude of correlation coefficients (vertical bars) between functional groups of predators (piscivores) and their prey (planktivores) in nine Northwest Atlantic continental shelf fish communities for the period 1970–1994. The initial averaging periods were <15 yr for areas 7 and 9 due to the shorter data series available. Also shown are estimates of the annual depletion rates of cod (proportion of biomass removal), averaged over the same time interval as the predator–prey correlation coefficients.

Download figure to PowerPoint

The linear multiple regression, with the annual depletion rate and bottom temperature as independent variables, was:

  • image(2)

where the standard deviations in parentheses are based on the 1000 randomized simulations. This model accounted for 58% of the variance in the observed predator–prey correlations with a root mean square (rms) difference of 0.3 (Fig. 3). Given the temperature and annual depletion ranges of 7.6°C and 0.36 yr−1, both variables contribute approximately equally to the variation of the predator–prey correlation. There is a tendency for correlations within areas to be aggregated (particularly areas 3 and 8); relationships based on these areas alone would yield regressions quite different from those derived using results from all nine areas. However, our goal was to derive a more general relationship, and it was therefore necessary to explore a broad range of temperature (species richness) and annual depletion.

image

Figure 3.  Observed versus predicted values (dots) of the correlation based on the multiple regression (Eq. 2) between predator and prey functional groups from the nine geographic areas in the Northwest Atlantic (Fig. 1). The least squares linear fit is also shown (line).

Download figure to PowerPoint

For the idealized case in which variance is completely accounted for, the transition from one trophic state to another would occur at that combination of temperature and annual depletion rate at which the predator–prey correlation changes sign. However, because our model does not account for all of the variance, a precise determination of the transition line is impossible. It is likely within the area defined by the rms difference (0.3), between observed and fitted values. Accordingly, we chose to locate the probable upper and lower boundaries of the zone of transition at correlations of 0.3 and −0.3 (Fig. 4). So structured, the model indicates that warmer (6–8°C) species-rich areas are capable of sustaining annual depletion rates as high as 55% while retaining a positive (unperturbed predator/prey) dynamic state. In contrast, colder, less species-rich fish communities appear to be able to sustain this status only at much lower depletion rates (<35%). Of particular note is area 6, as it clearly shows the transition from positive correlations (green dots, r > 0.3), to the transition zone (−0.3 < r < 0.3), and finally to negative correlations (r < −0.3; see also Fig. 2).

image

Figure 4.  Correlation contours (colours) as a function of bottom temperature and annual depletion rates experienced by top predators. The root mean square difference [square root of the average of (model – observations)2] of the correlations was 0.3. This difference specifies the transition zone (grey) separating positive index values for trophic structure (> 0.3, green) from negative index values (< −0.3, red). The same colour scheme (colour bar) is used for the observed correlations, which are shown as dots. The horizontal banding of the dots illustrates the data decomposition into 15-yr time intervals and the progression from positive to transitional to a negative trophic structure index over time (particularly evident in area 6). The offset species richness is based on the strong correlation and linear regression with bottom temperature discussed in text.

Download figure to PowerPoint

To test the general applicability of the model, we examined predator–prey correlations developed for 10 additional, independent fish communities, nine from the North Atlantic and one from the Northeast Pacific. The mix of fish species in these communities differed from that examined in our nine base areas. Moreover, for these new areas, annual depletion rates were available for more than one species. In these cases, the depletion rates were averaged, weighted by species biomass (see Daan et al., 2005). The areas investigated, predator and prey species, correlations between time series, average annual depletions and temperatures are listed in Table 1 and plotted in Fig. 5a and b. Qualitatively, the results for these regions were consistent with model expectations: the signs of the correlations were the same between the observed and model predictions for eight of the 11 comparisons (note, there are two correlation coefficients for the North Sea) (Fig. 5a,b). The poorest agreements were for West Greenland (predicted −0.01, observed 0.8), the North Sea (NSa) (predicted 0.38, observed −0.18) and the Iberian Seas (predicted 1, observed 0.37).

image

Figure 5.  (a) Observed versus predicted predator–prey correlations from the multiple regression model for 10 independent marine fish communities: Faeroe Shelf (FS), North Sea (NSa, NSb; see Table 1), Barents Sea (BS), Iceland (I), West Greenland (WG), Flemish Cap (FC), Irish Sea (IrS), Baltic Sea (BlS), Iberian Seas (IbS) and Northeast Pacific (NEP). (b) Superimposition predator–prey correlations for the 10 additional areas on the response surface shown in Fig. 4. The enclosed part within the overall figure represents the area of validity of the model fit, i.e., the range of annual depletion and bottom temperature encompassed by the nine base areas (see Fig. 4). The sizes of the data points are proportional to the magnitude of the correlation coefficient. Colours denote a positive (green) or negative (red) trophic structure index.

Download figure to PowerPoint

Discussion and conclusions

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion and conclusions
  7. Acknowledgements
  8. References
  9. Appendices

The modelled responses we report provide initial guidance in setting annual exploitation rates for top predators that are necessary to prevent overfishing and/or promote restoration of balanced predator–prey relationships. The model also provides a basis for assessing the potential impact on fish community structure of temperature changes ranging from low frequency climate variability to longer term trends, such as those related to global warming. Importantly, this is true even for those ecosystems for which there is limited knowledge of predator and prey community dynamics.

It is important to note, however, that the model is built on correlations among variables and therefore does not prove causality. Moreover, although the time series that form the foundation of these relationships are 25 yr long, there are few degrees of freedom because of temporal autocorrelation. The model would benefit from more independent tests from regions in addition to those shown in Fig. 5 and from process-oriented studies of its underlying hypotheses. The independent verification exercise we undertook revealed the potential importance of heightened fish community responses to ocean climate variability in other regions and of considering fisheries in which high effort is directed towards both predator and prey species. We hope that our findings will stimulate such work.

Further refinement of the model requires addressing several problem areas. The first is the use of the annual depletion rate for cod to represent the rate for the predator group in our nine base areas. While this is a reasonable measure of the effects of fishing for areas 1–4 (Appendix A) where cod dominate, accounting on average for 68% of the groundfish landings, it is less obviously so for areas 5–9, where cod constitute only 31% of total landings. This uncertainty is compensated for somewhat by the fact that, in these more southerly regions, the combined landings of cod, haddock and pollock, the latter two species with similar habitat preference and vulnerability to exploitation, account on average for 59% of the total landings, varying from 51% (area 8) to 77% (area 6).

Clearly the use of the annual depletion rate for cod is flawed for a significant component of the species complex in the southern areas. In particular, elasmobranch species experience significantly different annual depletion rates but, at the same time, could play important roles in predator–prey interactions. For example, compensatory increases in an elasmobranch species on Browns Bank (area 8; Shackell and Frank, 2007) and on Georges Bank (area 9; Fogarty and Murawski, 1998) were observed following the reduction in gadoid abundances. For the Gulf of Maine, Mid-Atlantic Bight area, Rago et al. (1998) report that for 1993, the landings and discards for spiny dogfish were about 3% of their total biomass and estimate a natural mortality of 0.09 yr−1, well below the 0.2 yr−1 that we applied. It is evident that improved estimates of the predator group depletion rate are a priority for further studies of the resilience of fish communities in the face of exploitation and environmental change. Within the nine regions we investigated, this is currently possible only for a limited number of species that are routinely aged. For the remaining species, only relative measures of annual depletion can be attempted.

Annual depletion rate for predators in addition to cod were available for six of the 10 independent areas used to test the model. In an attempt to understand further the consequences of using the annual depletion rate of cod to represent the predator group, we compared estimates of annual depletion for cod with those determined for other species from these independent areas. The comparisons (Table 2) show that the cod rates are comparable and on average 12% higher than those of eight other species from the six areas. Moreover, the correlation between cod and other species rates is quite strong (r2 = 0.47). Although this comparison is not definitive, it does provide support for the use of cod depletion rates as an overall measure of fishing pressure within our nine base areas.

Table 2.   Comparison of annual depletion of cod with other species.
CodOther speciesDifferenceSpecies Areas
  1. BS, Barents Sea; I, Iceland; FS, Faeroe Shelf; NS, North Sea; FC, Flemish Cap; IrS, Irish Sea.

0.600.390.21Greenland halibutBS
0.610.530.08HaddockBS, I, FS, NS
0.620.510.11SaitheBS, I, NS
0.460.400.05RedfishFC
0.660.610.05WhitingNS
0.690.540.15Norway poutNS
0.670.520.15PlaiceNS, IrS
0.670.490.18SoleNS, IrS
0.620.500.12Average 

Additionally, given the above discussion, we believe the current model to be most applicable to a gadoid-dominated system or to regions where a single species dominates the predator group to an extent similar to that of cod in the northernmost areas we investigated. Consider, then, the northernmost of our nine regions, where cod constitute 69% of the groundfish landings. Assuming a correlation of 0.3 to be a conservative upper bound for the transition from an unaltered to altered state, the transition point for area 1, the region with the coldest temperature (1.2°C), would be expected to occur at an annual depletion rate of 29%. This is 11% greater than the assumed pre-collapse northern cod annual natural mortality of 18% (mean = 0.2 yr−1) and was well within the recommended management strategy that called for exploitation not to exceed natural mortality (Walters and Maguire, 1996). However, the realized annual depletion rate of northern cod prior to collapse was 54%, twice the maximum annual depletion rate that our model indicates would be sustainable while maintaining a positive predator/prey balance and a positive trophic structure index (Fig. 4). Predator–prey correlations in area 1 during the pre-collapse period were consistently negative (= −0.56, SD = 0.14). Strict adherence to the management advice (Rivard and Maguire, 1993; Sinclair, 1996) in place at the time might well have prevented the collapse of top predators and the resulting trophic imbalance that occurred in this cold water fish community.

A second point requiring more critical evaluation is the use of single species rather than functional groups in some current, broad-scale studies of predator–prey correlations (e.g., Worm and Myers, 2003). To determine the potential bias on the outcome introduced by this reliance on single species, we recalculated the correlations using the dominant single predator and prey species time series for the nine Northwest Atlantic areas we studied. Single species and functional group-based correlations were very similar in the colder, relatively species-poor northern regions (areas 1–6) but differed substantially in the warmer, species-rich southern areas (area 7: functional group = −0.02, single species = 0.28; area 8: = 0.70 versus −0.28, area 9: = 0.44 versus −0.33) (Fig. 6). The differences generally increased as temperature and species richness increased toward the south, highlighting the stabilizing role of alternate predators in species-rich fish communities. This analysis highlights the importance of basing the correlations used to assess fish community status on functional groups rather than individual species, and the importance of species richness (inherent in analyses based on species groups) to stability and resilience of fish populations.

Finally, it is important to note that the model we developed does not account for the exploitation of prey species. Figure 4 implies that a simple reduction of fisheries exploitation of predators, and hence of the overall depletion rate, would allow recovery. However, it is easier to drive a fish community into an imbalance than return it to its pre-existing condition. Despite reductions or even elimination of harvesting of top predators in the areas we studied, many have not recovered (Frank et al., 2007). In a study of southern Gulf of St. Lawrence (area 3) fisheries, Swain and Sinclair (2000) demonstrated that the annual prerecruit survival rate of juvenile cod (based on the ratio of recruitment to spawning stock biomass, log transformed) was strongly, inversely related (= −0.76; Table 3) to the combined biomass of mackerel and herring – the former prey of cod in the southern Gulf of St. Lawrence. They suggested that this arose because of the consumption of cod eggs and larvae by adult herring and mackerel. We conducted a similar analysis for areas 1, 2 and 4–9 using cod spawning stock biomass and recruitment data available in Shelton et al. (2006) and NEFSC (2005) and functional group prey abundance (Table 3). Whereas the uncertainty of the relationships was high because of autocorrelation of the time series, the signs of the correlations between cod juvenile survival rates and functional group prey abundance were consistently negative for the five areas where the cod stocks had collapsed and failed to recover. Conversely, the correlations were positive for two of the three areas in which cod collapses did not occur. These results are consistent with the predator–prey role reversal hypothesis.

Table 3.   Analysis of the effect of functional group prey abundance and juvenile cod survival rates in nine areas throughout the northwest Atlantic.
AreaPredation effect P valueNo. of yrStatus of cod stock
  1. Negative correlations suggest that the former prey have become predators, acting to prevent stock recovery through predation impacts on prerecruit survival. Shaded rows highlight those areas where collapsed stocks have failed to recover.

  2. Based on the effective number of degrees of freedom (Chelton, 1983).

  3. There was a general lack of relationship between juvenile cod survival rates and spawning stock biomass, except in area 6. In this case, the spawning stock biomass effect was removed by examining the residuals of the regression between survival rate and SSB.

1−0.400.2125Collapsed/no recovery
2−0.230.4119Collapsed/no recovery
3−0.76Swain and Sinclair (2000)24Collapsed/no recovery
4−0.210.4518Collapsed/recovered
5−0.170.4325Collapsed/no recovery
6−0.230.2425Collapse/no recovery
7−0.210.5613No collapse
8+0.100.7612No collapse
9+0.170.5117No collapse

While reduced fishing pressure on predators did not bring about a rapid recovery, an acceptance of the importance of role reversal implies that the culling of prey species could promote recovery of the predator complex. In fact, Lessard et al. (2005) suggested that there will be situations when ecosystem management may require the use of direct, active controls to hasten recovery and/or reduce extinction risks. In a recent example, Persson et al. (2007) report that removal of large Arctic charr (Salvalinus alpinus), the prey species, allowed brown trout (Salmo trutta), the predator depleted by overfishing, to recover in a large Norwegian lake. In this case, the governing factor was not role reversal but size-selective feeding by the trout. However, the work of Lessard et al. (2005) and Yodzis (2001) highlight the complexity of ecosystems and their response to external forcing such as a cull.

A further confounding factor for the correlation approach we employed could arise from fisheries directed towards prey species. For the time frame we examined in our nine base areas, the only significant prey species fished is herring in areas 3 and 8. In contrast, significant prey fisheries operate in the North Sea (instantaneous fishing mortalities, F (yr−1), of 0.6 for herring and sandeel), Iceland (F = 0.32, herring), the Irish Sea (F = 0.52, herring), the Baltic Sea (F = 0.25, herring; 0.31, sprat), and the Iberian Seas (F = 0.27, sardines). Although simultaneous fishing of predator and prey species could maintain balance in the fish community, it could also lead to positive correlations even though both groups were being dangerously depleted.

Rapid, large changes in fishing effort could also give rise to misleading correlations. Idealized models of multi-species fisheries predict positive correlations as prey fisheries are initiated, changing to negative correlations as predator and prey populations adjust to new equilibria (e.g., Fig 4 in May et al., 1979). The magnitude and time scales of the adjustment depend on the growth rates of the populations under consideration. However, in the nine base areas we studied initially, 86% of the year-to-year changes in instantaneous fishing mortality are within 0.1 (see Fig. 2). We conclude that such changes were not a factor in our analyses.

As noted previously, three of the correlations derived from the independent areas used for verification [West Greenland, the Iberian Seas and the North Sea (1963–2007 series)] differed substantially from the predictions of the model. The West Greenland series was the shortest (15 yr) and the correlation was dominated by a recent surge in Greenland halibut in 2004 and cod in 2005 (Sünksen and Jørgensen, 2007). The predator–prey correlation based only on the cod and shrimp time series (32 yr, 1976–2007) was −0.18 (model predicted value, −0.01), in much better agreement with the model. The Iberian Seas bottom temperature is substantially outside the range of bottom temperatures used to develop the model and therefore represents a significant extrapolation. Consider the longest time series of Table 1: the North Sea, 1963–2007. The predator–prey correlations in 15-yr moving windows show low frequency variability similar to areas 3 and 6 (Figs 2 and 7a). The shift from negative to positive correlations occurs around 1990; the shift in annual depletion of predators occurred in 2002 (within the 1988–2002 window, the 1988–2001 average was 0.6 and the 2002 value was 0.43), which shows up as a gradual decline beginning with the 1988–2002 block. Several studies of the North Sea ecosystem have indicated a regime shift occurring sometime between 1983 and 1997, most likely between 1991 and 1998, depending on the variables examined (Kenny et al., 2006). The predator–prey correlation is quite coherent with North Sea bottom temperature (Fig. 7b,c; temperature courtesy of A. Kenny, CEFAS, Lowestoft, UK), suggesting a strong link to the physical environment. A comparison of the 15-yr moving window, observed predator–prey correlations and those predicted by the model (Fig. 7d) shows that they are strongly related (r2 = 0.55); however, the range of the predicted correlations is considerably smaller than the range seen in the data (Fig. 7d). Possible reasons for this disparity include a different response to climate variability for the species mix occurring in the eastern Atlantic relative to that found in our nine NW Atlantic areas. Fisheries exploitation of prey species, a minor factor and therefore neglected in the model’s nine base areas, was, however, clearly an important factor for the North Sea and other eastern Atlantic regions.

The ability to predict and avoid fish community imbalances, resulting from the differential exploitation of top predators relative to their prey, would constitute a major advance in the science of the management of marine resources. The model we report provides a first approximation to this end and illustrates the potential of an ecosystem approach to resource management that is based on easily derived input variables. Moreover, the demonstration that temperature, whether operationally independent, as a surrogate for, or in combination with quantitative measures of species richness, can be used as an effective input to the model, makes it readily applicable even to emerging fisheries. We suspect that the temperature dependence of fish community status revealed by the model relates both to its effect on metabolic/demographic rates and on species richness (Allen et al., 2002). In this context, the increased resiliency of the more southern Northwest Atlantic fish communities probably results from both higher growth rates (a consequence of the temperature dependence of growth; Brander, 2007) and the greater species richness at the higher trophic levels. This increased species richness allows alternate top predators to assume the niche vacated by heavily or over-exploited top predator species, thereby maintaining positive predator/prey balances. The compensatory increases in elasmobranch species on Browns (area 8) and Georges (area 9) Banks following the reduction in gadoid abundance there, are an important examples of this effect (Fogarty and Murawski, 1998; Shackell and Frank, 2007).

Finally, primarily because of the potential for predator–prey role reversals, marine ecosystems appear to be less amenable to rehabilitation than terrestrial ones (Griffith et al., 1989; Ripple and Beschta, 2003; Kauffman et al., 2007; Seddon et al., 2007). For this reason, an even more accurate determination of the critical limits of fish community transformation than we have been able to provide is vital to the progress of the evolving science of marine ecosystem management. The maintenance of fish community balances over periods of prolonged exploitation or environmental change, demands that the balance between functional groups remain within the bounds that permit renewal (successful reproduction) and regulation (keeping prey in check). Our model provides an initial framework for the identification of these boundaries, and insight into the ecological basis of, and geographical variability in, the presence/absence of profound fish community restructuring and recovery in response to exploitation and environmental variability.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion and conclusions
  7. Acknowledgements
  8. References
  9. Appendices

We thank Dr M. Sinclair, M. Fogarty, C. Hannah, J. Fisher and I. Perry for critical feedback on earlier versions of the manuscript. Mr. R. Pettipas and L. Petrie provided significant technical support. D. González-Troncoso, C. Fernández, H. Hovgaard, O. Jørgensen, A. Kenny, M. Kingsley and K. Sünksen provided useful advice on the location and quality of fisheries time series, in one case (C.F.) by providing a time series in advance of publication. We thank the three external referees for their insightful, useful advice that improved the paper considerably. This research was supported by Fisheries and Oceans Canada and the Natural Sciences and Engineering Research Council of Canada Discovery Grant program awards to K.T.F. and W.C.L.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion and conclusions
  7. Acknowledgements
  8. References
  9. Appendices
  • Allen, A.P., Brown, J.H. and Gillooly, J.F. (2002) Global biodiversity, biochemical kinetics, and the energetic-equivalence rule. Science 297:15451548.
  • Bailey, D.M., Ruhl, H.A. and Smith, K.L. Jr (2006) Long-term change in benthopelagic fish abundance in the abyssal northeast Pacific Ocean. Ecology 83:549555.
  • Bascompte, J., Melian, C.J. and Sala, E. (2005) Interaction strength combinations and the overfishing of a marine food web. Proc. Natl. Acad. Sci. USA 102:54435447.
  • Berger, J., Stacey, P.B., Bellis, L. and Johnson, M.P. (2001) A mammalian predator–prey imbalance: grizzly bears and wolf extinction affect avian neotropical migrants. Ecol. Appl. 11:947960.
  • Bergström, B. (2007) Results of the Greenland Bottom Trawl Survey for Northern shrimp (Pandalus borealis) off West Greenland (NAFO Sub area 1 and Division 0A), 1988–2007. NAFO SCR Doc 07/71, 44p.
  • Bowyer, R.T., Person, D.K. and Pierce, B.M. (2005) Detecting top-down versus bottom-up regulation of ungulates by large carnivores: implications for conservation of biodiversity. In: Large Carnivores and the Conservation of Biodiversity. J.C.Ray, K.H.Redford, R.S.Steneck & J.Berger (eds) Covelo, CA: Island Press, pp. 342361.
  • Brander, K.M. (2007) The role of growth changes in the decline and recovery of North Atlantic cod stocks since 1970. ICES J. Mar. Sci. 64:211217.
  • Bundy, A. (2005) Structure and functioning of the eastern Scotian Shelf ecosystem before and after the collapse of groundfish stocks in the early 1990s. Can. J. Fish. Aquat. Sci. 62:14531473.
  • Bundy, A. and Fanning, L.P. (2005) Can Atlantic cod (Gadus morhua) recover? Exploring trophic explanations for the non-recovery of the cod stock on the eastern Scotian Shelf, Canada. Can. J. Fish. Aquat. Sci. 62:14741489.
  • Chelton, D. (1983) Effects of sampling errors in statistical estimation. Deep-Sea Res. 30:10831103.
  • Choi, J.S., Frank, K.T., Leggett, W.C. and Drinkwater, K. (2004) Transition to an alternate state in a continental shelf ecosystem. Can. J. Fish. Aquat. Sci. 61:505510.
  • Daan, N., Gislason., H., Pope, J.G. and Rice, J.C. (2005) Changes in the North Sea fish community: evidence of indirect effects of fishing? ICES J. Mar. Sci. 62:177188.
  • Fernández, C., Cerviño, S., Saborido-Rey, F. and Antonio Vázquez, A. (2008) Assessment of the Cod Stock in NAFO Division 3M. NAFO SCR Doc 08/26, 79p.
  • Fogarty, M.J. and Murawski, S.A. (1998) Large-scale disturbance and the structure of marine systems: fishery impacts on Georges Bank. Ecol. Appl. 8(Suppl.):S6S22.
  • Frank, K.T. and Leggett, W.C. (1994) Fisheries ecology in the context of ecological and evolutionary theory. Annu. Rev. Ecol. Syst. 25:401422.
  • Frank, K.T. and Shackell, N.L. (2001) Area-dependent patterns of finfish diversity in a large marine ecosystem. Can. J. Fish. Aquat. Sci. 58:17031707.
  • Frank, K.T., Petrie, B., Choi, J.S. and Leggett, W.C. (2005) Trophic cascades in a formerly cod-dominated ecosystem. Science 308:16211623.
  • Frank, K.T., Petrie, B., Shackell, N.L. and Choi, J.S. (2006) Reconciling differences in trophic control in mid-latitude marine ecosystems. Ecol. Lett. 9:10961105.
  • Frank, K.T., Petrie, B. and Shackell, N.L. (2007) The ups and downs of trophic control in continental shelf ecosystems. Trends Ecol. Evol. 22:236242.
  • Gabriel, W.L. (1992) Persistence of demersal fish assemblages between Cape Hatteras and Nova Scotia, Northwest Atlantic. J. Northw. Atl. Fish. Sci. 14:2946.
  • González-Troncoso, D. and Paz, X. (2007) Some ecological indices in Flemish Cap derived from the surveys conducted by EU between 1988 and 2006. NAFO SCR Doc 07/65, 26p.
  • Griffith, B., Scott, J.M., Carpenter, J.W. and Reed, C. (1989) Translocation of species as a conservation tool: status and strategy. Science 245:477480.
  • Halliday, R.G., Peacock, F.G. and Burke, D.L. (1992) Development of management measures for the groundfish fishery in Atlantic Canada. Mar. Policy 16:411426.
  • Hjermann, D.O., Ottersen, G. and Stenseth, N.C. (2004) Competition among fishermen and fish causes the collapse of Barents Sea capelin. Proc. Natl. Acad. Sci. USA 101:1167911684.
  • Horwood, J., O’Brien, C. and Darby, C. (2006) North Sea cod recovery? ICES J. Mar. Sci. 63:961968.
  • Huggett, A.J. (2005) The concept and utility of ‘ecological thresholds’ in biodiversity conservation. Biol. Cons. 124:301310.
  • Hutchings, J.A. (1996) Spatial and temporal variation in the density of northern cod and a review of hypotheses for the stock’s collapse. Can. J. Fish. Aquat. Sci. 53:943962.
  • Hutchings, J.A. and Reynolds, J.D. (2004) Marine fish population collapses: consequences for recovery and extinction risk. Bioscience 54:297309.
  • Jackson, J.B.C., Kirby, M.X., Berger, W.H. et al. (2001) Historical overfishing and the recent collapse of coastal ecosystems. Science 293:629637.
  • Kauffman, M.J., Varley, N., Smith, D.W., Stahler, D.R., MacNulty, D.R. and Boyce, M.S. (2007) Landscape heterogeneity shapes predation in a newly restored predator–prey system. Ecol. Lett. 10:690700.
  • Kenny, A.J., Kershaw, P., Beare, D. et al. (2006) Integrated assessment of the North Sea to identify the relationship between human pressures and ecosystem state changes – implications for marine management. Rep. ICES Regional Group for the North Sea, (REGNS) 36p. http://www.ices.dk/iceswork/workinggroups (last accessed 27 February 2009).
  • Kingsley, M.C.S. (2007) A provisional assessment of the shrimp stock off West Greenland in 2007. NAFO SCR Doc 07/70, 25p.
  • Kumar, A.B. and Deepthi, G.R. (2006) Trawling and by-catch: implications on marine ecosystems. Curr. Sci. 90:922931.
  • Leibold, M.A., Chase, J.M., Shurin, J.B. and Downing, A.L. (1997) Species turnover and the regulation of trophic structure. Annu. Rev. Ecol. Syst. 28:467494.
  • Lessard, R.B., Martell, S.J.D., Walters, C.J., Essington, T.E. and Kitchell, J.F. (2005) Should ecosystem management involve active control of species abundances? Ecol. Society 10:pp. 23.(online) http://www.ecologyandsociety.org/vol10/iss2/art1/ (last accessed 27 February 2009).
  • Levin, P.S., Holmes, E.E., Piner, K.R. and Harvey, C.J. (2006) Shifts in a Pacific Ocean fish assemblage: the potential influence of exploitation. Conserv. Biol. 4:11811190.
  • Macpherson, E. (2002) Large-scale species-richness gradients in the Atlantic Ocean. Proc. R. Soc. Lond. B Biol. Sci. 269:17151720.
  • Mahon, R., Brown, S.K., Zwanenberg, K.C.T. et al. (1998) Assemblages and biogeography of demersal fishes of the east coast of North America. Can. J. Fish. Aquat. Sci. 55:17041738.
  • Mandelman, J.W. and Farrington, M.A. (2007) The estimated short-term discard mortality of a trawled elasmobranch, the spiny dogfish (Squalus acanthias). Fish. Res. 83:238245.
  • May, R.M., Beddington, J.R., Clark, C.W., Holt, S.J. and Laws, R.M. (1979) Management of multispecies fisheries. Science 205:267277.
  • De Melo, A.A., Saborido-Rey, F. and Alpoim, R. (2007) An XSA based assessment of Beaked Redfish (S. mentella and S. fasciatus) in NAFO Division 3M. NAFO SCR Doc 07/47, 43p.
  • Murawski, S.A. (2000) Definitions of overfishing from an ecosystem perspective. ICES J. Mar. Sci. 57:649658.
  • Myers, R.A. and Cadigan, N.G. (1995) Was an increase in natural mortality responsible for the collapse of northern cod? Can. J. Fish. Aquat. Sci. 52:12741285.
  • Myers, R.A. and Worm, B. (2003) Rapid worldwide depletion of predatory fish communities. Nature 423:280283.
  • NEFSC. (2005) Assessment of 19 Northeast Groundfish Stocks through 2004. In: 2005 Groundfish Assessment Review Meeting (2005 GARM), Northeast Fisheries Science Center Reference Document 05–13, 15–19 August, 2005. R.K., Mayo & M., Terceiro (eds) Woods Hole, Massachusetts: Northeast Fisheries Science Center, pp. 1221.
  • Oguz, T. and Gilbert, D. (2007) Abrupt transitions of the top-down controlled Black Sea pelagic ecosystem during 1960–2000: evidence for regime-shifts under strong fishery exploitation and nutrient enrichment modulated by climate-induced variations. Deep-Sea Res. 54:220242.
  • Olsen, E.M., Heino, M., Lilly, G.R. et al. (2004) Maturation trends indicative of rapid evolution preceded the collapse of northern cod. Nature 428:932935.
  • Persson, L., Amundsen, P.-A., De Roos, A.M., Klemetsen, A., Knudsen, R. and Primicerio, R. (2007) Culling prey promotes predator recovery – alternative states in a whole-lake experiment. Science 316:17431746.
  • Pikitch, E.K., Santora, C., Babcock, E.A.et al. (2004) Ecosystem-based fishery management. Science 305:346347.
  • Post, D.M., Palkovacs, E.P., Schielke, E.K. and Dodson, S.I. (2008) Intraspecific variation in a predator affects community structure and cascading trophic interactions. Ecology 89:20192032.
  • Quinn, J.F., Wing, S.R. and Botsford, L.W. (1993) Harvest refugia in marine invertebrate fisheries: models and applications to the red-sea urchin, Strongylocentrotus franciscanus. Am. Zool. 33:537550.
  • Rago, P.J., Sosebee, K.A., Brodziak, J.K.T., Murawski, S.A. and Anderson, E.D. (1998) Implications of recent increases in catches on the dynamics of Northwest Atlantic spiny dogfish (Squalus acanthias). Fish. Res. 39:165181.
  • Richardson, A.J. and Schoeman, D.S. (2004) Climate impact on plankton ecosystems in the northeast Atlantic. Science 305:16091612.
  • Ripple, W.J. and Beschta, R.L. (2003) Wolf reintroduction, predation risk, and cottonwood recovery in Yellowstone National Park. Forest Ecol. Manage. 184:299313.
  • Rivard, D. and Maguire, J.-J. (1993) Reference points for fisheries management: the eastern Canadian experience. In: Risk Evaluation and Biological Reference Points for Fisheries Management. S.J.Smith, J.J.Hunt & D.Rivard (eds) Can. Spec. Publ. Fish. Aquat. Sci. 120:3157.
  • Roff, D.A. (1992) The Evolution of Life Histories: Theory and Analysis. London: Chapman & Hall.
  • Rohde, K. (1992) Latitudinal gradients in species diversity: the search for the primary cause. Oikos 65:514527.
  • Roy, K., Jablonski, D., Valentine, J.W. and Rosenberg, G. (1998) Marine latitudinal diversity gradients: tests of causal hypotheses. Proc. Natl Acad. Sci. USA 95:36993702.
  • Savenkoff, C.F., Castonguay, M., Chabot, D., Hammill, M.O., Bourdages, H. and Morissette, L. (2007a) Changes in the northern Gulf of St Lawrence ecosystem estimated by inverse modeling: evidence of a fishery-induced regime shift? Estuar. Coast. Shelf Sci. 73:711724.
  • Savenkoff, C.F., Swain, D.P., Hanson, J.M. et al. (2007b) Effect of fishing and predation in a heavily exploited ecosystem: comparing periods before and after the collapse of groundfish in the southern Gulf of St Lawrence (Canada). Ecol. Modell. 204:115128.
  • Schmitz, O.J., Maback, P.A. and Beckerman, A.P. (2000) Trophic cascades in terrestrial systems: a review of the effects of carnivore removal on plants. Am. Nat. 155:141153.
  • Seddon, P.J., Armstrong, D.P. and Maloney, R.F. (2007) Developing the science of reintroduction biology. Conserv. Biol. 21:303312.
  • Shackell, N.L. and Frank, K.T. (2007) Compensation in exploited marine fish communities on the Scotian Shelf, Canada. Mar. Ecol. Prog. Ser. 336:235247.
  • Shelton, P.A. and Lilly, G.R. (2000) Interpreting the collapse of the northern cod stocks from survey and catch data. Can. J. Fish. Aquat. Sci. 57:22302239.
  • Shelton, P.A., Sinclair, A.F., Chouinard, G.A., Mohn, R. and Duplisea, D.E. (2005) Fishing under low productivity conditions is further delaying recovery of Northwest Atlantic cod (Gadus morhua) Can. J. Fish. Aquat. Sci. 63:235238.
  • Shin, Y.-J. and Cury, P. (2004) Using an individual-based model of fish assemblages to study the response of size spectra to changes in fishing. Can. J. Fish. Aquat. Sci. 61:414431.
  • Sinclair, A. (1996) Recent declines in cod species stocks in the Northwest Atlantic. NAFO Sci. Coun. Studies 24:4152.
  • Sinclair, A.F. (2001) Natural mortality of cod (Gadus morhua) in the southern Gulf of St Lawrence. ICES J. Mar. Sci. 58:110.
  • Sissenwine, M. and Murawski, S. (2004) Moving beyond ‘intelligent tinkering’: advancing an ecosystem approach to fisheries. Mar. Ecol. Prog. Ser. 274:291295.
  • Soulé, M.E., Estes, J.A., Miller, B. and Honnold, D.L.. (2005) Strongly interacting species: conservation policy, management, and ethics. Bioscience 55:168176.
  • Steele, J.H. and Collie, J.S. (2005) Chapter 22. Functional diversity and stability of coastal ecosystems. In: The Sea: Vol. 13. The Global Coastal Ocean. A. Robinson & K. Brink (eds). Harvard University Press, pp. 785820.
  • Steingrund, P. and Gaard, E. (2005) Relationship between phytoplankton production and cod production on the Faeroe Shelf. ICES J. Mar. Sci. 62:163176.
  • Sünksen, K. (2007) A preliminary estimate of Atlantic cod (Gadus morhua) biomass in West Greenland offshore waters (NAFO Subarea 1) for 2007 and recent changes in the spatial overlap with Northern shrimp (Pandalus borealis). NAFO SCR Doc 07/73, 10p.
  • Sünksen, K. and Jørgensen, O.A. (2007) Biomass and abundance of demersal fish stocks off West Greenland estimated from the Greenland Shrimp Survey, 1988–2006. NAFO SCR Doc 07/28, 31p.
  • Swain, D.P. and Sinclair, A.F. (2000) Pelagic fishes and the cod recruitment dilemma in the northwest Atlantic. Can. J. Fish. Aquat. Sci. 57:13211325.
  • Swain, D.P., Sinclair, A.F. and Hanson, M.J. (2007) Evolutionary response to size-selectivity mortality in an exploited fish population. Proc. Biol. Sci. 274:10151022.
  • Terborgh, J., Lopez, L., Nuñez, P. et al. (2007) Ecological meltdown in predator-free forest fragments. Science 294:19231926.
  • Turner, R.G., Gatehouse, C.M. and Corey, C.A. (1987) Does solar energy control organic diversity? Butterflies, moths and the British climate. Oikos 48:195205.
  • Turner, R.G., Lennon, J.J. and Lawrenson, J.A. (1988) British bird species distributions and the energy theory. Nature 335:539541.
  • Walters, C. and Maguire, J.-J. (1996) Lessons for stock assessment from the northern cod collapse. Rev. Fish Biol. Fish. 6:125137.
  • Worm, B. and Myers, R.A. (2003) Meta-analysis of cod-shrimp interactions reveals top-down control in oceanic food webs. Ecology 84:162173.
  • Yodzis, P. (2001) Must top predators be culled for the sake of fisheries? Trends Ecol. Evol. 16:7884.
  • Yom-Tov, Y., Yom-Tov, S., MacDonald, D. and Yom-Tov, E. (2007) Population cycles and changes in body size of the lynx in Alaska. Oecologia 152:239244.

Appendices

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion and conclusions
  7. Acknowledgements
  8. References
  9. Appendices

Appendix A

Virtual population analysis based on commercial landings and fishery-independent surveys provide the input data to estimate the total mortality (Z) which typically is decomposed into fishing (F) and natural (M) mortality. Total mortality of cod, instead of fishing mortality alone, was chosen as the measure of impact on the predator trophic level for two reasons: first, the trophic structure is affected by total mortality, regardless of its source, within a functional group; secondly, in the available analyses, natural mortality of cod was assumed to be constant (M = 0.2 yr−1). This assumption has been vigorously debated, with no general agreement having been reached (Myers and Cadigan, 1995; Shelton and Lilly, 2000; Sinclair, 2001). Sinclair (2001) demonstrated that M for cod in the southern Gulf of St. Lawrence (area 2, Fig. 1) was greater than 0.2 during a moratorium on fishing, after the resident cod stock had collapsed. He suggested that M may have increased before the stock collapsed but the evidence was weak. Myers and Cadigan (1995), in their analysis of M for northern cod (area 1, Fig. 1), found no evidence that M had increased just prior to stock collapse in the early 1990s. Some authors have found that the models used to reconstruct the population dynamics of cod fit better when M is allowed to vary (Shelton and Lilly, 2000). Despite this debate, it is reasonable to conclude that most of the estimated total mortality is attributable to fishing. During the mid-1980s when cod stock biomasses were high, fishing mortality was in excess of the fishery management target levels and increasing. Fishery managers were uncertain about the necessity to reduce F and ad hoc measures were adopted to maintain or increase landings (Rivard and Maguire, 1993; Sinclair, 1996).

Appendix B

Without exception, the dominant groundfish species in each of the nine areas examined was Atlantic cod, whose average annual contribution to total groundfish landings ranged from ∼70% in the northern areas to ∼20% in the south (Table A1). There was also a high degree of similarity between the time series of landings of all groundfish species (made up of several different species) and that of cod (Table A1, Fig. A1); moreover, in areas 6 and 8, a combined quota for cod, haddock and pollock was instituted in 1989 because of the strong spatial overlap among these species (Halliday et al., 1992). The principal gear type used in these commercial fisheries is the bottom trawl (Table A1), a gear widely known to have low selectivity (Kumar and Deepthi, 2006). This is illustrated by a comparison of bottom trawl catch composition relative to that of fixed gear (hook and line) on the eastern Scotian Shelf (area 6 of Fig. 1) from 1998 to 2003. Bottom trawl deployments over this interval landed 72–110 species compared to 39–50 species for fixed gear. Figure A2 illustrates the strong spatial overlap of cod with most other species in the predator functional group. These data support our use of cod as a surrogate for and indicator of the overall index of disturbance to the top predator level in the nine northwest Atlantic areas investigated. Figure A3 shows the spatial distributions of species within the prey functional group.

The strong overlap of species and the low selectivity of gear implies that mortality of non- or lightly fished species would be similar to that of cod. The non- or lightly fished species would constitute bycatch and die after being discarded. However, recent work indicates that fish that do not have swim bladders exhibit greater variability in their survival rate after discarding and that some species of flatfish appear to have relatively good chances of post-capture survival (Suuronen, 2005). Moreover, Mandelman and Farrington (2007) indicated that the mortality rate of trawled dogfish was well below the 50% bycatch discard mortality rate used in current fishery models. However, they acknowledged that discard mortalities may increase rapidly as catch weights rise above 200 kg, i.e., as the cod ends become more heavily packed.

Table A1.   Statistics of the commercial fisheries for cod in nine Northwest Atlantic areas (1–9, Fig. 1). Combined percentage of total landings for cod, haddock and pollock are shown in parentheses for areas 6–9.
AreaPercentage of cod (C-H-P) in total groundfish landings during 1978–1990Temporal correlation between annual cod landings and total groundfish landings from 1978 to 1990Percentage of total groundfish captured using bottom trawls
  1. Pre-1978 landings data are considered of poor quality; only minor amounts of haddock and pollock are landed in areas 1–5.

1690.8468
2680.9674
3840.7266
4730.9451
5310.9094
632 (45)0.6690
720 (47)0.3878
830 (78)0.8858
938 (59)0.8190
  • image(A1)

[  Annual landings of cod (solid bar) and total groundfish (solid + open bar) from nine areas in the NW Atlantic. Correlation between the two time series for each area was: area 1: = 0.84, area 2: 0.96, area 3: 0.72, area 4: 0.94, area 5: 0.90, area 6: 0.66, area 7: 0.38, area 8: 0.88 and area 9: 0.81. ]

  • image(A2)

[  Spatial distribution of each species in the predator functional group based on scientific survey data collected from 1970 to 1994. ]

  • image(A3)

[  Spatial distribution of each species in the prey functional group based on scientific survey data collected from 1970 to 1994. ]