Biodiversity and Conservation

, Volume 20, Issue 5, pp 1125–1132

American bullfrog invasion in Argentina: where should we take urgent measures?


    • Laboratorio de Herpetología y Animales Venenosos, Centro de Zoología Aplicada, Facultad de Ciencias ExactasFísicas y Naturales, Universidad Nacional de Córdoba, Rondeau 798
    • Museo Patagónico de Ciencias Naturales
  • Mauricio S. Akmentins
    • Centro de Investigaciones Básicas y Aplicadas (CIBA)Universidad Nacional de Jujuy
    • Instituto de Bio y Geociencias del NOA (IBIGEO)Universidad Nacional de Salta
  • Romina Ghirardi
    • Instituto Nacional de Limnología (CONICET_UNL)Ciudad Universitaria
  • Nicolás Frutos
    • Instituto de Altos Estudios Espaciales Mario Gulich, Comisión Nacional de Actividades Espaciales (CONAE)Universidad Nacional de Córdoba (UNC)
  • Gerardo C. Leynaud
    • Laboratorio de Herpetología y Animales Venenosos, Centro de Zoología Aplicada, Facultad de Ciencias ExactasFísicas y Naturales, Universidad Nacional de Córdoba, Rondeau 798
Brief Communication

DOI: 10.1007/s10531-011-0014-3

Cite this article as:
Nori, J., Akmentins, M.S., Ghirardi, R. et al. Biodivers Conserv (2011) 20: 1125. doi:10.1007/s10531-011-0014-3


Argentina is the country with the most geographically extended biological invasion of the American bullfrog (Lithobates catesbeianus) in South America after Brazil. Here, we used a maximum entropy ecological niche modeling algorithm (using records of the native range of American bullfrog) to project the model onto the whole of Argentina. We determined the most suitable habitats for this invasive alien species and where we consider urgent measures should be taken. Our projections showed good agreement with known feral populations of American bullfrog in Argentina. By implementing the “Multivariate Environmental Similarity Surface” analysis, we be able to determine that factors such as low precipitations or highest altitudes could be limiting the species’ ability to invade the west and south of the country. We suggest that strategies should focus on detecting established feral populations of the American bullfrog and preventing further introductions or range expansion of feral populations in the northeast portion of the country. Lastly, we report a new feral population of bullfrogs in Argentina.


ArgentinaEcological modelingInvasionLithobates catesbeianusMAXENTMultivariate Environmental Similarity Surface


Over the past three decades, the North American-native American bullfrog, Lithobates catesbeianus, has become a part of an increasingly centralized and expanded trade of live animals for food within South America, and between South America and the United States of America (Mazzoni et al. 2003; Hanselmann et al. 2004). As a consequence of this trade, there are currently feral populations of the American bullfrog in the majority of South American countries. Brazil and Argentina have reported the most extensive biological invasion of the American bullfrog, with feral populations reported in a large extent of their territories (Giovanelli et al. 2008; Akmentins and Cardozo 2010). In Argentina, the biggest expansion of bullfrog captive-breeding facilities started in 1983 and remained without survey or control by provincial or national government agencies up to 1993. In 1994, the Argentine Fish and Agriculture Agency (Dirección Nacional de Pesca y Acuicultura) began a regulation program on the introduction of exotic aquatic organisms. This year, 69 bullfrog farming facilities were reported in nine of the 23 provinces of the country (Luchini 1995). At present, only a few frog farms are still in production since a great majority where closed and dismantled (Akmentins per. obs.). Akmentins and Cardozo (2010) suggest that the lack of effective control over captive-breeding programs of exotic animals is of serious concern and the main source of feral populations of the American bullfrog in Argentina. They described the need to develop a predictive ecological niche model to access the outcome of this biological invasion.

Recently, researchers have developed some techniques to allow modelling species distribution based on ecological niche dimensions (Peterson 2001). One of the applications is to predict ecologically suitable areas for the establishment of invasive species in regions where the species is not found yet (Peterson 2003; Giovanelli et al. 2008; Elith et al. 2010). This methodology builds an ecological niche model based on the values of environmental variables (niche dimensions) at know localities for the target species (Ron 2005). Some global predictive models for the potential distribution of invasive American bullfrog populations have been proposed (e.g. Ficetola et al. 2007; Giovanelli et al. 2008), however, these works were mostly aimed at predicting bullfrog invasions on Brasil and Europe, without details of Argentina.

Any attempt to limit further spread of the bullfrog requires extensive knowledge of their current distribution. Considering the increasing number of feral populations reported in Argentina (Akmentins and Cardozo 2010), there is an urgent need to identify candidate habitats for future monitoring (Rödder et al. 2009). In this study, we used the ecological niche model based on distributional data from the native range of the bullfrog, to determine the most suitable areas for this species in Argentina and compared our results with the reported feral populations of American bullfrog in this country. Additionally, we determined the most limiting variables for the species in relation to their native range of distribution. Lastly, we reported a new feral population of American bullfrog from central Argentina.

Materials and methods

We performed field sampling in an area of Central Argentina (Los Sauces River, 3 km north of Villa Dolores, Córdoba province, Argentina). The specimens were collected by hand or fishing net and euthanized with an overdose of Lidocaine 2%. They were then fixed with formalin 10%, transferred to 70% ethanol after 3 days and deposited as part of the herpetological collection in the Centro de Zoología Aplicada (Córdoba, Argentina).

In order to obtain an ecological niche model for the native distribution of American bullfrog, we used 897 geo-referenced individual records from North America, obtained from the Global Biodiversity Information Facility (2010) and HerpNet (2010) databases, including occurrences in Mexico, USA, and Canada (Fig. 1a). Georeferencing was conducted when necessary with the Alexandria Digital Library Gazetteer ( All data were checked in the DIVA-GIS software for bias and errors (Hijmans et al. 2002).
Fig. 1

a Predicted potential geographic distribution for native range of Lithobates catesbeianus. White circles: localities within the specie native range (obtained from HerpNet and Global Biodiversity Information Facility). Inset: area of environmental calibration for the purpose of projecting the habitat suitability model. b Projection map for the potential distribution of Lithobates catesbeianus in Argentina and zoom of the area with highest probability values showing the river system. Black star: New reported locality. White cruxes: real occurrence localities in Argentina. Black lines: rivers. c Limiting factors based on MESS analysis showing the most dissimilar variables (MoD) in each pixel outside the calibration area

We used 19 bioclimatic layers and 1 topographic layer as predictors. The bioclimatic variables resulted from global land area interpolation of climate point data (1950–2000) at a spatial resolution of 10 arc-min (Hijmans et al. 2005; available at The altitude variable was derived from USGS, at the same spatial resolution (available at

The realized ecological niche was modeled using MAXENT 3.3.3a (Phillips et al. 2006; MAXENT uses the maximum entropy principle: in the estimation of an unknown probability distribution, the least biased solution is the one that maximizes its entropy, subject to some constraints that reflect available information. In the case of ecological niche modeling, the target is to calculate a probability distribution (the niche model) for a species over the given geographical space, using information from the observed association of the species’ localities with environmental variables. MAXENT calculates this maximum entropy distribution using the observed association between the species’ and environmental layers to set the following constraint: the expected value (expectation) of each ‘‘feature’’ (which is either an independent variable itself or one derived from it) under the estimated distribution must be similar to its observed average over sample locations (Phillips et al. 2006).

Occurrence data were divided into training data (75% of occurrence point data used for model prediction) and test data (25% of occurrence point data used for model validation). Several runs were performed and ten replicates were used as the input parameter and were “cross-validated” as replicated run type. The resulting average model was evaluated with the Receiver Operating Characteristics Curve (ROC), calculating the area under the curve (AUC), a threshold independent index widely used in ecological studies. A ROC plot was created by plotting the sensitivity values, the true-positive fraction against 1-specificity, and the false positive fraction for all available probability thresholds (further reading: Fielding and Bell 1997; Manel et al. 2001). The AUC is a measure of the area under the ROC ranging from 0.5 (random accuracy) to a maximum value of 1.0 (perfect discrimination).

The cumulative output of MAXENT is a continuous map which allows fine distinctions to be made between the modeled suitability of different areas (scale range: 0–100). To generate a prediction map, we used a cumulative threshold value that balances training omission, predicted area, and threshold value, i.e., balancing commission and omission errors (Phillips et al. 2006). The resulting model was projected onto all of Argentina to assess the potential geographic distribution of the species. Then, we plotted the actual species occurrence points in Argentina on the predicted distribution map. To avoid spurious projections, we calibrated the models using the native range of the species (represented by the inset, Fig. 1a). By implementing the “multivariate environmental similarity surface” (MESS) (implemented in MAXENT 3.3.3a) across this analysis, we can know how similar a point is to a reference set of points (native range), with respect to a set of predictor variables. Negative values discriminate areas where at least one variable has a value that is outside the range of environments of the reference set. Information on which variable is driving the MESS value at points outside the calibration area were extracted and mapped with the aim to know the most dissimilar (or limiting) variables (MoD) for the species (Fig. 1c) (For details see Elith et al. 2010).

Results and discussion

Climate matching and numerous introduction attempts are the basis for a successful biological invasion (Kolar and Lodge 2001; Bomford et al. 2009). This is clearly the case of the recent American bullfrog invasion in Argentina. In fact, Argentina is the country with the most geographically extended biological invasion of the American bullfrog in South America after Brazil (Akmentins and Cardozo 2010) (Fig. 1b). Although there are less than 20 known bullfrog populations in Argentina (Table 1), all were reported over the past 5 years, adding to the number of frogs’ farms installed in the country, which leads us to think there are several unreported feral populations. In this study, we report a new feral population of American bullfrog, located at Los Sauces river, 3 km North of Villa Dolores, Córdoba province, Argentina (between 31°55′31″S, 65°08′38″W and 31°54′47″S, 65°07′61″W). This is the second feral population reported for this province, 158 km southwestern from the first one. We documented a well established feral population (adults, tadpoles and metamorphs) of American bullfrog in a lotic environment and collected 5 of them deposited at herpetological collection of Centro de Zoología Aplicada: CZA-H 150-154.
Table 1

Real occurrence localities of Lithobates catesbeianus in Argentina






Buenos Aires

Bahía Blanca



IABIN (2010)

Buenos Aires

9 de Julio district—permanent semi-natural pond



Barraso et al. (2009)

Buenos Aires

9 de Julio district—irrigation channel system



Barraso et al. (2009)

Buenos Aires

9 de Julio district—rural zone



Barraso et al. (2009)


Aguas de las Piedras



Akmentins et al. (2009)


Villa Dolores



Present study


Guaraní—El Soberbio



Pereyra et al. (2006)


San Pedro—RB Yabotí: Saltos del Moconá



Pereyra et al. (2006)


San Pedro—RB Yabotí: Moconá Provincial Park



Pereyra et al. (2006)


San Pedro—Tobunas



Pereyra et al. (2006)


La Candelaria



Akmentins & Cardozo (2010)

San Juan

Calingasta departament



Sanabria et al. (2005)

Bold letters indicate new reported locality

In the MAXENT model, using 25% random test points out of the data set for testing (224 for testing and 663 for training), we received very good AUC values (Test: 0.821; Training: 0.901) suggesting high predictive power. Model probabilities below the threshold value of 1.26 were classified as predicted absence and transformed to null values.

The projection of the native distribution model onto Argentina showed that the predicted occurrence included all but one (in northwestern Argentina, at San Juan province) of the feral populations of the species in the country. Based on our model, we consider that this is a clear example of human assisted introduction of an alien species in a poorly suitable environment. The future of this feral population is unclear because it is located in an arreic basin and may provide a clear chance to begin an effective eradication program (Adams and Pearl 2007).

Our results suggest that all areas in the northeastern portion and eastern coastal of Argentina are suitable areas for American bullfrog (Fig. 1b). The predicted areas of the Atlantic Forest (northeastern portion of Argentina) are in total agreement with the predictive distribution map of Giovanelli et al. (2008) for SE Brazil. Similarly, we conclude that the presence of American bullfrogs in this biome is of special concern because the Atlantic Forest is a biodiversity hotspot with a large number of endemic species which can be severely affected though predation and competition. Additionally, detected feral populations and the projected most suitable habitats are connected by Argentina’s mayor river systems belonging to the Paraná and Uruguay rivers basins with a large latitudinal extension. This, added to the American bullfrogs’ capacity to successfully colonize lotic systems (Clarkson and De Vos 1986; Kupferberg 1997), raises concern for the possible rapid range expansion of the species.

Most of the Patagonia and High Andean biomes at south and west of Argentina were outside the model’s calibrated range for the species native distribution. An examination of the MoD in these areas shows that the main differences between these areas and the native range of the species are associated with precipitations and altitude (Fig. 1c). The same variables were the most influential in the model predictions (results not shown), so we can affirm that factors such as low precipitations, or highest altitudes could be limiting the species’ ability to invade the west and south of the country.

In agreement with our results, the areas with high values of suitability for American bullfrog include the zoogeographics regions of Eastern Subtropical and Littoral-Mesopotamian batrachofaunas, which are the most anuran diverse regions in Argentina (Cei 1980). Thus, we consider that urgent measures need to be taken in order to detect new feral populations in these areas. Also, control and eradication programs should be set up focusing on the four reported feral populations of Misiones province (Table 1) (reported populations in the areas with highest values of suitability) to prevent further range expansions. Lastly, the use of American bullfrog for economical activities, in particular in the center-eastern and northeastern portion of the country, should be strictly regulated and monitored by government entities with the aim of avoiding future accidental invasions.


We are grateful to María Eugenia Periago who provided invaluable support on field work and improved the English style. Also Ernesto Verga and María Lucrecia Herrero for collaboration in the field and Gerardo Coria for providing important information. Lastly, we thank Julián Lescano, Eduardo Sanabria, Nicolás Urbina-Cardona and Enrique Martinez Meyer for suggestions on an earlier version of the manuscript.

Copyright information

© Springer Science+Business Media B.V. 2011