1903

Invasive potential of common carp (Cyprinus

carpio) and Nile tilapia (Oreochromis niloticus) in
American freshwater systems
Luis Zambrano, Enrique Martínez-Meyer, Naercio Menezes, and
A. Townsend Peterson

Abstract: Nonnative fish introductions disrupt ecosystem processes and can drive native species to local extinction.
Two of the most widespread, introduced species are the common carp (Cyprinus carpio) from Eurasia and the Nile
tilapia (Oreochromis niloticus) from Africa. In North and South America, these introductions stem from aquaculture
facilities, as well as historical introductions for recreational angling. An emergent field of ecological niche modeling
provides robust predictions of the geographic potential of alien species to better understand their capacity to become
established at broad scales. We used this modeling approach to produce spatially explicit predictions of the invasive
potential of common carp and Nile tilapia in the Americas. Model predictions were tested using occurrence data for
established populations in their native area and in the Americas. Results indicated that predictive power of niche mod-
els was high. Distributional potential of common carp in the Americas covers most temperate regions and high moun-
tain tropical aquatic systems, whereas that of Nile tilapia is focused in the tropics and coast areas. The consequences
of the potential establishment of these exotic species can be profound on native aquatic faunas, particularly on highly
diverse regions such as the Amazon Basin and central Mexico.
Résumé : Les introductions de poissons non indigènes perturbent les processus écosystémiques et peuvent causer
l’extinction locale d’espèces indigènes. Deux des espèces les plus généralement introduites sont la carpe commune
(Cyprinus carpio) d’Eurasie et le tilapia du Nil (Oreochromis niloticus) d’Afrique. En Amérique du Nord et du Sud,
ces introductions proviennent des établissements d’aquaculture, ainsi que des empoissonnements passés pour la pêche
sportive. Un domaine en émergence de la modélisation des niches écologiques permet de faire des prédictions robustes
sur le potentiel géographique d’espèces exotiques et ainsi de comprendre leur capacité à s’établir sur de larges espaces
géographiques. Cette approche de modélisation nous a permis de faire des prédictions explicites du point de vue spatial
concernant le potentiel d’invasion de la carpe commune et du tilapia du Nil dans les Amériques. Nous avons testé les
prédictions du modèle à l’aide de données de présence de populations établies dans leur aire d’origine et dans les
Amériques. Les résultats montrent que le pouvoir prédictif des modèles de niches sont élevés. Le potentiel de réparti-
tion géographique de la carpe commune dans les Amériques englobe la plupart des régions tempérées et les écosystè-
mes aquatiques tropicaux de haute montagne; en revanche, le tilapia du Nil se concentre surtout dans les tropiques et
les régions côtières. L’établissement potentiel des ces espèces exotiques peut avoir de graves conséquences sur les fau-
nes aquatiques indigènes, particulièrement dans les régions de forte diversité, telles que le bassin de l’Amazone et le
centre du Mexique.

[Traduit par la Rédaction] Zambrano et al. 1910

Introduction such as escape from aquaculture facilities or via shipping

ballast water. Consequences of introductions can be pro-
Introductions of non-indigenous fishes can reduce diver- found; species diversity of native fish assemblages can be
sity and modify local community dynamics in freshwater greatly reduced (Rahel 2000; Jackson 2002), and food web
systems (Minns and Cooley 2000). Introductions can be in- function can be disrupted (Vander Zanden et al. 1999). Pop-
tentional, such as in extensive aquaculture or with introduc- ulation viability can be lowered because of predator–prey
tions of top predators for recreational fishing, or incidental, interactions, as in the case of sea lamprey (Petromyzon

Received 13 July 2005. Accepted 27 March 2006. Published on the NRC Research Press Web site at http://cjfas.nrc.ca on
25 July 2006.
J18791
L. Zambrano1 and E. Martínez-Meyer. Instituto de Biologia, Universidad Nacional Autonoma de Mexico, Ciudad Universitaria,
Mexico DF, Mexico.
N. Menezes. Museu de Zoologia, Univeridade de Sao Paulo, Caixa Postal 42694, 004299–970 Sao Paulo, Brasil.
A.T. Peterson. Natural History Museum and Biodiversity Research Center, The University of Kansas, Lawrence, KS 66045, USA.
1
Corresponding author (e-mail: [email protected]).

Can. J. Fish. Aquat. Sci. 63: 1903–1910 (2006) doi:10.1139/F06-088 © 2006 NRC Canada
1904 Can. J. Fish. Aquat. Sci. Vol. 63, 2006

marinus) introductions to the Laurentian Great Lakes of duction atlas (Carta Nacional Pesquera 2004) and literature
North America (Mills et al. 2003). (see Acknowledgements). In all, 54 points were available for
Most research on fish introductions has focused on site- common carp and 76 for Nile tilapia on their native ranges,
specific processes (e.g., loss of species, food web disrup- and 747 and 43 points in the Americas, respectively. All oc-
tions). A clear focus has been on top-down predator effects currences were georeferenced with Internet-based electronic
stemming from species introductions, such as the role of gazetteers (http://www.fallingrain.com/world/index.html) to
Nile perch (Lates niloticus) in the destruction of the native the nearest 0.01° of latitude and longitude.
cichlid fauna of Lake Victoria (Barel et al. 1985). A broader Ecological niches were modeled using the Genetic Algo-
perspective is necessary to better assess areas vulnerable to rithm for Rule-set Prediction (GARP) (Stockwell and Peters
fish introductions, for example to set up policies for prevent- 1999). This algorithm relates ecological characteristics of
ing or mitigating negative effects of introductions. known occurrence points to those of points randomly sam-
Common carp (Cyprinus carpio) and Nile tilapia (Oreo- pled from the rest of the study region, seeking — in an
chromis niloticus) introductions in North and South America evolutionary-computing environment — to develop decision
have been extensive, but may not yet represent the full distri- rules that best summarize factors associated with the species’
butional potential of both species. This point becomes rele- presence. Previous tests of GARP’s ability to predict geo-
vant because these species have important effects in aquatic graphic distributions accurately have been successful (Peter-
ecosystems (Tapia and Zambrano 2003). Common carp in- son 2001; Peterson and Vieglais 2001; Peterson 2003),
crease suspended solids in the water column (Zambrano et including previous applications to predicting fish distribu-
al. 2001), reduce water transparency (Pinto et al. 2005) and tions (Wiley et al. 2003; Iguchi et al. 2004; McNyset 2005).
macrophyte coverage (Zambrano and Hinojosa 1999), and In GARP, occurrence points are divided twice evenly into
decrease habitat heterogeneity for native species (Perrow et training and test data sets — that is, an initial 50% of the
al. 1999). Nile tilapia alter trophic web structure by compet- data points are set aside for a completely independent test of
ing with other fish and preying on juveniles of other fish model quality (extrinsic testing data); of the remaining
(Morgan et al. 2004) and amphibians (L. Zambrano, unpub- points, half are used for developing models (training data)
lished data). and half are used for tests of model quality internal to GARP
The emerging field of ecological niche modeling has proved processing (intrinsic testing data). GARP works in an itera-
useful in developing robust predictions of the distributional tive process of rule selection, evaluation, testing, and incor-
potential of alien species (Peterson 2003). This approach fo- poration or rejection; a method is chosen from a set of
cuses on modeling the geographic manifestation of species’ possibilities (e.g., logistic regression, bioclimatic rules), ap-
ecological niches (i.e., the conjunction of ecological condi- plied to the training data, and a rule is developed or evolved.
tions within which a species is able to maintain populations Predictive accuracy is then evaluated based on 1250 points
without immigration) (Grinnell 1917; Hutchinson 1957). resampled (with replacement) from the interface testing data
Software applications that relate georeferenced occurrence and on 1250 points sampled randomly from the study region
points to geographic information systems’ (GIS) layers, rep- as a whole. Rules may evolve by a number of means that
resenting environmental conditions of the landscape, have mimic DNA evolution: point mutations, deletions, crossing
been developed to characterize the ecological niches of spe- over, etc. Changes in predictive accuracy from one iteration
cies and produce predictive distribution maps (Stockwell and to the next are used to evaluate whether particular rules
Noble 1992; Stockwell and Peters 1999). With recent evi- should be incorporated into the model, and the algorithm
dence that niches are highly conserved over evolutionary runs either 1000 iterations or until convergence.
time periods (Peterson et al. 1999; Martínez-Meyer et al.
All modeling in this study was carried out on a desktop
2004) and several successful applications (Peterson 2003), it
implementation of GARP available for public download
is now clear that this approach offers excellent predictivity
(http://www.lifemapper.org/desktopgarp). This program offers
regarding species’ geographic potential on landscapes be-
excellent flexibility in choice of predictive environmental–
yond their native ones (Peterson and Robins 2003; Iguchi et
ecological GIS data coverages. In this case, initially, we
al. 2004; Peterson et al. 2004).
used 15 data layers summarizing elevation, slope, aspect,
Herein, we present an analysis of the invasive potential of
flow accumulation, flow direction, and topographic index at
carp and tilapia in the New World, based on ecological niche
a native pixel size of 1 arc-second (~1 km2, obtained from
models from their native ranges. Detailed evaluations of the ca-
the US Geological Survey’s Hydro-1K data set; http://
pacity of particular fish species for establishing and extending
edcdaac.usgs.gov/gtopo30/hydro). We also used aspects of
populations in new environments would help decision-makers
climate, including 1961–1990 annual means representing di-
evaluate risks and benefits of particular introductions. Indeed,
urnal temperature range; freeze days; mean annual precipita-
under some circumstances, potential benefits of aquaculture
tion; solar radiation; maximum, minimum, and mean annual
could be overshadowed by costs of control or eradication of
temperatures; vapor pressure; and wet days (all at 0.5° native
alien species causing ecological damage.
resolution — ~2500 km2), obtained from the Intergovern-
mental Panel on Climate Change (http://www.ipcc.ch). All
Materials and methods environmental coverages were resampled to an intermediate
resolution of 0.1° (10 km2) prior to analysis. This set of en-
Ecological niche models were based on georeferenced vironmental variables has been seen to be appropriate and
occurrence points drawn from diverse sources, including adequate for modeling distributions in several studies (Peter-
museum specimen records, the World Wide Web, a fish pro- son and Vieglais 2001; Peterson 2003).

© 2006 NRC Canada

Zambrano et al. 1905

Fig. 1. Common carp (Cyprinus carpio) native distribution across Eastern Europe and Asia. Circles represent collection localities, and
potential distribution is represented by shaded areas; light gray indicates low agreement (1–5), dark gray indicates medium agreement
(6–8), and black indicates high agreement (9–10) among the 10 best-subset models.

To reduce environmental coverage sets to just those curacy of the models. To choose best subsets of models, we
coverages that provide highest predictive accuracy, we used a (i) selected the 20 models with lowest omission error based
series of jackknife manipulations (Peterson and Cohoon 1999). on independent test points, (ii) calculated the median area
In general, we ran multiple iterations (1–20) of models omit- predicted present among these low-omission models, and
ting each coverage or each suite of coverages systematically. (iii) identified the 10 models that were closest to the median
We then examined correlations between inclusion or exclusion overall average area predicted present. A final map was cre-
of each coverage (binary coded) and omission error (percent- ated as the sum of these 10 best models (consensus map).
age of extrinsic test presence data not predicted as present). Model predictions were validated on native distributional
Correlations on the order of r > 0.1 were considered indicative areas (i.e., Africa for Nile tilapia, Asia for common carp) via
of strong detrimental contribution of a particular coverage to the extrinsic testing data set. A χ 2 test was used to compare
model quality, and such coverages were removed from further observed success in predicting distributions of test points
analyses. It is important to note that the jackknife manipula- with that expected under random models (the product of pro-
tions were done solely on the native distribution of each spe- portional area predicted present and number of extrinsic test
cies and so do not detract from the independent nature of the presence points provides an estimate of occurrence points
invaded-range predictions and tests presented herein. correctly predicted were the prediction to be random with
To optimize model performance, we developed 100 repli- respect to the distribution of the test points). Positive results
cate models of each species’ ecological niche based on ran- in these tests would establish that models developed had suf-
dom 50–50 splits of available occurrence points and ficient predictive ability to be able to predict distributional
followed a modified procedure for extracting a “best subset” phenomena for these species.
of the replicate models for further consideration (Anderson Projecting these models to the Americas provided predic-
et al. 2003). The procedure is based on the observations that tions of potential geographic distributions of each species on
(i) models vary in quality, (ii) variation among models in- its invaded range. We validated model predictions via over-
volves an inverse relationship between errors of omission laying known occurrence points for non-indigenous popula-
(leaving out true distributional area) and commission (in- tions on the projections of the native-range models to the
cluding areas not actually inhabited), and (iii) best models invaded distributional areas, using the same χ 2 approach as
(as judged by experts blind to error statistics) are clustered described above. To provide a binary prediction for these
in a region of minimum omission of independent test points tests, we used 90% model agreement as a threshold for pre-
and moderate area predicted (an axis related directly to com- diction of presence. To provide a more conservative test, we
mission error). The position of the cloud of points relative to evaluated proportional area predicted present within a buffer
the two error axes provides an assessment of the relative ac- of 750 km around the area predicted present.

© 2006 NRC Canada

1906 Can. J. Fish. Aquat. Sci. Vol. 63, 2006

Fig. 2. Common carp (Cyprinus carpio) potential and known distribution in the Americas. Circles represent localities where this fish
has been collected, and potential distribution is represented by shaded areas; light gray indicates low agreement (1–5), dark gray indi-
cates medium agreement (6–8), and black indicates high agreement (9–10) among the 10 best-subset models.

Results populations were significant both for North American popula-

tions (predictive success = 87.43%, χ 2 = 292.25, df = 1, P <
Common carp distribution 0.0001) and South American populations (predictive success
The native distribution of common carp covers a large = 53.33%, χ 2 = 9.93, df = 1, P < 0.002).
area from Eastern Europe eastward across Russia and China
(Fig. 1). In most of this area, temperature in winter is cold Nile tilapia distribution
enough to freeze lakes, but in summer water reaches 25 °C, The native distribution of Nile tilapia covers most of cen-
which is the temperature needed for carp reproduction tral Africa, limited to the north by the Sahara Desert and to
(Matiland and Campbell 1992). Our models predicted distri- the south in northern South Africa, where the limiting factor
butional areas for carp across broad swaths of Asia (Fig. 1). might be low air and water temperatures (Sifa et al. 2002).
Statistical tests indicate that models had much higher Our models predicted potential distributional areas across
predictivity than random expectations (predictive success = most of Africa except for the Sahara and the Kalahari
92.64%, χ 2 = 28.51, df = 1, P < 0.0001). deserts, but including the Mediterranean coast (Fig. 3).
Projection of native-range ecological niche models for Model predictivity was high and statistically significant (pre-
common carp to the Americas shows broad areas with suit- dictive success = 87.48%, χ 2 = 17.5, df = 1, P < 0.001).
able conditions for the species. In North America, the pre- Projection of the best-subset models to the Americas indi-
dicted potential range includes most of Canada, the United cates suitable conditions for Nile tilapia establishment from
States, and Mexico. Only the Sonoran Desert appeared in- the southeastern United States south along the coastal low-
appropriate for carp (Fig. 2). In South America, areas prone lands of Mexico and Central America (Fig. 4). In South
to invasion are located principally in the south (Argentina, America, however, the species’ shows a broader geographic
southern Brazil, and southern Chile; Fig. 2), but areas in potential: from central Brazil to central Argentina, and from
the Andes suggest that opportunities exist even in tropical the eastern slopes of the Andes to the Atlantic Ocean, as
areas. Tests of model predictions for non-indigenous carp well as parts of the coast of Venezuela and Guyana (Fig. 4).

© 2006 NRC Canada

Zambrano et al. 1907

Fig. 3. Nile tilapia (Oreochromis niloticus) native distribution across the African continent. Circles represent localities where this fish
has been collected, and potential distribution is represented by shaded areas; light gray indicates low agreement (1–5), dark gray indi-
cates medium agreement (6–8), and black indicates high agreement (9–10) among the 10 best-subset models.

Tests of model predictions for non-indigenous Nile tilapia (Fitzsimmons 2001). Only the Sonoran Desert, northern Chile,
populations were significant both in North American popu- and southeastern Argentina are likely to be unsuitable for
lations (predictive success = 59.37%, χ 2 = 33.17, df = 1, P < both species.
0.001) and in South America populations (predictive success = The resulting invasive potential of these species in the
77.27%, χ 2 = 54.25, df = 1, P < 0.001). Americas is a spatial representation of the appropriate envi-
ronmental conditions (i.e., ecological niches) modeled in
their native ranges. Hence, the veracity of the results de-
Discussion pends on how well the GARP system was able to represent
The area that common carp or Nile tilapia may potentially the niche of species, which in turn depends directly on the
inhabit in Asia, Africa, and the Americas indicates broad in- quality of the input data. For this study, we did not have
vasive potential. According to our results, common carp and available range-wide aquatic environmental data on factors
Nile tilapia have a geographically extensive, invasive poten- that directly determine the presence of fish species (e.g., wa-
tial in the Americas, owing to their broad ecological niches. ter temperature, pH, turbidity, etc.). Instead, we used atmo-
Common carp has the potential to establish in temperate sys- spheric variables as proxies that have some limitations and
tems, in both subtropical and temperate regions, and in the may bias the results, as can be observed in the Nile tilapia,
highlands within the tropics. Nile tilapia, on the other hand, which was not predicted present in the Sahara. However,
has the potential to invade almost all tropical regions, partic- populations can be established there when water is available.
ularly in the lowlands; indeed, this species is presently es- Ecological consequences of invasion and establishment of
tablished in virtually every country in the Americas these species in local systems can be serious. Direct impacts

© 2006 NRC Canada

1908 Can. J. Fish. Aquat. Sci. Vol. 63, 2006

Fig. 4. Nile tilapia (Oreochromis niloticus) distribution in the Americas. Circles represent localities where this fish has been collected,
and potential distribution is represented by shaded areas; light gray indicates low agreement (1–5), dark gray indicates medium agree-
ment (6–8), and black indicates high agreement (9–10) among the 10 best-subset models.

of introduced common carp and Nile tilapia in natural sys- General considerations for aquaculture
tems include population depletion and even local extinctions In recent years, common carp and Nile tilapia aquaculture
of native species (Cahn 1929; Ogutu-Ohwayo 1990; Cano- has increased in many American countries as a consequence
nico et al. 2005). Furthermore, these species are known to of a crisis in the fisheries resulting from overfishing and re-
affect the physical conditions of aquatic systems via increas- duction in shrimp culture (Aiken et al. 2002; Alceste et al.
ing turbidity, changing concentrations of nutrients and sus- 2001). The economic strategy has tended towards a switch
pended solids, and reducing spatial heterogeneity, from species of high monetary value but high production
particularly in rooted plants (Zambrano and Hinojosa 1999), cost (e.g., shrimp, Litopenaeus vannamei) to species of lower
altering the whole system structure and dynamics. economic value but very low production cost (e.g., common
The Americas hold an impressive fish diversity, including carp, C. carpio and Nile tilapia O. niloticus). Under this
>5400 freshwater species (Reis et al. 2003). Endemism reaches scheme, common carp and Nile tilapia production has to be
30% in the United States, Mexico, and Chile. Together with high to generate acceptable profit, increasing risk of negative
Brazil, with its immense Amazonian diversity, these coun- impacts on the system. In some regions, this production sys-
tries have seen the highest increases in Nile tilapia produc- tem has been economically and socially successful; however,
tion in recent years (Fitzsimmons 2001). Nile tilapia can in many others, ecological consequences were counter-
establish in systems dominated by American fish families productive (Tapia and Zambrano 2003).
such as Poecilidae (216 species), Cichlidae (406 species), Concern about potential negative effects of introductions
and Characidae (952 species), which together include >1574 of alien fish species is minimal in most countries. Awareness
species (Reis et al. 2003). Similarly, carp invasion may jeop- of dangers of introductions and the potential for dire ecolog-
ardize even entire endemic American families (e.g., ical and biodiversity consequences has developed only rela-
Goodeidae, >40 species). tively recently, after decades of promotion of aquaculture as

© 2006 NRC Canada

Zambrano et al. 1909

a solution to fulfill protein needs for human populations. Martínez-Meyer, E., Peterson, A.T., and Hargrove, W.W. 2004.
Once these species are established in a system, eradication is Ecological niches as stable distributional constraints on mammal
extremely expensive and in many cases impossible. Hence, species, with implications for Pleistocene extinctions and cli-
before releasing alien species in a system, it is necessary to mate change projections for biodiversity. Glob. Ecol. Biogeogr.
estimate the potential success and counterbalance the eco- 13: 305–314.
logical aftermath. This analysis should be mandatory in re- Matiland, P.S., and Campbell, R.N. 1992. Freshwater fishes. Harper
gions such as the Amazon Basin and central Mexico, where Collins Publishers, London, UK.
the potential for major loss of fish diversity is higher than McNyset, K. 2005. Use of ecological niche modeling to predict
any potential economical benefits. distribution of freshwater fish species in Kansas. Ecol. Freshw.
Fish, 14: 243–255.
Mills, E., Casselman, J., Dermott, R., Fitzsimos, J., Gal, G., Holeck,
Acknowledgements K., Hoyle, J., Johannsson, O., Lantry, B., Makarewicz, J., Millard,
E., Munawar, I., Munawar, M., O’Gorman, R., Owens, R.,
We thank the following institutions for their generous shar- Rudstam, L., Schaner, T., and Stewart, T. 2003. Lake Ontario:
ing of data: Museu de Zoología da Universidade de Sao food web dynamics in a changing ecosystem (1970–2000). Can.
Paulo, Universidade Federal da Paraiba Brazil, Tulane Uni- J. Fish. Aquat. Sci. 60: 471–490.
versity, The University of Kansas, Harvard University, Scripps Minns, C.K., and Cooley, J.M. 2000. Intentional introductions: Are
Institute of Oceanography, Los Angeles County Natural His- the incalculable risks worth it? In Nonindigenous freshwater or-
tory Museum, University of Michigan Museum of Zoology, ganisms. Lewis Publishers, New York. pp. 57–60.
Texas A&M University, Cornell University, Canadian Mu- Morgan, D.L., Gill, H.S., Maddern, M.G., and Beatty, S.J. 2004.
seum of Nature, Field Museum, California Academy of Sci- Distribution and impact of introduced freshwater fishes in West-
ences, South African Institute for Aquatic Biology, Fort ern Australia. N.Z. J. Mar. Freshw. Res. 38: 511–523.
Hays State University, Museum National d’Histoire Naturelle, Ogutu-Ohwayo, R. 1990. The reduction in fish species diversity in
lakes Victoria and Kyoga (East Africa) following human exploi-
Naturhistoriska Rlksmuseet, University of Alabama, Univer-
tation and introduction of non-native fishes. J. Fish Biol. 37:
sity of Washington, University of Southern Mississippi, Mis-
207–208.
sissippi Museum of Natural Science, Royal Ontario
Perrow, M.R., Jowit, A.J.D., Leigh, S.A.C., Hindes, A.M., and
Museum, and University of Florida. Rhodes, J.D. 1999. The stability of fish communities in shallow
lakes undergoing restoration: expectations and experiences from
the Norfok Broads (U.K.) Hydrobiologia, 408/409: 85–100.
References
Peterson, A.T. 2001. Predicting species’ geographic distributions
Aiken, K.A., Morris, D., Hanley, F.C., and Manning, R. 2002. based on ecological niche modeling. Condor, 103: 599–605.
Aquaculture in Jamaica. Naga, 25: 10–15. Peterson, A.T. 2003. Predicting the geography of species’ invasion
Alceste, C.C., Illingworth, J.A., and Jory, D.E. 2001. Tilapia farm- via ecological niche modeling. Q. Rev. Biol. 78: 419–433.
ing industry in Ecuador. Aquac. Mag. 27: 77–82. Peterson, A.T., and Cohoon, K.C. 1999. Sensitivity of distribu-
Anderson, R.P., Lew, D., and Peterson, A.T. 2003. Evaluating pre- tional prediction algorithms to geographic data completeness.
dictive models of species’ distributions: criteria for selecting op- Ecol. Model. 117: 159–164.
timal models. Ecol. Model. 162: 211–232. Peterson, A.T., and Robins, C.R. 2003. Using ecological-niche model-
Barel, C., Dorit, R., Greenwood, P., Fryer, G., Hughes, N., Jackson, ing to predict Barred Owl invasions with implications for Spotted
P., Kawanabe, H., Lowe-McConnell, R., Nagoshi, M., Ribbink, Owl conservation. Conserv. Biol. 17: 1161–1165.
A., Trewavas, E., Witte, F., and Yamaoka, K. 1985. Destruction of Peterson, A.T., and Vieglais, D.A. 2001. Predicting species inva-
fisheries in Africa’s Lakes. Nature (London), 315: 19–20. sions using ecological niche modeling. Bioscience, 51: 363–371.
Cahn, A.R. 1929. The effect of carp on small lake, the carp as a Peterson, A.T., Soberón, J., and Sánchez-Cordero, V. 1999. Conser-
dominant. Ecology, 10: 271–274. vatism of ecological niches in evolutionary time. Science (Wash-
Canonico, G., Arthington, A., Mccray, J., and Thieme, M. 2005. The ington, D.C.), 285: 1265–1267.
effects of introduced tilapias on native biodiviersity. Aquatic Con- Peterson, A.T., Scachetti-Pereira, R., and Hargrove, W.W. 2004.
servation Marine Freshwater Ecosystems, 15: 463–483. Potential geographic distribution of Anoplophora glabripennis
Carta Nacional Pesquera. 2004. Carta Nacional Pesquera. Diario (Coleoptera: Cerambysidae) in North America. Am. Midl. Nat.
Oficial de la Federación. 151: 170–178.
Fitzsimmons, K. 2001. Tilapia production in the Americas. INFO- Pinto, L., Chandrasena, N., Pera, J., Hawkins, P., Eccles, D., and
FISH, Kuala Lumpur, Malaysia. Sim, R. 2005. Managing invasive carp for habitat enhancement at
Grinnell, J. 1917. Field tests of theories concerning distributional Botany Wetlans, Asustralia. Aquatic Conservation Marine Fresh-
control. Am. Nat. 51: 115–128. water Ecosystems, 15: 447–462.
Hutchinson, G.E. 1957. Concluding remarks. Cold Spring Harbor Rahel, F.J. 2000. Homogenization of fish faunas across the United
Symposium. Quant. Biol. 22: 415–427. States. Science (Washington, D.C.), 288: 854–856.
Iguchi, K., Matsuura, K., McNyset, K., Peterson, A.T., Scachetti- Reis, R., Kullander, S.O., and Ferraris, C.J.J. 2003. Check list of
Pereira, R., Powers, K.A., Vieglais, D.A., Wiley, E.O., and Yodo, the freshwater fishes of South and Central America. Edipucrs,
T. 2004. Predicting invasions of North American basses in Japan Porto Alegre, Brazil.
using native range data and a genetic algorithm. Trans. Am. Fish. Sifa, L., Chenhong, L., Dey, M., Gagalac, F., and Dunham, R. 2002.
Soc. 133: 845–854. Cold tolerance of three strains of Nile tilapia, Oreochromis
Jackson, D.A. 2002. Ecological effects of Micropterus introduc- niloticus, in China. Aquaculture, 213: 123–129.
tions: the dark side of black bass. Am. Fish. Soc. Symp. 31: Stockwell, D.R.B., and Noble, I.R. 1992. Induction of sets of rules
221–232. from animal distribution data: a robust and informative method of

© 2006 NRC Canada

1910 Can. J. Fish. Aquat. Sci. Vol. 63, 2006

analysis. Mathematics and Computers in Simulation, 33: 385– Wiley, E.O., McNyset, K., Peterson, A.T., Robins, C.R., and Stew-
390. art, A.M. 2003. Niche modeling and geographic range predictions
Stockwell, D.R.B, and Peters, D.P. 1999. The GARP modeling sys- in the marine environment using a machine-learning algorithm.
tem: problems and solutions to automated spatial prediction. In- Oceanography, 16: 120–127.
ternational Journal of Geographical Information Science, 13: Zambrano, L., and Hinojosa, D. 1999. Direct and indirect effects of
143–158. carp (Cyprinus carpio L.) on macrophyte and benthic communities
Tapia, M., and Zambrano, L. 2003. From aquaculture goals to real in experimental shallow ponds in central Mexico. Hydrobiologia,
social and ecological impacts: carp introduction in rural central 408/409: 131–138.
Mexico. Ambio, 32: 252–257. Zambrano, L., Scheffer, M., and Martinez-Ramos, M. 2001. Cata-
Vander Zanden, J., Casselman, J., and Rasmussen, J. 1999. Stable strophic response of lakes to benthivorous fish introduction. Oikos,
isotope evidence for the food web consequences of species inva- 94: 344–350.
sions in lakes. Nature (London), 401: 464–467.

© 2006 NRC Canada