te

b

Mardell Buryniuk a, Royann J. Petrell a,*, Susan Baldwin a, K. Victor Lo b

Aquacultural Engineering 35aChemical and Biological Engineering, University of British Columbia, 2360 East Mall,

Vancouver, BC, Canada V6T1Z3bCivil Engineering, University of British Columbia, BC, Canada

Received 7 April 2005; accepted 18 August 2005

Abstract

A new method consisting of a screen-like device to trap and manage solid waste from below a fish cage is proposed. To

examine its effectiveness, a mathematical model was developed to predict the amount of waste and its degradation over time

under low-current conditions. It was also used to examine the effects of fish stocking, feed conversion ratio, screen size, mesh

size and harvest rate on the total amount accumulated and time required to degrade the waste after harvest. The characteristics of

waste and fish feed used to develop the model were experimentally determined as they degraded in a tank of oxygenated 8 8Csaline water. As the solid waste degraded, the carbon (%) and COD (mg/(L g dry weight)) remained constant as N (%) increased

and C/N decreased. Bacteria degradation consisted of activities related to both mineralization and the physical breakdown of the

waste into tiny particles. After 3040 days in cold and saline water, approximately 50% of the waste matter disappeared from the

3 mm mesh screen ( p < 0.001). The experimental waste degradation rate (kg m2 day1) increased with increasing specificarea of waste (kg m2) (r2 = 0.97). Model simulations indicated that staggering fish harvests was the most effective method forreducing waste loads and the period for total waste removal after fish harvest. Future work will focus on the fate within the

environment of the tiny particles released by the degradation process and the effect of current on waste erosion rates.

# 2005 Elsevier B.V. All rights reserved.

Keywords: C/N ratio; Chemical oxygen demand; Benthic impacts; Environmental effects; Salmon farming

1. Introduction

Waste produced by fish farms contains carbon,

phosphorus and nitrogen in dissolved and suspended

solids (Ackefors andEnell, 1990;Naylor et al., 2000) as

well as metals, such as zinc and copper (Beveridge,

1996; Edwards, 1998; Kempf et al., 2002). In cage

farming, suspended solids that are the result of uneaten

food and faeces can accumulate beneath the cages

especially under low-current conditions (McGhie et al.,

2000; Naylor et al., 2000). High accumulation may

affect benthic fauna, sediment chemistry, degradation

rate and environmental quality (Clarke and Phillips,

1989). Through the process of decomposition, oxygen

* Corresponding author. Tel.: +1 604 822 3475;

fax: +1 604 822 5407.

E-mail address: rpetrell@chml.ubc.ca (R.J. Petrell).

0144-8609/$ see front matter # 2005 Elsevier B.V. All rights reserved.doi:10.1016/j.aquaeng.2005.08.008Accumulation and natural disin

on a screen suspendedgration of solid wastes caught

elow a fish farm cage

www.elsevier.com/locate/aqua-online

(2006) 7890

chemical oxygen and degradation rate, and to use this

ral Ein and above the sediments can become depleted, and

under anoxic conditions gases, such as nitrogen, carbon

dioxide, methane and hydrogen sulphide can be

generated. As well, waste degradation rates could

decrease as anaerobic decomposition generally pro-

ceeds at a slower rate than aerobic decomposition does.

High accumulation ofwaste solids beneath fish farming

sites can affect the caged fish since oxygen depletion

and formation of chemical species toxic to the fish can

lead to self-pollution and a reduction in production

(Folke and Kautsky, 1989).

Typically, the solution to solidwaste accumulation is

a period of production inactivity, so that the solidwastes

on the sea floor have time to naturally degrade or erode.

This process is called site fallowing. The Canadian

Aquaculture Industry Alliance (CAIA) has promoted

this practice to allow benthic recovery, and states that

below most farming sites the benthic community

recovers after 69 months post harvest (CAIA Salmon

Facts, 2003). Recovery appears to depend on site

characteristics and farmingoperation (e.g.Brooks et al.,

2003, 2004). Peer-reviewed studies do not exist that

could be used to suggest an appropriate fallowing time

valid for all farming conditions. Disadvantages of

fallowing include lost production potential, possible

negative impacts on fauna and negative public

perception. Due to these disadvantages, the develop-

ment of an alternative waste management system is

required for a more sustainable industry.

The research described herein involves the devel-

opment of a newwaste management method that relies

on a screen-like device below fish farm cages.

Conceptually, the screen would trap the solids for

natural breakdown and/or on-site treatment. Land-

based fish farming operations have used stationary

screens as an effective pre-treatment step to remove

particulates from effluent (e.g. Makinen et al., 1988;

Bergheim and Forsberg, 1993; Bergheim et al., 1993).

Clogging rapidly occurs if concentrations of sus-

pended particles are too high or as with the case of fish

waste, the material tends to be adhesive (Wheaton,

1985; Cripps and Bergheim, 2000). Below a sea cage,

the clogging that is a problem for land-based treatment

would be used to an advantage to enhance solids

retention on the screen-like device.

It is anticipated that the new method would be

compatible with current infrastructure and could cover

M. Buryniuk et al. / Aquacultuthe entire affected zone. The required screen areainformation to develop a model that could be used to

predict accumulation and removal on a screen below a

typical fish farm. Promising results would help

generate an interest in a field trial wherein the benthic

recovery rates below screened and non-screened areas

could be compared.

2. Solid waste characterization and

degradation experiments

The purpose of these experiments was to track the

degradation and chemical composition of solid waste

and fish feed pellets on a screen suspended in water

under conditions similar to those found at a fish farm

with low-current conditions (i.e. a condition that would

favor degradation by microorganisms over physical

breakdown and dispersion by current). This would

provide information relating to a worst case scenario.

2.1. Solid waste material

Solid waste was collected every 23 h on three

Atlantic salmon fish farms using 1 m diameter funnels

suspended 15 m below the surface in fall and early

winter during three trips, each one lasting 4 days.

Samples were collected from three cages (sizedwould be site specific and dependant upon depth

below the cage, currents and bottom topography. In

general, screen area increases with suspension depth

below a cage. Locating the screens too close to the

cage bottom, however, could result in production

decline due to the outputs from waste degradation and

competition for dissolved oxygen.

Degradation rates on the screen are expected to be

higher than benthic rates, since the solid waste surface

area exposed to dissolved oxygen is higher on the screen

(both sides of the screen) and dissolved oxygen and

water temperature just under the cage are generally

greater than at benthic levels. The potential benefits of

the proposed method include compatibility with in-

place infrastructure, collection of dispersed waste and

an increase in degradation rates and recovery times as

compared to the conventional fallowing periods.

The aim of the research was to characterize the

solid waste by measuring particle size, composition,

ngineering 35 (2006) 7890 7933 m 33 m in area) at once on each farm. The

ural Eamount collected, due possibly to current or low

appetite as a result of seasonally cold water, was not

large yet sufficient to conduct basic waste character-

ization and degradation tests. Samples from the

various cages were combined to form different

composite samples. Material was drained over number

18 US standard sieve mesh (nominal open-

ing = 1.00 mm) and immediately frozen on dry ice.

An undetectable amount of particles passed through

the mesh. In the laboratory, they were stored in a

freezer at 10 8C.

2.2. Particle size distribution

Some of the collectedwastewas size fractionated on

site following a slightly modified method by Cripps

(1995). In the Cripps method of particle sizing, a

sample is poured over filter paper, then the filtrate is

poured over the next smaller size of filter paper and so

on. Residues on the filters are dried and measured. We

used small sieves (0.07525 mm) instead of filter paper.

2.3. Solid waste characterization

Solid waste dry mass (kg m3) and ash analyseswere completed as per procedures (#s 208A and

208E) outlined in the Standard Methods for the

Examination of Water and Wastewater (American

Public Health Association, 1998). The measurement

of total carbon, nitrogen, and sulphur content of

samples was done by the UBC Earth and Ocean

Sciences Laboratory with a Carlo Erba N2A-1500

Analyzer. The test for chemical oxygen demand

(COD) was performed on the total dried residue using

a pre-packaged mercury-free standard range digestion

kit (Lamotte Company). COD results were normalized

to the dry mass (g) present in the tested dilution.

Carbohydrates and lipids were assumed to be the

nitrogen free extract (NFE) (sample minus ash and

protein). Protein content was calculated from total

nitrogen content after multiplying the total nitrogen

content by 6.25, it was assumed that nearly all of

nitrogen was in the form of amino acids (FAO, 2003).

2.4. Degradation rate experiments

The rate of solid waste degradation on a screen

M. Buryniuk et al. / Aquacult80under environmentally relevant conditions was mea-sured over 21 days. Frozen waste samples were placed

on 6 pre-weighed number 6 US stainless-steel

standard sieves (nominal opening of mesh =

3.36 mm). This mesh size was chosen because this

mesh could capture before the screen clogs over 65%

of the particle on the salmon farms (see Section 4,

Particle size determination). The structural properties

of the solid waste did not appear to be affected by

freezing and thawing. For the sake of comparison, two

screens containing fish food were also degraded. Each

15 cm 15 cm screen was suspended approximately0.1 m below the surface in a tank (0.82 m 0.67 m 0.41 m deep). To represent a worst casescenario, each tank was filled with artificial seawater

(instead of natural water) at salinity between 25 and

30% prepared from Coralife scientific grade marinesalt (Energy Savers Unlimited, Inc.) and distilled

water. The tanks were kept in the dark at 8 8C (seeModel Development for rationale). Each tank was

equipped with air diffusers to promote a gentle mixing

action and maintain O2 concentration above 80%

saturation. pH (meter). Ammonia levels (LaMotte test

kit) were also regularly checked to ensure that the

water was maintained at a level required for salmonid

production (Shepherd and Bromage, 1988).

On the test screens, three amounts of dry waste

matter in units of specific area (kg m2) were tested.These amounts represented the daily waste production

of 3.5 kg fish at a stocking density of 18 kg m3 tonearly four times the daily waste production of 6.4 kg

fish at a stocking density of 20 kg m3 as perBergheim et al. (1991) or approximately 4 days of

accumulated waste at this time. The highest amount of

waste reflects a case where the waste material does not

spread evenly over the screen but accumulates in a

pocket or fold (generated by current or net billowing).

A 23 g (wet basis) sample was removed from

randomly selected spots on each screen every 4 days

and analyzed for total carbon, nitrogen, sulphur and

COD. Test samples were centrifuged to concentrate

the solids before analysis. The test period of 20 days

was chosen to ensure a measurable and accurate

weight lost. Awaste removal rate, R, was calculated as

the difference between the initial and final masses

divided by the trial period. In the calculation, we also

considered the weights removed for chemical ana-

lyses. For this, the following equation for calculating

ngineering 35 (2006) 7890mass, Mt, at the various sampling times was copied

ral EAmodel was developed for simulating waste input,

accumulation and removal on a screen suspended

below a salmon fish pen. The major assumptions used

to develop the model were: (1) uneaten feed is

negligible because it can be managed via camera or

other waste prevention method (Juell et al., 1993; Ang

and Petrell, 1997). We did, however, check for the

effect of uneaten feed by comparing the rates of

decomposition of waste and food materials (see

results). (2) Bacterial action is the major degrading

force. The other mechanisms of solid waste break-

down, such as erosion and dispersion (that require

current) that can quickly breakdown solid waste, were

not considered in order to study the worst case or

maximum accumulation scenario. This would provide

information that would be most useful for low-current

sites. (3) The biomass of the bacteria growing on the

waste is negligible as compared to the waste itself

(Boyd, 1995). (4) To be similar to water just beneath a

salmon farm in British Columbia, the water tempera-

ture at the screen level was set equal to 8 8C (Ang andPetrell, 1998). (5) Freshly produced waste falls on topinto a column in a spreadsheet program.

Mt Mt1 RDt Ms;t;

where Mt1 is the mass at time t 1 and Ms,t is themass removed for sampling at time t. Using this

equation, a value for R was sought that would yield

a final weight equal to the measured final weight.

2.4.1. Statistical analysis

To compare initial and final waste compositions for

waste andfish foodpellets, t-tests and/orANOVAswere

conducted. The maximum p-value for determining

statistical significancewas p = 0.05. Thewaste removal

rates found via experiment as described above were

regressed against specific screen area, kg m2 todetermine if there was a dependence on the initial

amount of waste on a screen (see Eq. (7), Section 3.3

Waste production generator under Model develop-

ment). To estimate the parameters r, and n, in Eq. (7),

non-linear least squares regression was carried out.

3. Model development

M. Buryniuk et al. / Aquacultuof older waste from the previous days, and theresulting layers do not mix. This would provide

information useful for a worst case no replenishment

of nutrients condition. (6) Dissolved oxygen is not

limiting (meaning that the screen is suspended just

below the cage). Furthermore, oxygen probably would

not limit degradation in bottom waste layers. Our need

for water changes near the beginning of our

experiments (see Section 4.3) indicated that bacterial

action probably would be much higher near the time of

deposition when the layer is well exposed to oxygen

then after several days when remaining residue would

be covered with other layers.

3.1. Conceptual model

The three elements of the compartment screen

waste degrader model are the fish biomass predictor,

the waste production generator and the solid waste

degrader. The system equation used to calculate the

mass on the screen at a given time (t) is as follows.

dWsdt

I B P I R (1)

whereWs is the solid waste accumulated on the screen,

dry mass (kg m2), I the daily solid waste input(kg m2 day1), B the bacterial degradation of solidwaste (kg m2 day1), P the physical disintegration ofsolid waste due to bacteria action (kg m2 day1) andR = (B + P) is the removal of solid wastes from thescreen (kg m2 day1).

3.2. Fish biomass predictor

Waste output depends in part on fish size; therefore,

we required a model that could accurately estimate

fish growth. Fortunately, available to the researchers

were sets of actual farm production data that showed

how Atlantic salmon (Salmo salar) weight changed

over time on a British Colombian fish farm at different

water temperatures. After evaluating a number of

existing growth models (Buryniuk, 2004), the data

were best represented by the following thermal growth

coefficient (TGC) model (Stigebrandt, 1999) (Fig. 1):

M1 M1=30 TGCt1 t0Tm3 (2)where M1 is the fish mass (weight) after time interval

ngineering 35 (2006) 7890 81(t1 t0), M0 the initial fish mass (weight) (g), Tm the

ural Engineering 35 (2006) 7890mean temperature during the time period (t1 t0) (8C)and TGC is the thermal growth coefficient (kg1/3 day1 8C1).

The use of this TGC equation implies many

assumptions. One relates to t, the inverse temperature

scale, that is comparable in nature to the Q10 scale and

was used to calculate TGC in the original research for

Atlantic salmon in Norway (Stigebrandt, 1999).

Stigebrandts value for t equal is 0.080 8C1. Thismeans that the biochemical metabolic rates of fish

double for every temperature increase of 8.6 8C.Another major assumption used is TGC is constant

and equal to:

TGC M1=32 M1=31

100

PTmt2 t1

(3)

M. Buryniuk et al. / Aquacult82

Fig. 1. The thermal growth coefficient (TGC) growth model fit the

available data.3.3. Existing model review

Several waste output models exist in the literature.

Unfortunately, farms do not have measures of actual

waste output from which we could have used to select

an appropriate model for predicting waste output. We,

therefore, reviewed a number of models, and selected

a modern one that utilized readily available para-

meters and produced average waste loads (Buryniuk,

2004). Details follow.

One of the models we inspected is the empirically

derived Liao and Mayo (1974) model; it represents the

relationship between feeding rates and suspended

solids production for trout between 10 and 15 8C and amaximum stocking density of 28.4 kg m3:

SS 0:52F (4)where SS is suspended solids production (kg SS

(100 kg fish)1 day1) and F is the feeding rate(kg food (100 kg fish)1 day1).

We also examined the Stigebrandts (1999)

monitoring-on growing fish farms-modelling

(MOM) model. In this model solid waste production

or maximum faecal loss, FLdw, in units of dry weight

of faecal matter per fish per day is expressed as a

fraction of the maximal food ration per fish (g day1)or appetite, APP as follows:

FLdw FLAPP (5)In this expression the unassimilated feed fraction

FL equals:

FL 1 ApEp 1 AlEl 1 AcEc;and requires measures of the assimilated fraction of

proteinAp, lipid Al and carbohydratedAc, as well as the

fractions of food supply by proteins Ep, lipid El and

carbohydrates Ec. The maximal appetite, APP, was

calculated using the following equation from Stigeb-

randt (1999):

APP aMg aCfMb

ed

etT

where M equals fish mass, T equals water temperature

in 8C, a, g, t, a and b are empirical constants, d equalsthe specific energy content of the feed,

e* = (1 BC EL) (0.15EpAp),Cf Cf 0:15CpPp, BC equals the fraction ofenergy required for biochemical breakdown of feed,

Cf is the specific energy content of the fish (species and

age dependent), Cp equals the specific energy of

protein and Pp equals the protein content of fish

(%). The necessary parameters for running the model

for salmon are given in Table 1.

A simplified version of the previously described

model by Cromey et al. (2002) was also reviewed. In

this model or the DEPOMOD (deposition model)

model, solid waste production in dry weight is

calculated as:Wfae Fc1 Fdig

where Wfae is the solid waste production (faecal

material) (kg day1 cage1) and Fdig is the digestibleproportion of feed (dimensionless). Dry weight of the

feed ingested by the fish, Fc is calculated from:

Fc F1 Fw1 Fwasted

where Fc is the feed consumption (kg day1 cage1),

F the feed input (kg day1 cage1), Fw the watercontent proportion of feed (dimensionless) and Fwastedis the proportion of wasted feed (dimensionless).

Finally, we reviewed Bergheim et al. (1991)

relationship between suspended dry matter loadings

and feed conversion ratios (FCR: weight of feed given

per weight gain of fish) in a land-based tank

production system for Atlantic salmon. The relation-

ship was developed for salmon from 30 g to 2 kg in

size (1 years growth in seawater) between 4 and

15 8C. Suspended dry matter loadings according tothis model or the BTM model (1991) is:

SDM 0:20100:49FCR (6)where SDM is the suspended dry matter loading (g

(kg fish)1 day1), and FCR is feed conversion ratio(kg dry feed/kg fish weight gain).

3.3.1. Waste-model selection

To determine which of the waste output models

would be used in our screen degradation model, we

applied all but the DEPOMOD model to generate

curves representing the increase in solid waste over

time on a salmon farm (Fig. 2). The DEPOMOD

model was not applied because it requires parameters

M. Buryniuk et al. / Aquacultural Engineering 35 (2006) 7890 83

Table 1

Parameters used in the monitoring-on growing fish farms-modelling

model (MOM) and the waste degrader model

MOM

Specific energy content of protein, Cp (cal/g) 5650

Assimulated fractions

Ap 0.89

Al 0.05

Ac 0.5

Food energy fractionsEp 0.39

El 0.57

Ec 0.037

Protein fraction per mass of fish, Pp 0.18

Specific energy content of feed, d (cal/g) 5470.5

Specific energy of fish, Cf (cal/g) 2413

Energy fraction for biochemical feed treatment, BC 0.13

a (cal day1 g0.8) 11g 0.8

t 0.08

a 0.028

b 0.67

Screen waste degrader model

Thermal growth coefficient, TGC (kg1/3 day1 8C1) 0.21Time step (day) 1

Mean water temperature (8C) 9.19Mean feed conversion ratio, FCR 1.24

Screen capture efficiency (%) 100

Initial fish number 441895

Initial fish mass (kg) 0.72

Mortality (% day1) 0.0003Screen surface area (m2) 3716including amount of ingested feed and digestibility of

that feed, which was not available to us. For the

simulations, fish mass (kg) was required, and it was

obtained by using the fish biomass predictor (Eq. (2)).

For the BTSmodel, FCRs were also required and these

were supplied by same salmon farm that had supplied

the fish growth data for the fish growth predictor

equation.

An inspection of the resulting curves (Fig. 2) shows

jagged lines that relate to drops in the temperature data

as supplied by the salmon producers. As one might

expect from older feed compositions and rations, the

model by Liao and Mayo (1972) gave waste loads far

above the two more recent models. We selected the

BTC model as the waste-production generator for the

compartment screen waste degrader model because it

Fig. 2. Comparison of different solid waste predictors: Liao and

Mayo (1974), monitoring-ongrowing fish farms (MOM) model, andBergheim et al. (1991) BTM model.

M. Buryniuk et al. / Aquacultural E84represented the worst case scenario of the two more

recent models. Comparably, the MOM model under-

estimated solid waste output, because the FCRs it

predicted were distinct from the actual farm records.

This model might be used as the best-case or ideal

scenario.

3.4. Solid waste degrader model

There are several waste degradation models

available from the literature. In one commonly used

model, the rate of solid waste removal is mathema-

tically expressed as follows:

dW

dt R rWn (7)

The removal rate constant, R, and the parameters r

and n are experimentally determined. The following

general equation (determined after integrating the

above equation with n 6 1) indicates how solid wastedecreases over time:

Wt rt1 n W 1n0 1=1n (8)where Wt is the mass of solid waste remaining at time

(t) and W0 is mass of solid waste at time zero.

In a Multi-G model as developed by Berner (1980),

a first order model (n = 1 in equation (18)) is assumed,

and the organic material is differentiated into

fractions, each with a corresponding negative reaction

rate constant (decay rate constant), k:

GT Xn0

Gi

where, GT is total concentration of organic matter

(kg m3) and Gi is concentration of organic matterin ith group (kg m3).

dGidt

kiGiwhere ki is decay rate constant for ith organic matter

group (day1). The fractions are differentiated by C:Nratio, each with a corresponding decay rate constant.

The degradation expression for the total organic mat-

ter is a sum of the individual fractions.

dGT Xn

kiGi

dt

0Aquaculture waste is heterogeneous due to

distinct components. The Multi-G model can be

used for an accurate representation of waste

degradation, and is very useful for examining the

degradation of recalcitrant materials within the

waste. Multi-G modelling has been carried out in

pond aquaculture, where various components of the

waste can remain within the system for long time

periods (Jamu, 1998). Unlike in a pond, in our screen

removal system, solid waste materials that are less

degradable than others do not remain in the system

(i.e. on the screen) but break down into small

particles, dissolved particles or dissolved waste

products that then slowly settle/diffuse into a large

deep body of water for wide dispersal. For this reason

Eqs. (7) and (8) were chosen over the Multi-G model

as our choice for representing waste degradation.

The necessary parameters, r and n, were experi-

mentally obtained as previously mentioned (see

Section 2.4 degradation rate experiments).

3.5. Model solution

The screenwaste degradermodelwas used to predict

daily waste accumulation on the screen in the following

way. Fish mass was calculated from the thermal growth

coefficient model (Eq. (2)). The total daily number of

fish on the farm was calculated assuming a mortality

rate. Solid waste production loss per fish was then

evaluated using theBTSmodel developed byBergheim

et al. (1991) (Eq. (6)). This waste was assumed to fall

over a specific area of screen on top of the waste from

the previous day.Waste materials in each layer decayed

according to Eq. (8). In that equation, the parameterW0represented the initial amount of waste in each new

layer, and the parameters r and n were experimentally

determined as previously mentioned. W had units of

specific area or kg of waste per square meter of screen.

Accumulationwas determined by adding the daily solid

waste input (I) to the amount of waste remaining on

each waste layer. For the solution, an accumulation

matrix was developed using Excel (Buryniuk, 2004). A

waste layer was assumed to be gone when it reached

0.0086 kg (dry basis) m2, as then due to buoyancy, it isnearly immeasurable. Output variables included days

after harvest for complete breakdownof solidwaste, the

amount of waste on each layer at a given time, as well

ngineering 35 (2006) 7890as, the total amount of waste at any given time.

produce the waste that falls onto a screen,

too. The screen would function best for fish over 1 kg

in size as the fraction of large particles (>3 mm) in asample appeared to be quite low for 1 kg fish. The

effect of the particles and clogging rate would have to

be empirically determined in a field study.

4.2. Solid waste characterization

The solid waste characteristics were compared with

those of a typical fish feed pellet in Table 2. TheC andN

contents of the solid waste were significantly less

( p < 0.05) that those found in the feed.Thesevalues arewithin the range of those reported previously for solid

fish waste and feed (Bergheim et al., 1993). However,

the C:N ratio was significantly greater in the waste (11)

versus that in the feed (5.8). Higher COD values were

obtained for the solid waste despite its lower %C

content. This suggests that the waste contains more

reactive components that are potentially more harmful

ral EThe results from 12 size fractionation experiments

are presented as cumulative fraction trapped (Fig. 3),

and indicate that approximately half of the solid waste

can be retained on a screen with an 8 mm mesh

opening; and more than 60% of the solid waste4.1.coResults and discussion

Particle size determination4. a bench scale experiment where the disappearanceof a measurable amount of material is recorded over

time.

The third option was chosen because of the

difficulty of accurately weighing in water tiny waste

pellets of nearly equal density as water (a requirement

in the first two options) and of obtaining a suitable

amount of sample material. The last option guaranteed

a sufficiently measurable mass within the water with

the disadvantage being that only the solid waste

degrader part of the compartment model would be

validated. The experimental set up was similar to the

method used for the degradation experiments with the

exception that no samples were removed for COD,

etc., the initial mass represented 4 days of accumu-

lated waste material from a fish cage near the end of a

fish production period, and the test period was 45 days.

During the test period, the submerged weight of a

screen plus the solid waste were measured every 4

days with a hanging scale (capacity: 2.7 kg, accuracy:

2 g), converted to above water dry weights using

published values of specific gravity and measured

moisture content (Chen, 1991), and compared to

model expectations. This validation experiment was

repeated twice.3.6. Model validation

Three possible options for model validation were

examined, and one was chosen based on accuracy and

technical feasibility. The options included:

a bench scale experiment where pieces of solidwaste are daily dropped onto a screen,

a pilot scale experiment where live fish are used to

M. Buryniuk et al. / Aquacultullected can be retained on a screen with a 5 mmmesh opening. These results assume no clogging of

the screen apparatus used to size fraction the waste.

Although a smaller mesh opening would support more

particles, it might be too heavy and costly to deploy in

the field. The optimummesh opening for a screen used

in the field would depend on the following factors: the

unknown fate in the environment of the tiny particles

that resulted from bacterial action, mooring system

design and actual clogging rate. Fish size is a factor

ngineering 35 (2006) 7890 85

Fig. 3. Solid waste particle size. Different symbols represent dif-

ferent fish sizes and the line is a trend line through the average at

each mesh opening.on the environment than uneaten feed pellets.

ural E

20 da

)

6)

.4 1)

0)*

8)***

renthe4.3. Suspended screen waste degradation

experiments

During the degradation experiments, ammonia

concentrations increased requiring the tank water to

be changed (24 times depending on the amount of

waste within the tank). In addition, bubbles and the

formation of foam were noted. The water was brown

for most of the study period, and clear near the end of

the experiment. From the beginning a tiny layer of

small particles started to accumulate on the sides and

bottoms of the tanks.

The solid waste characteristics measured at the end

M. Buryniuk et al. / Aquacult86

Table 2

Mean solid waste and fish feed characteristics at time = 0 and after

Solid waste initial Final

C (% dry weight) 37.5 0.7a (3) 40.3 2.3 (5N (% dry weight) 3.4 0.4a (3) 4.8 0.6b (5)S (% dry weight) 0.3 0.1 (3) 0.7 0.5 (5)C/N 11.1 1.3a (3) 8.5 0.6b (5)Ash (% dry weight) 29.1 3.8a (25) 32.0 3.4 (2COD (mg L g1 dw) 9.6 105 4.2 105a (16) 7.7 105 2NFE (% dry weight) 49.0 2.0a (4) 36.5 3.8b (5Moisture content

(% dry weight)

79.2 4.9 (8)* 77.0 1.5 (181.7 3.1 (13)** 78.9 4.2 (1

Error bars represent one standard deviation. Sample size, n, is in pa

different.* From a centrifuged sample.** From a frozen sample.*** Not centrifuged.**** Fresh feed pellets.of the 20 days degradation experiments revealed that

the C:N (from 11 to 8.5) and NFE content (from 48.6

to 36.5% dry wet) of the solid waste decreased

significantly. In contrast, for the fish feed pellet

degradation, only COD mg L1 g1 dry weightchanged significantly (from 2.9 105 to 1.2 106).We expected the mass specific COD for the solid waste

to also decrease with time as the easily degradable

material mineralizes and the ash content remains

constant. However, this was not the case since no

significant change in COD solids was observed. We

believe that ammonia formation specially near the

beginning, bubbling and the decreasing C:N with

decreasing total carbon and increasing fraction of total

N (%) were due to bacterial activity resulting in solid

waste breakdown. The fine particles accumulated on

the tank surface were probably transported there

through the screen by sedimentation or by convectioncurrent in the tank. These particles would be the most

likely particles to be transported away from an actual

site. In future investigations, the amount and nature of

the particles transported from the screen and their

potential effect on the benthos needs to be addressed.

Our data suggests that carbon may be limiting the

degradation rate of the solid wastes. The C:N ratios of

both the degraded feed and the collected solid waste

material were well below the suggested ratio of 30

generally considered optimum for organic matter

degradation (Boyd, 1995; Hamoda et al., 1998;

Avnimelech, 1999). As bacterial cells have a C:N

ratio of roughly 45:1, they require a feed source with

ngineering 35 (2006) 7890

ys of degradation at 8 8C in saline water

Fish feed initial Final

49.3 0.4b (2) 51.8 2.1 (2)8.5 0.6b (2) 8.4 0.0 (2)0.5 0.2 (2) 0.3 0.4 (2)5.8 0.4b (2) 6.2 0.3 (2)7.7 0.7b (11) 9.8 1.0 (10)

05 (12) 2.9 105 9.3 104b (8) 1.2 106 7.3 105a (4)39.4 4.1b (2) 37.4 0.3 (2)5.4 1.2 (5)**** 62.5 2.0 (3)*

66.1 4.3 (5)***

sis. Means with common superscipt (a and b) are not significantlya much higher C/N ratio in order for the available C to

meet both growth and other metabolic requirements

(Avnimelech, 1999; Boyd, 1995; Gaudy and Gaudy,

1980). Protein can be metabolized to meet energy

requirements under carbohydrate limitation but

growth would be slower. When the C:N ratio of the

degrading material is smaller than the optimal ratio of

30, excess organic nitrogen is mineralized; carbon

from other sources is required for continual degrada-

tion of the nitrogen material (Boyd, 1995; Avnime-

lech, 1999). To promote heterotrophic bacterial

breakdown, Avnimelech (1999) suggested an input

of carbon into the feed that would result in waste

production by the fish with a C:N ratio closer to 30.

Since it was not possible to distinguish between

mineralization, biological breakdown and transport by

sedimentation of the waste from the screen, overall

degradation rates (R + P) as in Eq. (1) were

ral Engineering 35 (2006) 7890 87

Fig. 5. Output from the model fell within the range of the experi-determined from the data. Degradation rates calcu-

lated from all experiments were plotted against area

density (Fig. 4) and used to fit an exponential kinetic

model (Eq. (7)). The order of solids degradation rate

with respect to initial areal concentration on the mesh

or n in Eq. (7) was found to be 1.19. The rate constant

(r) was 0.026 day1 (Eq. (7), Fig. 4). The coefficient

M. Buryniuk et al. / Aquacultu

Fig. 4. The waste removal rate depended on the amount of waste

material on the screen.of determination of the non-linear least squares

regression curve was 0.97. Since the order of reaction

was not much greater than 1. Linear regression was

also attempted, however, the resulting coefficient of

determination was less at 0.93. The coefficients

determined from the non-linear regression were,

therefore, used in Eq. (8), the waste degrader model.

4.4. Model validation

Fig. 5 is a plot of areal concentration of solids or

specific area (kg dry wet m2) versus time (days) fortwo experiments with solid waste. Over 3045 days

there was a significant drop in mass on the screen from

0.35 to 0.100.15 kg dry weight m2 as compared tothe initial 5 days ( p < 0.001). Simulations from thesolid waste degradation rate equation (Eq. (8)) model

fell within experimental values. Experimental error

bars in the figurewere generated based on the accuracy

of the scale employed in the trials (see validation

experiment, Section 3.6).5. Model applications

In this section the compartment screen waste

degrader model was used to simulate different

scenarios. After inputting parameters from actual

farm data (Table 1), and assuming a 100% capture

efficiency and no erosion due to current, the model

simulation of waste accumulation and recovery

indicated that the solid waste produced from a typical

farm operation and collected on a screen-like device

would be completely degraded within 8 months of a

total fish harvest, and 90% degradation would be

achieved within 4.5 months (Fig. 6). Eight months is in

line with site fallowing times as suggested by the

industry. A site with sufficient current to erode and

transfer away the waste on the screen may not

experience waste accumulation and therefore need to

mental values. The bars represent the accuracy of the scale used to

measure the submerged solid waste.be left fallow after harvest. A pilot scale investigation

could be used to show this.

Fig. 6. Output from the model indicated that after harvesting, the

screen would be empty of waste in 4.5 months.

McGhie et al. (2000) studied the degradation of

evice without current for eroding the waste could be

aster than benthic recovery at a site with current.

M. Buryniuk et al. / Aquacultural Engineering 35 (2006) 789088

Fig. 7. Effect of initial fish numbers (IFN) on solid waste accumu-

lation and breakdown on a screen-like device.ig. 9. Effect of capture efficiency (CE) on solid waste accumula-

on and breakdown on a screen-like device.solid waste from an Atlantic salmon farm in Australia.

To compare the predictions of this model to their data,

the following parameters were changed in the model.

The number of fish on the farm was set at 15,000 with

an initial weight of 1 kg. The production period was

273 days. This resulted in stocking densities approxi-

mately the same as those reported by McGhie et al.

(2000). All the remaining parameters were kept the

same as before (Table 1). Model simulations suggest

that the solid waste on a screen below this fish farm

would have been completely degraded in 169 days or

approximately 6 months. McGhie et al. (2000) found

evidence through fatty acid analysis that the sediments

were still affected by the fish farm 12 months after the

start of fallowing. This comparison suggests, although

a direct comparison is not possible due to the fate and

effect of the small particles from the screen-like

device is not known, that the recovery on a screen-likeFig. 8. Effect of feed conversion ratio (FCR) on solid waste

accumulation and breakdown on a screen-like device.

ig. 10. Effect of screen area on solid waste accumulation andFThe biomass of fish on the farm and the feed

conversion ratio of the fish directly affect the input of

solid waste onto the screen-like devices. A change in

these parameters (25% initial fish numbers and10% FCR) did not, however, greatly affect the timerequired to degrade the waste (Figs. 7 and 8). In terms

of fish farm management, this suggests that the

biomass of fish on the farm or the feed conversion ratio

will not greatly affect the engineering parameters of

the screen-like device and that the device, provided the

fish produce a sufficiently large waste particle, will be

of use over a greater range of applications.

The capture efficiency of the screen below the fish

farm greatly affects the simulated accumulation of

solid waste particles on the screen (50% captureefficiency) (Fig. 9). The size (spread of the solid

waste) over the screen has an effect on accumulation

but little on the time required for complete wasted

f

F

tibreakdown on a screen-like device.

M. Buryniuk et al. / Aquacultural Engineering 35 (2006) 7890 89staggered harvest will greatly affect the accumulation

of solid waste on the net and the time to completely

degrade the accumulated waste (Fig. 11). This is an

example of a farmmanagement practice that can affect

accumulation and degradation of solid waste regard-

less of the engineered design of the screen-like device.

Over 25% less waste would accumulate on the screen-

like device, potentially significant in terms of

engineering, and the time to degrade the waste

completely would be shorter by approximately 2

months. This is a farm management strategy that may

also be valid with current fish farm infrastructure.

Before the impacts of farm management strategies

such as the staggered harvest are examined, further

study is first required to determine the effect of a

screen on the benthic community.

Acknowledgmentsremoval (Fig. 10). This is important as depending on

the stiffness of the screen-like device, the waste could

pile up in one area. Fish farm practices such as a

Fig. 11. Effect of a staggered harvest on solid waste accumulation

and breakdown on a screen-like device.Appreciation for their financial support is extended

to Stolt Sea Farm and the National Science and

Engineering Research Council of Canada.

References

Ackefors, H., Enell, M., 1990. Discharge of nutrients from Swedish

fish farming to adjacent sea areas. Ambio 19, 2935.

Ang, K.P., Petrell, R.J., 1997. Control of feed dispensation in

seacages using underwater video monitoring: effects on growth

and food conversion. Aquacult. Eng. 16, 4562.

Ang, K.P., Petrell, R.J., 1998. Pellet wastage, and subsurface and

surface feeding behaviours associated with different feedingsystems in sea cage farming of salmonids. Aquacult. Eng. 18,

95113.

American Public Health Association, 1998. Standard Methods for

the Examination of Water and Wastewater. APHA, Washington,

DC, USA, 1192 pp.

Avnimelech, Y., 1999. Carbon/nitrogen ratio as a control element in

aquaculture systems. Aquaculture 176, 227235.

Bergheim, A., Forsberg, O., 1993. Attempts to reduce effluent

loadings from salmon farms by varying feeding frequencies

and mechanical effluent treatment. In: Barnabe, G., Kestemont,

P. (Eds.), Production, Environment and Quality. European Aqua-

culture Society, Special Publication No. 18, pp. 115124.

Bergheim, A., Sanni, S., Indrevik, G., Holland, P., 1993. Sludge

removal from salmonid tank effluent using rotating microsieves.

Aquacult. Eng. 12, 97109.

Bergheim, A., Tyvold, T., Seymour, E.A., 1991. Effluent loadings

and sludge removal from landbased salmon farming tanks.

Aquacult. Environ. 14, 29.

Berner, R.A., 1980. Early Diagenesis: A Theoretical Approach.

Princeton University Press, New Jersey, 241 pp.

Beveridge, M.C.M., 1996. Cage Aquaculture. Fishing News, Cam-

bridge, MA, USA, 346 pp.

Boyd, C.E., 1995. Bottom Soils, Sediment, and Pond Aquaculture.

Chapman and Hall, New York, 348 pp.

Brooks, K.M., Stierns, A.R., Backman, C., 2004. Seven year reme-

diation study at the Carrie Bay Atlantic salmon (Salmo salar)

farm in the Broughton Archipelago, British Columbia Canada.

Aquaculture 239, 81123.

Brooks, K.M., Stierns, A.R., Mahnken, C.V.W., Blackburn, D.B.,

2003. Chemical and biological remediation of the benthos near

Atlantic salmon farms. Aquaculture 219, 355377.

Buryniuk, M., 2004. Solid waste degradation on a screen in a cold,

oxygenated, saline environment: applications for salmon farm-

ing. Masters of Science Thesis. University of British Columbia,

135 pp.

CAIA, 2003. Canadian Aquaculture Industry Alliance. , www.a-

quaculture.ca.

Chen, S., 1991. Theoretical and experimental investigation of foam

separation applied to aquaculture. Ph.D. Thesis. Cornell U.,

Ithaca, NY, 239 pp.

Clarke, R., Phillips, M., 1989. Environmental impacts of

salmon aquaculture. Bull. Aquacul. Assoc. Canada 89-4, 24

31.

Cripps, S., 1995. Serial particle size fractionation and characteriza-

tion of an aquacultural effluent. Aquaculture 133, 323339.

Cripps, S., Bergheim, A., 2000. Solids management and removal for

intensive land-based aquaculture production systems. Aquacult.

Eng. 22, 3356.

Cromey, C.J., Nickell, T.D., Black, K.D., 2002. DEPOMOD

modelling the deposition and biological effects of waste solids

from marine cage farms. Aquacult. 214, 211239.

Edwards, P., 1998. A systems approach for the promotion

of integrated aquaculture. Aquacult. Eco. Manage. 2 (1),

112.

FAO, 2003. Food energy-methods of analysis and conversion fac-

tors. Report of a Technical Workshop. Rome, 36 December

2002, 100 pp.

Folke, C., Kautsky, N., 1989. The role of ecosystems for a sustain-

able development of aquaculture. Ambio 18, 234243.

Gaudy, A.F., Gaudy, E.T., 1980. Microbiology for Environmental

Engineers. McGraw-Hill, New York, 736 pp.

Hamoda, M.F., Abu Qdais, H.A., Newham, J., 1998. Evaluation of

municipal solid waste composting kinetics. Resour. Conserv.

Recycling 23, 209223.

Jamu, D.M., 1998. Modeling organic matter and nitrogen dynamics

in integrated aquaculture/agriculture systems: effects of cycling

pathways on nitrogen retention and system productivity. Ph.D.

Thesis. Office of Graduate Studies, Ecology, University of

Calfornia, Davis, CA.

Juell, J.E., Furevik, D.M., Bjordal, A., 1993. Demand feeding in

salmon farming by hydroacoustic food detection. Aquacult. Eng.

12, 155167.

Kempf, M., Merceron, M., Cadour, G., Jeanneret, H., Mear, Y.,

Mirammand, P., 2002. Environmental impact of a salmonid farm

on a well flushed marine site: II. Biosedimentology. J. Appl.

Ichthyol. 18, 5160.

Liao, P.B., Mayo, R.D., 1972. Salmonid hatchery water reuse

systems. Aquaculture 1, 317335.

Liao, P.B., Mayo, R.D., 1974. Intensified fish culture combining

water reconditioning with pollution abatement. Aquaculture 3,

6185.

Makinen, T., Lindgren, S., Eskelinen, P., 1988. Sieving as an effluent

treatment method for aquaculture. Aquacult. Eng. 7, 367

377.

McGhie, T., Crawford, C., Mitchell, I., OBrien, D., 2000. The

degradation of fish-cage waste in sediments during fallowing.

Aquaculture 187, 351366.

Naylor, R.L., Goldburg, R.J., Primavera, J.H., Kautsky, N., Bever-

idge, M.C.M., Clay, J., Folke, C., Lubchenco, J., Mooney, H.,

Troell, M., 2000. Effect of aquaculture of world fish supplies.

Nature 405, 10171024.

Shepherd, J., Bromage, N., 1988. Intensive Fish Farming. BSP

Professional Books, 404 pp.

Stigebrandt, A., 1999. Turnover of energy and matter by fisha

general model with application to salmon. Fisken og Havet 5, 29

pp.

Wheaton, F.W., 1985. Aquacultural Engineering. Wiley, New York,

NY, 708 pp.

M. Buryniuk et al. / Aquacultural Engineering 35 (2006) 789090

Accumulation and natural disintegration of solid wastes caught on a screen suspended below a fish farm cageIntroductionSolid waste characterization and degradation experimentsSolid waste materialParticle size distributionSolid waste characterizationDegradation rate experimentsStatistical analysis

Model developmentConceptual modelFish biomass predictorExisting model reviewWaste-model selection

Solid waste degrader modelModel solutionModel validation

Results and discussionParticle size determinationSolid waste characterizationSuspended screen waste degradation experimentsModel validation

Model applicationsAcknowledgmentsReferences