Page 6 - Cubadda2001

566 F. Cubadda et al. / Chemosphere 45 (2001) 561±569
the di€erence with the other stations did not reach the
level of statistical signi®cance.
4. Discussion
ANCOVA has been proposed and successfully em-
ployed by many authors to handle the e€ect of covari-
ates such as body size on metal concentrations in aquatic
organisms (Bendell Young and Harvey, 1991; Van
Hattum et al., 1991; Rainbow et al., 1998). The superi-
ority of the ANCOVA approach to other statistical
methods often used in biomonitoring has been stressed
by some authors (Mastala et al., 1992; Evans et al.,
1993).
In the present study the inclusion of body weight as
covariate in the statistical analysis explained from 81%
to 99% of metal variability within organisms (Table 5).
The percentage of variability explained by body weight
itself ranged from 8% to 75%. The picture provided by
ANCOVA appears more distinct compared to that ob-
tained in a previous study in which only individuals of
M. turbinata and P. caerulea belonging to a narrow size
range were analysed (Campanella et al., 2001). In that
case, for instance, contrasting results were obtained for
Cu in the two species, with P. caerulea showing the
highest concentration in station 1, where M. turbinata
had the lowest one. Here, analysis of a larger number of
individuals with di€erent sizes and adjustment to a
common body weight over all sites led to somewhat
di€erent results for P. caerulea. In fact, the individuals for the selected species using the soluble and total metal
from station 5 were found to contain the highest level of
Cu and station 1 turned out to be in the range of the
other four stations. Therefore, by means of ANCOVA
we could achieve improvements in terms of both re-
moval of biases in estimating mean metal concentrations
and power to detect di€erences among sites.
However ANCOVA has also some drawbacks, as a
number of conditions must be met for its application,
notably the existence of a linear metal-to-size relation-
ship and parallelism of regression lines. In this study on
two occasions (Pb in M. turbinata, Cr in P. caerulea) the
regression lines for di€erent stations did not show a
common slope. This result could be due to either the
restricted number of individuals sampled at some sta-
tions or to an inherent variability of the biological spe- worth noting that, even though in the calculation of CF
cies. This matter has been discussed by Boyden (1974,
1977), who in his pioneer studies reported many cases in
which the regression slopes were constant, but recog-
nized the existence of variations in relation to environ-
mental factors in some species. Cossa et al. (1980)
challenged the idea that slopes are generally constant for tissues are substantially in¯uenced by metals accumu-
a given metal and species, observing that the regression
slopes for a number of metals in Mytilus edulis were
variable with season and location and were in¯uenced
by metabolic changes associated with sexual maturation.
The problems which arise in the use of ANCOVA
when either slopes are variable or the metal-to-size re-
lationships are curvilinear (not observed in this investi-
gation) have been debated by many authors, and
alternative statistical methods have been proposed
(Tremblay et al., 1998). However, despite these dicul-
ties, our results undoubtedly con®rm that a regression
approach involving selection of several individuals over
a large size range and subsequent analysis of metal-to-
size relationship is the most appropriate tool in moni-
toring studies. In this way organisms sampled in distinct
locations can be compared meaningfully even though
mean sizes di€er. Secondly, this approach may provide
considerable insight into the responses of biomonitors
under variable environmental conditions.
One of the goals of the present study was to provide
background levels against which to perform intraspeci®c
comparison within the Mediterranean area. Obviously,
when dealing with biomonitors, the notion of `back-
ground levels' must be viewed with caution. In fact, even
in an unpolluted area, speci®c geochemical or hydro-
dynamic conditions may always shift `natural' contents
of one or more metals. However, when compared to the
values reported in the literature (Bargagli et al., 1985;
Nicolaidou and Nott, 1990; Ozretic et al., 1990; Nico-
laidou, 1994; Kozanglou and Catsiki, 1997), the levels
found in the current study for Cr, Cu, Pb and ± to a
smaller extent ± for Zn, do appear very low. Neverthe-
less, high concentration factors (CFs) can be calculated
concentrations in seawater reported in our previous
study (Campanella et al., 2001) (Table 6). These data
clearly indicate that all the four gastropods are very
ecient Cd accumulators, with P. caerulea and P. lusi-
tanica exhibiting the highest CFs. The species of the
genus Monodonta are also strong net accumulators of
Cu and Zn. The very high Cd CFs obtained for the
species of the genus Patella are remarkable as these
species are consumed as seafood in many Mediterranean
countries.
Accumulation of high levels of metals even in un-
polluted areas must be related to feeding habits on the
one hand, and to morphological and physiological fea-
tures on the other hand. As regards the former issue it is
ratios metal concentrations in seawater were employed,
the aqueous medium is only one of the sources of metals
for the organisms studied here. As all the species con-
sidered in this investigation are herbivorous, it can be
supposed that the metal concentrations in their soft
lated in the algae on which they graze.
Irrespective of their actual uptake mechanisms, the
four selected species provided a rather univocal picture
of bioavailable metal loads at the di€erent sites of the
experimental area (Tables 2 and 3). As may be expected,