Scenario testing of fisheries management strategies using a high resolution ERSEM–POM ecosystem model
Hellenic Centre for Marine Research, PO Box 2214, Iraklio, Crete GR71003, Greece
Correspondence to G. Petihakis: tel: +30 281 0337755; fax: +30 281 0337822; e-mail: pet{at}her.hcmr.gr
Petihakis, G., Smith, C. J., Triantafyllou, G., Sourlantzis, G., Papadopoulou, K-N., Pollani, A., and Korres, G. 2007. Scenario testing of fisheries management strategies using a high resolution ERSEM–POM ecosystem model. – ICES Journal of Marine Science, 64: 1627–1640.
Ecosystem models are just beginning to be considered as management tools. In terms of fishery impacts, dynamic ecosystem models provide an opportunity to make advances because they can both evaluate the state of the system and make predictions about the ecosystem under various fishing scenarios. In the framework of the Cost Impact project, a complex high-resolution (500 x 500 m grid) ecosystem model was implemented in Iraklion Bay, Crete. Several management scenarios were simulated to investigate the impacts of trawling on this particular ecosystem (reductions in fishing area, effort, and mortality). Introducing trawling impacts into the model led to increases in pelagic production. All scenarios also resulted in net increases in pelagic production, the level of which, and the degree of spatial variability, was dependent on the particular scenario. Changes in pelagic variables were often noted in areas well away from trawled areas. It was also clear that for pelagic variables and processes, depth of trawling is more important than reduction in trawling area, i.e. a scenario banning fishing in waters shallower than 100 m seems to lead to less change in the pelagic system than a scenario that reduces direct mortality to the benthos.
Keywords: ecosystem model, management scenarios, trawling impact
Received 14 August 2006; accepted 21 September 2007.
 |
Introduction |
|---|
 Top Introduction Methods and dataResultsDiscussionReferences |
|---|
Historically, fisheries have been assessed and managed as single species and stocks. As our understanding of predator–prey relationships improved, multispecies assessments were developed. The ecosystem-based approach to fisheries management (EAF) has taken this further, by acknowledging that fisheries are part of the environment and cannot be managed in isolation (Cury et al., 2005). Frid et al. (2005), in considering the needs for EAF, advocated applying environmental impact assessments to produce ecosystem-based fisheries advice. They noted that the standard approach involves six steps: (i) assessment of the current environment and resources; (ii) description of the consequences of fishing activity; (iii) assessment of the significant effects; (iv) selecting and evaluating mitigation strategies; (v) taking a decision on appropriate management action; and (vi) monitoring and reviewing. The difficulty for scientists in this approach is to acquire and assimilate the data from the complexity of parameters and linkages that represent the ecosystem. Primarily, the difficulties lie in the consequences of fishing activities and the assessment of what constitutes significant effects. The direct impacts of trawling are reasonably well understood; the net directly removes, destroys, and damages a number of organisms per unit area (target and non-target), and the trawl gear physically disturbs the sediment surface (Hall, 1999; Kaiser and de Groot, 2000; Frid et al., 2005). However, it is the indirect consequences of trawling in pelagic and benthic systems that are the most difficult to assess, e.g. changes in predator–prey relationships, food availability, fluxes of chemicals, resuspension, and sedimentation.
Several models are available to evaluate the ecosystem effects of fishing, and they come in various levels of complexity and utility (reviewed by Robinson and Frid, 2003). Two commonly used models that incorporate fish stocks and inter-relationships are Ecopath and Ecosim (Walters et al., 1997; Pauly et al., 2000; Christensen and Walters, 2004). Linked, the two models describe the ecosystem through balancing flows between trophic groups, and through dynamic simulation, they allow the prediction of how fisheries will alter ecosystems. Although the models incorporate both benthic and pelagic variables, covering a large percentage of the higher trophic groups, they lack important extrinsic drivers, such as climate variation, which is fundamental for interpreting community patterns and dynamics (Hall, 1994). Another important limitation is the high level of information on the stocks needed for accurate modelling (virtual population analysis, VPA, unreported catches, and discards; Zeller and Reinert, 2004). Other models deal in detail only with parts of the ecosystem. For example, the size-based model used by Duplisea et al. (2002) focused on the impacts of trawling on benthic biomass and production (17 size groups of benthos).
Models have been used in a number of ways in the past few years to make predictive estimates ultimately for management purposes. With effective ecosystem simulation, management scenarios can be applied through key parameters (e.g. a scenario of reduction in fishing effort, parameterized through a reduction in benthic mortality parameters), and the impact can be seen in changes to other parameters (e.g. to patterns of primary productivity). Many different scenarios can be tested through different parameterizations, mimicking, for example, effort fluctuations, or the extent of area or time limitations on spatio-temporal closures. Being able to visualize the implications of a scenario allows better formulation of future management strategies. Dinmore et al. (2003) investigated the impact of seasonal closures on the benthos of the North Sea. Hiddink et al. (2006) linked the Duplisea et al. (2002) benthic model with modelled data from vessel monitoring systems (VMS track data for fishing vessels, identifying microscale distribution of effort) to predict impacts after fishery effort reduction and/or after closing parts of the North Sea (including redirection of effort). Single-species and multispecies spatial dynamic models for fishing have been used also to explore the impacts of area closures on Georges Bank (Holland, 2003). Impacts of spatial closures and fishing effort restrictions have been modelled in the Faroe Islands marine ecosystem (Zeller and Reinert, 2004). Cury et al. (2005) noted many of the high-level difficulties in the EAF approach and pointed out some of the doubts in currently available ecosystem data and models (data availability and quality, parameter uncertainty, dynamism, spatio-temporal scales, and realism). However, data availability is improving, and ecosystem models are being developed all the time, and will continue to play a role in scenario testing in the future.
Mediterranean fisheries are characterized by a number of distinctive features (e.g. relative extension of national vs. international waters, the large number of artisanal fishing activities, scarcity of scientific information), with important implications for the conservation policy under the Common Fisheries Policy (CFP; CEC, 2002). Local fisheries tend to be multispecies, and landings are diffuse, creating problems for enforcement, control, and catch reporting. Also, fisheries statistics for stock assessments are still largely incomplete (CEC, 2005). Management is currently based on effort control and technical measures, including gear and vessel limitations, spatio-temporal closures, and minimum sizes (there are no quotas in the Mediterranean other than for bluefin tuna, Thunnus thynnus). Although management advice provided by the General Fishery Council of the Mediterranean (GFCM) is slowly moving from stock- to fisheries-orientated, and motions are under development for the implementation of an EAF in the Mediterranean (Tudela, 2004; FAO, 2005), management decisions will continue to be based on effort control, in line with the 2002 Community Action Plan for the sustainable exploitation of Mediterranean fisheries resources under the CFP (CEC, 2002).
As in other parts of the world, models can be used to assess how changes in effort affect the ecosystem. Lack of data on fisheries components and linkages makes the implementation of ecosystem models such as Ecopath and Ecosim difficult for most areas (with only one exception, the rocky sublittoral community of the Bay of Calvi, Corsica, western Mediterranean; Pinnegar and Polunin, 2004). However, a variety of dynamic models have received much attention in the Mediterranean over the past 10 years, most focusing on lower trophic levels (Hall, 1994; Tusseau et al., 1997; Triantafyllou et al., 2000, 2003b; Chifflet et al., 2001; Crispi et al., 2001; Blackford et al., 2004). One of these models is the European Regional Seas Ecosystem Model (ERSEM) (Baretta et al., 1995), a dynamic, highly complex model, originally developed for the North Sea, and which, because of its generic nature, has also been used in other parts of the world (http://web.pml.ac.uk/ecomodels/ersem.htm).
In the extreme oligotrophy of the eastern Mediterranean, production in the euphotic zone is primarily supported by regenerated nutrients, and sedimentation of cells is not an important pathway for the export of carbon to deeper water. In winter there is intense mixing, and nutrients are transported away from the euphotic zone as dissolved and particulate organic matter, and imported in dissolved inorganic form. Such mixing events have a major influence on biogeochemical fluxes, because the availability of nutrients for phytoplankton and bacterial growth is under hydrodynamic control in winter and under ecological control for the rest of the year (Tselepides and Polychronaki, 2000). To simulate such a system, models that take account of all trophic levels and that can consider offshore gradients in vertical-mixing regimes, light and nutrient availability, as well as grazing pressure on primary and bacterial production seem to be the only option in reproducing major qualitative aspects of the experimental system (Thingstad et al., 1999).
Here, we examine the impacts of trawling (parameterized through specific benthos components) with the help of a simulation model focusing on the lower trophic levels of both benthic and pelagic subsystems. Our main hypothesis is that trawling impacts a wide range of ecosystem components beyond its initial impact, in both time and space. Trawling impacts are parameterized reflecting existing conditions in the study area. With different parameterizations, different management scenarios (closures, reduced effort, reduced mortality) are investigated, with a secondary hypothesis that different scenarios may alter the degree and the extent of the impact.
|
Methods and data |
|---|
|
TopIntroductionMethods and dataResultsDiscussionReferences |
|---|
Model description
The model used in this application is a three-dimensional, on-line coupled system comprising two dynamic models, the Princeton Ocean Model (POM; Blumberg and Mellor, 1978) providing the physics, and a version of the ERSEM (Baretta et al., 1995) describing the biogeochemical processes. The development of the three-dimensional ecosystem models for the Cretan Sea was initiated under the projects Mass Transfer and Ecosystem Response (MATER), Mediterranean Forecasting System Pilot Project (MFSPP), and its follow-up Mediterranean Forecasting System Towards Environmental Predictions (MFSTEP; http://www.bo.ingv.it/mfstep/). The model has been customized, tuned, and extensively validated against climatological datasets as well as high-frequency in situ data from the marine observing station M3A (Petihakis et al., 2002; Triantafyllou et al., 2003a, b). As described elsewhere in the text, the model is nested to the coarser, but still high resolution, shelf model of the Cretan Sea, developed in the framework of the MFSPP. The nesting approach was considered important for conserving the important properties between the fine and the coarse-grid models.
POM is a primitive equation, time-dependent, 
-coordinate, free surface, split mode time-step model solving prognostic variables, i.e. the three components of velocity Ui = (U, V, W), temperature T, and salinity S. The POM model is one of the most described and applied physical models (Blumberg and Mellor, 1983; Oey et al., 1985a, b; Galperin and Mellor, 1990a, b; Mellor and Ezer, 1991; Lascaratos and Nittis, 1998).
The biogeochemical model is a generic complex model including organisms through a functional group approach, in order to keep the system simple and at the same time to provide the most necessary information for the description and analysis of carbon cycling. The most important processes operating both at physiological and community level, such as growth, respiration, lysis, excretion, mortality, and grazing are included, and the dynamics of nutrients are variably coupled (variable C:N:P ratios) to carbon dynamics. The foodweb includes the major groups present in the pelagic and benthic environment of the Cretan Sea, as shown in Figure 1, with subgroup divisions based mainly on size differentiation. Diatoms (P1) are large cells with an equivalent spherical diameter (ESD) of 20–200 µ, with an affinity for silica. Nanophytoplankton (P2) has an ESD of 2–20 µ, and picophytoplankton (P3) of 0.2–2 µ is associated with regenerated production. The final group, the flagellates (P4), represents cells with the same size as diatoms but without silica requirements, utilizing mainly nitrate.
|
Pelagic decomposers include the large group of free-living heterotrophic aerobic bacteria (B1), utilizing both dissolved and particulate organic matter and often competing with phytoplankton for nutrients. Consumers include four groups of zooplankton ranging from the small heterotrophic nanoflagellates to large carnivorous mesozooplankton (Z6–Z3), utilizing the complete spectrum of available food according to a food preference factor (0
factor 1; Table 1). The benthic system is relatively simple (Ebenhoh et al., 1995; Ruardij and van Raaphost, 1995; Blackford, 1997). The two systems are coupled through detrital deposition, benthic filter-feeder grazing, and nutrient exchange. Filter-feeders (Y3) make up a large functional group consisting of immobile organisms that feed directly from the pelagic system by filtering suspended particles from the water column (typical organisms are the molluscs). Apart from phytoplankton and detritus, they may also feed on benthic detritus and bacteria. Below the benthic surface in the oxic layer, there is a group of deposit-feeders (Y2), consisting of all organisms whose diet consists of benthic detritus and smaller organisms belonging to the meiobenthic community. The meiobenthos group (Y4) consists of a complex ensemble of heterotrophs with a size of ~1 mm and a relatively small impact on the sediment distribution through bioturbation and bioirrigation. Meiobenthos are the main predators of bacteria, which are the benthic decomposers. Bacteria are divided into aerobic (H1), which need oxygen for their functional dynamics, and anaerobic (H2) which combine the functionality of both the nitrate and sulphate reducers.
|
The benthic foodweb is assumed to have two coupled food chains, each containing standard predator–prey dynamics. The first represents the surface route of recycling via suspension-feeders, and the second represents the subsurface route involving bacteria, meiobenthos, and deposit-feeders.
Heterotrophs
There are three types of benthic heterotroph (Yi), the meiobenthos, the deposit-feeders, and the suspension-feeders, all of which share the same physiological processes. These include the uptake of food, assimilation, respiration, excretion, natural mortality, and predation, and can be written in text equation as:
|
| (1) |
Units are in mgC m–2 d–1 for carbon, and mmol m–2 d–1 for the nutrients N and P.
Uptake. As heterotrophs exploit more than one food source, the actual uptake is a sum of all different prey available. This can destabilize a foodweb because a scarce, slow-growing prey species may be driven to extinction if another prey species supports an expanding population of a common predator. To overcome this risk, a feeding threshold is adopted (luYi$), and for each prey species, a preference factor (p) is assigned. The dollar symbol used in the formulae denotes fixed parameter values. Therefore, the available prey (i) is given by:
|
| (2) |
The partial uptake flux from each prey is calculated by a Michaelis–Menten function controlled by a half saturation constant of uptake (huYi$):
|
| (3) |
where Avail_preyt is the total available food. Uptake_rate is calculated as the per capita maximum uptake rate (suYi$) multiplied by the consumer density (Yic) multiplied by one or more limitation factors:
|
| (4) |
where eT is the temperature limitation and eO the oxygen limitation factor. In the case of suspension-feeders, there is an extra limitation factor eC, mimicking the effect of crowding on the sea floor, and formulated as a Michaelis–Menten function with a threshold below which there is no shading effect. However, as interference is operational only above densities of 2500 mgC m–2 in the case of oligotrophic systems such as the Cretan Sea, the value of eC remains equal to 1.
In the case of suspension-feeders, there are two modifications. When uptake is from pelagic food sources there is a conversion in the units from mgC m–3 to mgC m–2. Also, as resuspension is not explicitly modelled, there is a parameter determining the thickness of the layer filtered (dwatY3$).
Assimilation and excretion. From the total food flux (fluxt) a fixed fraction is excreted (fluxex) in the form of faecal pellets, with a distinction between living prey (pueYi$) and detritus (pueQ6Yi$):
|
| (5) |
This flux produces particulate organic matter which is directed to detritus. As the quality of this material is inferior, containing less nutrients, a nutrient dilution factor (pudilYi$) is used in the calculation of the flux of phosphorus and nitrogen (but not with silicate).
The part of total flux not excreted is considered as assimilated:
|
| (6) |
Respiration. Respiration is modelled as carbon flux, consuming oxygen and producing carbon dioxide. Respiration has two components: the basal or rest respiration, which is a fixed proportion (srYi$) of the biomass, depending on temperature:
|
| (7) |
and the activity respiration, which is proportional to assimilated food:
|
| (8) |
Mortality. This includes non-predation mortality, and it can be attributed to causes such as age, stress, low oxygen, or trawling. With each cause there is a fixed mortality proportion:
|
| (9) |
|
| (10) |
|
| (11) |
where sdmO2Yi$ is the maximum rate of mortality attributable to oxygen stress, sdcYi$ the rate of cold-induced mortality at 0°C, ETW the ambient temperature, xdcYi$ the coefficient (slope) of cold mortality response, and sdYi$ the extra mortality (d–1) attributable to causes other than oxygen stress and cold.
The flux has the same composition as the Yi separated into two parts, leading to particulate and dissolved organic matter (DOM) (Q6 and Q1).
Nutrient excretion. As the heterotrophs ingest food with varying C:N:P ratios and at the same time lose carbon and nutrients through respiration, they must readjust their internal C:N:P ratio. Therefore, if there is a surplus of nutrients (most common), these are released directly to the dissolved inorganic pools. If there is nutrient shortage, the excess carbon is respired.
Predation. This loss is modelled as the predator uptake term already described, and is therefore a process associated not with the organism being analysed, but with its predator.
Decomposers
The concept of the standard animal is also used for decomposers, which are described by the function H and include two functional types: aerobic and anaerobic bacteria. These share the same processes, but differ in their food sources and parameterization, following the differential equation:
|
| (12) |
Units are again in mgC m–2 d–1 for carbon, and mmol m–2 d–1 for the nutrients N and P.
Decomposition. There are three decomposition pathways for detritus. The first is fast decomposition of particulate organic matter and is dependent on the quality of food. Therefore, the composition of the flux is different from the composition of the detritus, because the nutrient-rich fraction is decomposed first.
|
| (13) |
where suQ6fHi$ is the per-day detritus decomposition rate constant, Qavail the available detritus, H the bacteria biomass, and eT, eOx, and eN the temperature, oxic layer, and nutrient limitation factors, respectively. The latter is calculated as the ratio of the nutrient:carbon quota for the detrital food source over the nutrient:carbon quota for the bacteria for both nitrate and phosphate. The second is the slow particulate decomposition of slowly degradable (Q6) and refractory (Q7) detritus. This flux differs from the fast decomposition in that there is no nutrient limitation, whereas the detritus decomposition rate constant suQ6sHi$ is significantly smaller.
|
| (14) |
The last decomposition pathway is DOM, which follows the same procedure as detritus, with the only exception being that all DOM (Q1) is available:
|
| (15) |
The total decomposition flux is the sum of all partial fluxes. As a part of the particulate detritus is decomposed by extracellular enzymes, there is a flux back to DOM. This flux is simply a fraction of the particulate fluxes:
|
| (16) |
where pue6HiQ1$ and pue7HiQ1$ are the corresponding fixed fractions for slowly degradable (Q6) and refractory (Q7) detritus, respectively.
The difference between the decomposition flux and the DOM-excretion flux: is the assimilated flux:
|
| (17) |
Respiration. Respiration is similar to the heterotrophs separated into two components, the basal:
|
| (18) |
and the activity respiration, which is the proportion of the factor purHi$ to the assimilated food:
|
| (19) |
Mortality. Mortality represents non-predation mortality, and it is modulated by a fixed maximum daily mortality rate sdHi$ modified by the layer available for inhabitation. Therefore, if the oxic or anoxic layer decreases, mortality increases:
|
| (20) |
The mortality flux is directed into both particulate and dissolved detritus according to a fixed fraction pdHiQ1$.
Nutrient uptake/excretion. Unlike the consumers, benthic decomposers can take up inorganic nutrients, and under nutrient-limiting conditions, they can compete with primary producers. Therefore, they take up nutrients according to the decomposition flux and to the external inorganic nutrient concentration, though where there is a surplus, nutrients are excreted directly into inorganics.
Finally, processes such as bio-irrigation and bioturbation are also included in the model.
As the biogeochemical model is a zero-dimension model, the transformation of the variables into the three-dimensional domain is achieved through the equation:
|
| (21) |
Units are in mgC m–3 d–1 or mmol m–3 d–1, U, V, and W represent the velocity field, AH the horizontal mixing coefficient and KH the vertical eddy mixing coefficient provided by the physical model, and 
BF is the total biochemical reaction rate, calculated by the ecosystem model, for each variable.
For further details of the Cretan Sea application, the reader is referred to the detailed description given in Petihakis et al. (2002), and Triantafyllou et al. (2003a, b, 2007).
Model set-up
Although local trawling activities are concentrated inside Iraklion Bay between the north coast of Crete and the island of Dia, the model domain was extended farther north owing to the numerical instabilities of some pelagic state variables close to Dia Island caused by the reflection and/or distortion of passing disturbances through the boundary. The coordinates were 25°01'E–25°385'E and 35°33'N–35°495'N (Figure 2), with a horizontal resolution of 0.3 min (~500 x 500 m), developing a lattice of 76 x 34 grid boxes in the spherical coordinate system. As the modelled area is rather small, with significant open boundaries in the north, east, and west, detailed information on the boundary conditions is absolutely necessary. Therefore, the high-resolution model for Iraklion Bay (Petihakis et al., 2002; Triantafyllou et al., 2003a) was one-way nested with the coarser Cretan Sea three-dimensional ecosystem model, through 24 vertical layers.
|
The model's bathymetry was obtained from naval maps, enriched with detailed local observations made by RV "Philia" during the past decade, and for the coastline the NOAA-NGDC coastline extractor site (http://rimmer.ngdc.noaa.gov/coast/) was used. The method of Objective Analysis (Cressman, 1959; Barnes, 1964) was used on the bathymetric database (longitude, latitude, and depth) in order to fill all grid boxes, and a Shapiro (1970) third-order filter was used for smoothing.
The surface forcing fields of the model (monthly climatological windstresses, upward heat flux, net shortwave radiation, and evaporation rates) were derived from the 1979–1993 ECMWF 1 x 1 re-analysis 6-h atmospheric dataset, except for the precipitation fields, which were derived from Jaeger monthly climatology (Korres and Lascaratos, 2003). Additionally, the heat and freshwater surface fluxes were corrected using relaxation terms on the surface monthly climatologies.
Finally, the lateral boundary conditions at the three open boundaries of the model were provided by the coarse Cretan Sea model in the form of 10-d averaged prognostic fields of velocity, temperature, and salinity, having previously been analysed objectively and smoothed by a first-order Shapiro filter. The initial conditions for the biogeochemical variables were taken from the coarse Cretan Sea model once the fields had been analysed objectively to filter out spatial noise and interpolated to the grid points of the fine-resolution model where data were missing.
The optimal parameter sets for heterotrophs and decomposers in the model are given in Tables 2 and 3, respectively.
|
|
Experimental set-up
The trawling period in Greek national waters lasts from 1 October until the end of May (240 d) each year, without any quotas on landings, or any other restrictions on fishing effort. Therefore, the model was run for 10 years in order to reach quasi-steady state, and initialized with the 1 October model fields for a further simulation year. First, a reference run (no fishing effort) was performed, and the results stored in order to have a basis for comparison. Trawling paths, as observed in the field, were marked on the grid area (Figure 3). Of the total area covered, 370 boxes represent land, and of the remaining 2214 boxes, 389 have been trawled actively (17.6%). Although this picture is informative on concentration of fishing effort, there are no data on how often each box is visited during the trawling season and whether there are any particular patterns in these visits (east to west, etc). Based on the number of trawlers fishing, approximate number of days fishing, swept area, and box size, we considered that each box was trawled only once per year in a random manner. More specifically, every 2 d we randomly find one box and impose trawling on this box plus another two adjacent boxes (i.e. a total of three boxes every 2 d).
|
Trawl impacts were introduced into the model in the form of impact on benthic functional groups. Macrofaunal biomass data (AFDW) were taken from a trawl impact study in the same area (Smith et al., 2000), and trophic group mortalities attributable to trawling were calculated for 55% mortality of deposit-feeders, 97.8% of suspension-feeders, and an estimated 50% of benthic bacteria. In reality, benthic bacteria are not killed by trawling, but because of the mechanical action of the trawl, are suspended in the overlying water column. To simulate this, 50% of the bacterial biomass is removed from the benthic bacteria group and added to the pelagic bacterial biomass. To balance the budget, dead organic matter was added to benthic detritus in the form of POC, PON, and POP.
Although many scenarios can be examined with this model, here, apart from the reference run (no fishing) and current practice (no change), just three scenarios were tested (summarized in Table 4), following the existing trawl season. The scenarios represented relevant management measures with:
- Introduction of less-impacting gears (reduced mortality): regular mortality to the affected benthic groups reduced by 50 and 30% (the 30% level being equivalent to that reported by Kaiser et al., 2006).
- Restrictions to allowed trawling area: closure to trawling of the areas shallower than 100 m. In Iraklion Bay, 33% of the existing commercially trawled area (Figure 3, white areas) was closed, and all existing effort was concentrated in deeper water.
- Restrictions to fishing effort: total trawling activity reduced by 50% (equivalent to a 50% cut in the number of trawlers).
|
|
|
Results |
|---|
|
TopIntroductionMethods and dataResultsDiscussionReferences |
|---|
Simulation results expressed as differences between the scenario simulations minus the reference simulation are shown in Figures 5 and 6. As expected, suspension-feeders are the group most damaged by trawling. Following scenario 1a, most of the biomass present in the area (~90%) is removed. Scenario 1c indicates the least impact, but still results in a loss of ~30%. It is interesting that even if trawling is limited to the area >100 m (scenario 2a), >60% of suspension-feeders are removed, indicating that the major controlling factor is the direct mortality rate, though it is also coupled with the slow intrinsic growth of the group. Under all scenarios, trawling stops at day 240, but the decline continues to the end of the simulation period. This result is in contrast with that for deposit-feeders, which start to recover soon after fishing stops. Deposit-feeders decrease during the trawling season under all scenarios. In the two low-mortality scenarios (1c and b), they manage to exceed the reference conditions by 25 and 5%, respectively. It could be argued that such effects would be expected in scenarios 1a and 2a, as suspension-feeders are removed and more food becomes available for deposit-feeders (aerobic bacteria, and benthic and pelagic detritus). Under scenario 1a, as soon as trawling stops, the deposit-feeders show the expected increase, but as their biomass is very low, they never manage to attain their reference level. For scenario 2a, deposit-feeders are coupled not only to benthic resources, but also to pelagic variables. Therefore, the food produced in the pelagic system by trawling eventually settles and, in the absence of suspension-feeders, becomes available to the deposit-feeders in the form of detritus. When trawling is limited to deeper areas (scenarios 2a and b), the deposit-feeders remain at low levels without any tendency to increase. This is attributed to the deeper water column there, where the very low-light level fails to promote significant primary or bacterial production, as shown in Figure 6. The Cretan Sea is oligotrophic, with low-annual productivity of 120–300 mgC m–2 d–1 at the coast and 64–231 mgC m–2 d–1 offshore (Psarra et al., 2000), peaking between late winter and early spring because of intense mixing and a subsequent supply of nutrients to the euphotic zone. Simulated primary production (Figure 7) is close to these values, exhibiting a winter low and spring and autumn maxima.
|
|
|
Meiobenthos is indirectly affected, because it is not killed by trawling, exhibiting increased production under all scenarios. Schratzberger and Jennings (2002) suggested low mortality of meiofauna attributable to trawling. Duplisea et al. (2002), investigating size classes within bottom fauna, modelled trawling impacts on several size classes of meiofauna, and predicted that they would be largely unaffected (particularly the smaller sizes). A clear difference can be seen between scenarios 2a and b and the other scenarios, once trawling stops (Figure 5). Although one would expect that the more the predators are removed (in this case the deposit-feeders), the more noticeable the increase would be for meiobenthos, this is not always the case. Therefore, under scenario 1c, with the least removal of deposit-feeders, the meiobenthos exhibits two rapid increases (above the reference run), during the first 90 day and from day 240 to day 300, before returning to zero at the end of the experiment. This behaviour is a synergistic effect of top-down and bottom-up controls. As grazing pressure starts to be reduced by trawling, and although more food is added to the detrital pool, meiobenthos cannot take advantage of the situation and remains at very low levels. It is only when the predators drop below a certain biomass that the meiobenthos can exploit the food resources, exhibiting rapid growth (days 50–90), then remaining stationary at higher levels of biomass subsequently (days 90–200). A further decrease in deposit-feeders initiates a second increase in the meiobenthos, which exploits available food resources heavily (days 200–300). As trawling pressure ceases (day 240), the deposit-feeders cannot initially control the meiobenthos, which continues to grow, taking advantage of the freshly deposited organic matter. Within ~60 d, the deposit-feeders manage to "catch-up", exhibiting a classical predator–prey relationship. In the deeper areas (scenarios 2a and b), although meiobenthos shows behaviour similar to the other runs during the first 210 d, it decreases for the remaining period, affected by the relatively lower production in the water column and the subsequent decrease in detritus supply.
The simulation of benthic detritus at the beginning of the experimental period exhibits approximately the same trend for all scenarios, with an initial increase attributable to the continuous addition from trawling. However, as soon as the meiobenthos responds to this increased supply of food, a differentiation begins, with the different distributions being modulated through supply (trawling and sedimentation) and consumption/remineralization (meiobenthos and bacteria). A prominent characteristic for scenarios 1a–c at the end of the trawling period is the average reduction in detritus of ~260 mgC m–2.
Overall in the benthic system, trawling seems to promote a shift from organic carbon towards meiobenthos and benthic detritus, with the former extending all year round and the latter lasting for some five months. Another group positively affected by trawling is the bacteria, enhancing both aerobic and anaerobic mineralization.
An important part of this study was to examine what the possible effects of all the scenarios might be on the pelagic system. Irrespective of scenario, the pelagic system shows a significant response for all variables only after 200 d of trawling. This is explained by the fact that trawling activities mainly take place during late autumn, winter, and spring, when the system is under hydrodynamic control. Therefore, the products of increased benthic mineralization are transported into the upper water layers where, with favourable conditions (light and temperature), a significantly higher spring bloom than in the reference run is depicted by the chlorophyll results. As already mentioned, scenarios 2a and b are very close to the reference run, because the deeper water column limits nutrient transfer to surface layers. The increase in chlorophyll as a result of trawling under reduced mortality (scenarios 1b and c) and reduced effort (scenario 3) are rather similar to scenario 1a, indicating that even with 50 or 30% mortality, or with 50% decrease in effort, bottom trawling has a significant effect on the overlying water column processes.
Generally, there is good coupling between chlorophyll (and subsequently phytoplankton biomass), meiobenthos, and aerobic benthic bacteria, indicating the important role of the latter in the regeneration process. An interesting result comes from the evolution of bacterial biomass under scenarios 1a–c, where instead of an increased production, as for phytoplankton, bacteria decrease significantly soon after trawling stops. In oligotrophic systems, bacteria often compete for nutrients with phytoplankton, taking advantage of their smaller size and favourable surface:volume ratio (Thingstad and Rassoulzadegan, 1995). However, when the system is enriched with nutrients, it moves towards a classical foodweb, with larger phytoplankton cells and more energy directed towards higher trophic levels. Therefore, when conditions become favourable in the model, large phytoplankton take up the available nutrients and outcompete bacteria. However, soon after this initial phase, the available organic substratum (POC, POP, and PON) increases and, in conjunction with their very fast growth potential, pelagic bacteria also increase. As expected, the evolution of zooplankton biomass is tightly coupled to the available food, i.e. phytoplankton and bacteria biomass.
As stated above, the ecosystem in the study area is expected to be influenced significantly by the hydrodynamic fields transporting chemical and biological components both horizontally and vertically. Field measurements and model runs show a dominant circulation pattern in a west-to-east direction, transporting water from the central part of the bay to the west/central coastal areas (Figure 8). To explore the contribution of the physics in the ecosystem dynamics, two-dimensional plots were produced with the average, depth-integrated, daily, primary, and bacterial productivity (Figure 9).
|
|
Under scenarios 1a–c and 3, there is increased primary production in the coastal areas of the grid and in particular in the southwest, exceeding the reference run by as much as 50 mgC m–2, which is ~20–25% of the field values measured (Psarra et al., 1997; Turley et al., 2000). Bacterial productivity exhibits a much wider spatial increase, particularly in coastal areas. However, it is significantly lower than that of phytoplankton, reaching as much as 2.1 mgC m–2 year–1 higher than the reference run. This value is ~1.5% of the in situ value (Turley et al., 2000; Van Wambeke et al., 2000).
Scenarios 2a and b have relatively small effects on primary production, which is not localized in the coastal area, but covers almost all the model domain with the exception of the northwest. Bacterial productivity, on the other hand, seems to follow the circulation patterns. The hydrological structure of the Cretan Sea is complex (characterized by mesoscale spatial and temporal variability), and is the most likely reason for the patchiness.
|
Discussion |
|---|
|
TopIntroductionMethods and dataResultsDiscussionReferences |
|---|
We attempted to apply an existing complex dynamic model based on ERSEM to understand trawling impact and to investigate fisheries management scenarios with respect to a changing impact. In particular, two hypotheses were investigated: that trawling affects a wide range of ecosystem components beyond the initial impact; and that different management scenarios will alter the degree and extent of the impact. Both these hypotheses were true. All the impact scenarios presented resulted in a net increase in primary production, an increase that was spatially variable and strongly related to hydrological features (Figure 8). Pelagic production in shallow waters was strongly affected (increased) by trawling in shallow water, whereas pelagic production in deeper water was less affected. Perhaps more important, there was an effect on variables away from the trawled areas (several kilometres), indicating that the impacts were far from localized. Experimental work by Durrieu de Madron et al. (2005) in the western Mediterranean revealed that trawling causes a significant addition of porewater nutrients to the water column (particularly nitrate and nitrite) and that this can generate a local bloom, especially in nutrient-poor areas. The Cretan Sea is generally considered to be oligotrophic, on the basis of its low nutrient levels and poor productivity, but it presents an anomalously high nitrogen-to-phosphorus ratio, with the latter being the dominant limiting nutrient for phytoplankton and bacterial growth (Tselepides and Polychronaki, 2000). The underlying mechanisms for the shift in N:P ratio and apparent P-limitation of the Mediterranean Sea are unclear. Proposed hypotheses range from nitrogen fixation by planktonic and benthic algae to the removal of phosphorus by sedimentation of airborne iron-rich particles, scavenging phosphate from the Mediterranean pelagic system (Dolan et al., 1995). It is likely that detrital mineralization caused by trawling and the subsequent nutrient release, driven by the circulation patterns, allows the increase of primary production away from the trawl areas.
Duplisea et al. (2002) modelled trawling impacts on different size classes of bottom fauna, and predicted that the meiofauna would be largely unaffected (particularly smaller sizes). Schratzberger et al. (2002), in a direct impact study, reported that meiofauna show very low mortality from trawling. The ERSEM model showed an overall increase in meiofauna biomass in response to trawling. This was as a result of removal of macrofaunal disturbance (deposit-feeders) rather than trawling disturbance, and the meiofauna only decreased in the closed trawling season as deposit-feeders recovered.
Three major scenarios were examined, representing realistic management options: reduction in mortality/impact (e.g. through the use of less damaging gears), reduction in fishing area (e.g. spatial closures, strips, boxes, or depth-/distance-defined closures), and reduction in fishing effort (e.g. temporal limitations). There is already a management policy included in the model in terms of the closed fishing season from the end of May to the end of September. From the modelled scenarios for the area studied, depth of trawling was more important in terms of pelagic variables than reduction in mortality (reduction in trawling area had an order of magnitude less impact on the pelagic variables). In other words, as a control measure, a ban on fishing in waters shallower than 100 m would seem to be more beneficial, because it would be less disruptive to the pelagic system than a reduction in mortality/impact.
In judging between the three scenarios presented (reduced mortality/impact, reduced area, and reduced effort), it would seem that the reduced trawling area would be the more conservative/safe option, and that this could be implemented simply in the management process through spatial closures. Modelling spatial closures and effort restrictions with Ecopath and Ecosim, Zeller and Reinert (2004) noted that closures were beneficial for demersal stocks, but that effort reduction was only partially beneficial even at drastic levels (different stocks were affected differently). The results of Dinmore et al. (2003), modelling benthos impacts, suggest that closures cause redistribution of effort away from high production areas into less disturbed ones, where the impact on the benthos may be greater. They argue that where seasonal closures may cause increased impact on more-sensitive habitats, effort reductions or permanent closures should be considered. Newly opened areas tend to attract greater fishing effort than when they have been open for some time. At another level, Holland (2003), linking fish models with bioeconomics, noted that closures were unlikely to increase net revenues significantly, because again effort tends to be redistributed away from areas of high productivity. In Greek waters, the existing seasonal closure covers all trawling activities, so there is no redistribution of effort, and even with the scenario of shallow water closure (<100 m), the redistribution of total effort into deeper water had the least impact on pelagic variables. From an ecosystem perspective, stability in applying closures is important, because this would imply stability in the level of impacts.
The scenario of reduced mortality/effort used mortality values obtained in the study area, but also looked at lower mortality levels consistent with otter and beam trawling elsewhere (Kaiser et al., 2006), and it also posed the question of what would happen if we could introduce less damaging gears (same effort, but less mortality). With an input reduction in mortality of 50 and 30%, these were the only two scenarios that allowed the deposit-feeders to recover significantly, but there was no overall difference in primary production in the Bay across the scenario.
Given the results of the ERSEM model and the range of scenarios that could be run, we find the model to be a powerful tool for management. Interpretation of the results can, however, be very difficult. If, for example, just the integrated results were considered, then the trawling scenarios give an overall positive result (in the oligotrophic eastern Mediterranean) of increased primary production. It must be stressed that this was in all cases at a cost to the benthos, as a proxy for trawling impact. Other processes can also be impacted or weighted by impact. In using the model for "what-if" scenarios, it is therefore important to have an understanding of both the system and its balance, to be able to ask the right questions, and to understand the importance of all results (and inputs). Models cannot stand alone in the decision-making process because they cannot provide all the answers, and any management decision-making process should also take into account socio-economic considerations. The model in Iraklion Bay is working realistically and will be used in the future to investigate a more detailed suite of scenarios, including both spatial and temporal closures. At present, though, it is set up for a variety of functional groups and processes, which could be improved, particularly with the inclusion of fish trophic and functional groups.
|
Acknowledgements |
|---|
This paper is dedicated to co-author George Sourlantzis, who passed away tragically at an early age in 2005. We thank Capt. Manolis Kokos of the RV "Philia" for providing information on the location of trawling activities in Iraklion Bay, and to Bert Brinkman and an anonymous reviewer for their constructive approach and suggested improvements to the draft manuscript. The work was carried out under the EU-funded project COST-IMPACT: Costing the impact of demersal fishing on marine ecosystem processes and biodiversity, FP5, Quality of Life: Q5RS-2001-00 993. Co-financing was provided by the General Secretariat for Research and Technology, Ministry of Development, Greece.
|
References |
|---|
|
TopIntroductionMethods and dataResultsDiscussionReferences |
|---|
-
Baretta J. W., Ebenhoh W., Ruardij P. The European Regional Seas Ecosystem Model, a complex marine ecosystem model. Netherlands Journal of Sea Research (1995) 33:233–246.
Barnes S. L. A technique for maximizing details in numerical weather map analysis. Journal of Applied Meteorology (1964) 3:396–409.[CrossRef]
Blackford J. C. An analysis of benthic biological dynamics in a North Sea ecosystem model. Journal of Sea Research (1997) 38:213–230.[CrossRef]
Blackford J. C., Allen J. I., Gilbert F. J. Ecosystem dynamics at six contrasting sites: a generic modelling study. Journal of Marine Systems (2004) 52:191–215.[CrossRef][Web of Science]
Blumberg A. F., Mellor G. L. A coastal ocean numerical model. In: Proceedings of an International Symposium. Mathematical Modelling of Estuarine Physics—Holtz J., Sunderman K., eds. (1978) Berlin: Springer. 203–214. In.
Blumberg A. F., Mellor G. L. Diagnostic and prognostic numerical circulation studies of the South Atlantic Bight. Journal of Geophysical Research (1983) 88:4579–4592.
CEC. Communication from the Commission to the Council and the European Parliament: laying down a Community Action Plan for the conservation and sustainable exploitation of fisheries resources in the Mediterranean Sea under the Common Fisheries Policy. (2002) Brussels, 28.5.2002. COM (2002) 186 final.
CEC. Report of The Scientific, Technical and Economic Committee for Fisheries. Review of Scientific Advice for 2005 (2005) Brussels, 17.2.2005. SEC (2005) 266.
Chifflet M., Andersen V., Prieur L., Dekeyser I. One-dimensional model of short-term dynamics of the pelagic ecosystem in the NW Mediterranean Sea: effects of wind events. Journal of Marine Systems (2001) 30:89–114.[CrossRef][Web of Science]
Christensen V., Walters C. J. Ecopath with Ecosim: methods, capabilities and limitations. Ecological Modelling (2004) 172:109–139.[CrossRef][Web of Science]
Cressman G. P. An operational objective analysis scheme. Monthly Weather Review (1959) 87:329–340.
Crispi G., Mosetti R., Solidoro C., Crise A. Nutrients cycling in Mediterranean basins: the role of the biological pump in the trophic regime. Ecological Modelling (2001) 138:101–114.[CrossRef][Web of Science]
Cury P. M., Mullon C., Garcia S. M., Shannon L. J. Viability theory for an ecosystem approach to fisheries. ICES Journal of Marine Science (2005) 62:577–584.
Dinmore T. A., Duplisea D. E., Rackham B. D., Maxwell D. L., Jennings S. Impact of large-scale area closure on patterns of fishing disturbance and the consequences for benthic communities. ICES Journal of Marine Science (2003) 60:371–380.
Dolan J. R., Thingstad T. F., Rassoulzadegan F. Phosphate transfer between microbial size-fractions in Villefrance Bay (NW Mediterranean Sea). France in autumn 1992. Ophelia (1995) 41:71–85.
Duplisea D. E., Jennings S., Warr K. J., Dinmore T. A. A size-based model of the impacts of bottom trawling on benthic community structure. Canadian Journal of Fisheries and Aquatic Sciences (2002) 59:1785–1795.
Durrieu de Madron X., Ferre B., Le Corre J., Grenz C., Conan P., Pujo-Pay M., Buscail R., et al. Trawling-induced resuspension and dispersal of muddy and dissolved elements in the Gulf of Lion (NW Mediterranean). Continental Shelf Research (2005) 25:2387–2409.[CrossRef][Web of Science]
Ebenhoh W., Kohlmeier C., Radford P. J. The benthic biological sub-model in the European Regional Seas Ecosystem Model. Netherlands Journal of Sea Research (1995) 33:423–452.
FAO. GFCM Report of the Seventh Session of the Scientific Advisory Committee. FAO Fisheries Report, 789 (2005) 96 pp.
Frid C. L. J., Paramor O. A. L., Scott C. L. Ecosystem-based fisheries management: progress in the NE Atlantic. Marine Policy (2005) 29:461–469.[CrossRef][Web of Science]
Galperin B., Mellor G. L. A time-dependent, three-dimensional model of the Delaware Bay and River. 2. Three-dimensional flow fields and residual circulation. Estuarine Coastal and Shelf Science (1990) 31, a. 255–281.[CrossRef]
Galperin B., Mellor G. L. A time-dependent, three-dimensional model of the Delaware Bay and River. 1. Description of the model and tidal analysis. Estuarine Coastal and Shelf Science (1990) 31, b. 231–253.[CrossRef]
Hall S. J. Physical disturbance and marine benthic communities: life in unconsolidated sediments. Oceanography and Marine Biology. An Annual Review (1994) 32:179–239.
Hall S. J. The Effects of Fishing on Marine Ecosystems and Communities. (1999) Oxford: Blackwell Science. 274. pp.
Hiddink J. G., Hutton T., Jennings S., Kaiser M. J. Predicting the effects of area closures and fishing effort restrictions on the production, biomass, and species richness of benthic invertebrate communities. ICES Journal of Marine Science (2006) 63:822–830.
Holland D. S. Integrating spatial management measure into traditional management systems: the case of the Georges Bank multispecies fishery. ICES Journal of Marine Science (2003) 60:915–929.
Kaiser M. J., Clarke K. R., Hinz H., Austen M. C. V., Somerfield P. J., Karakassis I. Global analysis of response and recovery of benthic biota to fishing. Marine Ecology Progress Series (2006) 311:1–14.[CrossRef][Web of Science]
Kaiser M. J., de Groot S. J. The Effects of Fishing on Non-Target Species and Habitats: Biological, Conservation and Socio-Economic Issues. (2000) Oxford: Blackwell Science. 399.
Korres G., Lascaratos A. A one-way nested eddy resolving model of the Aegean and Levantine basins: implementation and climatological runs. Annales Geophysicae (2003) 21:205–220.
Lascaratos A., Nittis K. A high-resolution three-dimensional study of intermediate water formation in the Levantine Sea. Journal of Geophysical Research (1998) 103C:18497–18511.[CrossRef]
Mellor G. L., Ezer T. An equation of state for numerical models of oceans and estuaries. Journal of Atmospheric Oceanic Technology (1991) 8:609–611.[CrossRef]
Oey L. Y., Mellor G. L., Hires R. I. A three-dimensional simulation of the Hudson–Raritan estuary. 1. Description of the model and model simulations. Journal of Physical Oceanography (1985) 15, a. 1676–1692.[CrossRef][Web of Science]
Oey L. Y., Mellor G. L., Hires R. I. A three-dimensional simulation of the Hudson–Raritan estuary. 2. Comparison with observation. Journal of Physical Oceanography (1985) 15, b. 1693–1709.[CrossRef][Web of Science]
Pauly D., Christensen V., Walters C. Ecopath, Ecosim and Ecospace as tools for evaluating ecosystem impact of fisheries. ICES Journal of Marine Science (2000) 57:697–706.
Petihakis G., Triantafyllou G., Allen J. I., Hoteit I., Dounas C. Modelling the spatial and temporal variability of the Cretan Sea ecosystem. Journal of Marine Systems (2002) 36:173–196.[CrossRef][Web of Science]
Pinnegar J. K., Polunin N. V. C. Predicting indirect effects of fishing in Mediterranean rocky littoral communities using a dynamic simulation model. Ecological Modelling (2004) 172:249–267.[CrossRef][Web of Science]
Psarra S., Tselepides A., Ignatiades L. Primary productivity in the oligotrophic Cretan Sea (NE Mediterannean): seasonal and interannual variability. Progress in Oceanography (2000) 46:187–204.[CrossRef][Web of Science]
Psarra S., Tselepides A., Ignatiades L., Dafnomili E. Primary production estimates in the oligotrophic Cretan Sea. 5th Oceanographic Symposium, Kavala (1997) 1:225–228.
Robinson L. A., Frid C. L. J. Dynamic ecosystem models and the evaluation of ecosystem effects of fishing: can we make meaningful predictions? Aquatic Conservation: Marine and Freshwater Ecosystems (2003) 13:5–20.[CrossRef][Web of Science]
Ruardij P., van Raaphost W. Benthic nutrient regeneration in the ERSEM ecosystem model of the North Sea. Netherlands Journal of Sea Research (1995) 33:453–483.
Schratzberger M., Dinmore T. A., Jennings S. Impacts of trawling on the diversity, biomass and structure of meiofauna assembladges. Marine Biology (2002) 149:83–93.
Schratzberger M., Jennings S. Impacts of chronic trawling disturbance on meiofaunal communities. Marine Biology (2002) 141:991–1000.[CrossRef]
Shapiro R. Smoothing, filtering, and boundary effects. Reviews of Geophysics and Space Physics (1970) 8:359–387.[Web of Science]
Smith C. J., Papadopoulou K-N., Diliberto S. Impact of otter trawling on an eastern Mediterranena commercial trawl fishing ground. ICES Journal of Marine Science (2000) 57:1340–1351.
Thingstad T. F., Perez M., Pelegri S., Dolan J., Rassoulzadegan F. Trophic control of bacterial growth in microcosms containing a natural community from northwest Mediterranean surface waters. Aquatic Microbial Ecology (1999) 18:145–156.[CrossRef][Web of Science]
Thingstad T. F., Rassoulzadegan F. Nutrient limitations, microbial food webs, and "biological C-pumps": suggested interactions in a P-limited Mediterranean. Marine Ecology Progress Series (1995) 117:299–306.[CrossRef][Web of Science]
Triantafyllou G., Korres G., Hoteit I., Petihakis G., Banks A. C. Data assimilation of biophysical observations into the Eastern Mediterranean Biogeochemical Flux Model. Ocean Science (2007) 3:397–410.
Triantafyllou G., Korres G., Petihakis G., Pollani A., Lascaratos A. Assessing the phenomenology of the Cretan Sea shelf area using coupling modelling techniques. Annales Geophysicae (2003) 21, a. 237–250.
Triantafyllou G., Petihakis G., Allen I. J. Assessing the performance of the Cretan Sea ecosystem model with the use of high frequency M3A buoy data set. Annales Geophysicae (2003) 21, b. 365–375.
Triantafyllou G., Petihakis G., Dounas C., Koutsoubas D., Arvanitidis C., Eleftheriou A. Temporal variations in benthic communities and their response to physicochemical forcing: a numerical approach. ICES Journal of Marine Science (2000) 57:1507–1516.
Tselepides A., Polychronaki T. The CINCS project: introduction. Progress in Oceanography (2000) 46:85–88.[CrossRef][Web of Science]
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. In: Studies and Reviews GFCM, No. 74 (2004) Rome: FAO. 44 pp.
Turley C. M., Bianchi M., Christaki U., Conan P., Harris J. R. W., Psarra S., Ruddy G., et al. Relationship between primary producers and bacteria in an oligotrophic sea—the Mediterranean and biogeochemical implications. Marine Ecology Progress Series (2000) 193:11–18.[CrossRef][Web of Science]
Tusseau M. H., Lancelot C., Martin J-M., Tassin B. 1D coupled physical—biological model of north—western Mediterranean Sea. Deep-Sea Research II (1997) 44:851–880.[CrossRef]
van Wambeke F., Christaki U., Bianchi M., Psarra S., Tselepides A. Heterotrophic bacterial production in the Cretan Sea (NE Mediterannean). Progress in Oceanography (2000) 46:205–216.[CrossRef][Web of Science]
Walters C., Christensen V., Pauly D. Structuring dynamic models of exploited ecosystems from trophic mass-balance assessments. Reviews in Fish Biology and Fisheries (1997) 7:139–172.[CrossRef][Web of Science]
Zeller D., Reinert J. Modelling spatial closures and fishing effort restrictions in the Faroe Islands marine ecosystem. Ecological Modelling (2004) 172:403–420.[CrossRef][Web of Science]
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||