Download Predicting the potential distribution of the vase

788
Predicting the potential distribution of the vase tunicate
Ciona intestinalis in Canadian waters: informing
a risk assessment
Thomas W. Therriault and Leif-Matthias Herborg
Therriault, T. W., and Herborg, L-M. 2008. Predicting the potential distribution of the vase tunicate Ciona intestinalis in Canadian waters:
informing a risk assessment. – ICES Journal of Marine Science, 65: 788 – 794.
A crucial step in characterizing the potential risk posed by non-native species is determining whether a potential invader can establish
in the introduced range and what its potential distribution could be. To this end, various environmental models ranging from simple
to complex have been applied to predict the potential distribution of an invader, with varying levels of success. Recently, in marine
waters, tunicates have received much attention, largely because of their negative impacts on shellfish aquaculture. One of these species
is the vase tunicate Ciona intestinalis, which recently has had a negative impact on aquaculture operations in Atlantic Canada and
could pose a risk in Pacific Canada. To inform the risk assessment of this species, we evaluated two different types of environmental
model. Simple models based on reported temperature or salinity tolerances were relatively uninformative, because almost all waters
were deemed suitable. In contrast, a more complex genetic algorithm for rule-set prediction (GARP) environmental niche model,
based on documented Canadian occurrence points, provided informative projections of the potential distribution in Canadian
waters. In addition to informing risk assessments, these predictions can be used to focus monitoring activities, particularly towards
vectors that could transport C. intestinalis to these favourable environments.
Keywords: Ciona intestinalis, environmental niche model, genetic algorithm for rule-set prediction, invasive species, risk assessment, vase
tunicate.
Received 10 June 2007; accepted 25 February 2008; advance access publication 16 April 2008.
T. W. Therriault and L-M. Herborg: Pacific Biological Station, 3190 Hammond Bay Road, Nanaimo, BC, Canada V9T 6N7. Correspondence to
T. W. Therriault: tel: þ1 250 7567394; fax: þ1 250 7567138; e-mail: [email protected].
Introduction
Non-indigenous species (NIS) pose an enormous risk to native
biodiversity and can compromise ecosystem function (Sala et al.,
2000). For marine ecosystems, non-indigenous tunicates have
become a global concern because they can displace native species
and foul aquaculture gear (Mazouni et al., 2001; Lambert and
Lambert, 2003; Blum et al., 2007). To characterize the potential
risk posed by a new invader or the spread of an existing invader
to additional waters, typically a formal risk assessment is conducted
where the potential for arrival, survival, reproduction, and spread
beyond the area of introduction is characterized (Mandrak and
Cudmore, 2004). The probability of arrival is a measure of a
species’ introduction potential and requires the presence of suitable
vectors and pathways (Colautti and MacIsaac, 2004). In contrast,
the probability of survival and reproduction is primarily associated
with the physiological tolerances of the species and characteristics
of the receiving environment (Guisan and Thuiller, 2005), or secondarily associated with community interactions (Shea and
Chesson, 2002). Environmental modelling is one approach used
to characterize locations where a species could survive, relative to
those where it could not, provided vectors and pathways are available. Therefore, predicting the potential distribution of a NIS is a
critical component of the risk assessment.
Predicting the potential distribution of organisms, based on
biological traits, habitat, and environmental and distributional
data, has been attempted for a variety of taxa (e.g. Goodwin
et al., 1999; Ruesink, 2005; Kaschner et al., 2006). However,
most studies of this type have suggested data limitations that
may compromise model performance. Therefore, more robust
approaches are sought, such as the development of environmental
niche models to predict suitable locations (Drake and
Bossenbroek, 2004; Iguchi et al., 2004; Guinotte et al., 2006;
Herborg et al., 2007). Despite the increasing availability of oceanographic data, surprisingly few marine taxa have been modelled,
notably sessile organisms such as tunicates. The vase tunicate
Ciona intestinalis is a tunicate species that could pose a risk to
Canadian waters. It is a large, solitary tunicate that is found in
dense aggregations, often as a dominant member of the epibenthic
community, especially in enclosed or semi-protected marine
embayments such as harbours or marinas (Monniot and
Monniot, 1994; Lambert and Lambert, 2003; Ramsay et al., in
press). Although often considered a member of the subtidal deepwater fauna (Brunel et al., 1998), many populations have been
reported in shallow coastal waters, most notably fouling artificial
structures (Cohen et al., 2000; Lambert and Lambert, 2003) or
aquaculture gear (Mazouni et al., 2001; Carver et al., 2003).
# 2008 International Council for the Exploration of the Sea. Published by Oxford Journals. All rights reserved.
For Permissions, please email: [email protected]
Potential distribution of Ciona intestinalis in Canadian waters: risk assessment
Ciona intestinalis is tolerant of a wide range of environmental
conditions, partially evidenced by its cosmopolitan distribution.
Reported temperature tolerance varies among populations, but
C. intestinalis is generally considered a cold-water or temperate
species, although temporary or transient populations have been
reported from tropical harbours (Monniot and Monniot, 1994).
Dybern (1965) identified 308C as an upper threshold for the
species, but Petersen and Riisgård (1992) noted that filtration
rates decreased .218C, suggesting thermal stress at higher temperatures. Generally, adult mortality increases ,108C, but in
Atlantic Canada, populations have survived for several months
at temperatures around 218C (Carver et al., 2003). As a euryhaline species, C. intestinalis is tolerant of a wide range of salinities
(12– 40), although most populations are found in water of ,30
salinity, because this species cannot withstand extended periods
in waters with salinity ,11 (Dybern, 1967), although it may be
able to withstand short-term fluctuations in salinity.
For NIS, it is beneficial if the native range is known, especially
for understanding the potential risks posed by an invader, including its potential distribution. However, the native range for
C. intestinalis has not been resolved and is the focus of continuing
debate. The species ranges from the Mediterranean Sea to
Scandinavia and Greenland (Millar, 1966), causing some to
suggest that the species should be classified as cryptogenic
(NIMPIS, 2002; J. T. Carlton, pers. comm.), at least until the
native range can be resolved. Huntsman (1912) listed C. intestinalis as a member of the benthic invertebrate community on the west
coast of Canada, but it is probable that specimens of Ciona
savignyi, an Asian native, were misidentified at that time
(Lambert, 2003). Ramsay et al. (in press) suggest that C. intestinalis
is non-indigenous to bays around Prince Edward Island, but the
species has been reported at other locations in eastern Canada
(Carver et al., 2003). Until further taxonomic and/or genetic
studies can confirm its native range, we consider C. intestinalis
to be cryptogenic in Atlantic Canadian waters and non-indigenous
in Pacific Canadian waters. The species has been widely introduced
to parts of South America, Africa, Oceania, and Asia (Millar, 1966;
NIMPIS, 2002; Figure 1), primarily by shipping as a hull-fouling
organism (Monniot and Monniot, 1994).
We evaluate different environmental models that could be used
to inform risk assessments (among other activities) by predicting
the potential distribution of C. intestinalis in Canadian waters.
789
Specifically, we compare simple models based on reported
environmental tolerances (temperature and salinity, see below)
and a genetic algorithm for rule-set prediction (GARP) environmental niche model, based on the known invaded distribution
in Canadian waters.
Material and methods
Environmental variables
Data on ocean temperature, salinity, and dissolved oxygen were
extracted from the National Oceanographic and Atmospheric
Administration’s World Ocean Database (www.nodc.noaa.gov/
OC5/indprod.html), to a depth of 50 m. Temperature and salinity
data were divided into four seasons based on sampling months
(January –March, April –June, July – September, and October–
December) using Ocean Data View 3.2.0. Seasonal temperature
and salinity, and annual oxygen data were converted into a
continuous data layer by calculating the mean value of each
variable over the 50 m using inverse distance weighting
(Geostatistical Wizard within ArcMAP 9.1). Annual surface
chlorophyll a data were obtained from the “Integrating Multiple
Demands on Coastal Zones with Emphasis on Aquatic
Ecosystems and Fisheries” project database (www.incofish.org).
The resulting ten environmental layers, January –March temperature, April –June temperature, July–September temperature,
October–December temperature, January –March salinity,
April –June salinity, July – September salinity, October–
December salinity, annual oxygen, and annual chlorophyll a, all
at 0.018 resolution, were then mapped for the simple models or
compared with occurrence data for C. intestinalis in the environmental niche model. For both coasts, predictions were made
from 458N to 608N for Canadian coastal waters, where depth
was bounded at the 200 m contour.
Predicting suitable environments
The potential range or distribution of C. intestinalis was predicted
using two approaches: simple models based on reported environmental tolerances and an environmental niche (GARP) model.
We used seasonal temperature and salinity data to identify potential
suitable habitats based on literature reports of temperature and
salinity tolerances: temperature between 218C and 308C, and
salinity between 12 and 40 (see above). These reported temperature
and salinity tolerances represent the maximum range from
Figure 1. Global distribution of Ciona intestinalis, showing countries where the species should be considered cryptogenic (grey) and countries
where the species has invaded (black).
790
populations around the globe, including cryptogenic and invaded
distributions. Also, we combined the simple temperature and
salinity models to create a combined model, for which any area
deemed unsuitable in either the temperature or salinity model for
each season was deemed unsuitable in the combined model.
In addition to the simple models, we explored the use of a more
complex environmental niche model, a GARP model, to predict
suitable environments based on the current Canadian distribution.
GARP models were constructed using species presence data from
the current Canadian range, and georeferenced environmental
data. As C. intestinalis has not been reported from Canadian
Pacific coastal waters, Canadian east coast data were used to
predict the potential west coast distribution, in addition to the
potential east coast distribution. Recent georeferenced occurrence
points (n ¼ 53) were available from 2004, based on continuing
monitoring and research activities conducted by Fisheries and
Oceans Canada (D. Sephton, B. Vercaemer, J. Martin, and
R. Bernier, pers. comm.).
All ten environmental layers were tested separately for their
contribution to the GARP models using multiple linear regression
following Drake and Bossenbroek (2004). The number of presence
points correctly predicted by GARP provided a measure of model
accuracy (Peterson and Cohoon, 1999). Variables positively
correlated with omission errors (false negatives) were rejected.
Environmental variables that contributed significantly to model
prediction accuracy were created following the best subset
procedure described by Anderson et al. (2003). GARP was set
for 100 predictions using a 0.001 convergence limit and a
maximum of 3000 iterations (per simulation) with a 5% “hard”
omission and a 50% commission threshold. Resulting predictions
were converted into a map of percentage environmental suitability
using the raster calculator in ArcMAP 9.1.
Evaluating model performance
Herborg et al. (2007) identified additional tests to validate GARP
model predictions further. Hierarchical partitioning measures
the relative contribution of each environmental layer, using all
possible combinations of environmental variables included
in the final GARP prediction, as well as its associated accuracy
(Peterson and Cohoon, 1999). In addition, we used the evaluation
strip method to determine the actual range of environmental conditions deemed suitable by the model, following the procedure
described by Elith et al. (2005). This approach is based on the
insertion of columns containing the full range of values for each
environmental parameter into each environmental layer. The
number of models that predict each value for the environmental
parameters deemed suitable can then be used to identify suitable
environmental ranges (Elith et al., 2005).
As an independent measure of model performance, we calculated the area under the receiver operating characteristic curve
(AUC), a widely used measure of the ability of a model to discriminate between sites where a species is present or absent (Hanley and
McNeil, 1982; Wiley et al., 2003; Elith et al., 2006). The AUC
measures predictive accuracy on a scale between 0 and 1, where
1 represents perfect prediction for both presence and absence
points, and 0.5 represents a no-better-than-random predictive
ability. The AUC was based on occurrence points used for the
GARP prediction and an equal number of randomly selected
absence points obtained from the area within the analysis mask.
The GARP score from the 100 best subsets used in the final
prediction were extracted for each of these points, and the AUC
T. W. Therriault and L-M. Herborg
was calculated using the “verification” package in R 2.5.0 software
(R Development Core Team, 2007). As C. intestinalis has been
reported only for the east coast of Canada, this test only applies
to east coast predictions. Further, we applied the AUC to the
temperature and salinity simple models and the combined
simple model based on reported temperature and salinity tolerances to compare GARP model performance with these models.
Results
Potential distribution in Canada
Based on literature accounts of environmental tolerances, the
potential distribution for C. intestinalis in Canadian waters was
extremely broad. Winter temperatures (January– March) in
Atlantic Canada determined the most unsuitable habitat in the
combined model. No additional unsuitable locations in eastern
Canadian waters were identified, based on the remaining three
seasonal temperature models or the four seasonal salinity
models. For Atlantic Canada, the combined model identified
unsuitable environments northwest of the Magdalen Islands to
the New Brunswick coast and through the northwest part of
Northumberland Strait between New Brunswick and Prince
Edward Island (Figure 2a). Additional pockets of unsuitable
environment existed in the northern Gulf of St Lawrence
(around Anticosti Island), the north coast of Labrador, waters
between Labrador and Newfoundland, and waters off the northeast and southeast coasts of Newfoundland (Figure 2a). On the
west coast of Canada, there were no unsuitable environments
identified in Canadian waters, based on temperature or salinity,
for any of the four seasons considered. The combined model
identified all coastal British Columbian waters as suitable
(Figure 2b).
Canadian sites in which C. intestinalis is currently found along
the Atlantic coast were used to generate GARP model predictions.
On the Atlantic coast, waters along the Atlantic coast of Nova
Scotia extending around Cape Breton Island and into the southern
Gulf of St Lawrence to the Gaspé Peninsula, including the
Magdalen Islands, are predicted to have high environmental suitability, as are the waters around the Bay of Fundy between Nova
Scotia and New Brunswick (Figure 3a). Favourable conditions
also exist on the north and west coasts of Prince Edward Island
and around the southwest corner of Newfoundland (Figure 3a).
A slightly lower environmental suitability exists for deeper
waters off the coast of Nova Scotia and the south coast of Prince
Edward Island (Figure 3a). On the Pacific coast, the highest
environmental suitability exists along the outer coast of British
Columbia’s central and north coasts, almost continuous from
Johnstone Strait to the border with Alaska (Figure 3b).
Moderate to high environmental suitability also exists along the
north and northeast coasts of the Queen Charlotte Islands
(Figure 3b). Ciona intestinalis has a very low environmental
match throughout the Strait of Georgia (Figure 3b).
Model performance
The AUC test for the simple combined temperature and salinity
model demonstrated that predictive capability was better than
random (score 0.7738; p , 0.001). The unsuitable environments
determined by the combined model were largely influenced by
the winter temperature layer for the east coast because, when the
AUC test was applied to the temperature-only model, results
were similar (score 0.7738; p , 0.001), whereas the salinity-only
Potential distribution of Ciona intestinalis in Canadian waters: risk assessment
791
Figure 2. Simple ecological niche models based on combined
seasonal temperature and salinity data used to identify suitable (dark
grey) and unsuitable (light grey) environments, based on reported
tolerances for (a) the Atlantic coast, and (b) the Pacific coast.
Figure 3. Environmental suitability predicted by GARP ecological
niche models for (a) the Atlantic coast, and (b) the Pacific coast.
Environmental suitability is displayed as the number (out of 100) of
models that predicted a particular location to be suitable.
model was not significant (score 0.7371; p ¼ 0.0951). The GARP
model had significantly greater predictive capacity than the
simple models based on reported tolerances, because the AUC
test for the east coast GARP model revealed a significant increase
in predictive capability (score 0.9694; p , 0.0001).
All environmental variables contributed significantly to the
overall GARP model accuracy. Except for the spring salinity
layer, the remaining nine environmental layers contributed relatively uniformly to the overall model, with contributions
between 8% and 16% (Table 1). Although there were slight seasonal differences, the values determined using the environmental
strips were consistent with reported literature tolerances, but did
not approach the maximum values reported (Table 1).
to guide management options for NIS (Drake and Bossenbroek,
2004; Roura-Pascual et al., 2004; Herborg et al., 2007). Further,
identifying potentially suitable habitat is a crucial component of
most risk assessment methodologies for NIS (Pheloung et al.,
1999). However, the choice of environmental model is critical,
especially when resources are limited. The extremely broad potential ranges that were projected, based on reported temperature and
salinity tolerances for C. intestinalis, were relatively uninformative.
These simple models depend on field observations or experiments
that do not necessarily mimic the environment in which the
species actually lives. Although we considered seasonal temperature and salinity, the simplistic approach does not capture other
environmental or biotic interactions that could be taking place.
In contrast, the GARP model develops its own environmental
niche definition based on occurrence points and the environmental conditions encountered there. This allows prediction of
environmental ranges, even for species whose physiological tolerances are not fully understood. Further, perhaps biotic interactions
Discussion
We applied simple and complex environmental niche models to
predict the potential distribution of C. intestinalis in Canadian
waters. The advantage of these models is that they can be used
792
Table 1. Environmental variables used to predict potential east
and west coast distributions of Ciona intestinalis.
Variable
Hierarchical Environmental
partitioning suitability
(%)
January
–March temperature (8C)
12.9
20.6–3.7
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . .
April
–June
temperature
(8C)
9.2
3.3–8.2
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . .
July
–September
temperature
(8C)
8.8
9.3–17.4
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . .
October–December
temperature
(8C)
15.4
7.2–12.7
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . .
January
–March
salinity
8.1
30.8
–32.4
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . .
April
–June
salinity
0.9
27.4
–31.3
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . .
July
–September
salinity
8.2
24.4
–32.1
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . .
October
–December
salinity
13.5
28.6
–31.8
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . .
–1
)
11.2
4.6–6.5
Annual
oxygen
(mg
l
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . .
Annual chlorophyll a (mg l – 1)
11.9
1 057 –1 957
Their relative contribution to prediction accuracy was determined by
hierarchical partitioning. The environmental suitability range for each
variable where model accuracy increased is also shown. Reported literature
tolerances were 218C to 308C, and salinity 12–40.
at these distribution points are influenced by the environmental
variables included in the GARP model. Therefore, the GARP
approach provides focus for potential management options by
identifying areas where monitoring efforts should be directed
and providing a basis for evaluating potential vectors and
pathways.
Environmental niche modelling in the marine environment
is relatively rare. This could be the result, at least in part, of
limited georeferenced occurrence data. Mapping large-scale
distributions of marine mammals, Kaschner et al. (2006) noted
that a lack of sightings data precluded application of standard
approaches, whereas a relative environmental suitability model
performed well. Similarly, when anemonefish data were limiting,
Guinotte et al. (2006) used environmental data for the host
anemone to map the potential distribution of the symbiont.
Using occurrence data for C. intestinalis here, we were able to
identify potentially suitable environments in both Atlantic
Canada, where the species is already present, and Pacific
Canada, where it has yet to invade. The advantage of combining
occurrence and environmental data is the ability to predict the
presence (or absence) of a species (Wiley et al., 2003), which is
important not only for biological invasions but also for the
conservation of marine organisms.
The simple environmental models, based on reported temperature and salinity tolerances, likely represent multiple genetic
strains of C. intestinalis (if they exist), because such data are
derived from a number of populations. The more complex
environmental niche model based on Canadian occurrence
points may or may not represent a common genetic strain.
However, given the available sample size and the spatial distribution of the occurrence points, it is likely that we have captured
the realized niche of this species, at least sufficiently to be useful for
making predictions. There are a number of examples where different distributions have been noted based on different genetic
strains. For example, it is believed that European green crab
(Carcinus maenas) introductions in North America represent
two genetically different strains (Roman, 2006). The initial introduction was a warmer-water strain, and the second appears to
be more cold-water tolerant and has spread farther north in
T. W. Therriault and L-M. Herborg
Atlantic Canada. Similarly, Therriault et al. (2004) identified a
fresh-water and brackish water race of Dreissena rostriformis. If
the reported tolerances encompass more than one strain or race,
then the potential distribution will represent the combined tolerances of all genetic strains or races, so potentially overestimating
the tolerances associated with a single strain or race. However,
this may represent a worst case scenario from a prediction perspective, because all strains or races have contributed to the realized
niche for the species and would have utility for risk assessments
because the genetic strain(s) of an invader often is (are)
unknown. Simple models based on environmental tolerances
within C. intestinalis’ global distribution would likely over-predict
the potential distribution in Canada. Alternatively, GARP models
based on the current invaded range in Canada identify suitable
environments for the introduced strain(s), but predictions will
need updating as the species spreads so as not to under-predict
the potential distribution. It is unclear how many potential
genetic varieties or races of C. intestinalis exist, but literature
accounts mention at least three forms (forma typica, forma
gelatinosa, and forma longissima), which could have different
environmental tolerances or habitat preferences, because forma
gelatinosa and forma longissima are reported as Arctic to Subarctic
forms (Millar, 1966).
Although C. intestinalis is considered to be cryptogenic in
Atlantic Canada, there is considerable potentially suitable habitat
along the Atlantic coast that is not currently known to support
populations. Much of this potential habitat exists along the
Atlantic coast of Nova Scotia, but it also includes additional
areas around Prince Edward Island and the southeastern corner
of Newfoundland. Ciona intestinalis has already been reported
as an aquaculture pest in Nova Scotia (Carver et al., 2003), and
it is a species of concern for the shellfish aquaculture industry
in Prince Edward Island (Ramsay et al., in press). Additional
spread of C. intestinalis in eastern Canada should be expected,
with potentially suitable environments being a priority for
monitoring, especially areas with increased densities of artificial
structures that can facilitate population establishment (Carver
et al., 2003; Lambert and Lambert, 2003). On the Pacific coast of
Canada, where C. intestinalis has not invaded, much of the suitable
environment exists between Johnstone Strait and the British
Columbia –Alaska border, consistent with the characterization of
C. intestinalis as a colder-water species (Brunel et al., 1998). The
Strait of Georgia, one of the most invaded habitats in British
Columbia (Levings et al., 2002), appears largely unsuitable for
C. intestinalis.
Distributional models of NIS help inform risk assessments that
can serve as a basis for decision-makers, managers, and policymakers, who have to manage, control, or mitigate the potential
impact of NIS (for the C. intestinalis risk assessment, see
Therriault and Herborg, 2008). Based on limited occurrence
data, we were able to apply an environmental niche model that
identified potentially suitable environments for C. intestinalis on
both coasts of Canada, with increased resolution over simple tolerance models. The refined predictive ability of this model should
make it easier for managers to allocate resources. It should be
expected that risk assessments will need to be revisited from
time to time as more information is gained and identified data
gaps are filled. For example, inclusion of additional data points
from its cryptogenic or invaded range might refine the actual
environmental niche of C. intestinalis. Nonetheless, it is unlikely
that the actual habitat suitable for C. intestinalis would be as
Potential distribution of Ciona intestinalis in Canadian waters: risk assessment
broad as that predicted by the simple environmental models
based on reported temperature or salinity tolerances.
Acknowledgements
D. Sephton, B. Vercaemer, J. Martin, and R. Bernier each graciously provided C. intestinalis occurrence points that allowed us
to conduct our GARP modelling. The work was supported by
funds from DFO’s AIS programme and funds from the Centre
of Expertise for Aquatic Risk Assessment (CEARA). We thank
the participants from the Charlottetown, Prince Edward Island,
tunicate meeting for helpful advice, comments, and discussion,
and the reviewers for their input, which greatly improved the
manuscript.
References
Anderson, R. P., Lew, D., and Petersen, A. T. 2003. Evaluating predictive models of species’ distributions: criteria for selecting optimal
models. Ecological Modelling, 162: 211– 232.
Blum, J. C., Chang, A. L., Liljesthröm, M., Schenk, M. E., Steinberg, M.
K., and Ruiz, G. M. 2007. The non-native ascidian Ciona intestinalis (L.) depresses species richness. Journal of Experimental Marine
Biology and Ecology, 342: 5 – 14.
Brunel, P., Bossé, L., and Lamarche, G. 1998. Catalogue of the marine
invertebrates of the estuary and Gulf of St Lawrence. Canadian
Special Publication of Fisheries and Aquatic Sciences, 126. 405 pp.
Carver, C. E., Chisholm, A., and Mallet, A. L. 2003. Strategies to mitigate the impact of Ciona intestinalis (L.) biofouling on shellfish
production. Journal of Shellfish Research, 22: 621 – 631.
Cohen, B. F., Currie, D. R., and McArthur, M. A. 2000. Epibenthic
community structure in Port Phillip Bay, Victoria, Australia.
Marine and Freshwater Research, 51: 689– 702.
Colautti, R. I., and MacIsaac, H. J. 2004. A neutral terminology to
define “invasive” species. Diversity and Distributions, 10: 135 – 141.
Drake, J. M., and Bossenbroek, J. M. 2004. The potential distribution
of zebra mussels in the United States. Bioscience, 54: 931 –941.
Dybern, B. I. 1965. The life cycle of Ciona intestinalis (L.) f. typica in
relation to the environmental temperature. Oikos, 16: 109– 131.
Dybern, B. I. 1967. Settlement of sessile animals on eternite slabs in
two polls near Bergen. Sarsia, 29: 137 – 180.
Elith, J., Ferrier, S., Huettmann, F., and Leathwick, J. 2005. The
evaluation strip: a new and robust method for plotting
predicted responses from species distribution models. Ecological
Modelling, 186: 280– 289.
Elith, J., Graham, C. H., Anderson, R. P., Dudı́k, M., Ferrier, S.,
Guisan, A., Hijmans, R. J., et al. 2006. Novel methods improve
prediction of species’ distributions from occurrence data.
Ecography, 29: 129 – 151.
Goodwin, B. J., McAllister, A. J., and Fahrig, L. 1999. Predicting
invasiveness in plant species based on biological information.
Conservation Biology, 13: 422– 426.
Guinotte, J. M., Bartley, J. D., Iqbal, A., Fautin, D. G., and Buddemeier,
R. W. 2006. Modeling habitat distribution from organism occurrences and environmental data: case study using anemonefishes
and their sea anemone hosts. Marine Ecology Progress Series,
316: 269 – 283.
Guisan, A., and Thuiller, W. 2005. Predicting species distribution:
offering more than simple habitat models. Ecology Letters, 8:
993– 1009.
Hanley, J. A., and McNeil, B. J. 1982. The meaning and use of the area
under a receiver operating characteristic (ROC) curve. Radiology,
143: 29 – 36.
Herborg, L. M., Mandrak, N. E., Cudmore, B., and MacIsaac, H. J.
2007. Comparative distribution and invasion risk of snakehead
and Asian carp species in North America. Canadian Journal of
Fisheries and Aquatic Sciences, 64: 1723– 1735.
793
Huntsman, A. G. 1912. Ascidians from the coasts of Canada.
Transactions of Canadian Institutes, 9: 111 – 148.
Iguchi, K., Matsuura, K., McNyset, K. M., Peterson, A. T.,
Scachetti-Pereira, R., Powers, K. A., Vieglais, D. A., et al. 2004.
Predicting invasions of North American basses in Japan using
native range data and a genetic algorithm. Transactions of the
American Fisheries Society, 133: 845– 854.
Kaschner, K., Watson, R., Trites, A. W., and Pauly, D. 2006. Mapping
world-wide distributions of marine mammal species using a
relative environmental suitability (RES) model. Marine Ecology
Progress Series, 316: 285 –310.
Lambert, C. C., and Lambert, G. 2003. Persistence and differential distribution of nonindigenous ascidians in harbors of the Southern
California Bight. Marine Ecology Progress Series, 259: 145 – 161.
Lambert, G. 2003. New records of ascidians from the NE Pacific: a new
species of Trididemnum, range extension and redescription of
Aplidiopsis pannosum (Ritter, 1899) including its larva, and
several non-indigenous species. Zoosystema, 25: 665 – 679.
Levings, C., Kieser, D., Jamieson, G., and Dudas, S. 2002. Marine and
estuarine alien species in the Strait of Georgia, BC. In Alien
Invaders in Canada’s Waters, Wetlands, and Forests, pp.
111– 131. Ed. by P. Nantel, R. Claudi, and E. Muckle-Jeffs.
Canadian Forest Service, Science Branch, Ottawa 320 pp.
Mandrak, N. E., and Cudmore, B. C. 2004. Risk assessment for Asian
carps in Canada. Canadian Science Advisory Secretariat Research
Document 2004/103. 52 pp.
Mazouni, N., Gaertner, J-C., and Deslou-Paoli, J. M. 2001.
Composition of fouling communities on suspended oyster
cultures: an in situ study of their interactions with the water
column. Marine Ecology Progress Series, 214: 93 – 102.
Millar, R. H. 1966. Ascidiaceae. Scandinavian University Books,
Oslo. 123 pp.
Monniot, C., and Monniot, F. 1994. Additions to the inventory of
eastern tropical Atlantic ascidians: arrival of cosmopolitan
species. Bulletin of Marine Science, 54: 71 – 93.
NIMPIS. 2002. Ciona intestinalis species summary. In National
Introduced Marine Pest Information System. Ed. by C. L. Hewitt,
R. B. Martin, C. Sliwa, F. R. McEnnulty, N. E. Murphy, T. Jones,
and S. Cooper. http://crimp.marine.csiro.au/nimpis (last accessed
24 December 2007).
Petersen, J. K., and Riisgård, H. U. 1992. Filtration capacity of the
ascidian Ciona intestinalis and its grazing impact in a shallow
fjord. Marine Ecology Progress Series, 88: 9 – 17.
Peterson, A. T., and Cohoon, K. P. 1999. Sensitivity of distributional
prediction algorithms to geographic data completeness.
Ecological Modelling, 117: 159– 164.
Pheloung, P. C., Williams, P. A., and Halloy, S. R. 1999. A weed
risk assessment model for use as a biosecurity tool evaluating
plant introductions. Journal of Environmental Management, 57:
239– 251.
Ramsay, A., Davidson, J., Landry, T., and Arsenault, G. Process of
invasiveness among exotic tunicates in Prince Edward Island,
Canada. Biological Invasions, in press.
R Development Core Team. 2007. R: a language and environment for
statistical computing. R Foundation for Statistical Computing,
Vienna, Austria. http://www.R-project.org.
Roman, J. 2006. Diluting the founder effect: cryptogenic invasions
expand a marine invader’s range. Proceedings of the Royal
Society of London, Series B, 273: 2453– 2459.
Roura-Pascual, N., Suarez, A. V., Gomez, C., Pons, P., Touyama, Y.,
Wild, A. L., and Peterson, A. T. 2004. Geographical potential of
Argentine ants (Linepithema humile Mayr) in the face of global
climate change. Proceedings of the Royal Society of London,
Series B, 271: 2527– 2534.
Ruesink, J. L. 2005. Global analysis of factors affecting the outcome
of freshwater fish introductions. Conservation Biology, 19:
1883– 1893.
794
Sala, O. E., Chapin, F. S., Armesto, J. J., Berlow, E., Bloomfield, J.,
Dirzo, R., Huber-Sanwald, E., et al. 2000. Global biodiversity
scenarios for the year 2100. Science, 287: 1770– 1774.
Shea, K., and Chesson, P. 2002. Community ecology theory as a
framework for biological invasions. Trends in Ecology and
Evolution, 17: 170 – 178.
Therriault, T. W., Docker, M. F., Orlova, M. I., Heath, D. D., and
MacIsaac, H. J. 2004. Molecular resolution of the family
Dreissenidae (Mollusca: Bivalvia) with emphasis on Ponto-Caspian
species, including first report of Mytilopsis leucophaeata in the Black
Sea Basin. Molecular Phylogenetics and Evolution, 30: 479–489.
T. W. Therriault and L-M. Herborg
Therriault, T. W., and Herborg, M. 2008. A qualitative biological risk
assessment for vase tunicate Ciona intestinalis in Canadian waters:
using expert knowledge. ICES Journal of Marine Science, 65:
000– 000.
Wiley, E. O., McNyset, K. M., Peterson, A. T., Robins, C. R., and
Stewart, A. M. 2003. Niche modeling and geographic predictions
in the marine environment using a machine-learning algorithm.
Oceanography, 16: 120– 127.
doi:10.1093/icesjms/fsn054