PII:
Marine Pollution Bulletin Vol. 42, No. 11, pp. 1007±1013, 2001
Ó 2001 Elsevier Science Ltd. All rights reserved
Printed in Great Britain
0025-326X/01 $ - see front matter
S0025-326X(01)00135-7
Viewpoint
Can Saltwater Toxicity be Predicted
from Freshwater Data?
KENNETH M. Y. LEUNG *, DAVID MORRITT , JAMES R. WHEELER , PAUL WHITEHOUSEà,
NEAL SOROKINà, ROBIN TOY§, MARTIN HOLT and MARK CRANE School of Biological Sciences, Royal Holloway, University of London, Egham, Surrey TW20 0EX, UK
àWRc-NSF, Henley Road, Medmenham, Marlow, Buckinghamshire SL7 2HD, UK
§Shell International Chemicals Ltd., Shell Centre, London SE1 7NA, UK
ECETOC, Av E van Nieuwenhuyse 4, Bte 6, B-1160, Brussels, Belgium
The regulation of substances discharged to estuarine and
coastal environments relies upon data derived from ecotoxicity tests. Most such data are generated for freshwater rather than saltwater species. If freshwater toxicity
data are related to saltwater toxic e€ects in a systematic
and predictable way, the former can be used to predict the
latter. This would have economic advantages due to a
reduction in toxicity testing of saltwater species. If toxicity data are plotted as species sensitivity distributions,
four theoretical relationships between freshwater and
saltwater can be envisaged. Examples show that each one
of these relationships is supported by empirical data.
These examples show that although there is considerable
potential for freshwater to saltwater prediction, species
parity and representativeness need to be examined for
each chemical substance to avoid bias. Ó 2001 Elsevier
Science Ltd. All rights reserved.
Keywords: species sensitivity distribution; parity; representativeness; freshwater to saltwater toxicity.
Introduction
European Union Directives and Regulations dealing
with chemical risk assessment for New and Existing
Substances are implemented using the European Union
System for Evaluation of Substances (EUSES). This
system focuses on risk assessment for freshwater and
terrestrial environments, but procedures for saltwater
are less well developed. A key step in chemical risk assessment for any environment is the estimation of a
Predicted No E€ect Concentration (PNEC). In practice,
*Corresponding author. Tel.: +44-1784-414-196; fax: +44-1784470-756.
E-mail address: [email protected] (K.M.Y. Leung).
risk assessors must extrapolate from a relatively small
dataset, usually containing information from less than
10 species, to estimate the PNEC (OECD, 1992). This is
usually achieved by applying safety factors to the lowest
e€ects concentrations from reliable studies (Whitehouse
and Cartwright, 1998).
There are generally fewer toxicity data available for
saltwater species than for freshwater ones, especially in
relation to organic compounds (Solbe et al., 1993). This
is largely because there are fewer standard test methods
for saltwater species and because aquatic risk assessments, at least in Europe, have tended to focus on
freshwater systems. Because of this paucity of data,
many saltwater PNECs rely on extrapolations based on
freshwater information. This `surrogate' approach assumes that freshwater species respond similarly to
marine species, and that the distributions of the sensitivities of the two groups of species are identical. These
assumptions remain untested, and have led to proposals
to add an additional safety factor of 10 to marine risk
assessments based upon freshwater data (S. Robertson,
Environment Agency of England and Wales, personal
communication). Some previous studies have shown
correlations between sensitivities of particular freshwater and saltwater invertebrates, ®sh and algae
(Dawson et al., 1977; Robinson, 1999). For example,
Hutchinson et al. (1998) concluded that freshwater to
saltwater toxicity could be predicted with greater con®dence for ®sh than for invertebrates. It is worth noting, however, that for ®sh the predictions were based
on comparisons within one class of chordates whereas
the invertebrate comparisons lumped data for several
higher taxonomic groups. No formal method has been
proposed that allows an objective decision to be made
as to whether or not to use freshwater data to predict
saltwater toxicity for a particular chemical under
review.
1007
Marine Pollution Bulletin
Species Sensitivity Distributions
Useful progress in comparing the responses of
freshwater and saltwater biota to chemicals can be
made by comparing the species sensitivity distributions
(SSDs) of freshwater and saltwater organisms exposed
to the same chemical. Essentially, this involves ranking
the sensitivity of organisms to a speci®c chemical
(based on either LC/EC50 or NOEC data) and plotting
these ranks against chemical concentrations. A distribution is then ®tted to the toxicity data for di€erent
species, and from which certain quantitative parameters
may be estimated, e.g. the concentration predicted to
a€ect only a small proportion (typically 5%) of species
(Aldenberg and Slob, 1993). A visual assessment of the
extent of either congruence or discrepancy between
freshwater and saltwater datasets can also be made. A
key advantage to this approach for comparing responses to chemicals is that it obviates the need for
`pairing' freshwater and saltwater species. Whilst it is
not contentious to compare responses at a low level of
taxonomic resolution (e.g. freshwater ®sh with saltwater ®sh), comparisons at higher levels of taxonomic
resolution are fraught with diculties. This is especially
the case at the species level where it would become
necessary to identify, for example, a marine equivalent
of Daphnia magna. Should this be done on the basis of
taxonomy (i.e. select a marine cladoceran), life history
characteristics or some other ecological or physiological criterion?
It is on this point that SSDs have a major advantage.
The approach allows comparison of the responses of
`assemblages' of freshwater and marine species exposed
to the same chemical. Usually, data for all taxa can be
incorporated in the same distribution, although for
chemicals with a speci®c mode of action this may not
always be advisable because a series of `sub-distributions' may exist (Solomon et al., 1996). In marked
contrast with previous approaches, however, the need to
identify marine species that can be regarded as counterparts of freshwater species (and vice-versa) is avoided.
If SSDs are used to compare freshwater and saltwater
species responses to de®ned chemicals, essentially four
di€erent outcomes of these comparisons can be envisaged, as illustrated in Fig. 1. In these examples, concentration is plotted (on the x-axis) against the number
of species a€ected (on the y-axis). In (a), SSDs for the
freshwater and saltwater species are indistinguishable
whereas in (b), there is a systematic shift re¯ecting the
greater sensitivity of freshwater species, although the
distributions of both these and saltwater ones have
similar slopes; in (c), the reverse applies, with saltwater
species showing greater sensitivity. Finally, in (d), both
the slope characteristics and intercept with the y-axis are
di€erent, perhaps denoting more fundamental di€er-
(a)
(b)
(c)
(d)
Fig. 1 Possible outcomes from comparing species sensitivity distributions for freshwater (solid line) and saltwater (broken line)
species (a), similarity, (b), freshwater consistently more sensitive than saltwater, (c), saltwater consistently more sensitive
than freshwater and (d), dissimilar responses at high and low
concentrations (distributions cross).
1008
Volume 42/Number 11/November 2001
ences between these groups of organisms in their responses to toxicants.
may be either missing or absent from most toxicity
datasets.
Parity and Representativeness
Examples of Freshwater to Saltwater
Comparisons
A closer examination of freshwater and saltwater datasets reveals consistent di€erences in their species
compositions, i.e. a lack of parity. Generally, greater
taxonomic diversity is seen in toxicity data for freshwater species. Under such circumstances, bias could
be introduced as a result of the presence of highly sensitive taxa, for example, insects and crustaceans that are
sensitive to organophosphate insecticides, in either the
freshwater or saltwater datasets. Given the greater taxonomic diversity in freshwater datasets, this may be
expected to give rise to a systematic shift resulting in the
apparently greater sensitivities of such species.
A related consideration is the representativeness of
the available datasets, i.e. the extent to which the species for which toxicity data are available re¯ect the
natural taxonomic diversity of freshwater and saltwater
biota. Both freshwater and saltwater toxicity datasets
are incomplete in this regard but those for the latter
taxa tend to be much less representative than their
former counterparts. Indeed, some major saltwater taxa
The US EPA's AQUIRE database (http//:www.epa.
gov/med/databases/databases.html#aquire) was used to
select acute toxicity data for four chemical substances to
illustrate the SSD approach, and to show the in¯uence
that parity and representativeness may have on toxicity
comparisons.
Cadmium
There is a close similarity in the distribution of
freshwater and saltwater responses to cadmium, as
shown in Fig. 1(a). Forty-two freshwater species from
nine higher taxonomic groups are compared with 31
saltwater species from six higher taxonomic groups in
Fig. 2. The SSDs indicate close agreement between
saltwater and freshwater assemblages. This similarity is
perhaps surprising in view of previous studies on cadmium toxicity in saline media. As salinity increases, so
the extent of cadmium chloro-complexing increases
(50% CdClo2 ; 30%CdCl‡ ; 20%CdCl3 ; Hunt, 1987),
Fig. 2 (a) Species sensitivity distributions for cadmium, based on
acute toxicity data for freshwater and saltwater species found
on AQUIRE, plus the taxonomic composition of (b), freshwater (FW) and (c), saltwater (SW) data used to construct
them.
1009
Marine Pollution Bulletin
with an apparent decrease in toxicity, but with a corresponding increase in magnesium and calcium cations
which may further ameliorate it. There are several
reports covering a range of test organisms in which
cadmium toxicity was reduced by increasing salinity.
The e€ect has been attributed to competition for active
sites by calcium (Pagano et al., 1982).
In this example, the conclusions are based on reasonably well-populated freshwater and saltwater datasets dominated by crustaceans (27% and 46%,
respectively) and ®shes (19% and 42%, respectively).
The saltwater dataset has a higher proportion of ®sh
and crustacean information, largely as a result of the
increased taxonomic diversity of freshwater biota,
which, for example, also includes insects, annelids,
platyhelminthes and ectoprocts. All of which are often
lacking for marine biota lists. Certainly, data for marine
annelids and platyhelminthes would be valuable. Ectoprocta (bryozoan) data come from one study of three
freshwater species exposed to a range of heavy metals,
and which is interesting as the phylum is primarily a
marine group. On the basis of these data, it is reasonable
to conclude that extrapolation from freshwater data
would be adequately protective for saltwater organisms,
although some uncertainties remain over parity and
representativeness.
Nickel
Nickel provides an example of where the distribution
of freshwater responses is to the left of the saltwater
ones, as shown in Fig. 1(b), suggesting that the former
species are generally more sensitive to this substance
than the latter. Eleven freshwater species from six higher
taxonomic groups are compared with nine saltwater
species from four higher taxonomic groups in Fig. 3.
These data for nickel show a large discrepancy between
freshwater and saltwater data, with greater sensitivity
consistently exhibited by the former species. Nickel undergoes broadly similar speciation in freshwaters and
saltwaters and this alone is unlikely, therefore, to account for the observed di€erences.
The interesting trend here is that there is a reversal in
the relative proportions of ®shes and crustaceans between freshwater (46% and 18%, respectively) and
saltwater (22% and 56%, respectively) datasets. There
are no marine annelid and platyhelminth data and no
freshwater mollusc data. Interestingly, there are no insect data for freshwaters. Clearly, despite some uncertainties about parity and representativeness for both
freshwater and saltwater datasets, extrapolation from
the former to the latter is likely to be protective with a
considerable (approximately an order of magnitude)
margin of safety.
Chlordane
The insecticide chlordane provides an example of
where the distribution of saltwater responses is to the left
of the freshwater ones, as shown in Fig. 1(c), suggesting
that the former species are generally more sensitive to
this substance than the latter. Twenty-®ve freshwater
Fig. 3 (a) Species sensitivity distributions for nickel, based on acute
toxicity data for freshwater and saltwater species found on
AQUIRE, plus the taxonomic composition of (b) freshwater
(FW) and (c) saltwater (SW) data used to construct them.
1010
Volume 42/Number 11/November 2001
Fig. 4 (a) Species sensitivity distributions for chlordane, based on
acute toxicity data for freshwater and saltwater species found
on AQUIRE, plus the taxonomic composition of (b) freshwater (FW) and (c) saltwater (SW) data used to construct
them.
species from ®ve higher taxonomic groups are compared
with eight saltwater species from two higher taxonomic
groups in Fig. 4. The freshwater data set contains a high
proportion of ®sh (60%) and crustacean (24%) data, but
meaningful comparison between this and the saltwater
dataset is dicult because of the small size of the latter.
Comparisons would be aided by acquiring more data for
annelids, rotifers and molluscs, for which there are no
saltwater data. In this case, there is evidence that freshwater data would not be protective of saltwater species.
However, parity between freshwater and saltwater datasets is low, and the representativeness of the latter is
poor, so that generation of further data may produce
greater congruence between SSDs. Interestingly, there
may not necessarily be a need for saltwater data, despite
low numbers. This is because ®sh dominated the freshwater dataset (and the lower tail of the freshwater SSD)
whereas crustaceans occupied this role in the saltwater
SSDs. Apparently greater sensitivity by saltwater organisms may thus be over-estimated because of the
preponderance of crustaceans, which are sensitive to
insecticides. This could be investigated by reinforcing the
freshwater dataset with either crustacean or insect data.
Potassium dichromate
Potassium dichromate provides an example in which
the distributions of freshwater and saltwater responses
cross each other, as shown in Fig. 1(d), suggesting that
the relative sensitivities of the two habitats' taxa di€er at
high and low concentrations. Regulatory attention
should, however, be drawn to the lower tails of the
distributions where data for the more sensitive species
lie. Eighty freshwater species from 10 higher taxonomic
groups are compared with 33 saltwater species from ®ve
higher taxonomic groups in Fig. 5. The most sensitive
freshwater species are a€ected to a greater extent than
the most sensitive saltwater ones. Di€erences in speciation are an unlikely cause of this as chromium exists
predominantly as an oxy-anion …Cr2 O27 † in both media.
Fishes, crustaceans and annelids are well represented
in both salt and freshwaters but there are few marine
mollusc data. The freshwater species composition is affected by the inclusion of one study on twelve species of
Protozoa, although they probably had little in¯uence on
the SSD as they exhibited intermediate sensitivities to
this substance. Despite some minor reservations with
regard to parity and representativeness, extrapolation
from freshwater data would likely be protective of
saltwater organisms for this compound, almost certainly
with a margin of safety.
Discussion
The SSD approach comparing freshwater and saltwater toxicity data, in combination with estimates of
parity and representativeness, is a simple visual method
1011
Marine Pollution Bulletin
Fig. 5 (a) Species sensitivity distributions for potassium dichromate,
based on acute toxicity data for freshwater and saltwater species found on AQUIRE, plus the taxonomic composition of (b)
freshwater (FW) and (c) saltwater (SW) data used to construct
them.
for summarising available information on the biological
e€ects of chemicals in both habitats. When using this
approach, the following features become apparent:
1. The distribution of toxicity values, and any particularly sensitive groups of biota.
2. Di€erences between freshwater and saltwater toxicity, and whether they occur in the important lower
tail of the distribution.
3. Obvious outliers from the general distribution that
may deserve further attention.
4. Taxonomic groups missing from either distribution,
and for which generation of toxicity data may be prudent.
Like any extrapolation procedure, care needs to be
exercised when interpreting the results of SSDs. Recent
evidence shows that these distributions may be in¯uenced strongly by a chemical's mode of toxic action,
which can in¯uence both the range and the complexity
of the distribution (Vaal et al., 1997). An understanding
of a chemical's mode of toxicity is thus important when
comparing the distributions of species sensitivities. Additionally, quantitative SSD models assume a random
selection of test organisms, which is clearly not the case
(OECD, 1992). Therefore a possible cause of any differences between freshwater and saltwater organisms'
sensitivity distributions could be di€erences in the taxonomic composition of the datasets. This could be due
either to di€erences in parity between freshwater and
saltwater data, or to the low representativeness of either
1012
habitats. As well as considering the distributions of
species sensitivities, it is important to consider the in¯uence that particular species have in de®ning them, and
the extent to which the responses of particular freshwater and saltwater ones are correlated. It is possible
that some species will be particularly sensitive to a wide
range of chemicals and will thus be found more frequently in the lower tail of the species sensitivity distributions, making them more important in de®ning the
PNEC.
We thank the CEFIC Long Range Initiative for funding and Professor
Brian Morton for helpful comments on a draft. KMYL was supported
by The Croucher Foundation, Hong Kong.
Aldenberg, T. and Slob, W. (1993) Con®dence limits for hazardous
concentrations based on log-logistically distributed NOEC toxicity
data. Ecotoxicology and Environmental Safety 25, 48±63.
Dawson, G. W., Jennings, A. L., Drozdowski, D. and Rider E. (1977)
The acute toxicity of 47 industrial chemicals to fresh and saltwater
®shes. Journal of Hazardous Materials 1, 303±318.
Hunt, D. T. E. (1987) Trace metal speciation and toxicity to aquatic
organisms ± a review. WRc Technical Report TR 247, Medmenham, UK.
Hutchinson, T. H., Scholz, N. and Guhl, W. (1998) Analysis of the
ECETOC aquatic toxicity (EAT) database IV ± comparative
toxicity of chemical substances to freshwater versus saltwater
organisms. Chemosphere 36, 143±153.
OECD. (1992) Report of the Organisation for Economic Co-operation
and Development (OECD) workshop on the extrapolation of
laboratory aquatic toxicity data to the real environment. OECD
Monograph No. 60, Organisation for Economic Co-operation and
Development, Paris.
Pagano, G., Esposito, A. and Giordano, G.G. (1982) Fertilization and
larval development in sea urchins following exposure of gametes
Volume 42/Number 11/November 2001
and embryos to cadmium. Archives of Environmental Contamination
and Toxicology 11, 47±55.
Robinson, P. W. (1999) The toxicity of pesticides and organics to
mysid shrimps can be predicted from Daphnia spp. Water Research
33, 1545±1549.
Solbe, J. F., de, L. G., Buyle, B., Guhl, W., Hutchinson, T., Laenge,
R., Mark, U., Munk, R. and Scholz, N. (1993) Developing hazard
identi®cation for the aquatic environment. Science of the Total
Environment Suppl 129, 47±61.
Solomon, K. R., Baker, D. B., Richards, R. P., Dixon, K. R., Klaine,
S. J., La Point, T. W., Kendall, R. J., Weisskopf, C. P., Giddings, J.
M., Giesy, J. P., Hall, L. W., Jr. and Williams, W. M. (1996)
Ecological risk assessment of atrazine in North American surface
waters. Environmental Toxicology and Chemistry 15, 31±76.
Vaal, M., van der Waal, J. T., Hoekstra, J. and Hermens J. (1997)
Variation in the sensitivity of aquatic species in relation to the
classi®cation of environmental pollutants. Chemosphere 35, 1311±
1327.
Whitehouse, P. and Cartwright, N. G. (1998) Standards for environmental protection. In Pollution Risk Assessment and Management,
ed. P. E. T. Douben, pp. 235±272, Wiley, Chichester.
1013