This article appeared in a journal published by Elsevier. The attachedcopy is furnished to the author for internal non-commercial researchand education use, including for instruction at the authors institution

and sharing with colleagues.

Other uses, including reproduction and distribution, or selling orlicensing copies, or posting to personal, institutional or third party

websites are prohibited.

In most cases authors are permitted to post their version of thearticle (e.g. in Word or Tex form) to their personal website orinstitutional repository. Authors requiring further information

regarding Elsevier’s archiving and manuscript policies areencouraged to visit:

http://www.elsevier.com/copyright

Author's personal copy

Shellfish-DEPOMOD: Modelling the biodeposition from suspended shellfishaquaculture and assessing benthic effects

Andrea M. Weise a,⁎, Chris J. Cromey b, Myriam D. Callier a,c,1, Philippe Archambault a,c,Jon Chamberlain d, Christopher W. McKindsey a,c

a Fisheries and Oceans Canada, Maurice Lamontagne Institute, 850 route de la mer, PO Box 1000, Mont-Joli (QC), Canada G5H 3Z4b Scottish Association for Marine Science, Dunstaffnage Marine Laboratory, Oban, Argyll, PA34 4AD, United Kingdomc Institut des Sciences de la Mer, Université du Québec à Rimouski, 310 allée des Ursulines, PO Box 3300, Rimouski (QC), Canada G5L 3A1d Fisheries and Oceans Canada, Institute of Ocean Sciences, 9860 West Saanich Road, Sidney (B.C.), Canada V8L 4B2

a b s t r a c ta r t i c l e i n f o

Article history:Received 1 August 2008Received in revised form 28 November 2008Accepted 1 December 2008

Keywords:ModellingShellfish aquacultureBiodepositionMusselsBenthic impactsMacrofaunal community structure

By predicting the dispersal of particulate aquaculture wastes around farm sites, numerical modelling canprovide an effective tool to assess the spatial extent of environmental effects. The present paper describeshow the aquaculture waste model DEPOMOD (Cromey, C.J., Nickell, T.D., Black, K.D. 2002a. DEPOMOD —

modelling the deposition and biological effects of waste solids from marine cage farms. Aquaculture 214,211–239.), originally developed for finfish aquaculture sites, was adapted and validated for suspendedshellfish aquaculture. Field data were collected for species-specific model input parameters (musselbiodeposition rates and particle settling velocities) and several finfish model parameters (farm representa-tion and calculation of aquaculture wastes) were adjusted for the shellfish scenario. Shellfish-DEPOMOD wastested at three coastal mussel Mytilus edulis farms with differing hydrodynamic regimes in Quebec, Canada.For each site, model predictions were compared to observed deposition measured in situ with sedimenttraps. Sedimentation rates under the three mussel culture sites were ca. two to five times those observed atcorresponding reference sites. Mussel biodeposits were predicted to accumulate within 30 m of the farms inthe shallow depositional sites while being dispersed more than 90 m in the deeper dispersive site. At thefarm site in Great-Entry Lagoon, model predictions agreed well with field data for the 0+ and 1+ musselcohorts when the maximum biodeposit production parameter was used. At the farm site in House-HarbourLagoon, model predictions did not agree with observed sedimentation rates, due most likely to theresuspension and advection of non farm-derived material and complex hydrodynamics. The model correctlypredicted the pattern of waste dispersal at the third farm site in Cascapedia Bay, although it underestimatedbiodeposition. Predicted fluxes may have been underestimated at this site because biodeposits frombiofouling communities were not included in the calculation of aquaculture wastes. The relationshipbetween modelled long-term biodeposition and benthic descriptors was assessed for the three farms.Alterations to the benthic community were observed at high biodeposition rates (N15 g m−2 d−1). At the mostdisturbed site, predicted fluxes were best correlated with the Infaunal Trophic Index (ITI) (R=−0.79,Pb0.001), followed by AZTI's marine disturbance index (AMBI) (R=0.64, Pb0.001). The potential applicationof Shellfish-DEPOMOD in terms of the management of shellfish aquaculture sites is discussed.

© 2008 Elsevier B.V. All rights reserved.

1. Introduction

Suspended shellfish culturemodifies pelagic-benthic energy fluxesand locally enhances the flux of organicmatter to bottom sediments incoastal ecosystems via filter-feeding and subsequent biodeposition offaeces and pseudofaeces, hereafter referred to as biodeposits, by theorganisms in culture. Increased sedimentation of organic matter

through biodeposition may lead to changes in sediment character-istics and benthic community composition (see review by Cranfordet al., 2006). Studies on the environmental impacts of shellfish aqua-culture have shown that these can range from little (Crawford et al.,2003; Danovaro et al., 2004), to slight (Baudinet et al., 1990; Grantet al., 1995), to severe (Dahlbäck and Gunnarsson, 1981; Stenton-Dozey et al., 2001). The degree of environmental impact is likelyrelated to both the site (background enrichment, sediment character-istics, and currents) and the husbandry practices (culture density anddepth) (Chamberlain et al., 2001; Hartstein and Stevens, 2005; Mironet al., 2005). The culture of shellfish is generally considered to haveless severe environmental effects than finfish aquaculture since

Aquaculture 288 (2009) 239–253

⁎ Corresponding author. Tel.: +1 418 775 0897; fax: +1 418 775 0718.E-mail address: andrea.weise@dfo-mpo.gc.ca (A.M. Weise).

1 Current address: University College Dublin, School of Biology and EnvironmentalScience, Marine Biology Ecology and Evolution (MARBEE), Belfield, Dublin 4, Ireland.

0044-8486/$ – see front matter © 2008 Elsevier B.V. All rights reserved.doi:10.1016/j.aquaculture.2008.12.001

Contents lists available at ScienceDirect

Aquaculture

j ourna l homepage: www.e lsev ie r.com/ locate /aqua-on l ine

Author's personal copy

shellfish are grown at comparatively lower biomass and no externalfeed is added. However, shellfish farms typically cover a much greaterarea than finfish farms and may thus result in very different dispersalpatterns. Moreover, the size of shellfish farms is continually increasingdue to improved technologies and industry consolidation. Sinceshellfish aquaculture is growing worldwide and farm sites are oftenlocated in shallow, low-energy coastal sites where waste material mayaccumulate, questions regarding the potential effect of such activitieson coastal ecosystems have arisen. There is thus a need for regulatorsto adequately assess the potential for environmental effects.

Numerical modelling provides an effective means to evaluate theinteractions between aquaculture activities and the ecosystem. Todate, modelling effort with regard to shellfish cultivation has focusedprimarily on predicting bivalve growth and production carryingcapacity rather than environmental interactions (see review byMcKindsey et al., 2006). The dynamics of biodeposition, including

biodeposit production rates and their potential for dispersal, are poorlyunderstood and are not well parameterized in manymodels. Althoughbiodeposits are included in a general benthic detritus compartment insomemodels (Bacher et al.,1995;Dowd, 2005;Grantet al., 2005), thesebox models are limited by their resolution and scale to accuratelypredict the area of benthic impact. Modelling the near-field effects ofshellfish aquaculture through biodeposition has received little atten-tion (Chamberlain et al., 2001; Hartstein and Stevens, 2005) and con-sequently there is a need for effective models to predict the organicflux from culture sites to the bottom (Henderson et al., 2001).

The potential “footprint” of shellfish farms, i.e. the spatial extent ofthe benthic impact, can be assessed with particulate waste dispersalmodels which may predict the dispersal of shellfish biodepositsthrough the water column (Hartstein and Stevens, 2005) as well astheir redistribution via resuspension on the sediment surface (Cromeyet al., 2002a,b; Giles 2006). The aquaculture waste model DEPOMOD

Fig.1. Schematic diagram showing the location of the threemussel farm sites: Great-Entry Lagoon (GE) and House-Harbour Lagoon (HH) in the Magdalen Islands, and Cascapedia Bay(CAS); acoustic Doppler current profiler deployments (⁎); and transect directions (black arrows). In GE, the farm was divided into 2 zones based on age classes: 0+ (less than 1 yearold) and 1+ (greater than 1 year old). The sampling sites for 2003 are indicated for: 0+ ( ), 1+ (●), and reference sites (○). The 1+ zonewas replaced by 0+mussels the following year.Replicate sediment traps were deployed along all transects and sediment cores were collected along the SW transect in GE, and NE transects in HH and CAS. See “Materials andmethods” for details. Only sections of interest weremodelled: a block of ninemussel backlines for each of the transects in GE and HH; a block of thirty mussel backlines for each of thethree transects in CAS.

240 A.M. Weise et al. / Aquaculture 288 (2009) 239–253

Author's personal copy

(Cromey et al., 2002a) was developed to predict the dispersal anddeposition of finfish aquaculture wastes and associated changes inbenthic communities. DEPOMOD has been validated for numerousfish farms in Scotland and has been applied across a range of differentscenarios in terms of environmental conditions, farm size, hydro-dynamic conditions and bathymetry (Cromey et al., 2002a). In Canada,the Department of Fisheries and Oceans Canada has been workingwith the industry and provincial agencies to test and validateDEPOMOD for use with British Columbia finfish farms (Chamberlainand Stucchi, 2007) and has shown an interest in adapting this modelfor use in the management of shellfish aquaculture sites (Chamberlainet al., 2006).

This paper presents the application of an amended version of theaquaculture waste model DEPOMOD, termed Shellfish-DEPOMOD, forsuspended mussel culture. Since there are key differences betweenmodellingfinfishand shellfish farms,weexamine themodel parametersthat require modification for the shellfish scenario, including musselbiodeposit production and settling rates, and farm representation.Shellfish-DEPOMOD was tested at three coastal mussel farms withdiffering hydrodynamic regimes in eastern Quebec, Canada. Theobjectives were to 1) predict the flux and dispersion of biodeposits,and 2) predict the benthic response using semi-empirical relationshipsbetween modelled fluxes and observed descriptors of macrobenthos.We examine the performance of the model through comparison ofpredicted and measured fluxes over 24-hour periods and indices ofbenthic condition over 1-month periods. To our knowledge, this is thefirst study to explicitly model predictions of shellfish aquaculture wastedispersal and benthic biological influence and validate model outputswith extensive field measurements.

2. Materials and methods

2.1. Farm sites

2.1.1. Great-Entry and House-Harbour Lagoons, Magdalen IslandsIn eastern Quebec, shellfish aquaculture activities have tradition-

ally taken place in the Magdalen Islands in Great-Entry Lagoon (GE)and House-Harbour Lagoon (HH) (Fig. 1). These lagoons are char-acterised by an average tidal range of 0.6 m (Canadian HydrographicService, 2008), receive no freshwater river run-off, and are covered byice during winter. Water depth ranges from 5 to 7 m at both farms.Mean currents are weak, e.g. typical speeds are 5 cm s−1, and occa-sionally increase to 10 cm s−1 during strong wind events in GE(Koutitonsky et al., 2002). Mussels are cultivated on suspended long-lines over a two-year grow-out cycle until mussels reach a commercialsize of 5 to 6 cm. The two farm sites studied have been in operationsince the 1980s and each produces about 180 t annually.

The farm in GE covers 2.5 km2 and is divided into two zones, onewith mussels in their first year of growth (0+ cohort) and the otherwith mussels in their second year of growth (1+ cohort), the latterbeing replaced by juveniles each fall following harvest. Each zonecontains 200, 91 m backlines, spaced 18 m apart, which are eitheroriented parallel or perpendicular to the shoreline (Fig. 1). Eachbackline supports 366 m of mussel sock as the mesh sleeves arelooped continuously along the backline (C. Eloquin, pers. comm.). Thefarm in HH covers 1.25 km2 and consists of 200, 76 m long backlines,spaced 12 m apart (Fig. 1) (Léonard, 2004). The actual length of ropeon a single HH backline is 244 m. Mussel lines are seeded at 575 and985 spat m−1 in GE and HH, respectively.

2.1.2. Cascapedia BayIn contrast to the Magdalen Islands, the mussel culture site in

Cascapedia Bay (CAS), Baie des Chaleurs, is located in a more exposedand energetic environment (Fig. 1). Tides are mixed and semi-diurnalwith a mean tidal range of 1.9 m (Canadian Hydrographic Service,2008). The freshwater discharge from the nearby Grande and Petite

Cascapedia rivers is ca. 70m3 s−1 during the summer (Bonardelli et al.,1993). Typical current velocities off Carleton are 9.8±4.9 cm s−1

(Bonardelli et al., 1993).The 1.4 km2 mussel culture site is located ca 3.5 km offshore in

20 m deep water. A total of 150 backlines, measuring 142 m long andspaced 30 m apart, are oriented parallel to the shoreline (Fig. 1). Eachbackline supports 1100 m of mussel line, seeded at a density of 820mussels m−1 (R. Allard, pers. comm.). Mussels are cultured in a three-year grow-out cycle and cohorts are rather arbitrarily distributedwithin the farm site.

2.2. Hydrographic measurements

Acoustic Doppler current profilers (ADCP, SonTek) were used tocollect current data at the three study sites between 2003 and 2005(Fig. 1). Deployment length was designed to capture typical conditionsfor the season of study. Two upward-looking ADCPs (500 kHz and1500 kHz) were deployed in GE from June to October 2003. Currentflow was measured in pulse-coherent mode in 20, 0.25 m thick cellsfrom 0.6 to 5.6 m above the sediment bottom. Measurements wereaveraged over 2 min at 20-minute intervals. From June to August2004, the 1500 kHz and 500 kHz ADCPs were programmed asdescribed above and deployed in GE and HH, respectively. In CAS, the500 kHz ADCP was programmed to measure currents in 25, 1 m thickcells from 1.4 to 25m above the sediment bottom. The instrument wasdeployed from July to August 2005.

Themussel farms inGE andHHwere subject to relativelyweakflowsduring the summermonths, with average speeds less than 6 cm s−1 andmaximumflows less than13 cms−1 (Table 1). In contrast, current speedsin CAS were at least double these values, with mean and maximumspeeds of 11.8±8.0 cm s−1 and 52.3 cm s−1, respectively.

2.3. Model description

The model consists of four modules including grid generation,particle tracking, resuspension, and benthic impact (Cromey et al.,2002a). First, a grid is generated detailing bathymetry of the area,positionsof the cages aswell as samplingstations. Particles, representingthewastematerial, are advected in themodelbasedon settlingvelocitiesof different particle groups and observed current velocities. The outputof the particle tracking module describes the initial deposition ofparticles in terms of a contour plot or “footprint” of the farm. Thedeposited material can be further advected through resuspensionprocesses in the resuspension module. Finally, the benthic impactmodule uses empirical relationships between the predicted accumula-tion of aquaculture wastes and observed changes in the benthiccommunity. The following model parameters required modificationfor the shellfish scenario: 1) farm structure configuration to representsuspended mussel lines rather than finfish cages, 2) aquaculture wastevalues as no “additional” feed is added to the system, and 3) sinking

Table 1Hydrographic data measured during a summer month (July) for Great-Entry Lagoon(GE) and House-Harbour Lagoon (HH) (depositional sites), and Cascapedia Bay (CAS)(dispersive site)

Site Period(1 month)

Heightabovebed (m)

Averagespeed(cm s−1)

Maximumspeed(cm s-1)

Speed≥9.5 cm s−1

(%)

Residualspeed(cm s−1)

Residualdirection(°)

GE 1+ July 2003 1 5.9±2.6 13.2 8 3.6 193GE 0+ July 2003 2 2.9±1.5 8.8 0 1.5 292GE 0+ July 2004 2 3.8±2.0 9.3 0 1.9 25HH 0+ July 2004 2 3.7±1.7 8.9 0 1.3 148CAS July 2005 10 11.8±8.0 52.3 53 4.3 237

For brevity, data are not presented for all the current layers that were measured nor forthe entire periods that the instruments were deployed. The abbreviations 0+ (less than1 year old) and 1+ (greater than 1 year old) refer to the age classes of the mussels. Thecurrent speed of 9.5 cm s−1 refers to the critical threshold for resuspension.

241A.M. Weise et al. / Aquaculture 288 (2009) 239–253

Author's personal copy

velocities of mussel biodeposits. Model input parameterswere based onin situmeasurements of biodeposit production, biodeposit sinking rates,and hydrographic data. The above-mentioned parameters are discussedin detail below.

2.3.1. Farm representationIn contrast to the original model, Shellfish-DEPOMOD offers a

larger grid domain of 500×500 cells. This is useful for modellingshellfish farms as these cover a much more extensive area (km) thanfinfish cages (m).

For the shellfish scenario, a “cage”mayconsistof thewhole farmorofindividual mussel backlines. To obtain the best resolution possible, i.e.scale of meters rather than hundreds of meters, individual musselbacklinesweremodelled rather than thewhole farm site as a single box.Modelling individual backlines also allowed us to assign differentbiodeposit production rates for each of the mussel cohorts within the

same farm site. We focused on specific sections of each farm whichallowed us to 1) use a finer grid cell resolution, 2) optimize calculationtimes, 3) focus on areas where mussels are present (i.e. have not beenharvested), and 4) validate the model in the section where sedimenttraps were deployed and sediment cores collected. For GE, a499 m×499 mmodel grid with a fine cell resolution of 1 mwas createdand sections with nine mussel backlines were modelled (Table 2). Forthe deeper site in CAS, the dispersion of mussel biodeposits wasexpected to be greater and thus a grid cell resolution of 3 mwas chosenwithin a 1497m×1497m grid. Two sections, eachwith thirty backlines,were modelled. The backlines are described in the model domain as arectangular boxwith awidth equal to the diameter of themussel sleeve,a length equal to that of the backline, and a depth equal to the totalheight of thewater column occupied by the loopedmussel sleeve. Sincethemodel assumes that “cages”, i.e. in this case backlines, are positionedat the water surface, the height of the water column above the sub-surface mussel lines was subtracted from the total depth modelled.Particles could thus not be transported above the top of mussel lines viavertical dispersion, a sound assumption since particles are negativelybuoyant. Given that the mussel backlines are anchored to the sediment

Table 4Biodeposit production measured in situ for 2 mussel cohorts (0+ and 1+) in Great EntryLagoon (GE) during 3 trial dates and for 2 mussel cohorts (1+ and 2+) in Cascapedia Bay(CAS) during 2 trial dates

Site Trial date Mussel size Biodeposit production

(cm) (mg mussel−1 d−1)

GE 0+ Aug 14–15 4.0±1.1 24–32Aug 18–19 4.5±0.3 25–75Aug 21–22 5.2±0.3 13–21

GE 1+ Aug 14–15 6.9±0.2 32–52Aug 18–19 6.7±0.2 65–126Aug 21–22 6.7±0.3 17–33

CAS 1+ July 6–7 5.7±0.3 29–58July 9–10 5.5±0.3 15–32

CAS 2+ July 6–7 6.7±0.2 45–95July 9–10 6.6±0.4 29–39

Average mussel shell length (cm), minimum andmaximum biodeposit production rates(mgmussel−1 d−1) are given for eachmussel cohort. Biodepositionwas calculated as theamount of material collected in sediment traps with mussels minus the averagesedimentation obtained in the corresponding shell controls. Each treatment had 3replicates on each trial date.Data from the GE site modified from Callier et al. (2006).

Table 3Literature review of several key parameters required for the model

Parameter Comment Reference

Biodeposit production18 to 114 mg DW g−1 tissue d−1 Field measurements for Mytilus edulis Callier et al. (2006)1 to 50 kg farm−1 d−1 Applied estimates for M. edulis Chamberlain (2002)b80 mg DW g−1 tissue d−1 Field measurements for M. edulis (estimated from figure) Cranford and Hill (1999)8±3 pellets mussel−1 h−1 (lab); 399±226 pellets m−2 h−1 (field) Laboratory and field measurements for Perna canaliculus faecal pellets Giles (2006)1 to 16 mg DW g mussel−1 (incl. shells) d−1 Field measurements for M. edulis Kautsky and Evans (1987)2 to 20 mg ash−free DW g DW mussel−1 d−1 Laboratory measurements for M. edulis Tenore and Dunstan (1973)13 to 126 mg DW mussel−1 d−1 Field measurements for Mytilus edulis This study

Biodeposit settling velocities0.3 to 1.8 cm s−1 Laboratory measurements for M. edulis biodeposits collected in situ Callier et al. (2006)0.1 to 1.8 cm s−1 Laboratory measurements for M. edulis biodeposits; mussels fed different diets Chamberlain (2002)0.1 to 4.5 cm s−1 Laboratory measurements for P. canaliculus biodeposits; mussels fed different diets Giles and Pilditch (2004)3.0±0.4 cm/s Laboratory measurements for P. canaliculus biodeposits Hartstein and Stevens (2005)

Critical erosion thresholdUz200=9.5 cm s−1 Finfish wastes Cromey et al. (2002b)u⁎=0.41 cm s−1

Uz10=12.3 to 19.6 cm s−1 P. canaliculus biodeposits (mussels fed a natural diet) Giles and Pilditch (2004)u⁎=0.36 to 0.67 cm s−1

Uz150=18 cm s−1 Resuspendable “fluff” layer in oyster grounds Van Raaphorst et al. (1998)u⁎=0.8 cm s−1

u⁎=1.70 to 1.77 cm s−1 Sediment resuspension around mussel farm Walker and Grant (in press)Uz10=20 to 25 cm s−1 Resuspendable seston in M. edulis beds Widdows et al. (1998)

Table 2Key input parameters for the grid generation and particle tracking modules for themussel farms (0+, 1+, and mixed mussel cohorts) in Great-Entry Lagoon (GE) andHouse-Harbour Lagoon (HH), and Cascapedia Bay (CAS)

Input parameters GE 1+(2003)

GE 0+(2003)

GE 0+(2004)

HH 0+(2004)

CAS(2005)

Grid size (m) 499×499 499×499 499×499 499×499 1497×1497Grid cell

resolution (m)1×1 1×1 1×1 1×1 3×3

Backlinedimensions(l×w×d) (m)

91×0.2×1 91×0.2×2 91×0.2×2 76×0.2×2.4 142×0.2×5.5

Depth below topof backlines (m)

1.5 3.5 3.5 3.5 12

Height of currentmeters abovebed (m)

1 1, 2, 3 1, 2, 3 1, 2, 3 2, 4, 6, 8, 10

Biodepositproduction(kg backline−1 d−1)

26.4 15.8 15.8 18.0 52.8 (1+),86.5 (2+)

Faeces settlingvelocity (µ, σ)(cm s−1)

1.0±0.3 0.8±0.3 0.8±0.3 0.8±0.3 0.8±0.1

Pseudofaecessettlingvelocity (µ, σ)(cm s−1)

0.2±0.02 0.2±0.02 0.2±0.02 0.2±0.02 NA

242 A.M. Weise et al. / Aquaculture 288 (2009) 239–253

Author's personal copy

bottom, the water depth was not adjusted for tidal amplitude as is thecase for finfish cages and shellfish rafts, which float at thewater surface.

2.3.2. Biodeposit productionThe model requires feed input and wastage rates to calculate the

quantity of aquaculture wastes that are produced. However, in contrastto finfish aquaculture, no feed is added to shellfish farms. We thereforechose to model the quantity of biodeposits produced by the mussels byinputting this value in the model as “feed input”with zero digestibility,which results in the ejection of all particles as biodeposits. Few studiesexist on shellfish biodeposit production and the few rates that havebeen reported are mainly derived from small numbers of mussels in

laboratory settings (Table 3). Here, the “feed input”, hereafter referred toas biodeposit production parameter, was based on the in situ biodepositproductionmeasured formussels in theMagdalen Islands (see details inCallier et al., 2006) and on field measurements made in the presentstudy for mussels in CAS. Briefly, biodeposition by the different musselcohorts was measured in situ by placing a fixed number of musselswithin cylindrical vexar cages fitted into the top of sediment trapsmadeof PVC tubing (10.2 cm diameter, 76.2 cm height). The number ofmussels used ensured that ca. 2/3 of the cage areawas covered by a layerof mussels. Sediment traps containing dead mussels were used ascontrols to measure background sedimentation rates. Each treatmenthad three replicates and was repeated on three trial dates in the

Fig. 2. a) Modelled biodeposition (g m−2 d−1) over a 24-hour period for a 1+ mussel cohort in Great-Entry Lagoon (GE) during August 13–14, 2003 with the resuspension moduleturned off. Biodeposition rates were modelled for nine mussel lines near each transect. Mussel lines are represented by white lines and sediment trap positions (■) are shown for thethree transects. b) Comparison between modelled and observed sedimentation rates. The dashed line represents the background sedimentation (19±5 g m−2 d−1) which was addedto model predictions to allow for comparisons.

243A.M. Weise et al. / Aquaculture 288 (2009) 239–253

Author's personal copy

Magdalen Islands (GE) and on two trial dates in CAS. Sediment trapswere deployed outside the mussel farm in GE and hung on emptybacklines in CAS. Sediment traps were retrieved after 24 h and thecontents filtered through pre-burned and pre-weighed glassfiber filters(Whatman GF/F, 0.7 µm). Filters were rinsed with ammonium formate,dried to constant weight, and weighed. Biodepositionwas calculated asthe amount of material collected in the sediment traps with musselsminus the average sedimentation obtained in the corresponding shellcontrols, and expressed as biodeposit production per individual mussel(mg mussel−1 d−1).

The biodeposit production (kg backline−1 d−1)was calculated for eachmussel backline by multiplying the biodeposit production per mussel bythe estimated number of mussels per backline. In GE, biodepositproduction ranged from 13 to 75, and from 17 to 126 mg mussel−1 d−1

for 0+ and 1+ mussel cohorts, respectively (Table 4). Biodepositproduction ranged from 15 to 58, and from 29 to 95 mg mussel−1 d−1

for 1+ and 2+ mussel cohorts, respectively, in CAS. Given the range ofvalues that were measured, we tested the effect of this parameter onmodel outputs by using the minimum, mean, and maximum values.

2.3.3. Particle characteristicsData on biodeposit sinking velocities is relatively scarce and has

only recently been addressed (Table 3). The sinking velocities ofmussel M. edulis faecal pellets were therefore measured in GE (seedetails in Callier et al., 2006). Briefly, fresh biodeposits produced byfive different size classes of mussels were collected in situ in sedimenttraps. The length and diameter of individual faecal pellets weremeasured, transferred to a settling column, and their sinking velocitymeasured. We thus assigned settling velocities of 0.8±0.3 and 1.0±0.3 cm s−1 for faecal pellets produced by 0+ and 1+ cohorts in GE andHH, respectively (Table 2). For pseudofaeces, we assigned a settlingvelocity of 0.2±0.02 cm s−1 based on personal observations (data notshown). Similarly, a faecal particle settling velocity of 0.8±0.1 cm s−1

was applied for the CAS modelling scenarios based on personalobservations (data not shown).

We assigned 67% of the filtered material rejected as faecal pelletsand 33% as pseudofaeces for the GE and HHmodelling scenarios basedon thework byWiddows et al. (1979) and Bayne et al. (1993).Widdowset al. (1979) observed thatM. edulis individuals produce pseudofaeceswhen seston concentrations exceed ca. 5mg l−1 and Bayne et al. (1993)noted that 40% of particles are rejected as pseudofaeces when sestonconcentrations range between 6 and 10 mg l−1. The average sestonconcentration measured south-east of the mussel culture site duringAugust 2003 inGEwas8.4±3.2mg l−1 (Callier et al., 2006). Althoughnoseston data was available for Cascapedia bay, previous studiesmeasured low concentrations, averaging 2.6±1.2 mg l−1 (Tamigneauxand Thomas, 2004). No pseudofaeces were modelled for CAS. Particleswere assigned random starting positions and were released con-tinuously throughout the modelling runs. To avoid model patchinessreported in a sensitivity analysis by Cromey et al. (2002a), a highnumber of particles (N5×106), a suitable particle trajectory evaluation(every 6 s), and a fine grid cell resolution (1 m or 3 m) were used inthe present study. Particle numbers were optimised by increasingthe number of particles until no further changes in depositionfootprints were detected. Since site-specific dispersion coefficientswere not available, the horizontal and vertical dispersion coefficients(kx, ky=0.1 m2 s−1, kz=0.001 m2 s−1) recommended by Gillibrand andTurrell (1997) were applied.

2.3.4. Resuspension parametersResuspension processes are complex and a range of critical erosion

thresholds have been reported in the literature (Table 3). InDEPOMOD, finfish wastes are resuspended at near-bottom (z≈2 m)current speeds of 9.5 cm s−1 (Cromey et al., 2002b). To assess theimportance of the resuspension, we compared predictions with theresuspension model enabled and disabled.

2.4. Model validation

2.4.1. Particle tracking model validationSediment trap studies were designed to validate the particle

tracking model by comparing predicted and observed biodeposition(g m−2 day−1) at the mussel culture sites over 24-hour periods. Thediver-deployed sediment traps were made from PVC tubing (50 cmheight, 5 cm diameter), giving an aspect ratio of 10:1 recommended bymany authors (Nodder and Alexander, 1999; Gust and Kozerski, 2000).The traps fit into basesmade of flat steel crosses with a plastic pipe capto allow for easy deployment and retrieval. In 2003, sediment trapswere set up along three transects extending away from the 1+ musselfarm towards the SW, SE, and NW in GE (Fig. 1). Paired sediment traps,separated by 4 m, were positioned at distances of 0, 3, 6, 9, 12, 15 and30 m along transects placed perpendicular to the edge of the musselfarm and to the mussel lines themselves. Three transects were againpositioned around the GE farm the following year (2004) with a 0+cohort replacing the 1+ cohort (Fig. 1). Paired sediment traps wereplaced at 0, 3, 6, 9, 15, and 30 m and additional traps were placed500 m SW, SE, and NW from the farm. Only two transects could bedeployed in HH due to the configuration of the site, one towards theNE and another towards the NW, with reference stations at 300 m tomaintain a similar water depth (Fig. 1). We examined the 0+ mussellines because the 1+ mussels had already been harvested at the timeof the study. In CAS, paired sediment traps were positioned atdistances of 0, 10, 20, 40, 60 and 90 m along transects towards the SW,SE, and NE (Fig. 1). Paired sediment traps were also placed beneath amussel line within the farm site (−40 m) and at four reference siteslocated approximately 500 m NW, SE, and NE as well as 1000 m NE ofthe farm site. During a second survey, additional replicate traps wereset between lines within the farm site. For all sites, sediment trapswere retrieved by scuba divers and the contents filtered on pre-weighed and pre-burned glassfiber filters (Whatman GF/F, 0.7 um) toobtain the mass of total solids. The % organic matter (OM) in the

Fig. 3. Scatterplots of measured current velocity (cm s−1) and direction during the24-hour sediment trap validation studies at themussel farms in a) Great-Entry Lagoon 1mabove seabed (Aug 13–14, 2003), b) Great-Entry Lagoon 2 m above seabed (July 26–27,2004), c) House-Harbour Lagoon 2m above seabed (June 27–28, 2004), and d) CascapediaBay 10 m above seabed (July 10–11, 2005). For brevity, data are not presented for all thecurrent layers that were used in the modelling runs.

244 A.M. Weise et al. / Aquaculture 288 (2009) 239–253

Author's personal copy

sedimented material was calculated as the weight loss of driedmaterial combusted at 450 °C for 5 h (Byers et al., 1978). Results wereexpressed in g m−2 d−1 to compare with model outputs.

Model performance was evaluated for each mussel farm byregressing observed vs. predicted fluxes (Piñeiro et al., 2008). To thisend, biodeposition rates were extracted from the model grid at eachsediment trap location. To adequately compare predictedwith observedsedimentation due only to mussel culture (i.e., biodeposition), theaveraged farm- and date-specific background sedimentation ratesmeasured at reference stations were subtracted from the sedimenttrap data. Only non-zero biodeposition rates (N0.5 g m−2 d−1) wereincluded in the analyses and two extreme outliers (see Fig. 4b) wereexcluded.

2.4.2. Benthic response model validationResults from extensive benthic surveys undertaken by Callier et al.

(2006, 2007, 2008) at the farm sites in GE and HH during the summers2003 and 2004 as part of a larger study of aquaculture impacts in theMagdalen Islands were used and compared with modelled predictions.Full methods and results for benthic sampling are given in Callier et al.(2007, 2008). In GE (2003), sediment cores were collected by scuba-divers in4 randomlychosen sites in eachof 2 farmzones (0+ and1+) anda reference zone (N500 m from farm). Samples were taken directly at 2positions in each site, under mussel lines (e.g. 0+under) and betweenmussel lines (e.g. 0+between). At each position, 3 replicateswere collectedusing a wedge corer (sampled area of 263 cm2 to a depth of ca. 15 cm)(Wildish et al., 2003). In GE and HH (2004), sediment samples were

Fig. 4. a) Modelled biodeposition (g m−2 d−1) over a 24-hour period for a 0+ mussel cohort in Great-Entry Lagoon (GE) during July 26–27, 2004 with the resuspension module turnedoff. Biodeposition rates were modelled for nine mussel lines near each transect. Mussel lines are represented by white lines and sediment trap positions (■) are shown for the threetransects. b) Comparison between modelled and observed sedimentation rates. The dashed line represents the background sedimentation (29±3 g m−2 d−1) which was added tomodel predictions to allow for comparisons.

245A.M. Weise et al. / Aquaculture 288 (2009) 239–253

Author's personal copy

collected at 7 stations located along a transect leading from the outer-most mussel lines of the 0+ portion of the Magdalen Islands musselfarms and in the direction of the main water current direction (i.e.towards the NE in HH and towards the SW in GE). Samples were thuscollected at 0, 3, 6, 9,15, 30m, and at a control station situated at 300 mand 500 m in HH and GE, respectively. Five replicate samples werecollected at each sampling station using sediment cores made of PVCtubing (sampled area of 79 cm2 to a depth of ca. 15 cm). In CAS (2005),five replicates were also sampled at each station along a transectextending towards the NE at distances of −40, 0, 10, 20, 40, 60, 90, 500,and 1000 m from the NE edge of the farm. All sediment samples weresieved through a 500 µmmesh and the retainedmaterial preserved in a5% formaldehyde-saline solution until further analysis. Identificationwas made to the lowest taxonomic level possible. Three sediment coresamples of the 0–2 cm surface layer were also taken at each samplingstation with a cut-off 10 ml syringe. The percent organic matter (% OM)in sediments was calculated as the weight loss of dried materialcombusted at 450 °C for 5 h (Byers et al., 1978).

Sediment core samples were characterized in terms of number ofspecies (S), total abundance (N), Margalef's species richness (d),Shannon Wiener diversity (H'), and Pielou's eveness (J'), which werecalculated using PRIMER (Clarke and Warwick, 1994). To avoid biasedindices, bivalve juveniles were not included in the calculations sincetheir presence did not reflect the sediment's condition but rather localhydrodynamic conditions. The tolerance of the benthic communities toorganic enrichmentwas evaluated using the Infaunal Trophic Index (ITI)(Word, 1978) and AZTI's Marine Biotic Index (AMBI) (Borja et al., 2000).The ITI describes the dominance of the benthic fauna by dividing theminto trophic groups (for details, see Mearns and Word, 1982). ITI scoresN50 indicate little effect on benthic conditions, 25–50 indicates an

enriched environment, and scores b25 indicate a degraded state. AMBIclassifies benthic species into ecological groups based on the sensitivity/tolerance of infauna to pollution (for details, see Borja et al., 2000, 2003).AMBI scores b1.2 indicate undisturbed, 1.2–3.3 slightly disturbed, 3.3–5.0 moderately disturbed, 5.0–6.0 heavily disturbed, and N6 extremelydisturbed conditions. Variations in indiceswere evaluated using ANOVAfollowed by Tukey multiple comparison tests with SYSTAT.

Modelledfluxeswere based onmeasured currents of themonth priorto benthic sampling. Since the sediment cores were collected near thesediment trap positions, biodeposition rates were predicted for thesepositions, i.e. 2 positions per distance along each transect. Biodepositionbeneath and between mussel backlines was also modelled for both theGE 0+ and 1+ cohorts in 2003 since benthic data was available fromCallier et al. (2007). A sectionof 9backlines, orientedparallel to theshore,wasmodelled for the 0+ cohort. Since 1+ backlineswere positioned bothparallel and perpendicular to shore (Fig. 1), two sections of 9 backlineswere modelled. Biodeposition rates were extracted beneath all lines(9 lines×2 replicates) and between lines (6 positions×2 replicates) sincesites were randomly chosen within each zone for the benthic cores.Semi-empirical relationships between the average biodeposition ratesand the observed benthic community descriptors described above wereevaluated using Pearson's correlation.

3. Results

3.1. Modelling biodeposit dispersion

Analysis of model outputs showed that scenarios based on theminimum or average biodeposit production resulted in the modelunder-predicting fluxes relative to measured sedimentation rates.

Fig. 5. a) Modelled biodeposition (g m−2 d−1) over a 24-hour period for a 0+ mussel cohort in House-Harbour Lagoon (HH) during June 27–28, 2004 with the resuspension moduleturned off. Biodeposition rates were modelled for nine mussel lines near each transect. Mussel lines are represented by white lines and sediment trap positions (■) are shown for thetwo transects. b) Comparison betweenmodelled and observed sedimentation rates. The dashed line represents the background sedimentation (31±4 g m−2 d−1) which was added tomodel predictions to allow for comparisons.

246 A.M. Weise et al. / Aquaculture 288 (2009) 239–253

Author's personal copy

Conversely, model predictions more accurately reflected observedfluxes when the maximum biodeposit production was used and thusthose were used in subsequent modelling runs. Issues with respect tothe biodeposit production parameter model settings are addressed inthe discussion.

3.1.1. Great Entry Lagoon (1+ mussel cohort)With the model parameter settings described in Table 2, the model

predicted high biodeposition rates, ca. 35 g m−2 d−1, directly beneaththe 1+ mussel lines and a strong deposition gradient with increasingdistance from the lines (Fig. 2a). Biodeposits were predicted todisperse furthest towards the SW and to a lesser extent towards theNW and SE, a reflection of the dominant currents during the 24-hour-validation study (Fig. 3a).

Predicted and observed sedimentation rates are shown in Fig. 2b.The background sedimentation rate was estimated at 19±5 g m−2 d−1

based on data collected from a total of 60 sediment traps deployed atfour reference sites on four dates between June and August 2003(Callier et al., 2006). High sedimentation rates, ca. 3 times greater thanthe background value, were measured directly beneath the 1+ mussellines with 58±3 g m−2 d−1 (SW transect), 43±4 g m−2 d−1 (NWtransect), and 66±25 g m−2 d−1 (SE transect), similar to the valuespredicted by the model (Fig. 2b). Sedimentation rates decreased withincreasing distance from the farm site edge. The model reproducedthis high to low gradient as well as the general dispersal pattern.Along the SW transect, the model reproduced the strong depositiongradient and predicted the dispersion of biodeposits to ca. 15 m fromthe lines (Fig. 2b). Biodeposits were predicted to disperse to a lesser

Fig. 6. a) Modelled biodeposition (g m−2 d−1) over a 24-hour period for mixed mussel cohorts in Cascapedia Bay (CAS) during July 10–11, 2005 with the resuspension module turnedoff. Biodeposition rates weremodelled for thirty backlines in the SE zone (NE transect) and SW zone of the lease (SE and SW transects). Dashed, grey, and white lines represent emptymussel lines, 1+ mussel cohorts, and 2+ mussel cohorts, respectively. Sediment trap positions (■) are shown for the three transects. b) Comparison between modelled and observedsedimentation rates. The dashed line represents the background sedimentation (65±6 g m−2 d−1) which was added to model predictions to allow for comparisons.

247A.M. Weise et al. / Aquaculture 288 (2009) 239–253

Author's personal copy

extent towards the NW and SE, which corresponds to the measuredfluxes.

3.1.2. Great Entry Lagoon (0+ mussel cohort)The same farm site was modelled for a 0+ mussel cohort the

following summer (Table 2). The predicted biodeposition rates directlybeneath the mussel lines were approximately one-fourth of thosemodelled for the 1+ cohort (2003), mainly due to the lower biodepositproduction by the0+mussel cohort (Fig. 4a). Biodepositswerepredictedto disperse mostly towards the NE, the direction of the dominantcurrents during the 24-hour-validation study (Fig. 3b). For the mussellines oriented in the same direction as the dominant current, i.e. alongthe SW-NE axis, the model predicted highest sedimentation ratesdirectly beneath the mussel lines (ca. 10 g m−2 d−1) (Fig. 4a).

Comparisons between predicted and observed sedimentation ratesat the 0+ mussel farm are given in Fig. 4b. The background sedi-mentation rate was estimated at 29±3 g m−2 d−1, based on measuredfluxes at the three reference stations located 500 m from the farm. Aspredicted by the model, observed fluxes were generally weak alongthe three transects (Fig. 4b). A very high sedimentation rate of 111.2±5.9 g m−2 d−1, almost six times greater than background values, wasmeasured directly beneath the mussel line of the SE transect and wasnot predicted by the model.

3.1.3. House-Harbour Lagoon (0+ mussel cohort)The model predicted that biodeposits would be dispersed mainly

towards the SW (Fig. 5a) based on the model parameter settings(Table 2) and the dominant SW currents (Fig. 3c). However, themodelled predictions did not reflect themeasuredfluxes (Fig. 5b). Highsedimentation rates were measured along the NE transect, averaging98.9±6.0 g m−2 d−1 from 0m to 30 m, while the background sedimen-tation rate was estimated at 31.5±3.5 g m−2 d−1 based on the tworeference stations located 500 m distant from the farm. Very highsedimentation rates were observed at all distances except for thecontrol site in the NE transect. No sedimentation patternwas observedin the sampling stations located along the NW transect except for ahigh flux of 75.7±36.0 g m−2 d−1 directly beneath mussel lines and anaverage of 34.0±1.6 g m−2 d−1 for the remaining stations along thistransect. A second 24-hour survey conducted one month later in thesame position also showed no difference in sedimentation rates withincreasing distance from the farm, i.e. sedimentation rates averaged39.8±13.4 g m−2 d−1 and 29.4±4.2 g m−2 d−1 along the NE and NWtransects, respectively (data not shown). During this second survey,background sedimentation rates were of 24.1±4.9 g m−2 d−1 at thereference stations located 100 m and 500 m northeast and northwestof the farm site.

3.1.4. Cascapedia Bay (mixed mussel cohorts)Current speeds in CAS were ca. twice those in GE and HH during the

24-hour-validation study (Fig. 3d), resulting in the dispersion of bio-deposits as far as 100 m NE from the edge of the mussel farm (Fig. 6a).Becausemussel cohortswere arbitrarily distributedwithin the CAS farmsite, backlineswith differing biodeposit production ratesweremodelledwithin a same run (Table 1). The importance of describing individualbacklines in the model domain is seen by the resulting footprint withhighest biodeposition rates of ca. 15 g m−2 d−1 beneath the 2+ musselcohorts (Fig. 6a).

Two 24-hour-validation studieswere conducted but only results fromthe second survey are shown since results were similar. The backgroundsedimentation rates were much greater in CAS than in the MagdalenIslands, averaging 60±11 g m−2 d−1 (July 8–9) and 65±6 g m−2 d−1

(July 10–11) at the four reference stations located400mand1000m fromthe farm. Sedimentation rateswere ca.1.5 times greaterwithin and at theedge of the farm, averaging 103±12 g m−2 d−1 and 98±6 g m−2 d−1,respectively (Fig. 6b). Biodeposits were mainly dispersed towards theNE, generally decreasingwith increasing distance from the farm sitewith

85±9 gm−2 d−190m from the farm and reaching background values at adistance of 400 m. Although predicting the general pattern of wastedispersal, the modelled fluxes were significantly lower than observedfluxes along the NE transect. The model predicted little dispersiontowards the SE and SW, which was reflected by the sediment trap data(Fig. 6b).

3.2. Model performance

Fig. 7 shows a fair relationship between predicted and observedbiodeposition for both mussel cohorts in GE (n=58) when adjusted toaccount for date-specific background sedimentation (i.e., 19±5 and 29±3 g m−2 d−1 were subtracted from 2003 and 2004 measured sedi-mentation data, respectively). The performance of the particle trackingmodel was poor for the farm in HH since other processes, e.g. theresuspension and advection of non-farm derived material, likelyconfounded the biodeposition rates from the mussel farm. For the CASfarm, biodeposition rates were only about 50% greater than backgroundsedimentationwithin the farm,making it difficult to detect a correlationbetween observed and predicted fluxes.

3.3. Benthic response

Longer-term biodeposition, occurring over a one-month period,was modelled and compared to benthic community and sedimentcharacteristics at each farm site. Themodel predicted a highly localizedeffect for the 1+ cohort in GEwith high deposition directly beneath themussel lines (27.2±5.0 g m−2 d−1) and ca. 4 times less betweenindividual mussel lines (6.2±1.0 g m−2 d−1) (Table 5). Moderatebiodepositionwas predicted for the 0+ cohort, averaging 9.1±0.9 gm−2

d−1 and 8.2±0.4 g m−2 d−1, under and between mussel lines,respectively. Biodeposition beneath 0+ mussel lines was predicted tobe twice as great in HH as compared to GE with predictedbiodeposition rates of 13.4±1.0 and 6.3±0.3 g m−2 d−1, respectively,in 2004 (Table 5). Disabling the resuspension module had little effecton the resulting model predictions for GE and HH since current speedsdid not exceed the critical erosion threshold for the 0+ cohorts andonly 8% of the time for the 1+ cohort in GE (Table 1) resulting in verylittle waste material being exported from the model grid. Conversely,current speeds exceeded the threshold 53% of the time in CAS and thusincluding the resuspension processes had a significant effect on theresulting model predictions for that site (Table 1). However, predictedbiodeposition within the farm was weak regardless of whether theresuspension module was enabled or disabled (b1 and b5 g m−2 d−1,respectively).

Benthic communities differed between and within farms. Allmeasures of diversity (S, H', J') were reduced in the 1+ zone in GE in2003 where predicted biodeposition was the greatest (Table 5). The

Fig. 7. Comparison between observed and predicted biodeposition (g m−2 d−1) for the1+ (■) and 0+ (□) mussel cohorts in 2003 and 2004 in Great-Entry Lagoon.

248 A.M. Weise et al. / Aquaculture 288 (2009) 239–253

Author's personal copy

dominance of the polychaete Capitella capitata in the 1+ zone,especially under the mussel lines, contributed to the low diversity.In contrast, S and H' were greatest in the 0+ zone which wasdominated by the polychaetes Pectinaria granulata, Polydora ciliataand Heteromastus filiformis and the gastropod Hydrobia minuta (seedetails in Callier et al., 2007, 2008). None of the biological indicesdiffered among distances along the transect of the 0+ cohort in GEduring 2004 (Table 5) with the gastropod Retusa canaliculatadominating at most stations. Conversely, indices differed significantlyamong distances in HH. Abundance was lowest directly beneath themussel lines (0 m) and the mean number of species increased withincreasing distance from the 0+ farm (Table 5). With the exception ofthe reference station, P. granulata was the dominant species at allsampling stations (see details in Callier et al., 2008). Shannon Wienerdiversity (H') and Margalef's d were greatest at the HH reference

station (Table 5). Richness and diversity indices did not differ betweensampling stations in CAS as individuals of the family Cirratulidaedominated at all stations.

AMBI classified the sediments beneath the 1+ mussel lines in GE as“heavily disturbed” in 2003 (5.4±0.5) while all other sampling stationsfromthe three study siteswere only “slightly”or “moderately disturbed”(scores between 1.4±0.8 and 4.1±0.2) (Table 5). Similarly, ITI classifiedsediments under the1+mussel lines inGE as “degraded” (8±9). Samples

Fig. 8. Relationship between predicted biodeposition (g m−2 d−1) and Infaunal TrophicIndex (ITI) for the three farm sites. Symbols represent individual sediment core replicatessampled inGreat-Entry Lagoon in 2003 (■) and 2004 (□), House-Harbour Lagoon (○), andCascapedia Bay (Δ). The resuspension module was enabled and data were transformed(biodeposition +1) so that stationswhere a fluxof zerowas predicted could be included onthe logarithmic scale. The vertical dashed line represents 15 g m−2 d−1.

Table 6Correlation matrix (Pearson's correlation coefficient) between predicted biodeposition,with resuspension enabled, and various biotic indicators

Flux AMBI ITI S N d J'

AMBI 0.21ITI −0.64 0.22S −0.32 0.35 0.57N −0.38 0.53 0.53 0.72D −0.12 0.18 0.41 0.78 0.23J' 0.07 −0.28 0.00 −0.17 −0.48 0.32H' −0.22 0.18 0.51 0.79 0.36 0.89 0.40

Data were pooled for the three farm sites. Statistically significant values are indicated inbold (Pb0.05). Abbreviations for the indices are the same as in Table 5.

Table 5Predicted flux of biodeposits (g m−2 d−1), with the resuspension module enabled, over a 1 month period and measured marine biotic indices and sediment characteristics (% OM) atthe mussel farms in Great-Entry Lagoon (GE) and House Harbour Lagoon (HH) in 2003 and 2004, and Cascapedia Bay (CAS) in 2005

Sites Pred. flux n S N J' H' d AMBI ITI % OM

GE-20031+ under 27.2±5.0 12 3±1c 634±484ab 0.6±0.2b 0.7±0.2c 1.0±0.3c 5.4±0.5b 8±9b 6.4±2.01+ between 6.2±1.0 12 4±1bc 231±52b 1.0±0.0a 1.4±0.2b 2.0±0.3b 3.3±1.0a 42±16a 4.8±0.60+ under 9.1±0.9 12 9±2a 881±243ab 0.9±0.1ab 1.9±0.2a 2.8±0.4a 2.8±0.3a 55±5a 5.0±2.40+ between 8.2±0.4 12 8±2a 1185±497a 0.8±0.1ab 1.6±0.2ab 2.2±0.2b 3.3±0.5a 51±15a 3.7±1.4N500 m 0 18 6±1b 629±365ab 0.8±0.1ab 1.3±0.1b 1.7±0.1b 3.7±0.2a 52±4a 3.4±0.6

GE-20040 m 6.1±0.3 5 3±2 866±472 0.9±0.1 1.3±0.2 1.6±0.5 2.6±1.2 47±12 3.8±0.33 m 5.0±0.2 5 3±1 1433±1160 0.8±0.2 0.8±0.4 1.0±0.5 1.8±0.5 41±10 3.7±0.36 m 3.4±0.1 5 3±2 1019±468 0.8±0.1 0.9±0.3 1.4±0.3 1.9±0.4 38±5 3.6±0.79 m 2.7±0.3 5 3±1 815±438 0.8±0.1 0.8±0.3 1.0±0.4 1.8±0.7 36±9 3.7±0.215 m 1.1±0.0 5 3±2 1401±830 0.8±0.1 1.1±0.2 1.2±0.2 1.4±0.8 39±6 3.9±0.630 m 0.3±0.0 5 4±2 1987±1068 0.7±0.2 0.8±0.4 0.9±0.5 1.6±0.3 38±6 4.6±1.2500 m 0 4 3±2 1631±1075 0.6±0.2 0.8±0.5 1.0±0.5 1.7±0.2 38±6 4.2±0.8

HH-20040 m 13.4±1.0 5 4±2b 1248±856b 0.8±0.1a 1.1±0.4b 1.3±0.4b 2.5±1.2ab 35±17 6.8±1.6c

3 m 11.6±1.0 5 6±2b 5172±838a 0.6±0.1b 1.1±0.2b 1.2±0.4b 3.3±0.3b 39±3 5.9±0.8bc

6 m 9.5±0.3 5 6±1ab 5223±974a 0.7±0.1b 1.2±0.1b 1.5±0.2b 3.1±0.3b 39±1 5.3±1.0bc

9 m 6.7±0.1 5 7±1ab 5427±672a 0.6±0.1b 1.2±0.2b 1.6±0.4b 3.2±0.2b 39±1 5.1±1.2bc

15 m 3.8±0.0 5 7±0ab 6471±480a 0.6±0.0b 1.2±0.1b 1.5±0.1b 3.1±0.2b 41±2 5.3±0.4bc

30 m 0.9±0.1 5 7±2ab 5987±901a 0.7±0.1ab 1.3±0.2b 1.5±0.4b 2.8±0.3b 43±4 4.0±0.7ab

300 m 0 4 10±3a 4427±2155a 0.9±0.1a 1.9±0.3a 2.5±0.6a 1.8±0.3a 45±4 2.3±0.6a

CAS-2005−40 m 0.1±0 5 7±1 6599±2746bc 0.6±0.1b 1.2±0.2 1.6±0.4 4.1±0.2b 61±3a 6.4±0.2ab

0 m 0 5 8±2 5605±1441c 0.7±0.1ab 1.4±0.3 1.9±0.5 3.7±0.3ab 58±4ab -10 m 0.1±0 5 8±2 11083±1494a 0.7±0.1ab 1.3±0.2 1.5±0.4 4.0±0.1b 54±2b 6.2±0.4ab

20 m 0 5 9±1 9274±1691abc 0.7±0.1ab 1.5±0.1 1.8±0.2 3.7±0.1ab 57±1ab 6.5±0.4ab

40 m 0 5 9±2 9962±1717ab 0.7±0.1ab 1.6±0.2 1.8±0.4 3.7±0.2ab 56±3ab 6.5±0.4ab

60 m 0 5 8±1 9605±1498abc 0.7±0.0ab 1.4±0.1 1.7±0.3 3.9±0.1ab 57±2ab 6.9±0.4b

90 m 0 5 7±2 10752±3218a 0.6±0.1b 1.2±0.3 1.4±0.3 3.9±0.2ab 58±2ab 6.7±0.3ab

500 m 0 5 7±2 8051± 2278abc 0.7±0.0a 1.5±0.2 1.5±0.4 3.4±0.6a 55±3ab 5.7±0.4a

1000 m 0 5 9±1 9248±1338abc 0.7±0.1ab 1.4±0.1 1.8±0.3 3.8±0.1ab 59±3ab 6.2±0.1ab

Modelled fluxes were based on currents measured during a one-month period prior to benthic sampling and exclude background sedimentation rates. Number of cores (n) persampling site, number of species per sediment core (S), and number of individuals m−2 (N). Pielou's eveness (J'), ShannonWiener diversity (H'), Margalef's species richness (d) indices,as well as Marine Biotic Index (AMBI), and Infaunal Trophic Index (ITI) (mean±SD) and % organic matter (% OM). Different superscript letters indicate significant differences (Pb0.05)between sampling stations.

249A.M. Weise et al. / Aquaculture 288 (2009) 239–253

Author's personal copy

collected along the transects in GE and HH during 2004 were identifiedasbeing from “enriched conditions” (ITI scores in the35–47 range)whileAMBI considered these “slightly disturbed” (AMBI scores between 1.4and 3.3). AMBI scores were greater in HH compared to GE due to thegreat abundance of P. granulata, which is classified as a second-orderopportunist, in the latter site.AMBI and ITI classified samples fromCASatdifferent disturbance levels. The ITI scores N50 along the transectindicated that benthic community structure was not affected whileAMBI classified communities as “moderately disturbed”.

Predicted biodeposition, pooled from the three study sites, wasbest correlated with ITI (R=−0.64, Pb0.001) (Table 6). Only weakcorrelations were observed between predicted biodeposition and S(R=−0.32, Pb0.001) and N (R=−0.38, Pb0.001) (Table 6). Correlationsat the most perturbed site, GE in 2003, showed predicted fluxes to bebest correlated with ITI (R=−0.79, Pb0.001), followed by AMBI(R=0.64, Pb0.001). These indices were significantly correlated toeach other (R=−0.88, Pb0.001).

Despite the great natural variability among replicate benthiccommunity samples for a given sampling station (Table 5), the ITIscores tended to decrease with increasing predicted fluxes (Fig. 8).

Sediment % OM did not differ significantly among sampling stationsalong transects in GE with values ranging between 3.6 to 4.6% but de-creased significantly with increasing distance from the farm in HH(F6,14=10.562, Pb0.001), with a maximum of 6.8±1.6% directly under thelines to aminimumof 2.3±0.6% at the reference sites (Table 5). Similarly, %OM averaged 3.4±0.6% at reference sites in GE in 2003 and reached 6.4±2.0% for sediments under 1+ mussel lines.

4. Discussion

4.1. Model predictions and field data

Overall, the model performed well considering the low measuredsedimentation rates and natural variability in sediment trap and benthicfauna data. Predictions were mostly consistent with observed trendsalthough the model did not predict the natural variability betweensediment traps, despite using detailed information and fine gridresolution and particle time step. One reason for the variation betweenpredicted and observed fluxes is likely that the true exit position of theparticles from mussel lines varies because of mussels clumping on thelines. Another may be fine scale complexity of water currents. In GE,model predictions compared favourably with the observed sedimenta-tion rates both in terms of flux and extent of dispersion. The model alsocorrectly predicted the biodeposition pattern from the mixed musselcohorts at the farm site in CAS, although generally underestimatingfluxes. As discussed in the next section (model parameter uncertainties),this is likely due in part to underestimating the particulate wasteproduced by mussels and associated epibiota on the lines.

The fine grid cell resolution (up to 1 m2) that can be applied inShellfish-DEPOMOD is clearly advantageous for modelling shellfishfarms. Effects from shellfish farms areusuallywithin the scale of tens ofmeters, thus fine grid cell resolution is important. For example, ourstudy showed clear differences in benthic effects beneath mussel linesvs. betweenmussel lines, only 10mdistant. Callier et al. (2008) showedthat communities directly under mussel lines differed from those only3 m away in HH. This type of resolution could not be obtained withlarger scale models.

Although several aquaculture waste models are capable of predict-ing the flux of waste material originating from farms, few relate thesefluxes to benthic effects. Generalized relationships between predictedfluxes and benthic descriptors are often problematic given the greatlevel of natural variability in replicates within sampling stations andamong farm locations. Despite this, our study identified some trendsand potentially useful indices. DEPOMOD uses ITI as the defaultbenthic community descriptor as it has been shown to be a reasonableindicator of effects for finfish farms (Cromey et al., 2002a). However,

other indices (e.g. Llansó et al., 2002; Simboura and Zenetos, 2002;Warwick andClarke,1998;Weisberg et al.,1997) could easily be used inlieu of ITI, although these too remain to be fully validated.

In the present study, both the ITI and AMBI indices performed wellclassifying the benthic community beneath the 1+ mussel lines in GE,i.e. sediments with the highest flux rates, as “degraded” and “heavilydisturbed”, respectively. It is interesting to note that although % OM inCAS sediment samples was similar to that measured in samplesbeneath the 1+ GE mussel lines, the benthic community structurediffered and was classified as only “moderately disturbed” (AMBI) and“not affected” (ITI). This was mainly due to the high abundance ofCirratulidae individuals in the CAS samples. Being located in an areawith high current speeds, one might expect a higher abundance ofsuspension feeders and fewer opportunistic species due to the physicaldisturbance at such a site.

Both ITI and AMBI suggest that the transition in benthic status from“slightly disturbed conditions” to “degraded” or “heavily disturbed”conditions (ITIb25, AMBI N5) occurs in the 15 to 30 g m−2 d−1 range.Indeed, all faunal sampleswhere biodeposition rateswereN15gm−2 d−1

were characterized by the dominance of the pollution tolerantpolychaete C. capitata. This species is considered to be a good indicatorof bivalve farm-related disturbance (Mattsson and Lindén, 1983;Tsutsumi, 1990; Weston, 1990) and its dominance in organicallyenriched areas may be explained by its resistance to hypoxia and highsulphide concentrations (Cuomo, 1985; Diaz and Rosenberg, 1995).Callier (2008) examined the influence of differing levels of musselbiodeposition (to the equivalent of 0, 127, 255, 382, 510, 637, and 764mussels m−2) on natural benthic communities within sediment cores inGE. This in situ “benthocosm” study showed that increasing biodeposi-tion rates led to decreasedmacrofaunal abundance and species richness.Furthermore, important increases in the abundance and biomass ofopportunistic species such as C. capitatawere observed in the treatmentwith the greatest deposition (16.8 g m−2 d−1). These patterns mirrorthose found in the present study. Thus, a value of ca.15 gm−2 d−1 seemsto be a useful indicator when a detectable impact might be expected atshellfish farms. Further testing of the model at farm sites with highermussel densities could provide the necessary data to identify thethreshold where the tolerance for organic enrichment has beenexceeded and local benthic conditions may not support diverse benthiccommunities.

4.2. Model parameter uncertainties

The primary aim of a model is to reproduce environmental ob-servations in the most non-complex manner possible. To assist modelusers, capabilities and limitations resulting from sensitivity analysesand validation studies should be clearly defined. Discrepancies areusually due to a combination of missing processes in the model,uncertainties in model parameterizations, and variability in environ-mental observations which cannot be easily reproduced by the model.These aspects are discussed below.

4.2.1. Biodeposit production parameterThe model under-predicted fluxes for many simulations even

when the maximum biodeposit production rate was used. This couldbe due to a number of reasons. First, the present study showed thatthere was great variation in mussel biodeposit production rates,increasing the uncertainty with respect to the biodeposit productionparameter. Model outputs were sensitive to this parameter and greatlyaffected model predictions, in proportion to the values used. Forexample, fluxes at the 1+ farm site in GE were reduced by 58% whenthe biodeposit production parameter was reduced from 26.4 kgbackline−1 d−1 to 10.9 kg backline−1 d−1, based on the average ratherthan the maximum biodeposit production. Similarly, fluxes werereduced by 55% when the biodeposit production parameter wasreduced from 15.8 kg backline−1 d−1 to 8.2 kg backline−1 d−1 for the 0+

250 A.M. Weise et al. / Aquaculture 288 (2009) 239–253

Author's personal copy

cohort. Although a great body of literature exists on mussel feeding,few studies exist on biodeposit production. Generalizations are diffi-cult given that this rate is dependent on a range of complex andinterrelated factors including the quantity and quality of suspendedmaterial, the feeding behaviour of mussels, and the biomass of mus-sels. Chamberlain (2002) attempted to apply a feed load and con-version factor for mussels but was unable to produce results that wererepresentative of the field data. In this study, the biodeposit pro-duction parameter is based on the in situ biodeposit production ratesmeasured for mussels in the Magdalen Islands by Callier et al. (2006)and for mussels in CAS with values obtained in the present study.Although using a single biodeposit production rate is a simplificationof a variable factor, this study has shown that the model can performwell, for example in GE, to simulate both the flux and dispersal patternaround the farm site.

Fluxes may have been underestimated at certain sites becausebiodeposition from epibiota on the mussel lines and detritus from thefarm structures were not included. Several studies have shown thatfarm structures provide a new substrate for the settlement of epibiota(Stenton-Dozey et al., 2001; Tenore and Gonzales,1976). In GE, Richardet al. (2006) found that the associated fauna on the mussel linesplayed an important role in nutrient dynamics in the water column,but also noted that the fouling community was not as developed asreported from other studies. These authors reported that the numberand composition of associated fauna depended on line age, with adominance of epifauna (mostly mussel spat) on the 0+ lines andinfauna (mainly the sedentary polychaete Amphitrite sp.) on 1+ lines.Callier et al. (2006) observed that polychaetes, starfish and hydro-zoans were more abundant on 1+ than on the 0+ mussel lines. Thefouling community on the mussel lines was much greater in CAS(author's pers. obs.) than in the Magdalen Islands and may be due tothe longer immersion time of the mussel lines. Mussels obtaincommercial size after only 18–22 months of culture in the MagdalenIslands compared to 30–36 months in CAS. Detritus and biodepositsproduced by fouling communities on farm structures have beensuggested to significantly contribute to sedimentation (Stenton-Dozeyet al., 2001). To estimate the additional material settling under the CASfarm, we can subtract the modelled mussel biodeposition from theobserved biodeposition. This suggests that ca. 30 g m−2 d−1 of thematerial settling beneath the farm may be due to biodeposition fromepibiota and detritus originating from the farm structures. This iswithin the ranges reported by Giles et al. (2006) who estimated that91 g m−2 d−1 of the material settling beneath Perna canaliculusmusselfarms, i.e. 86% of additional material settling under the farms, was notdue to mussel faeces deposition. This highlights the need for furtherwork on biodeposition, including that produced by the cultivatedspecies and that produced by their epibionts.

4.2.2. Resuspension processesThe erosion and transport of sedimented material are complex

processes and may vary according to differences in the type, size,density, cohesiveness, and porosity of the particulate material. Theresuspension module could not be extensively tested because two ofthe three farm sites were depositional (GE and HH) and fluxes wereweak at the third site (CAS). Few studies have examined the erosionrates of shellfish biodeposits (Table 3). Giles and Pilditch (2004) foundthat mussel P. canaliculus biodeposits, produced on a “natural” diet,began to erode at a critical threshold (u⁎) of 0.36 cm s−1 and that 90%were eroded at 0.67 cm s−1. Extrapolation of these velocity profiles tothe height used in DEPOMOD (1.8 m above the seabed), shows that10% of the faecal pellets are eroded at current speeds between 12.6and 15.0 cm s−1 (H. Giles, pers. comm.). The near-bed current speedrequired to resuspend finfish wastes is set at 9.5 cm s−1 in DEPOMOD,which could potentially result in the over-estimation of resuspensionand advection of material away from shellfish sites. This would beparticularly true for exposed and energetic sites, such as off-shore

sites (e.g. in New Zealand or in the Rias in Spain). Further, biodepositsmay bemore easily degraded in energetic sites, thereby reducing theirsettling velocity and thus modifying their dispersal potential. Furtherwork on resuspension processes could improve confidence in modelpredictions for such sites. That being said, as shellfish farms are oftenlocated in sheltered coastal sites, simulating resuspension processesmay not be critical, as was the case for GE and HH.

4.3. Model limitations

Shellfish-DEPOMOD predicts the dispersal and resuspension offarm-derived material but does not simulate the advection of otherresuspended material. In HH, the high sedimentation rates recordedalong the NE transect suggest that material was advected from moreshallow areas towards the farm site. Themodel could not be applied atthis farm site where the predicted fluxes of farm-derived materialwere confounded by high background sedimentation rates. To modelthe area more accurately, a system-wide approach would be needed,including the use of a hydrodynamic model to predict transportthroughout the whole area. Although the particle tracking modulewas not validated by the 24-hour-sedimentation study, the benthicresponse in HH could be related to the longer-term flux predictions.

Because Shellfish-DEPOMOD assumes a homogeneous horizontalflowfield, i.e. currents donot vary spatially throughout the grid, the usermust have a good understanding of the study area before beginning themodellingwork.GEhas beenextensively studied and three-dimensional(3-D) numerical modelling has shown that currents are horizontallyhomogenous throughout the mussel lease (Koutitonsky, 2005a). In HH,however, recent 3-D model simulations of currents showed that thecentral zone of this lagoon may be affected by eddies and gyres(Koutitonsky, 2005b). In thepresent study, currentsweremeasurednearthe NE edge of the mussel farm in HH. The modelled currents maypossibly not reflect the actual currents at theNWedge of the farmwherethe second transect was deployed. The use of a second current metercould perhaps have been useful for this site, where hydrodynamicsappear to bemore complex than in the adjacentGE. The assumption of ahomogeneous horizontal flow field for the offshore site in CAS is valid aswork but Bonardelli et al. (1993) found that currents were primarilyalongshore and largely determined by tides.

The advantage of Shellfish-DEPOMOD assuming a uniform flowfield is that a simple but representative hydrodynamic regime can beused to model almost any site whereas accurate three-dimensionalhydrographic models exist for very few areas. There are often insuf-ficient data available for many systems to be modeled using the morecomplex ecosystem-type models. Moreover, some models are unableto correctly predict the currents. For example, Giles (2006) attemptedto model mussel biodeposit dispersal but their hydrodynamic modelwas unable to adequately predict currents near the farm. Further, thetime and cost involved for these types of models may at times beprohibitive and a simplified modelling approach may therefore bemore appropriate.

4.4. Application of the model

The present study demonstrates that DEPOMOD can be adapted forshellfish culture sites and may be a good tool to investigate the spatialextent of biodeposition. Shellfish-DEPOMOD is available throughECASA (http://www.ecasatoolbox.org.uk) and SAMS (http://www.sams.ac.uk). The shellfish model can predict near-field effects at ahigh resolution (meter-scale). Since shellfish culture sites are typicallylocated in shallow coastal areas, this type of resolution is important toadequately model the dispersion of waste material as dispersal ofbiodeposits may occur over fairly short distances. Larger scale models,such as finite element or ecosystem box models cannot makepredictions at such fine resolution. Shellfish-DEPOMOD can operateat different spatial resolutions but the 1 and 9 m2 used in the present

251A.M. Weise et al. / Aquaculture 288 (2009) 239–253

Author's personal copy

work is particularly suitable for farm level particulate dispersionmodelling. The model has the merit of being user-friendly and modeloutputs (contour plots) can be easily interpreted. Further, this is thefirst known validation of a benthic response model for the impact ofshellfish farms. Although there is an acknowledged need to more fullyunderstand shellfish farm waste production and resuspensionprocesses, the model presented here can be used to estimate thespatial extent of benthic effects. Shellfish-DEPOMOD, with itscapabilities of assessing near-field effects, in conjunction with othermodels/indices that focus on far-field effects (e.g. nutrient cycling,pelagic carrying capacity), can provide the industry and managementwith the tools to efficiently and comprehensively assess the effectsassociated with shellfish culture activities within an ecosystem-basedmanagement framework.

Acknowledgements

We would like to thank the Magdalen Island's MAPAQ and DFOpersonnel as well as the Bivalve Environmental Carrying CapacityStudies (BECCS) team: F. Hartog, M. Richard, P. Robichaud, and J.Tomac. We are grateful to R. Allard, M. Fournier, C. Eloquin, and theirassociates for their collaboration and access to the mussel farms. Wethank A. Drouin and R. Larocque for their scuba-diving expertise and S.Leblanc and G. Desmeules for their help with current meters andprobes. J. Grant, V. Koutitonsky, G. Tita, and T.R.Walker are thanked forcontributing to discussions on modelling and aquaculture-environ-ment interactions. Financial support for this project was provided byFisheries and Oceans Canada's Aquaculture Collaborative Researchand Development Program (ACRDP), the Société de Développementde l'Industrie Maricole (SODIM), and the Réseau Aquaculture Québec(RAQ) to P. Archambault, M.D. Callier, C. McKindsey, and A.M. Weise. J.Chamberlain was funded on a DFO contribution agreement withDalhousie University, Nova Scotia.

References

Bacher, C., Bioteau, H., Chapelle, A., 1995. Modelling the impact of a cultivated oyster popu-lation on the nitrogen dynamics— the Thau Lagoon case (France). Ophelia 42, 29–54.

Baudinet, D., Alliot, E., Berland, B., Grenz, C., Plante-Cuny, M., Plante, R., Salen-Picard, C.,1990. Incidence of a mussel culture on biogeochemical fluxes at the sediment waterinterface. Hydrobiologia 207, 187–196.

Bayne, B.L., Iglesias, J.I.P., Hawkins, A.J.S., Navarro, E., Heral, M., Deslous-Paoli, J.-M.,1993. Feeding behaviour of the mussel, Mytilus edulis: responses to variations inquantity and organic content of the seston. J. Mar. Biol. Assoc. UK 73, 813–829.

Bonardelli, J.C., Himmelman, J.H., Drinkwater, K., 1993. Current variability and upwellingalong the north shore of Baie des Chaleurs. Atmos-Ocean 31 (4), 541–565.

Borja, A., Franco, J., Perez, V., 2000. A Marine Biotic Index to establish the ecologicalquality of soft-bottom benthos within European estuarine and coastal environ-ments. Mar. Pollut. Bull. 40, 1100–1114.

Borja, A., Muxika, I., Franco, J., 2003. The application of a Marine Biotic Index to differentimpact sources affecting soft-bottom benthic communities along European coasts.Mar. Pollut. Bull. 46, 835–845.

Byers, S.C., Mills, E.L., Stewart, P.L., 1978. A comparison of methods of determiningorganic carbon in marine sediments, with suggestions for a standard method.Hydrobiologia 58, 43–47.

Callier, M.D., 2008. Influence de la mytiliculture (Mytilus edulis L.) sur les caractéris-tiques physico-chimiques du sédiment et sur les communautés macrobenthiques.PhD thesis, Université du Québec à Rimouski, Rimouski, pp. 248.

Callier, M.D., Weise, A.M., McKindsey, C.W., Desrosiers, G., 2006. Sedimentation rates ina suspendedmussel farm (Great-Entry Lagoon, Canada): biodeposit production anddispersion. Mar. Ecol., Prog. Ser. 322, 129–141.

Callier, M.D., McKindsey, C.W., Desrosiers, G., 2007. Multi-scale spatial variations inbenthic sediment geochemistry and macrofaunal communities under a suspendedmussel farm (Great-Entry Lagoon, Canada). Mar. Ecol., Prog. Ser. 348, 103–115.

Callier, M.D., McKindsey, C.W., Desrosiers, G., 2008. Evaluation of indicators used todetect mussel farm influence on the benthos: two case studies in the MagdalenIslands, Eastern Canada. Aquaculture 278, 77–88.

Canadian Hydrographic Service, 2008. April 18 2008 www.waterlevels.gc.ca.Chamberlain, J., 2002. Modelling the Environmental Impacts of Suspended Mussel

(Mytilus edulis L.) Farming. PhD Thesis, Napier University, UK, 200 pp.Chamberlain, J., Stucchi, D., 2007. Simulating the effects of parameter uncertainty on

waste model predictions of marine finfish aquaculture. Aquaculture 272, 296–311.Chamberlain, J., Fernandes, T.F., Read, P., Nickell, T.D., Davies, I.M., 2001. Impacts of

biodeposits from suspended mussel (Mytilus edulis L.) culture on the surroundingsurficial sediments. ICES J. Mar. Sci. 58, 411–416.

Chamberlain, J., Weise, A.M., Grant, J., Dowd, M., 2006. Modelling the effects ofbiodeposition from shellfish farms on the near field benthic environment. ModelingApproaches to Assess the Potential Effects of Shellfish Aquaculture on the MarineEnvironment. DFO Can. Sci. Advis. Sec. Res. Doc. 2006/032. iv + 50 p.

Clarke, K.R., Warwick, R.M., 1994. Change in Marine Communities: an Approach toStatistical Analysis and Interpretation. Plymouth Marine Laboratory, Plymouth.141 pp.

Cranford, P.J., Hill, P.S., 1999. Seasonal variation in food utilization by the suspension-feeding bivalve molluscs Mytilus edulis and Placopecten magellanicus. Mar. Ecol.,Prog. Ser. 190, 223–239.

Cranford, P., Anderson, R., Archambault, P., Balch, T., Bates, S., Bugden, G., Callier, M.D.,Carver, C., Comeau, L., Hargrave, B., Harrison, G., Horne, E., Kepay, P.E., Li, W.K.W.,Mallet, A., Ouellette, M., Strain, P., 2006. Indicators and thresholds for use inassessing shellfish aquaculture impacts on fish habitat. DFO Can. Sci. Advis. Sec. Res.Doc. 2006/034. 166 pp.

Crawford, C.M., Macleod, C.K.A., Mitchell, I.M., 2003. Effects of shellfish farming on thebenthic environment. Aquaculture 224, 117–140.

Cromey, C.J., Nickell, T.D., Black, K.D., 2002a. DEPOMOD-modelling the deposition andthe biological effects of waste solids from marine cage farms. Aquaculture 214,211–239.

Cromey, C.J., Nickell, T.D., Black, K.D., Provost, P.G., Griffiths, C.R., 2002b. Validation of afish farmwaste resuspension model by use of a particulate tracer discharged from apoint source in a coastal environment. Estuaries 25, 916–929.

Cuomo, M.C., 1985. Sulphide as a larval settlement cue for Capitella sp. I. Biogeochem-istry 1, 169–181.

Dahlbäck, B., Gunnarsson, L.Å.H., 1981. Sedimentation and sulfate reduction under amussel culture. Mar. Biol. 63, 269–275.

Danovaro, R., Gambi, C., Luna, G.M., Mirto, S., 2004. Sustainable impact of musselfarming in the Adriatic Sea (Mediterranean Sea): evidence from biochemical,microbial and meiofaunal indicators. Mar. Pollut. Bull. 49, 325–333.

Diaz, R.J., Rosenberg, R., 1995. Marine benthic hypoxia: a review of its ecological effectsand the behavioural responses of benthic macrofauna. Oceanogr. Mar. Biol. Annu.Rev. 33, 245–303.

Dowd, M., 2005. A bio-physical coastal ecosystem model for assessing environmentaleffects of marine bivalve aquaculture. Ecol. Model. 183, 323–346.

Giles, H., 2006. Dispersal and Remineralisation of Biodeposits: Ecosystem Impacts ofMussel Aquaculture. PhD Thesis, The University of Waikato, New Zealand, 150 pp.

Giles, H., Pilditch, C.A., 2004. Effects of diet on sinking rates and erosion thresholds ofmussel Perna canaliculus biodeposits. Mar. Ecol., Prog. Ser. 282, 205–219.

Giles, H., Pilditch, C.A., Bell, D.G., 2006. Sedimentation from mussel (Perna canaliculus)culture in the Firth of Thames, New Zealand: impacts on sediment oxygen andnutrient fluxes. Aquaculture 261, 125–140.

Gillibrand, P.A., Turrell, W.R., 1997. Simulating the dispersion and settling of particlesand settling substances from salmon farms. Aberdeen Marine Laboratory, ReportNo 3/97, Aberdeen AB11 9DB UK.

Grant, J., Hatcher, A., Scott, D.B., Pocklington, P., Schafer, C.T., Winters, G.V., 1995. Amultidisciplinary approach to evaluating impacts of shellfish aquaculture onbenthic communities. Estuaries 18, 124–144.

Grant, J., Cranford, P., Hargrave, B., Carreau, M., Schofield, B., Armsworthy, S., Burdett-Coutts, V., Ibarra, D., 2005. A model of aquaculture biodeposition for multipleestuaries and field validation at blue mussel (Mytilus edulis) culture sites in easternCanada. Can. J. Fish. Aquat. Sci. 62, 1271–1285.

Gust, G., Kozerski, H.-P., 2000. In situ sinking-particle flux from collection rates ofcylindrical traps. Mar. Ecol., Prog. Ser. 208, 93–106.

Hartstein, N.D., Stevens, C.L., 2005. Deposition beneath long-line mussel farms. Aquac.Eng. 33, 192–213.

Henderson, A., Gamito, S., Karakassis, I., Pederson, P., Smaal, A., 2001. Use ofhydrodynamic and benthic models for managing environmental impacts of marineaquaculture. J. Appl. Ichthyol. 17, 163–172.

Kautsky, N., Evans, S., 1987. Role of biodeposition by Mytilus edulis in the circulation ofmatter and nutrients in a Baltic coastal ecosystem.Mar. Ecol., Prog. Ser. 38, 201–212.

Koutitonsky, V.G., 2005a. Three-dimensional structure of wind-driven currents incoastal lagoons. Chapter 9.2: Case studies. In: Gonenc, E., Woflin, J. (Eds.), CoastalLagoons: Ecosystem Processes and Sustainable Use and Development. CRC Press,Florida, pp. 371–386.

Koutitonsky, V.G., 2005b. Modélisation numérique intégrée des courants, des vagues etdu transport des sédiments à l'entrée de la lagune du Havre-au-Maisons. Rapportde recherche LHE 05-2. Laboratoire d' hydraulique environnementale. Institut desSciences de la Mer, Rimouski, Québec, Canada.

Koutitonsky, V.G., Navarro, N., Booth, D., 2002. Descriptive physical oceanography ofGreat-Entry Lagoon, Gulf of St. Lawrence. Est. Coast. Shelf Sci. 54, 833–847.

Llansó, R.J., Scott, D.B., Hyland, J.L., Dauer, D.M., Russell, D.E., Kutz, F.W., 2002. Anestuarine benthic index of biotic integrity for the mid-Atlantic region of the UnitedStates. II. Index development. Estuaries 25, 1231–1242.

Léonard, M., 2004. Évaluation et caractérisation de la chute de moules (Mytilus edulis)dans la lagune de Havre-aux-Maisons, aux Îles-de-la-Madeleine, Québec. MScthesis ENITA Clermont-Ferrand, 63 pp.

Mattsson, J., Lindén, O., 1983. Benthic macrofauna succession under mussels, Mytilusedulis L. (Bivalvia), cultured on hanging long-lines. Sarsia 68, 97–102.

McKindsey, C.W., Thetmeyer, H., Landry, T., Silvert, W., 2006. Review of recent carryingcapacity models for bivalve culture and recommendations for research andmanagement. Aquaculture 261, 451–462.

Mearns, A.J., Word, J.Q., 1982. Forecasting effects of sewage solids on marine benthiccommunities. In: Mayer, G.F. (Ed.), Ecological Stress and the New York Bight:Science and Management. Columbia S Carolina Estuarine Research Federation,United States, pp. 495–512. Columbia, S Carolina.

252 A.M. Weise et al. / Aquaculture 288 (2009) 239–253

Author's personal copy

Miron, G., Landry, T., Archambault, P., Frenette, B., 2005. Effects of mussel culturehusbandry practices on various benthic characteristics. Aquaculture 250, 138–154.

Nodder, S.D., Alexander, B.L., 1999. The effects of multiple trap spacing, baffles and brinevolume on sediment trap collection efficiency. J. Mar. Res. 57, 537–559.

Piñeiro, G., Perelman, S., Guerschman, J.P., Paruelo, J.M., 2008. How to evaluate models:observed vs. predicted or predicted vs. observed? Ecol. Model. 216, 316–322.

Richard, M., Archambault, P., Thouzeau, G., Desrosiers, G., 2006. Influence of suspendedmussel lines on the biogeochemical fluxes in adjacent water in the Îles-de-la-Madeleine (Quebec, Canada). Can. J. Fish. Aquat. Sci. 63, 1198–1213.

Simboura, N., Zenetos, A., 2002. Benthic indicators to use in ecological qualityclassification of Mediterranean soft bottom marine ecosystems, including a newbiotic index. Med. Mar. Sci. 3/2, 77–111.

Stenton-Dozey, J., Probyn, T., Busby, A., 2001. Impact of mussel (Mytilus galloprovincialis)raft-culture on benthic macrofauna, in situ oxygen uptake, and nutrient fluxes inSaldanha Bay, South Africa. Can. J. Fish. Aquat. Sci. 58, 1021–1031.

Tamigneaux, E., Thomas, B., 2004. Étude des caractéristiques biophysiques de 12 sitesd'intérêtsmaricole dans les baies de Gaspé, de Tracadigache et de Cascapédia pendantl'automne 2000. Ministère de l'Agriculture, des Pêcheries et de l'Alimentation, 87 p.(Cahier d'information n° 146).

Tenore, K.R., Dunstan, W.M., 1973. Comparison of rates of feeding and biodeposition ofthe American oyster, Crassostrea virginica Gmelin, fed different species ofphytoplankton. J. Exp. Mar. Biol. Ecol. 12, 19–26.

Tenore, K.R., Gonzales, N.,1976. Food chain patterns in the Ria de Arosa, Spain: an area ofintense mussel aquaculture. In: Persoone, G., Jaspers, E. (Eds.), Proceedings of the10th European Symposium onMarine Biology, Ostende, Belgium, September 17–23,1975. Population Dynamics of Marine Organisms in Relation with Nutrient Cyclingin Shallow Waters, vol. 2. Universa Press, Wetteren, pp. 601–619.

Tsutsumi, H., 1990. Population persistence of Capitella sp. (Polychaeta; Capitellidae) on amud flat subject to environmental disturbance by organic enrichment. Mar. Ecol.,Prog. Ser. 63, 147–156.

Van Raaphorst, W.V., Malschaert, H., Van Haren, H., 1998. Tidal resuspension anddeposition of particulate matter in the Oyster Grounds, North Sea. J. Mar. Res. 56,257–291.

Walker, T.R., Grant, J., in press. Quantifying erosion rates and stability of bottomsediments at mussel aquaculture sites in Prince Edward Island, Canada, J. Mar. Syst.doi:10.1016/j.jmarsys.2008.07.009.

Warwick, R.M., Clarke, K.R., 1998. Taxonomic distinctness and environmental assess-ment. J. Appl. Ecol. 35, 532–543.

Weisberg, S.B., Ranasinghe, J.A., Schaffner, L.C., Diaz, R.J., Dauer, D.M., Frithsen, J.B., 1997.An estuarine benthic index of biotic integrity (B-IBI) for the Chesapeake Bay.Estuaries 20, 149–158.

Weston, D.P., 1990. Quantitative examination of macrobenthic community changesalong an organic enrichment gradient. Mar. Ecol., Prog. Ser. 61, 233–244.

Widdows, J., Fieth, P., Worrall, C.M., 1979. Relationships between seston, available foodand feeding activity in the common mussel Mytilus edulis. Mar. Biol. 50, 195–207.

Widdows, J., Brinsley, M.D., Salkeld, P.N., Elliott, M., 1998. Use of annular flumes todetermine the influence of current velocity and bivalves on material flux at thesediment-water interface. Estuaries 21 (4A), 552–559.

Wildish, D.J., Hargrave, B.T., MacLeod, C., Crawford, C., 2003. Detection of organicenrichment near finfish net-pens by sediment profile imaging at SCUBA-accessibledepths. J. Exp. Mar. Biol. Ecol. 285–286, 403–413.

Word, J.Q.,1978. The infaunal trophic index. Annual Report 1978. CoastalWater ResearchProject, pp. 13–39. El Segundo, California, USA.

253A.M. Weise et al. / Aquaculture 288 (2009) 239–253