Page 3 - m313p261

Sarà et al: Dispersal of fish farm waste 263
Table 1. Measurements of current velocity frequencies (%)
within 5 cm s–1 increments up to 20 cm s–1 and higher in the s–1: HYDRO 2), and high as measured in Capo d’Or-
study areas Gulf of Castellammare (HYDRO 1: mean velocity
of current [MVC] = 12.0 ± 7.5 cm s–1), Menfi Bay (HYDRO 2:
MVC = 21.0 ± 11.8 cm s–1), and Capo D’Orlando Bay
(HYDRO 3: MVC = 41.4 ± 18.4 cm s–1)
Current velocity increments
(cm s–1)
5 10 15 > 20
HYDRO 1 23 44 25 8
Subsurface (3 m) 22 27 23 27
Mean depth (10 m) 20 24 23 33
Bottom (30 m)
40 30 0 30
HYDRO 2 50 30 0 20
Subsurface (3 m) 43 37 0 20
Mean depth (10 m) 33 0 0 66
Mean depth (20 m)
Bottom (30 m) 14 26 17 44
5 24 22 49
HYDRO 3 4 17 19 60
Subsurface (3 m) 4 8 14 74
Mean depth (10 m)
Mean depth (20 m)
Bottom (30 m)
was immediately frozen at –20°C and stored until
analysis. About 10 specimens of cultivated fish were
sampled and muscle tissue dissected and, after an
overnight evacuation, fish faeces were collected using
pipettes. In addition, samples of pelletted foods used
in the fish farms were collected for carbon and nitro-
gen isotopic analysis. Water (POM) and sediments
(SOM) as well as muscles, faeces and pellets were
analysed to measure carbon and nitrogen isotope
ratios. All samples were acidified in 2 N HCl, dried at
60°C for at least 24 h, then ground. Isotopic analyses
were performed using a Finnigan Delta-Plus isotope
ratio mass spectrometer. Isotopic values were ex-
pressed in parts per thousand as well as deviations
from standards (Peedee belemnite limestone for δ13C,
and nitrogen in air for δ15N):
δ13C or δ15N = [(Rsample/Rstandard)] × 103
where R is 13C/12C or 15N/14N.
Sampling design and statistical analyses. The
control-impact sampling design enabled comparisons
between putatively impacted and control locations
over different distances. We used a series of partial
(0 vs. 500 m; 0 vs. 1000 m; 500 vs. 1000 m) 3-way
ANOVAs to evaluate the hypothesis that δ13C and
δ15N of POM and SOM varied as a function of dis-
tance from the cages. The 1st factor (hydrodynamics:
HYDRO) was treated as fixed and orthogonal. The 3
levels of hydrodynamics were low as measured in the
Gulf of Castellammare (MVC ~12 cm s–1: HYDRO 1),
intermediate as measured in Menfi Bay (MVC ~22 cm
lando Bay (MVC ~40 cm s–1: HYDRO 3). The control
location (Egadi Islands) was considered in the
ANOVA as the 4th level of the HYDRO factor, with a
MVC of ~20 cm s–1 (Lazzari et al. 1994). Distance was
the 2nd factor in the ANOVA (3 levels: DISTANCE).
Distance was treated as fixed and orthogonal. Sites
(2 for each distance) were treated as random (2 levels:
SITE) and nested in the interaction of the above
factors. Three replicates were randomly collected at
each site. Heterogeneity of variances was tested using
Cochran’s C-test, and appropriate means were com-
pared using Student-Newman-Keuls (SNK) test (Un-
derwood 1997).
We used the mixing model approach to infer the
most important carbon and nitrogen sources qualita-
tively responsible for the isotopic composition of POM
and SOM (Phillips & Gregg 2003, Sarà et al. 2003). We
designated POM and SOM carbon and nitrogen sig-
nals (as measured throughout the study locations) as
targets in the models. The end-members in the models
were pelletted food (PEL: δ13C = –22.6 ‰; δ15N = 7.8 ‰)
and fish faeces signals (EJE: δ13C = –23.1 ‰; δ15N =
10.7 ‰) as measured throughout the study locations.
The phytoplankton (PHY: δ13C = 19.0 ‰; δ15N = 6.5 ‰),
sand microflora (SM: δ13C = –15.0 ‰; δ15N = 6.5 ‰) and
terrestrial signals (TER: δ13C = –26.4 ‰; δ15N = 1.9 ‰)
were extrapolated from the literature (Dauby 1989,
Martinotti et al. 1997, Riera 1997, Middelburg & Nieu-
wenhuize 1998, Mazzola & Sarà 2001, Sarà et al. 2003,
Sarà et al. 2004).
Differences in percent contribution of each carbon
and nitrogen source (terrigenous, primary [e.g. phyto-
plankton or sand microflora], and fish farm waste) at
each combination of site and hydrodynamic level were
analysed by cluster analyses (CA). Bray-Curtis similar-
ity was used as a distance index. SIMPER procedure
(cut-off: 100%) was employed to determine the or-
ganic matter source, which contributed the most to
observed dissimilarities (Clarke 1993). Data were
square-root transformed. The GMAV 5.0 statistical
package (University of Sydney) and the PRIMER pack-
age (Plymouth Marine Laboratory) were used to per-
form uni- and multivariate analyses. The ISOSOURCE
(Phillips & Gregg 2003) software package was used to
run mixing equations.
RESULTS
Mean, minimum and maximum values of δ13C and
δ15N in POM and SOM, recorded within each distance
category at each location (including control sites
located at the Egadi Islands), are shown in Table 2.