Ž .
deployed in groups of five in a circle of 3-m radius at each of three plots A, B and C ca. 50 m apart. Samples of surface sediment were collected at each of these plots in
March 1998 and at one plot in March 1997. These samples were analysed for bulk Ž
. density, volatile organic matter content combustion at 4008C for 18 h , particulate
Ž nitrogen, and total organic carbon Perkin-Elmer 2400 CHN elemental analyzer, after
. fuming with HCl to remove carbonates . Concentrations of Cu and Zn in these samples
Ž .
were measured by flame AAS Perkin-Elmer 3110 after extraction with 2 M HCl Ž
. Williamson et al. 1995 .
Ž .
Ž Benthic fluxes of dissolved oxygen DO , N NH and NO rNO , the latter referred
4 2
3
. Ž
2y
. to as NO hereafter , and sulphide S
were determined at Sites 1 and 2, as well as at
x
Ž .
control sites using the chambers described by Burns et al. 1996 . In summary, these Ž
. consisted of unstirred, opaque sampling depth was below the euphotic zone , plastic
Ž
2
. basins volume — 8 l, area of sediment enclosed 0.07 m
sampled from the surface via a 3-mm ID semi-rigid nylon tube. There were two chambers at each of the three plots at
each site. At each sampling, an appropriate volume of the initial water drawn through the tube was discarded to avoid sampling residual water in the tube. Volume-compensa-
tion water entered through a small hole in the chamber wall as each sample was
Ž withdrawn. DO was measured immediately using a calibrated BOD bottle probe YSI
. Model 5730 . Separate samples of water were taken for nutrient and sulphide analyses
and the latter were fixed with zinc acetate. Concentrations of DO, nutrients, and S
2y
in compensation water were obtained from water collected just above the sediment with a
van Dorn sampler. After being gently lowered to the seabed, each chamber was left for 0.5–1 h to allow any sediment disturbed during deployment to settle. Chambers were
sampled at several intervals up to 24 h after deployment. Estimates of fluxes were obtained from time-plots of concentrations of the various chemical species in the
chambers. Concentrations of nutrients were measured by Alpkem continuous flow air segmented autoanalyser. Concentrations of sulphide were determined spectrophotometri-
Ž .
cally using the method of Rees et al. 1971 . DO concentrations in the water-column Ž
. were measured in situ by DO meter YSI Model 58 fitted with long cable and stirrer .
3. Results and discussion
3.1. Current Õelocities Ž
In January 1997, the minimum 2 h-average velocity 1-m above the seabed the .
y1
summary statistic used by Findlay and Watling, 1997 was 1.1 cm s at Site 1, but
y1
Ž .
values as low as 0.2 cm s were often recorded over shorter time-intervals Table 1 .
In March 1998, minimum 2-h average velocity was 2.0 cm s
y1
at Site 1 and minimum
y1
Ž values recorded over shorter intervals were ca. 0.4 cm s
. An earlier study NIWA, .
y1
unpubl. data reported current velocities of F 5 cm s at Site 2 and at a site on the
northern side of the bay, even during moderate winds and rough seas. These values suggest that it is reasonable to generalize current data from Sites 1 to 2. Maximum
velocity recorded during the present study was 15 cm s
y1
.
The model was not adapted for the difference in water-temperature between Findlay Ž
. and Watling’s study and the present study 158C vs. 128C for the following reason.
Ž . Maximum rates of benthic oxygen supply J for the model are derived from Fick’s first
law: J s D C y C
rZ Findlay and Watling’s Eq. 1, p. 151
Ž .
Ž .
` d
where D is the molecular diffusion constant of oxygen at ambient temperature, C and
`
C are the concentrations of oxygen in the bulk water and at the sediment–water
interface, respectively, and Z is the thickness of the diffusive boundary layer. Follow-
d
Ž ing Findlay and Watling, we used concentrations of oxygen confirmed by in situ
. measurements at saturation and zero for C
and C , respectively. Consequently, all
`
terms on the right hand side of the equation are multiplicative. The ratio of saturation concentration of oxygen at 128C to that at 158C is 1.07, and the ratio of D at 128C to
Ž that at 158C varies between 0.88 and 0.99 depending on the method of derivation Wilke
. and Chang, 1955; Wise and Houghton, 1966; Broecker and Peng, 1974 . Therefore, the
product of 1.07 and these values for the ratio of D varies about 1. Since the maximum change in J due to temperature is only 6, we chose to ignore this and note that the
model’s prediction of maximum rate of oxygen supply is probably robust to differences in ambient temperature within this range.
3.2. Rates of input of carbon at an operating fish farm Ž
. Rates of input of carbon to the sediments below the operating farm Site 2 , measured
by the sediment traps deployed 1 m above the seabed, ranged from 463–967 mmol C
y2 y1
Ž .
m d
Table 2 and ranges were similar for both years. Rates at the control sites in 1998 ranged from 46 to 53 mmol C m
y2
d
y1
. Traps at Site 2 were placed in a 5-m gap Ž
. between cages rather than directly underneath them and, at the time of deployment, the
cage on one side of each set of traps was stocked with an adult fish and that on the other side was empty. In 1998, trap-set 2 at Site 2 was positioned nearer to the empty cage
Ž than the stocked cage and received less sediment than trap-sets 1 or 3 by factors of 2.7
. and 1.6, respectively , indicating the degree of within-farm variation in input of waste to
Ž .
the seabed. Pridmore and Rutherford 1992 reported rates of sedimentation 5–10 times larger than ours at the same farm, but their sediment traps were placed 1 m below the
bottom of the cages and feeding techniques on the farm have become more efficient since the time of their study. Our rates of input of carbon lie within the ranges reported
Ž .
in many other studies reviewed by Hargrave 1994 . If the depth below the cages is 15
y1
Ž m, average rate of sinking of food pellets is 12 cm s
Gowen and Bradbury, 1987; .
y1
Ž .
Findlay and Watling, 1994 and average current velocity is 2.8 cm s Table 1 , waste
food will be dispersed 3.5 m beyond the cages and fecal material, being less dense, is Ž
. likely to travel further Findlay and Watling, 1994 . The data from the sediment traps
therefore probably provided a reasonable estimate of rates of input of waste to the seabed in our study area.
Ž .
We used the mean values from the traps at the operating farm Site 2 for each year Ž
y2 y1
. 820 and 638 mmol C m
d for 1997 and 1998, respectively and the maximum
Ž
y2 y1
. value for a single trap-set 967 mmol C m
d as summary statistics for rates of
Ž
y1
. Ž
input. These were combined with minimum 2-h average velocity 1.1 cm s and in a
. Ž
y1
. separate calculation unaveraged minimum velocity 0.2 cm s
to estimate rates of benthic oxygen supply and demand from Findlay and Watling’s regression equations
Ž Ž .
Ž . .
their Eqs. 3 and 4 , p. 155 : O supply mmol m
y2
d
y1
s 736.3 q 672.5log Õ
Ž .
Ž .
2 10
Ž
y1
. where Õ is current velocity cm s
, O demand mmol m
y2
d
y1
s 1.07x y 32.6
Ž .
2
Ž
y2 y1
. where x is carbon flux to the seabed mmol m
d .
Ž . If the ratio rate of O supplyrrate of O demand I is ca. 1, the Findlay–Watling
2 2
model predicts that sediments will be ‘‘moderately impacted’’, however, if I is - 1, Ž
sediments will be severely impacted i.e., azoic and Beggiatoa-type mats will be .
present . When I is much less than 1, waste material will accumulate on the seabed. Values of I for Site 2 in the present study were 0.90, 1.18, and 0.76 for a current
velocity of 1.1 cm s
y1
and rates of carbon input of 820, 638, and 967 mmol m
y2
d
y1
, respectively, suggesting that a severe impact would be expected, but not an accumula-
Ž
y1
. tion of waste. When the minimum observed current velocity 0.2 cm s
was used, for purposes of comparison, the equivalent values of I were 0.32, 0.41, and 0.27. In
contrast to values derived using the minimum 2-h average velocity, these values suggest severe impact and that, at the higher end of the observed range of carbon input, the
accumulation of waste is likely. This conclusion was supported by measurements of the depth of accumulated material on the bed below this farm. The depth ranged from 0 to
31.5 cm, the surface sediment was uniformly black and numerous patches of Beggiatoa-like growths were present. These observations of spatially variable accumula-
Ž .
tion of waste over scales of metres within plots and 10s of metres between plots corresponded with the measured amounts and variability of input of carbon recorded by
the sediment traps. Our estimates of I indicate the sensitivity of the model to the value Ž
. of current velocity used. Findlay and Watling 1997 pointed out that this sensitivity is
greatest at slower velocities. 3.3. RecoÕery of a disused salmon-farm site
Ž .
Mean depth of waste material on the seabed in March 1997 Table 3 varied from 32 Ž
to 39 cm among the three plots at Site 1 differences among plots were significant: .
Kruskal–Wallis test statistic s 6.196, P - 0.05 . The mean decrease in depth of the waste layer between 1997 and 1998 was 12.0, 12.5, and 5.6 cm at Plots A, B, and C,
Ž .
respectively Table 3 . Some of the stakes had been lost, probably due to entanglement in surface-marker lines. There was considerable variation in the change in depth at Plot
Ž .
Ž .
C ranges among the five stakes 0–10 cm relative to Plot A ranges 10–15 cm . Dividing the mean depth at each plot in March 1997 by the rate of decrease in depth
gives recovery periods from the cessation of stocking of the cages of 3.3, 3.1, and 7.5 Ž
. years for Plots A, B, and C, respectively Table 3 .
The total amount of carbon in the waste layer present at each plot at Site 1 in March Ž
y2
. 1997 mmol C m
was derived from measurements of depth of the waste layer, and of
D.J. Morrisey
et al.
r Aquaculture
185 2000
257 –
271
265 Table 3
Ž .
Values of sediment-related variables, depth of accumulated waste, and change in depth of waste between March 1997 and March 1998 at three plots A–C at the Ž
. Ž
. disused salmon farm Site 1 . Estimated periods of recovery mineralisation of accumulated waste of the three plots are also given. ‘BD’ — bulk density measured in
Ž
y1
Ž .
samples collected in 1998 a mean value of 0.29 g ml 0.02 SE obtained from these samples and others collected in 1997 was used to estimate changes in amount
. Ž
. of sediment at each plot ; ‘C’ and ‘N’ — percentage of C and N in the sediment dry weight basis , respectively; ‘MR’ — estimated rate of mineralisation of
carbon — MR1 is calculated from observed rate of decrease in depth of waste layer, MR2–5 from the Findlay–Watling model assuming minimum current velocities
y1
Ž .
of 0.2, 0.4, 1.1, and 2.0 cm s , respectively see text for details
y1
Ž .
Ž .
Plot BD
C N
Depth of waste cm Decrease cm yr
MR 1 Recovery
y1 y2
y1
Ž .
Ž .
Ž . g dry weight ml
March 1997 1997–1998
mmol m d
time yr Ž
. Ž
. Ž
. Ž
. 95 CI
95 CI 95 CI
95 CI Ž
. Ž
. Ž
. Ž
. A
0.30 13.0
2.4 36.2 29.3–43.1
12.0 8.6–15.4 1035 739–1332
3.3 1.8–4.7 Ž
. Ž
. Ž
. Ž
. B
0.25 14.6
2.6 32.3 27.5–37.1
12.5 y32.0–57.0 1206 y3084–5496
3.1 y10.3–16.5 Ž
. Ž
. Ž
. Ž
. C
0.29 12.8
2.4 39.3 36.1–42.5
5.6 0.2–11.0 476 12–939
7.5 0.9–14.1 Ž .
Ž . Plot
Amount C at 3.97 MR 2
Recovery time yr MR 3
Recovery time yr
y2 y2
y1 y2
y1
Ž .
Ž .
Ž .
Ž .
Ž .
mmol m mmol m
d 95 CI
mmol m d
95 CI Ž
. Ž
. A
1,140,584 279
11.2 9.1–13.3 468
6.7 5.4–8.0 Ž
. Ž
. B
1,138,082 279
11.2 9.5–12.8 468
6.7 5.7–7.7 Ž
. Ž
. C
1,219,440 279
12.0 11.0–12.9 468
7.1 6.6–7.7
y1
Ž .
Current velocity cm s 0.2
0.4 Ž .
Ž . Plot
Amount C at 3.97 MR 4
Recovery time yr MR 5
Recovery time yr
y2 y2
y1 y2
y1
Ž .
Ž .
Ž .
Ž .
Ž .
mmol m mmol m
d 95 CI
mmol m d
95 CI Ž
. Ž
. A
1,140,584 745
4.2 3.4–5.0 908
3.4 2.8–4.1 Ž
. Ž
. B
1,138,082 745
4.2 3.6–4.8 908
3.4 2.9–3.9 Ž
. Ž
. C
1,219,440 745
4.5 4.1–4.8 908
3.7 3.4–4.0
y1
Ž .
Current velocity cm s 1.1
2.0
Ž .
the bulk density and carbon content of the material Table 3 . Given the low current Ž
y1
. velocities at this site maximum recorded velocity during the study was 15 cm s
, Ž
resuspension and dispersal of the sediment is unlikely Amos et al. 1992, Maa et al. .
1993 . We therefore assumed that the reduction in depth of waste was due to decomposi- tion, although resuspension andror compaction may also have occurred, causing
overestimation of the rate of decomposition. Using the minimum current velocities Ž
. discussed above and Findlay and Watling’s Eqs. 2–4 , we estimated the maximum rate
Ž .
of mineralisation of carbon in the sediment at each plot assuming I s 1 . Dividing the Ž
. total mass of carbon present in the waste at each plot Table 3 by the maximum
predicted rate of mineralisation provided a second estimate of the time for recovery of the sediment. Recovery times for the three plots based on minimum 2-h average
Ž .
velocities ranged from 3.4 to 4.5 years Table 3 . Times based on minimum observed velocities ranged from 6.7 to 12.0 years.
Recovery times estimated from the model using minimum 2-h average velocities corresponded reasonably well with those calculated from observed rates of decrease in
the depth of the layer of waste. As with estimates of impact, predictions were sensitive to estimates of time-averaged current velocity, but minimum 2-h average velocities
appeared to give better predictions than unaveraged minima.
Rates of mineralisation of waste derived from the rate of decrease in depth of waste
y2 y1
Ž
y2 y1
. at Site 1, 5.7–14.5 g C m
d 476–1206 mmol C m
d , are comparable to
values based on minimum 2-h averaged current velocity of 1.1 cm s
y1
derived using the Ž
y2 y1
y2 y1
. Findlay–Watling model 8.9 g C m
d or 745 mmol C m
d . They are also
comparable to values reported in other studies. Rates of mineralisation of waste in the
y2 y1
Ž
y2 y1
. sediments at Site 1 were estimated at 5–6 g C m
d 417–500 mmol C m
d in
Ž .
Ž .
a study in the mid-1980s NIWA unpubl. data . Angel et al. 1995 estimated rates of
y2 y1
Ž
y2 y1
. mineralisation of 1.8–5.0 g C m
d 150–417 mmol C m
d below a farm in
the Red Sea. 3.4. Chemical fluxes from sediments
Fluxes of all the chemical species measured showed large variation between replicate chambers at the same plot, among replicate plots at the same site and, in the case of
Ž .
measurements at the disused farm Site 1 in 1998, between replicate times of sampling Ž
. 3 days apart at the same position Table 4 . Fluxes were similar between years and
Ž .
between the two farm-sites Sites 1 and 2 , apart from one plot at the operating farm Ž
. Site 2 that showed much larger fluxes than other plots. Fluxes at the farm-sites were
very much larger than at controls. These data indicate that there is large variability in sediment processes across the impacted sites and confirm the impacts relative to the
non-impacted controls. Fluxes of NH –N and NO –N were combined and multiplied by a conversion factor
4 x
Ž .
to provide estimates of rates of mineralisation of C. The conversion factor 4.286 was derived from the equation C H NO q 5O
™5CO q2H OqNH , in which 5 mol
5 7
2 2
2 2
3
Ž .
of C are mineralised for each mole of ammoniacal-N produced Fritz et al., 1979 . Fluxes of NH –N and NO –N were combined, based on the assumption that fluxes of
4 x
Ž NO were due to nitrification of NH –N mineralised from the sediments e.g., Black-
x 4
Table 4 Ž
. Rates of benthic chemical fluxes at the disused and operating salmon farms Sites 1 and 2, respectively and
Ž the control sites. At Site 1 in 1998, measurements were repeated, 3 days apart, at each plot chambers were
. lifted between times of sampling . Estimates of rates of mineralisation of C are also shown, based on fluxes of
2y 2y
Ž .
N, S , and NqS
. There were two chambers y1 and y2 , 5 m apart, at each of the three plots at each site. Locations of control plots were not the same in 1997 as in 1998 and are labeled differently. ‘nr’ — not
recorded
y2 2y
Plot- Date
Period DO g m NO –N NH –N
S mg
C minera- C minera- C minera-
X 4
y1 y2
y2 y2
y1
Ž . chamber
h d
mg m mg m
m d
lisation lisation
lisation
y1 y1
Ž . Ž .
Ž .
d d
N mmol S mmol
total mmol
y2 y1
y2 y1
y2 y1
m d
m d
m d
Site 1 A-1
1997 6.4
y0.21 724
3010 259
188 447
A-2 1997 21.4
y0.42 269
2760 96
172 268
A-1 1998
7.3 y0.46
y3.1 351
nr 127
5.8 y0.60
y4.5 250
nr 91
A-2 1998
7.5 y0.26
y0.6 443
nr 159
5.9 y0.88
y4.0 165
nr 61
B-1 1997
8.6 y0.60
196 1220
70 76
146 B-2
1997 22.3 y0.56
338 2910
121 181
302 B-1
1998 7.3
y0.51 y2.7
178 nr
65 6.0
y0.81 y2.9
96.1 nr
35 B-2
1998 7.4
y0.99 y2.3
144 nr
53 6.1
y0.71 y4.8
157 nr
58 C-1
1997 7.3
y0.81 429
2470 154
154 308
C-2 1997 22.3
y0.56 716
4310 256
269 525
C-1 1998
7.2 y0.62
y7.9 417
nr 152
6.0 y1.57
y5.6 474
nr 171
C-2 1998
7.3 y0.61
y9.4 476
nr 173
6.1 y0.99
y5.6 490
nr 177
Site 2 D-1
1998 6.3
nr nr
nr nr
D-2 1998
3.0 0.05
y8.6 460
nr 167
E-1 1998
2.8 y0.48
y2.6 622
nr 223
E-2 1998
6.4 y0.08
8.2 179
nr 67.0
F-1 1998
2.7 y0.85
1.5 2910
nr 1040
F-2 1998
2.8 y4.11
y14.9 30,300
nr 10,800
Control sites G-1
1997 5.3
y1.07 1.5
9.19 3.83
3.83 G-2
1997 5.3
y0.88 4.1
3.10 2.59
2.59 H-1
1997 5.1
y1.87 y3.2
8.01 nr
4.01 I-1
1997 26.8 nr
y4.2 9.01
10 4.70
0.63 5.33
I-2 1997
4.9 y0.94
y14.9 5.52
70 7.30
4.38 11.68
J-1 1998 21.7
y0.21 3.1
1.64 nr
1.70 J-2
1998 21.8 0.03
2.4 y1.63
nr 0.27
K-1 1998 21.4
0.05 0.5
y2.38 nr
y0.67 L-1
1998 0.9
y1.79 14.8
10.9 nr
9.19 L-2
1998 21.4 y0.45
1.8 10.1
nr 4.25
.
2y
burn, 1996 . Fluxes of S were converted to rates of mineralisation of C using a
Ž .
Ž .
2y
conversion factor 0.75 derived from the equation 2 n CH O q nSO ™2nCO q
2 4
2
2 nH O q nS
2y
, in which 2 mol of C are mineralised for each mole of S
2y
produced
2
Ž .
Dahlback and Gunnarsson 1981 .
¨
Although the use of unstirred benthic chambers will potentially underestimate the Ž
. nutrient fluxes across the sediment–water interface Glud et al. 1995 , their use at these
study sites was considered realistic because the minimum 2-h averaged current velocity was about 1 cm s
y1
. Effects of flow in stirred benthic chambers appear to be more pronounced at velocities above 2 cm s
y1
, where an increase in rate of flux of 25 may Ž
. occur Glud et al. 1995 . Consequently, our non-stirred chamber may be underestimating
the rate of mineralisation of carbon, but probably not by more than 25. Estimates of rates of mineralisation of C from benthic chambers at the disused farm
Ž .
2y
Site 1 were obtained by adding the rates based on N flux and those based on S flux
because these two processes are mutually exclusive, and therefore, additive. They ranged
y2 y1
Ž . Ž
. from 146 to 525 mmol m
d mean 333, 95 CI 191–474
Table 4 . These were comparable with, but generally smaller than, rates derived from measured decreases in
Ž the depth of the layer of waste at the farm and from the Findlay–Watling model Table
.
2y
3 . Estimates were based on fluxes measured in 1997, since data on fluxes of S were
not available for 1998. Rates of mineralisation of C based on fluxes of N in 1997 and Ž
y2 y1
1998 were not significantly different mean rate in 1997 — 159.30 mmol m d
, .
95 CI 74.50–244.11; mean rate in 1998 — 110.09, 95 CI 75.21–144.97 , and since fluxes of nitrogen species and sulphide were approximately 1:1 over the range of release
Ž .
rates measured Table 4 , this suggests that rates were likely to be similar at both times Ž
. of sampling. Earlier studies at this site
NIWA, unpubl. data estimated rates of
mineralisation of C to be 417–500 mmol m
y2
d
y1
, 83–167 mmol m
y2
d
y1
of which was derived from efflux of methane, while 188 mmol m
y2
d
y1
was by release of S
2y
and 158 mmol m
y2
d
y1
was by sediment oxygen demand. These values lie within the Ž
y2 y1
. range 146–525 mmol m
d obtained with the benthic chambers in the present
Ž .
study Table 4 . Methane production was not measured in the present study. Adding an Ž
. equivalent proportional contribution 19–32 by methanogenesis from the earlier study
to the estimated of rates of mineralisation based on nutrient fluxes in this study gives values of 170–700 mmol C m
y2
d
y1
. 3.5. Potential effects of heaÕy metal contaminants on recoÕery of farm sites
Mineralisation of accumulated waste is the first stage in the process of recovery, facilitating recolonisation by benthic organisms. It implies that the impact of marine
farms on the benthic environment is a reversible process. Recolonisation may be impaired, however, by the presence in the sediments of persistent contaminants, such as
Ž .
heavy metals. Studies of sediments below farms in Scotland Edwards, 1998 and Ž
. Canada British Columbia Environmental Assessment Office, 1997 have reported high
Ž .
concentrations of copper derived from antifoulant on the cages; Beveridge, 1996 and Ž
. zinc derived from food; British Columbia Environmental Assessment Office, 1997 .
Ž .
y1
Concentrations of copper up to 725 mg g dry weight sediment and of zinc 1150 mg
Ž .
y1
Ž .
g dry weight sediment were found below cages in Scotland Edwards 1998 . In the
present study, the concentration of zinc in sediments from the three plots at Site 1 was Ž
.
y1
Ž 665 mg g dry weight sediment
19.5 SE, n s 5; mean concentration at nearby Ž
.
y1
Ž .
control sites: 18 mg g dry weight sediment 2.1 SE, n s 3 . Concentrations of
copper were generally similar to nearby control sites, although a few individual samples contained higher concentrations. If copper in the sediments is derived from particles of
antifoulant paint dislodged from the cages, locally high but spatially variable concentra- tions are likely in the period after first use of the paint.
The zinc concentrations exceed the sediment quality criteria proposed by Long et al. Ž
. Ž
1995 at which adverse biological effects of the contaminant are likely to occur 410 Ž
.
y1
. Ž
. mg g dry weight sediment
. Watzin and Roscigno 1997 found differences in recruitment of benthic invertebrates between control sediments and sediments experi-
Ž Ž
.
y1
. mentally spiked with zinc to concentrations 646 mg g dry weight sediment
, similar Ž
. to those measured at the disused farm site Site 1 in the present study. These differences
included reduced numbers of individuals of several taxa in the zinc-spiked sediment. Bioavailability of heavy metals in sediments is controlled by the relative abundance of
Ž metal-binding phases, such as organic matter and sulphides Casas and Crecilius, 1994;
. Ankley, 1996 , and is therefore, likely to be relatively low in organically enriched,
anoxic sediments. As marine farm sites recover, in terms of mineralisation of accumu- lated waste, concentrations of metal-binding phases will decrease and heavy metal
contaminants will become more bioavailable and potentially toxic. Consequently, recov- ery of the benthic communities may be impaired and the assumption that benthic
impacts of farms are completely reversible may not be appropriate.
4. Conclusions