OUP user menu

Native-Range Ecology and Invasive Potential of Cricetomys in North America

A. Townsend Peterson , Monica Papeș , Mary G. Reynolds , Neil D. Perry , Britta Hanson , Russell L. Regnery , Christina L. Hutson , Britta Muizniek , Inger K. Damon , Darin S. Carroll
DOI: http://dx.doi.org/10.1644/05-MAMM-A-133R3.1 427-432 First published online: 6 June 2006

Abstract

African giant pouched rats (Cricetomys) are native to tropical Africa, where they range from Senegal and Gambia east across West Africa and the Congo Basin to the Indian Ocean coast of East Africa. Ecological niche models show that Cricetomys species differ in their invasive potential. Although neither of the presently recognized Cricetomys species appears to have genuinely broad distributional potential in North America, models predict that C. emini would have extremely restricted distributional potential, whereas C. gambianus would have a broader potential across the southeastern United States.

Key words
  • African giant pouched rat
  • Cricetomys
  • ecological niche modeling
  • Genetic Algorithm for Rule-set Prediction
  • invasive species
  • rodent ecology

Invasive species represent a serious, ongoing challenge to the integrity of natural systems, with threats including competition with and exclusion of native species, introduction of novel pathogens, transformation of native habitats and ecosystems, and so on (National Academy of Sciences 2002). As such, development of a predictive understanding of the invasion process has become a particular focus of research in ecology and quantitative geography in recent years (Benning et al. 2002; Carlton 1996; Grosholz 1996; Johnson and Carlton 1996; Lawton and Brown 1986); the challenge, nonetheless, is complex (National Academy of Sciences 2002), and has only recently begun to yield useful predictions regarding geographic and ecological behavior of invasive species (Peterson 2003).

African giant pouched rats (Cricetomys) are native to tropical Africa, ranging from Senegal and Gambia east across West Africa and the Congo Basin to the Indian Ocean coast of East Africa. Members of this genus are large-bodied (>1-kg) murid-rodents that are mostly terrestrial, and that shelter in self-excavated burrows (Kingdon 1997; Rosevear 1969). Two species are currently recognized (C. emini and C. gambianus), although some doubt exists as to the validity of this arrangement (Wilson and Reeder 1993). Because both species are popular in the United States exotic pet trade, accidental or intentional releases of these species in North America must be considered as probable events. In fact, an established, stable population of C. gambianus, presumably the result of released exotic pet stocks, has recently been found on Grassy Key, in extreme southern Florida (Perry et al. 2006).

The existence of this established introduced population, along with the high probability of additional releases, justifies further investigation into the potential for dispersal and establishment of this genus in North America. Cricetomys was one of the exotic mammals linked to the 2003 North American monkeypox virus outbreak (Centers for Disease Control and Prevention 2003), and has previously been associated with monkeypox outbreaks in the Democratic Republic of the Congo (Hutin et al. 2001). Cricetomys is linked to other known and potentially zoonotic pathogens (Gretillat et al. 1981; Herder et al. 2002; Hutin et al. 2001; Machancu et al. 2004) and, as such, represents a concern regarding establishment of novel zoonotic diseases in the United States (Anderson et al. 2003; Enserink 2003; Langkop et al. 2003; Perkins 2003; Reed et al. 2004). Finally, given the nebulous nature of the presently accepted species limits, light shed on their similarity or difference in different character sets would be welcome. As such, we applied the new tools of ecological niche modeling (Peterson 2003; Peterson et al. 2002) to evaluate the invasive potential of these rodents in North America and to test whether ecological differences are detectable between the 2 species.

Materials and Methods

Input data.—We accumulated native-range point-occurrence data for the 2 currently recognized species in the genus (51 for C. emini and 125 for C. gambianus), based on label data associated with specimens housed in natural history museums (Appendix I). Total numbers of unique localities for the 2 species were 51 and 125, respectively. All occurrence data were georeferenced to the nearest 0.1′ of latitude and longitude based on reference to Internet-based gazetteers (United States Geological Survey 2005).

Environmental data sets (in the form of digital maps, or “coverages”) employed to describe ecological landscapes included 14 coverages summarizing aspects of topography (elevation, slope, aspect, topographic index, flow accumulation, and flow direction from the United States Geological Survey's Hydro-1K data set—United States Geological Survey 2001; native resolution 0.01 × 0.01°) and climate (annual means of daily temperature range, frost days, wet days, vapor pressure, precipitation, and maximum, minimum, and mean temperatures for 1960–1990 from the Intergovernmental Panel on Climate Change, http://ipcc-ddc.cru.uea.ac.uk/cru_data/examine/HadCM2_info.html; native resolution 0.5 – 0.5°). To minimize conflicts in scale between topographic and climatic data, we conducted analyses at an intermediate resolution (0.1 – 0.1°).

Niche modeling.—This approach was based on the idea of modeling ecological niches, which have been shown to constitute long-term stable constraints on the potential geographic distributions of species (Martinez-Meyer et al. 2004; Peterson 2003; Peterson et al. 1999; Raxworthy et al. 2003). Ecological niches are herein defined as the set of conditions under which a species is able to maintain populations without immigration (Grinnell 1917, 1924). Several avenues of research have demonstrated widespread evolutionary conservatism in niche characteristics, allowing accurate predictions of the potential distributions of invasive species (Enserink 2003; Peterson 2003; Peterson et al. 2003a, 2003b; Peterson and Robins 2003; Peterson and Vieglais 2001). This method consisted of 3 steps: model ecological niche requirements of species based on known occurrences on native distribution areas; test accuracy of native-range predictions; and project niche model onto other regions to identify areas putatively susceptible to invasion.

The software tool used for niche modeling was the Genetic Algorithm for Rule-set Prediction (GARP—Stockwell 1999; Stockwell and Noble 1992; Stockwell and Peters 1999). GARP uses a genetic algorithm to search for nonrandom associations between environmental variables and known occurrences, as contrasted with environmental characteristics of the overall study area. Previous tests of GARP have shown successful predictions of distributional phenomena for numerous taxa and regions (Anderson et al. 2003; Feria and Peterson 2002; Levine et al. 2004; Peterson and Cohoon 1999), including many applications to predicting the distributional potential of invasive species (see above).

Within the GARP program, available occurrence points are re-sampled with replacement to create a population of 1,250 presence points; an equivalent number of points is resampled from the population of grid squares (“pixels”) from which the species has not been recorded. These 2,500 points are divided equally into training (for creating models) and testing (for evaluating model quality) data sets.

Models are composed of a set of conditional rules developed through an iterative process of rule selection, evaluation, testing, and incorporation or rejection (all within the GARP program's processing, and not subject to user options). First, a method is chosen from a set of possibilities (e.g., logistic regression, bioclimatic rules, etc.) and applied to the training data set. Then, a rule is developed by a number of means mimicking DNA evolution (point mutations, deletions, crossing over, etc.) to maximize predictive accuracy. Rule accuracy is evaluated via the testing data, as a significance parameter based on the percentage of points correctly predicted as present or absent by the rule. The change in predictive accuracy from one iteration to the next is used to evaluate whether a particular rule should be incorporated into the final rule set (i.e., did accuracy increase or decrease?). The algorithm runs either 1,000 iterations or until addition of new rules has no effect on predictive accuracy (“convergence” most models converge at 100–120 iterations, so the maximum iteration parameter is never reached in practice). The final rule set, or ecological niche model, is then projected onto a digital map to identify a potential geographic distribution.

Spatial predictions of presence and absence can hold 2 types of error: omission (predicted absence in areas of known presence) and commission (predicted presence in areas of known absence—Fielding and Bell 1997). Because GARP is a random-walk procedure, it does not produce unique solutions; consequently, we followed recently published best-practices approaches to identify an optimal subset of resulting replicate models (Anderson et al. 2003). For each species, we developed 100 replicate models; of these models, we retained the 20 with lowest omission error. Finally, we retained the 10 models with moderate commission error (i.e., we discarded the 10 models with area predicted present showing greatest deviations from the overall median area predicted present across all models). This “best subset” of models was summed to produce final predictions of potential distributions.

To validate the model predictions, we compared their ability to predict independent sets of test points with that expected under null models of no association between test points (assumed independent of one another) and model prediction. In trial runs, we split available occurrence data randomly into 2 equal sets, and used one for model development and the other as an independent test of model quality. Models were validated via chi-square tests that incorporate dimensions of correct prediction of both presences (based on independent test data) and absences (based on expected frequencies). Expected frequencies under the null model were calculated as the product of the proportional area predicted present and the number of test presence points. Observed frequencies of correct and incorrect predictions of presence were then compared with expected frequencies using a chisquare statistic with 1 d.f.

Results

The predicted native distributions of the 2 Cricetomys species overlapped across tropical Africa (Fig. 1). In general, C. gambianus was predicted to occur broadly, both geographically and ecologically, including diverse tropical wet and dry habitats, whereas C. emini was predicted more narrowly in the humid rain-forest regions of central and western Africa. Compared with random null models, all models predicted independent subsets of known occurrences better than null models of no association between test points and predictions (P < 0.05).

Fig. 1

Summary of native range and potential North American predictions for Cricetomys emini (top), C gambianus (middle), and the combined occurrences of these 2 species in a single niche model (bottom). The inclusion of all African occurrence points for Cricetomys yields a more robust data set that allows for the visualization of the worst-case scenario regarding the invasive potential of this genus in North America. In each case, the native range is depicted with known occurrences (white squares), and the distributional predictions are shown with darker shades of gray indicating greater model agreement in predicting presence or potential presence (white = 0 models predict presence, black = 10 of 10 models predict presence).

By projecting models developed on native distributional areas for the 2 species onto North America, the geographic extent of the native-range niche in North America can be predicted (Fig. 1); these maps can be interpreted as representing the spatial extent of conditions modeled as appropriate for the species on their native ranges. In general, C. emini had a restricted potential North American distribution, with high levels of prediction only in a few sites in peninsular Florida and Louisiana. C. gambianus, on the other hand, showed a broader potential distribution, including peninsular Florida and the Gulf Coast west to Texas, as well as parts of the Pacific Northwest; much less agreement between models was evident, as some areas were predicted in the Great Lakes region and along the east coast, but these predictions do not have high model agreement.

Because both Cricetomys species are part of the United States exotic pet trade and current distributional data regarding Cricetomys species may often be based on incorrect identifications, we combined the occurrences of the 2 species and created a niche model for Cricetomys as a whole (Fig. 1, bottom panel). Although combining distinct niches within a single model is likely to cause overestimation of niche dimensions and over-prediction of distributional areas (Peterson and Holt 2003; Stockwell and Peterson 2002), the inclusion of all African occurrence points for Cricetomys yields a more robust data set that allows for the visualization of the worst-case scenario regarding the invasive potential of this genus in North America.

Finally, visualizing the niches of the 2 species, a curious relationship is evident. The modeled niche of C. gambianus is much larger than that for C. emini—3,649 environmental combinations for C. gambianus compared to 1,018 for C. emini. However, the niche of C. emini is almost completely nested within that of C. gambianus, with only 3% of its modeled niche not also being modeled as within the niche of C. gambianus. Visualizing the niches of the 2 species in 2 climatic dimensions (Fig. 2), the nested nature of their niches is clear, as is the concentration of C. emini in areas of high precipitation.

Fig. 2

Visualization in 2 dimensions (annual mean temperature and annual mean precipitation) of the ecological niche models developed for Cricetomys gambianus (white diamonds) and C. emini (gray diamonds) in relation to the availability (dots) of conditions across tropical sub-Saharan Africa.

Discussion

Invasive potential.—The methods used in this paper are relatively new, and may be unfamiliar to many readers. They are founded on the concept of an ecological niche and its influence on the geographic potential of species, and on the empirical observations that ecological niches are highly conserved over evolutionary time periods (Huntley et al. 1989; Martinez-Meyer et al. 2004; Peterson 2003; Peterson et al. 1999; Raxworthy et al. 2003; Ricklefs and Latham 1992). Additionally, the ecological behavior of a species is fairly invariant with respect to community context (Peterson 2003). Given these 2 points, species are expected to obey the same ecological rules when placed in a novel geographic setting (i.e., as an alien or invasive species), which would offer considerable predictivity of geographic phenomena related to invasive species. This expectation has been confirmed in numerous studies that document the predictive nature of the geography of various invasive species (Peterson 2003).

The observation that Cricetomys species have successfully colonized a thus-far limited area of Florida suggests that these animals can persist if introduced to areas outside of their native African range. Although neither of the presently recognized Cricetomys species appears to have genuinely broad distributional potential in North America, these models predict that C. emini would have extremely restricted distributional potential, whereas C. gambianus would have broader potential in the southeastern United States. Currently, the only confirmed introduction of Cricetomys in North America is the population, apparently C. gambianus, established in Florida; it is not unlikely, however, that other Cricetomys populations have been introduced or could be introduced to other locations (Perry et al. 2006).

The species-level identifications in these analyses were drawn from museum specimens collected throughout Africa and identified based on traditional morphological characters (Kingdon 1997; Rosevear 1969), and it is possible that some of these identifications are correct only to the generic level. Additionally, some phenotypic overlap exists in features used to distinguish between these species in some portions of their geographic ranges (Rosevear 1969). As a consequence, we considered the genus Cricetomys as a whole and found that the resultant geographic potential of the genus in North America is only slightly expanded (Fig. 1) beyond that predicted for C. gambianus.

Species limits.—Descriptions of the ecology of Cricetomys describe C. emini as an obligate forest species and C. gambianus as being found in “grass woodlands” and disturbed forest settings (Rosevear 1969). The latter environment is the setting in which both species are said to often occur in sympatry (Rosevear 1969). The ecological characteristics of the niches modeled for the 2 Cricetomys species may have bearing on their taxonomic status. The broadly overlapping, nested nature of the distributions of these species suggests 2 potential explanations. First, Cricetomys could represent yet another example of ecological similarity persisting through speciation events (Huntley et al. 1989; Martínez-Meyer et al. 2004; Peterson et al. 1999; Raxworthy et al. 2003). Although this pattern is common, the Cricetomys example is not entirely typical; usually, ecologically similar and closely related taxa are not sympatric. Also, the marked difference in niche breadth (about 3-fold) would suggest either marked specialization or marked generalization from the ancestral niche in one of the daughter lineages.

The alternative possibility, of course, is that the currently recognized species limits are flawed. One possibility is that there are phenotypic markers that are either expressed or evolved in rain-forest Cricetomys populations, and thus some of these characteristics used to distinguish C. emini from C. gambianus result in incorrect assignments of species-level identifications. This has been alluded to in previous works that have questioned current species limits in Cricetomys on the basis of inconsistent diagnostic characters (Rosevear 1969; Wilson and Reeder 1993). The ecological characters considered herein (e.g., see Fig. 2) show no evidence of independent evolution in the C. emini lineage, and as such might suggest that populations of Cricetomys in high-rainfall areas often get identified as C. emini, regardless of biological reality.

Clearly, detailed molecular studies based on carefully vouchered specimen material are in order. The Cricetomys distribution patterns in this preliminary work have proven quite complex, so a final answer will have to await more detailed taxonomic and phylogenetic studies.

Conclusions.—Cricetomys hosts a variety of pathogens with the potential to cause human infections or infections in native fauna. Our results are particularly relevant given the recent appearance of monkeypox in the United States (Anderson et al. 2003; Enserink 2003; Langkop et al. 2003; Perkins 2003; Reed et al. 2004). Clearly, a priority should be the development of a detailed understanding of the systematics and ecology of this genus and clarification of which species has established in the United States. As a result of the monkeypox outbreak, federal policies were established to limit importation of these and other African rodent species that could potentially harbor such zoonotic diseases (Centers for Disease Control and Prevention 2003). Additional priorities should be the elimination of non-native Cricetomys populations, monitoring their potential to transmit zoonotic infections to humans, and monitoring their potential to cause harm to native fauna by active displacement or through the spread of disease.

Acknowledgments

We thank the natural history museums listed in Appendix I for their assistance in assembling the data sets that form the basis of our analyses. This research was supported by a contract from the United States Department of Defense. This research was partially funded by Cooperative Agreement 0491130808CA, between the United States Department of Agriculture, Animal and Plant Inspection Service, and the University of Georgia. We thank C. R. Okraska for field assistance with the trapping on Grassy Key, Florida, as well as P. Frank, the United States Fish and Wildlife Service National Key Deer Refuge, D. Setton, and C. Faast for logistical support and advice. The findings and conclusions in this report are those of the authors and do not necessarily represent the views of the Centers for Disease Control and Prevention or the United States Government.

Appendix I

Natural history collections used in gathering Cricetomys distributional data.—The Royal Museum for Central Africa, Tervuren, Belgium; The Natural History Museum, London, United Kingdom; Museum für Naturkunde, Berlin, Germany; Muséum National d'Histoire Naturelle, Paris, France; and collections served via the Mammal Networked Information System (MaNIS; Museum of Vertebrate Zoology, University of Washington Burke Museum, Los Angeles County Museum of Natural History, James R. Slater Museum, Field Museum, Michigan State University Museum, University of Kansas Natural History Museum, University of Minnesota Bell Museum of Natural History, Royal Ontario Museum, Museum of Natural Science, University of New Mexico Museum of Southwestern Biology, and Museum of Texas Tech University).

Footnotes

  • Associate Editor was Douglas A. Kelt.

Literature Cited

View Abstract