Biol Invasions DOI 10.1007/s10530-009-9667-1

ORIGINAL PAPER

Risk assessment of potential invasiveness of exotic imported to south Florida

Ikuko Fujisaki • Kristen M. Hart • Frank J. Mazzotti • Kenneth G. Rice • Skip Snow • Michael Rochford

Received: 5 February 2009 / Accepted: 16 November 2009 Ó Springer Science+Business Media B.V. 2009

Abstract The recent explosion of exotic reptiles in predict establishment success of 33 reptiles that were south Florida requires effective management strate- most frequently imported through Miami and St. gies. The objective of this study is to bring knowledge Petersburg ports from 2000 to 2005 and two additional of ecological correlates and quantitative modeling reptiles of concern in Florida, we identified eight methods into management by providing the foundation and four as potentially successful for a screening procedure that will identify potentially invaders. We further assessed adverse impacts associ- invasive and assess adverse impacts associated ated with potential invaders, should they become with these species. We considered 17 variables and, established, by identifying species that are (1) danger- based on model selection procedures, we identified the ous to humans, (2) dangerous to the ecosystem (upper following significant predictors of establishment suc- trophic-level predators), and (3) rapidly spreading. cess: taxonomic order, maximum temperature match Controlling exotic reptiles can be expensive and labor between a species’ native range and Florida, intensive once they are established. Information on sale price, and manageability (defined as a species’ which species are potential invaders based on screen- maintenance cost, aggressiveness, proneness to ing procedures and what impacts these species might escape, and venomousness). Applying the models to cause will be a valuable contribution to the develop- ment of proactive management strategies.

I. Fujisaki (&) F. J. Mazzotti M. Rochford Ft. Lauderdale Research and Education Center, University Keywords Climate match Establishment of Florida, 3205 College Ave., Davie, FL 33314, USA Exotic species Prediction e-mail: ikuko@ufl.edu

K. M. Hart Florida Integrated Science Center Davie Field Office, U.S. Introduction Geological Survey, 3205 College Ave., Davie, FL 33314, USA Global trade of live reptiles has facilitated introduc- tion and establishment of exotic reptiles in many K. G. Rice Florida Integrated Science Center Gainesville, U.S. locations around the world. Located at the southern Geological Survey, 7920 NW 71 Street, Gainesville, FL extent of the continental United States, south Florida 32653-3701, USA (south of Lake Okeechobee) has a history of success- ful establishment of a variety of herpetofauna origi- S. Snow Everglades National Park, 40001 State Road 9336, nating from tropical and temperate zones (Meshaka Homestead, FL 33034, USA et al. 2004). Although early introductions were via 123 I. Fujisaki et al. accidental import from shipping cargo (Butterfield in Australia and has also been used in New Zealand et al. 1997), increased popularity of exotic pets (Pheloung 1995; Pheloung et al. 1999). Daehler et al. has led to a currently booming trade of exotic reptiles, (2004) applied a screening procedure to identify pest and intentional or accidental release by pet owners plants in Hawaii and other Pacific Islands. and dealers became the major cause of recent Despite increasing concern about adverse impacts introductions (Hoover 1998; Kraus 2003). Exotic posed by exotic reptiles, relatively little effort has reptiles can cause adverse impacts in a number of been made to assess the risk of potential invasiveness ways. Some species are dangerous to humans and of exotic reptiles. Bomford et al. (2005) completed an domestic due to their large size and extensive literature review on ecological correlates, (Bomford et al. 2005; Reed and Rodda 2009). and Bomford et al. (2008) conducted the leading Introductions of exotic reptiles may have ecological quantitative risk assessment of reptiles and found impacts such as alterations to native food webs and climate match, , and number of jurisdictions to ecosystem functions through changes in competition be significant predictors of invasiveness. However, a and . Socioeconomic impacts are also pos- number of other factors may also be effective sible, such as reduced agricultural productivity, high predictors of successful establishment. Regarding control costs, and elevated consumer concerns (Bom- the adverse impacts potential invaders may have ford et al. 2005). Regulatory rules that consolidate once established (e.g., harm to humans and ecosys- most laws pertaining to non-native species became tems), there is also relatively little empirical infor- effective January 1, 2008 in Florida. These rules mation available. Reed and Rodda (2009) conducted prohibit release of non-native species and the keeping a risk assessment of nine constrictor species in the of certain species as pets. Under the rules, six exotic United States, rating each species (low, medium, reptiles including Python molurus bivittatus, Python high) on probability of establishment and conse- reticulatus, Python sebae, Morelia amethistina, quences of establishment. Their study indicated that Eunectes murinus and Varanus niloticus were iden- the majority of the assessed species had either high tified as ‘‘reptiles of concern’’ (ROC) and a permit is probability of establishment or high risk of estab- now required to possess these species. lishment consequences, or both, and thus are organ- In early 1980s, conservation concerns led a isms with risk potential. Scientific Committee on Problems of the Environ- In this study, we developed optimal quantitative ment (SCOPE) invasion program to identify three models to predict establishment success of exotic questions to guide future studies in invasion biology: reptiles, which can be incorporated into management (1) identifying factors that determine successful strategies. Our goal was to provide a foundation for invasion, (2) characterizing environments that are screening procedures to assess whether newly either vulnerable or resistant to invasions, and (3) imported species are likely to establish in Florida, incorporating knowledge obtained by addressing and if so, what potential adverse impacts may be these questions into effective management strategies associated with these species. The results of this (Davis 2006). Ecological correlates of the invasion study may provide a means to address the current process were identified in various taxa from early explosion of exotic reptiles in south Florida. works in invasion biology (Elton 1958; Baker and Stebbins 1965). Recent studies tend to focus on quantitative models of establishment success or the Methods invasion process (Hayes and Barry 2008). Various modeling methods were applied for this purpose, Data while concerns about inconsistency among different methods were pointed out (Hayes and Barry 2008). We used the Florida and Wildlife Conservation To improve invasive species management in the Commission (FFWCC) website (http://myfwc.com/ United States, the Ecological Society of America nonnatives/) and the U.S. Geological Survey Nonin- recommended the introduction of screening proce- digenous Aquatic Species (USGS NAS) database to dures for imported species (Lodge et al. 2006). A gather information on exotic reptiles in Florida. We similar screening tool was developed for exotic plants also added supplemental information derived from 123 Risk assessment of potential invasiveness of exotic reptiles

Bomford et al. (2005), Meshaka et al. (2004), and taxonomic order, that is, Crocodilian (crocodilians), consultations with local herpetologists and USGS -Lacertilia (lizards), Squamata-Serpentes NAS staff. We used the definition of establishment (snakes), and Testudines (turtles) instead of family success employed by USGS NAS, that the ‘‘popula- and genus. We did not include previous establishment tion is reproducing and over-wintering,’’ because it history, as the majority of recently imported species is inclusive of the definition from the FFWCC. had no record of attempting to establish elsewhere. Although we were initially interested in assessments Propagule pressure (number of arrived/released indi- of exotic reptiles in south Florida, to obtain sufficient viduals) was found to be an important factor for sample sizes for analysis we included all exotic rep- establishment success across different biological tile species that were either established or were groups including reptiles (Bomford et al. 2008; Hayes known to be not yet established (considered ‘‘failed’’) and Barry 2008). Whereas arrival numbers may be in Florida. The majority of species on our list were reflected by import records, release numbers cannot be distinct at the species level except for a few cases for defined (Bomford et al. 2008). Manageability and price which a particular was known to be are two variables we introduced to reflect release observed in Florida, such as P. molurus bivittatus. likelihood of each species from pet owners based on Also, we included a few species that are native to the assumptions that inexpensive pets and difficult pets United States and were observed in locations in may be more likely to be released. Florida, far from their native range. We excluded For each selected species, we characterized total species with an uncertain establishment status, even body size, primary diet, clutch or litter size, reproduc- though they may now be present in Florida. The final tive frequency, number of years to reach reproductive list of species included one crocodilian (established), maturity, and parthenogenesis by referencing books, 50 lizards (37 established and 13 failed), six snakes journal articles, and online databases, and by consult- (three established and three failed), and 11 turtles ing with herpetologists. Representative literature (one established and 10 failed). included Boulenger (1893, 1965),Pope(1935), Prit- There are many biotic and abiotic factors that are chard and Trebbau (1984), Stebbins (1985), Mehrtens generally thought to affect establishment success (1987), Ernst and Barbour (1989), Conant and Collins (Bomford et al. 2005), but some are difficult to define (1991), Schwartz and Henderson (1991), De Lisle (e.g., human commensalisms) or are dependent on (1995), Branch (1998), Cogger and Zweifel (1998), condition of the individual animal and release events Rodrı´guez-Schettino (1999), Cogger (2000), Henkel (e.g., parasite load, release location). Therefore, we and Schmidt (2000), Lee (2000), Stafford and Meyer only considered those attributes that could be defined (2000), Arnold (2002), Daniel (2002), Savage (2002), (either quantitatively or categorically) based on avail- Spawls et al. (2002), Hutchins (2003), Meshaka et al. able information. The variables we included in our (2004), Bonin et al. (2006), and Baha El Din (2006). analysis are taxonomic order, total body size, number For those variables defined by references, missing data of years required to reach reproductive maturity, clutch were scattered across species and variables. We size (i.e., number of eggs per reproductive bout), substituted for missing data the mean value for animals frequency of reproduction (i.e., number of bouts per of the same genus, the same upper taxonomic category, year), possible parthenogenesis, native geographic or a similar species based on expert opinions. To range size (in km2), climate match, import quantity, estimate geographic range size (km2), we hand-digi- manageability (i.e., relative difficulty to manage as a tized a native range polygon for each species using pet), and sale price. Bomford et al. (2008) found that georeferenced distribution maps and computed the family and genus were important predictors of estab- resulting area using a geographic information system. lishment success. However, as they stated that many Among a number of climate match methods previously families and most genuses have not been introduced used (Bomford et al. 2008; Rodda et al. 2009), we anywhere, these are not in our training data. This chose to use CLIMEX 3.0, which quantifies climate deficit poses a particular challenge in that establish- match level between two point locations from zero (no ment success of a newly introduced species whose match) to one (exact match), to derive climate match family and genus are not in the training data-set is not score between Florida and the native range of each predictable. To mitigate this problem, we included species (Sutherst et al. 2007). We created three raster 123 I. Fujisaki et al. layers of climate match scores of native range polygons distribution is met, logistic regression and RPART do of each animal per each climate attribute (minimum, not depend on this assumption (Johnson 1998). maximum, and average temperatures, rainfall pattern, Logistic regression usually has a binary response rain total, relative humidity, and soil moisture), with variable and predicts probability of an event occur- three locations in Florida (north, central, and south rence. An advantage of logistic regression over Florida) with 0.5 degree spatial resolution. Climate discriminant analysis may be ease of interpreting match scores for each species were defined as the mean model parameters. In contrast, RPART is a tree-based match scores of each raster layer. We obtained import classification method that iteratively splits trees and quantities of exotic reptiles at Miami and St. Petersburg decides on the best split at each node. Ease of ports from 2000 to 2005 through U.S. Fish and Wildlife interpreting results of tree-based models is appealing, Service’s (USFWS) Law Enforcement Management but a disadvantage is that the method tends to favor Information System (LEMIS). These records include training data. We used a stepwise procedure for over 90% of reptiles imported into Florida, the majority discriminant analysis and logistic regression for of which derive from the pet trade. We used import variable selection. Variable selection with RPART quantity of animals as a model input. Additionally, we was based on a cross-validation error rate and a cost- obtained data on animal price by visiting 89 dealer sites complexity measure (Venables and Ripley 2002). online and taking the median price for each species (excluding hybrids and animals with unique colors and Prediction of establishment and assessment patterns, sometimes referred to as ‘‘designer reptiles’’). of risks We independently verified price for each species by visiting a reptile show and conducting informal We created a list of frequently imported species using interviews with dealers. We defined pet ‘‘manageabil- LEMIS data including three crocodilians, 10 lizards, 10 ity’’according to mean scores within a five-point rating snakes, and 10 turtles. We also included two species, system (1 = very easy, 2 = easy, 3 = intermediate, E. murinus and M. amethistine, which are listed as ROC 4 = difficult, and 5 = very difficult), based on four in Florida. We then used establishment models to factors: cost of food and equipment, aggressiveness, predict establishment success of these exotic reptiles. proneness to escape, and venomousness. Ratings were Among those predicted to be potential invaders, we scored by three expert herpetologists (Kenneth Dodd, identified species that are (1) dangerous to humans, (2) USGS-retired; Kenneth Krysko, Florida Museum of dangerous to the ecosystem, and (3) fast spreading or Natural History; and Kevin Enge, FFWCC). dispersing. We identified species that may be dangerous to humans based on past records of the species attacking Predictive models of establishment success a person. We defined invaders as dangerous to the ecosystem if their diet indicates that they are higher A goal of the analysis was to select a set of variables trophic-level predators. We characterized spread by that was most relevant to predicting establishment reproductive potential, defined as number of eggs per success. There have been several methods used to reproductive bout 9 nesting frequency. analyze establishment success of exotic species, yet a single standard method has not yet emerged (Hayes and Barry 2008). We used discriminant analysis, Results logistic regression, and recursive partitioning and regression trees (RPART) that were used in a number Among the three regions of Florida for which we of representative studies in this field (Rejmanek and calculated climate match scores with the native range Richardson 1996; Kolar and Lodge 2002; Bomford of exotic reptiles, south Florida tended to have higher et al. 2008). Discriminant analysis classifies obser- climate match scores in assessed climate variables vations using a linear discriminant function based on (minimum, maximum, and mean temperatures, an assumption of multivariate normal populations. rain total, rainfall pattern, soil moisture, and relative We standardized variables to meet this assumption. humidity) compared to north and central Florida; Whereas discriminant analysis is known to be best at however, we elected to use climate match scores with classifying if the assumption of multivariate normal north Florida since our training data included species 123 Risk assessment of potential invasiveness of exotic reptiles that established in northern portions of the state and successful), herbivores (25% or 1 out of 4 successful), these values would be more conservative. Preliminary and carnivores that specialized in vertebrates (60% or exploratory analysis with t-tests and chi-square tests 3 out of 5 successful). This result is similar for adults: indicated that established and failed species differed success rates were 88% (29 out of 33), 43% (3 out of significantly on several variables (Table 1; Figs. 1, 2) 7), 31% (5 out of 16), and 60% (3 out of 5), 2 including taxonomic order (v2 = 16.34; P \ 0.01), respectively, for insectivores, herbivores, omnivores, 2 juvenile diet (v3 = 13.06; P \ 0.01), adult diet and carnivores that specialized in vertebrates. Estab- 2 (v3 = 11.57; P \ 0.01), number of years to reach lished reptiles typically required fewer years to reach reproductive maturity (t31 =-3.51; P \ 0.01), and reproductive maturity (x=1.6; SD = 1.1) than species manageability (t38 = 4.09; P \ 0.01); these variables that failed to establish (x = 3.4; SD = 2.5); estab- (except for juvenile diet) were significantly different lished reptiles also had higher manageability score between established and failed species after Bonfer- (more difficult to manage), (x = 4.2; SD = 0.6) than roni’s adjustment. Turtles failed to establish (91% or 9 failed species (x = 3.3; SD = 1.0). No other vari- out of 11 failed) more frequently than lizards (26% or ables tested were significantly different between 13 out of 50 failed) and snakes (50% or 3 out of 6 established and failed species. failed). Among juveniles, insectivores had the highest The final models from the three analysis methods establishment rate (88% or 30 out of 34 successful), as included slightly different sets of variables (Table 1). compared to omnivores (33% or 6 out of 18 For discriminant analysis (Wilk’s k = 0.73, F2,65 =

Table 1 Variables that are significantly different between and with Bonferroni correction (**) and selected variables in established and failed species with t-test (numerical variables) the final model (***) using discriminant analysis (DA), logistic and chi-square test (categorical variables) at a level of 0.05 (*) regression (LR), and classification tree (CT) p DA LR CT

Taxonomic ordera ** b *** Minimum temperature match Maximum temperature match *** Mean temperature match Rainfall total match Rainfall pattern match Soil moisture match Relative humidity match Native range size Juvenile diet * bb Adult diet ** bb Number of eggs Parthenogenesis b Number of years to reach reproductive maturity ** Import quantity Price *** Manageability ** *** *** *** Error rate (%) 27.1 23.9 38.2 Effective sample size 68 67 67 We used establishment probability 0.5 for cut-off with logistic regression. Error rates are based on cross validation for discriminant analysis and logistic regression, and root node error for classification tree a Crocodilians were not included in the chi-square test, logistic regression, and classification tree since there is only one exotic crocodilian in the training data b Variables are not included in analysis

123 I. Fujisaki et al.

123 Risk assessment of potential invasiveness of exotic reptiles bFig. 1 Box-plots of numerical variables by establishment status of 68 exotic reptiles (42 established and 26 failed) in Florida. Means of each variable by establishment status are above each plot

11.91, P \ 0.001), selected variables were sale price and manageability. Logistic regression identified taxonomic order and manageability as important predictive vari- ables. Finally, for RPART, selected variables were maximum temperature and manageability (Fig. 3). Manageability is the common variable included in final models of all three methods. Cross-validation error rates of final models for discriminant analysis and logistic regression were 27.1 and 23.9%, respectively. Smaller error rates of logistic regression may suggest suitability Fig. 3 Final model with classification tree results. The of this method over discriminant analysis. Root node numbers in each node indicate the proportion of established error for the final model by classification tree was 38.2%. to failed species The final model by logistic regression was as follows:

where pi is probability of establishment success, logit(piÞ¼7:25 þ 3:02 Xlizard;i þ 2:12 Xsnake;i Xlizard,i and Xsnake,i are binary variables as associated þ 1:35Xmanage;i with lizards and snakes (we used turtles as a base),

and Xmanage,i is manageability of ith exotic reptile species. Based on this final model, probability for establishment success increases by 0.05 and 0.02 for lizards and snakes, respectively (compared to turtles) if animals are easy to manage (manageability = 1). Similarly, probability of establishment success increases by 0.55 and 0.46 for lizards and snakes, respectively (compared to turtles) if animals are difficult to manage (manageability = 5). With the RPART model, over-fitting with data was prevented using the cost-complexity measure. In our model, the next possible split variable (i.e., average temperature match) was not included in the final model due to marginal improvement observed. According to the LEMIS database, over five million reptiles were imported through the Miami and St. Petersburg ports from 2000 to 2005 (Table 2). We selected three crocodilians and 10 lizards, 10 snakes, and 10 turtles that were most frequently imported and are neither established nor have failed to establish in the wild in Florida (Table 3). We did not include imported reptiles that were known only at the genus level or with ambiguous names. Therefore, actual import quantities of species examined could be higher than those presented (Table 3). Among fre- Fig. 2 Bar-plots of categorical variables including taxonomic quently imported species, Physignathus cocincinus order, parthenogenesis, and adult and juvenile diet by establish- ment status of 68 exotic reptiles (42 established and 26 failed) in stands out in import quantity, with over 200,000 Florida animals of this species imported over the six-year 123 I. Fujisaki et al.

Table 2 Summary of import records (number of species and animal quantity) of reptiles through Miami and St. Petersburg ports from 2000 to 2005 Known species Unknown species Total Species Quantity Species Quantity Species Quantity

Crocodile 7 11,409 1 60 8 11,469 314 3,570,738 116 808,062 430 4,378,800 229 650,749 67 11,869 296 662,618 Turtle 106 124,505 35 24,694 141 149,199 Unknown – – 3 50,116 3 50,116 Total 656 4,357,401 222 894,801 878 5,252,202

time period. This was also one of the most frequently establishment success of exotics, the focus of our sold lizards among the 87 pet dealers from which we study was to incorporate this knowledge into man- obtained price information (12 dealers sold the agement practice as a screening tool. In the model species). Overall, lizard species were imported in development and variable selection part of our study, larger quantities compared to other reptiles. we found additional evidence of ecological correlates Although each model has strengths and weak- of exotic reptiles previously discussed by Meshaka nesses, we used all models to predict successful et al. (2004) and Bomford et al. (2005). Bomford establishment of frequently imported reptiles to min- et al. (2008) reported that genus was a significant imize false negative predictions. Discriminant analy- predictor of establishment success, and we found that sis and logistic regression resulted in similar even order can be a useful predictor. Using order- predictions, although variables included in the final level information may be more realistic for screening models were slightly different (Table 3). The RPART purposes because the majority of imported species resulted in predicting six species with a potential belong to previously unsampled (i.e., neither estab- to establish, four of which were also predicted by lished nor failed to establish) genera. Confirming the discriminant analysis and logistic regression to suc- findings of Bomford et al. (2008), climate match, in cessfully establish. Although no crocodilians or turtles particular maximum temperature, was a strong pre- were predicted to successfully establish, there were dictor. This is not an unexpected result because most eight lizards and four snakes that were predicted to be exotic reptiles in Florida are tropical or subtropical in potential invaders in Florida by at least one model. derivation, and the majority of them were established Among these 12 species, P. sebae, arietans, in south Florida. Our study also explicitly demon- E. murinus, and M. amethistina are potentially strated that human-related factors are important pre- dangerous to humans (Table 4). P. sebae, B. arietans, dictors of animals’ establishment success. Specifically, E. murinus, and M. amethistina consume vertebrates successfully established species were characterized by and thus may become upper predators in the ecosys- lower sale prices and difficulty managing as pets, tem. sexlineatus and Chamaeleo senegal- implying that they would be more likely to be released. ensis are potentially fast-spreading species that could In sum, our final models showed that a combination of produce large numbers of eggs annually. biotic (order), abiotic (climate match), and human- related (price and manageability) factors can serve as a strong predictor of an exotic species’ likelihood of Discussion establishing in south Florida. This finding does not imply that other factors are insignificant, because Model development and variable selection associations between variables (e.g., and diet) clearly exist. The more variables included in the Whereas a number of previous studies have examined model, the better the predictions; however, given the ecological correlates and analytical methods to model large number of imported reptile species in Florida,

123 Risk assessment of potential invasiveness of exotic reptiles

Table 3 Predicted establishment success of 33 frequently Table 4 Species that are dangerous to humans, dangerous to imported reptiles and additional two reptiles of concern which ecosystem, and fast spreading species among frequently have not yet been reported as invaders in Florida imported species which were predicted to successfully estab- lish (denoted by an asterisk (*)) Species Import quantity DA LR CT Species Danger to Danger to Fast Crocodylia (crocodilians) human ecosystem spread Paleosuchus palpebrosus 2,011 – Physignathus Paleosuchus trigonatus 1,587 – cocincinus Crocodylus niloticus 146 – Takydromus * Squamata-Lacertilia (lizards) sexlineatus Physignathus cocincinus 210,007 * * * Uromastyx 88,885 * * * maliensis Uromastyx maliensis 62,463 * * Sceloporus Sceloporus malachiticus 50,168 * * malachiticus Chamaeleo senegalensis 40,462 * * Chamaeleo * senegalensis Pogona vitticeps 39,847 Tropidurus Tropidurus hispidus 25,092 * * hispidus armata 22,230 Coleonyx mitratus Coleonyx mitratus 21,753 * * * Chamaeleo dilepis Chamaeleo dilepis 16,754 * * * Python sebae ** Squamata-serpentes (snakes) Bitis arietans ** Corallus caninus 8,321 Eunectes murinus ** Python curtus 5,476 Morelia ** Epicrates cenchria 3,000 amethistina Bitis gabonica 2,761 Bitis nasicornis 2,013 selecting the strongest subset of predictor variables Python sebae 1,736 * provides a method to efficiently collect information Morelia viridis 1,373 required for screening. pullatus 1,229 We recognize that our data are not without oxycephalum 971 limitations. Our training data-set excludes species Bitis arietans 933 * * that have not yet been reported; it only provides Eunectes murinus –a * information on those that were already established Morelia amethystine 246 * and those that are known to have failed to establish. Testudines (turtles) Thus, unsuccessful establishment is probably under- Testudo horsfieldii 18,761 reported. Further monitoring and documentation Geochelone carbonaria 11,728 efforts can complement risk assessments by detecting Rhinoclemmys pulcherrima 11,547 presence of exotic species, and such efforts should be Cuora amboinensis 11,342 initiated where possible. We also should point out Kinixys homeana 9,021 that some variables may not be reflective of the Pelomedusa subrufa 6,815 establishment status of several species. For example, Cyclemys dentate 4,756 our recent import data may not be reflective of import Geochelone denticulate 4,264 composition or quantity around the time of early Rhinoclemmys punctularia 5,164 establishment of certain exotic reptiles. Through our Pelusios castaneus 2,264 data collection process, we found that acquiring even basic biological and life-history information for many Methods of predictions are discriminant analysis (DA), logistic regression (LR), and classification tree (CT). Asterisks (*) species in our training database was difficult. Revis- indicate that the species was predicted to successfully establish ing current training data to reflect newly available a Import quantity is uncertain due to ambiguous name in the information on these species will lead to improve- import data ments in our models, and likely our predictions. 123 I. Fujisaki et al.

Prediction of establishment success snakes and lizards to electrical infrastructure (Fritts and Leasman-Tanner 2001; Kraus and Cravalho 2001; Although most previous studies adopted a single Huriash 2008) and to the poultry industry (Fritts and modeling method to predict invasiveness, we fol- McCoid 1991) have been reported in Florida and other lowed Kolar and Lodge (2002), who used two states. The explosion of exotic reptiles in south Florida methods to predict establishment of fish species, in is a relatively new issue, so it is uncertain what specific using a multiple modeling approach. We applied impacts these species will cause in the future. How- three methods that have been used frequently in ever, information on which species have the greatest recent literature on invasion biology. Such a multiple potential to become invasive and what types of modeling approach can be an effective method to adverse impacts these species might have will help reduce false negative predictions ( II errors). to inform management strategies. The reliability of predictive models in invasion biology has been called into question, largely because Recommendations and future study release and establishment are highly uncertain because chance and stochastic events are not incorporated into We recommend extending this study to assess the the model (Gilpin 1990; Williamson and Fitter 1996). establishment success of all reptile species imported to In our study, discriminant analysis and logistic Florida. Managing and controlling exotic reptiles can regression resulted in relatively low cross-validation be expensive and labor intensive once species are error rates (24–27%), while that of RPART was established (Lodge et al. 2006); the availability of comparably higher (38%). Although Hayes and Barry information such as that provided by our models could (2008) discussed a concern about inconsistent results greatly expedite this process. Applying this risk when using different methods, our three models assessment model to screen imported species would all predicted similar potential invaders, suggesting allow us to develop a list of ‘‘risky reptiles’’ for which consistency among models. Moreover, one of the trade would be restricted to eliminate risk of estab- potential invaders we predicted, P. sebae, was recently lishment. A similar risk assessment has been used as a found in a suburban area west of Miami (R. Reed, U.S. screening tool for plants (Pheloung 1995, Pheloung Geological Survey, personal communication) which et al. 1999, Daehler et al. 2004). Applying this supports reliability of our predictions. approach to imported reptiles could focus attention on relatively few species in a confined geographic area Adverse impacts associated with potentially where the climate matches that of their native range. invasive species Such efforts both to identify imported species that are likely to establish and to prevent release of the Several species identified by our models as potential identified species may reduce risks and costs associ- invaders are capable of causing harm to humans and ated with future invasions. domestic animals and adverse ecological impacts. Currently, imperfect documentation of species Fatal attacks on humans, livestock, and pets by large or entering Florida’s ports hampers our ability to track venomous snakes, such as P. sebae, M. amethistina, which species are actually coming into our ports. E. murinus, and B. arietans, were reported in their Detailed information on imports would help in devel- native range (Branch and Hacke 1980; Rivas 1998; oping a list of incoming species to be evaluated. Lack Luiselli et al. 2001; Spawls et al. 2002; Mallow et al. of basic life history information (e.g., clutch size, 2003). Notably, a large number of these species frequency of nesting) for many of the imported species (nearly or more than 1,000 animals of each species is also of concern and limits usefulness of the models. from 2000 to 2005) were imported through Miami and A better understanding of the life history characteris- St. Petersburg ports. These upper trophic-level species tics of imported species would not only improve the may also impact native food webs (Meshaka et al. reliability of predictive models but would also increase 2004; Bomford et al. 2005). Although not well confidence in the models among stakeholders. established in the literature, there are growing con- We note that this study does not provide a risk cerns about the economic damage caused by exotic assessment of the overall environment (i.e., identifying reptiles (Bomford et al. 2005). Damage caused by environments that are vulnerable or resistant to 123 Risk assessment of potential invasiveness of exotic reptiles invasions), which is one of the guiding principles for Butterfield BP, Meshaka WE Jr, Guyer C (1997) Nonindige- invasion biology suggested by the SCOPE invasion nous amphibians and reptiles. In: Simberloff D, Schmitz DC, Brown TC (eds) Strangers in paradise: impact and program. Compared to risk assessments of species, a management of non-indigenous species in Florida. Island relatively smaller number of such geographic risk Press, Washington, pp 123–137 assessments have been conducted in invasion biology. Cogger HG (2000) Reptiles and amphibians of Australia. Combining information from geographic risk assess- Ralph Curtis Books, Sanibel Island Cogger HC, Zweifel RG (1998) Encyclopedia of reptiles and ments with that obtained through species-based assess- amphibians. Academic Press, San Diego ments would allow us to identify high-risk areas for Conant R, Collins JT (1991) A field guide to reptiles and potential invaders. Such information would be valu- amphibians, eastern and central North America. Houghton able for prioritizing management strategies to control Mifflin Company, Boston Daehler CC, Denslow JS, Ansari S, Kuo H (2004) A risk- the spread of invasive species, especially on protected assessment system for screening out invasive pest plants lands. from Hawaii and other Pacific Islands. Conserv Biol 18:360–368 Daniel JC (2002) The book of Indian reptiles and amphibians. Acknowledgments This study was funded by the South Oxford University Press, New York Florida Water Management District and the U.S. Geological Davis MA (2006) Invasion biology 1958–2005: the pursuit of Survey Priority Ecosystem Science Program. We acknowledge science and conservation. In: Cadotte MW, McMahon Kenneth Dodd, Kenneth Krysko, Kevin Enge, Mac Kobza, and SM, Fukami T (eds) Conceptual ecology and invasion Michael Cherkiss for their professional advice, Robert Reed biology: reciprocal approaches to nature. Springer, The and Gordon Rodda for insightful reviews, Amanda Rice, Netherlands Hardin Waddle, Laura Brandt, Nicola Hughes, Pam Fuller, and De Lisle HF (1995) The natural history of monitor lizards. Wellington Guzman for their assistance with data collection, Krieger Publishing Company, Malabar and Rebecca Harvey and Jen Frost for editorial help. We also Elton CS (1958) The ecology of invasions by animals and appreciate the comments provided by Editor-in-Chief Dan plants. Methuen, London Simberloff and an anonymous reviewer, which helped to Ernst CH, Barbour RW (1989) Turtles of the world. Smithso- clarify and contextualize the manuscript. The use of trade nian Institution Press, Washington names does not imply endorsement by the U.S. government. Fritts TH, Leasman-Tanner D (2001) The brown tree snake on Guam: how the arrival of one invasive species damaged the ecology, commerce, electrical systems, and human health on Guam: a comprehensive information source. References Available online at: http://www.fort.usgs.gov/resources/ education/bts/bts_home.asp Arnold EN (2002) Reptiles and amphibians of Europe. Fritts TH, McCoid MJ (1991) Predation by the brown tree Princeton University Press, Princeton snake Boiga irregularis on poultry and other domesticated Baha El Din SBE (2006) A guide to the reptiles and amphib- animals in Guam. Snake 23:75–80 ians of Egypt. American University in Cairo Press, Cairo Gilpin M (1990) Ecological prediction. Science 248:88–89 Baker HG, Stebbins GL (1965) The evolution of colonizing Hayes KR, Barry SC (2008) Are there any consistent predictors species. Academic Press, New York of invasion success? Biol Invasions 10:483–506 Bomford M, Kraus F, Braysher M, Walter L, Brown L (2005) Henkel F, Schmidt W (2000) Amphibians and reptiles of Risk assessment model for the import and keeping of Madagascar and the Mascarene, Seychelles, and Comoro exotic reptiles and amphibians. Bureau of Rural Sciences Islands. Krieger Publishing Company, Malabar for the Department of Environment and Heritage Hoover C (1998) The U.S. role in the international live reptile Bomford M, Kraus F, Barry SC, Lawrence E (2008) Predicting trade: Amazon tree boas to Zululand dwarf chameleons. establishment success for alien reptiles and amphibians: a TRAFFIC North America, Washington role for climate matching. Biol Invasions. doi:10.1007/ Huriash LJ (2008) Iguana causes power outage for 20,000 in s10530-008-9285-3 Broward Coutny, Sun-Sentinel 8 May 2008, sun-sen- Bonin F, Devaux B, Dupre´ A (2006) Turtles of the world. John tinel.com/news/local/broward/sfl-0508iguana,0,5475336. Hopkins University Press, Baltimore story Boulenger GL (1893) Catalogue of the snakes in the British Hutchins M (2003) Grzimek’s animal life encyclopedia. museum, vol 1. Longmans & Co, London Thomson, Detroit Boulenger GL (1965) Catalogue of the lizards in the British Johnson DE (1998) Applied multivariate methods for data museum, vol 1, 2. British Museum, London analysis. Duxbury Press, Pacific Grove Branch B (1998) Field guide to snakes and other reptiles of Kolar CS, Lodge DM (2002) Ecological predictions and risk southern Africa. Ralph Curtis Books, Sanibel Island assessment for alien fishes in North America. Science Branch WR, Hacke WD (1980) A fatal attack on a young boy 298:1233–1236 by an African rock python (Python sebae). J Herp 14:305– Kraus F (2003) Invasion pathways for terrestrial vertebrates. 307 In: Carlton J, Ruiz G, Mack R (eds) Invasive species:

123 I. Fujisaki et al.

vectors and management strategies. Island Press, Wash- Reed RN, Rodda GH (2009) Giant constrictors: biological and ington, pp 68–92 management profiles and an establishment risk assessment Kraus F, Cravalho D (2001) The risk of Hawaii from snakes. for nice large species of pythons, anacondas, and the boa Pac Sci 55:409–417 constrictor: U.S. Geological Survey Open-File Report Lee JG (2000) A field guide to the amphibians and reptiles of 2009-1202, 302 pp the Maya world. Cornell University Press, New York Rejmanek M, Richardson DM (1996) What attributes make Lodge DM, Williams S, MacIsaac HG, Hayes KR, Leung B, some plant species more invasive? Ecology 77:1655–1661 Richard S, Mack RN, Moyle PB, Smith M, Andow DA, Rivas JA (1998) Predatory attack of a green anaconda (Eunectes Carlton JT, McMichael A (2006) Biological invasions: murinus) on an adult human. Herp Nat Hist 6:157–159 recommendations for U.S. policy and management. Ecol Rodda GH, Jarnevich CS, Reed RN (2009) What parts of the Appl 16:2035–2054 US mainland are climatically suitable for invasive alien Luiselli L, Angelici F, Akani G (2001) Food habits of Python sebae pythons spreading from Everglades National Park? Biol in suburban and natural . Afr J Ecol 39:116–118 Invasions 11:241–252 Mallow D, Ludwig D, Nilson G (2003) True vipers: natural Rodrı´guez-Schettino L (1999) The iguanid lizards of Cuba. history and toxinology of old world vipers. Krieger Pub- University Press of Florida, Gainesville lishing Company, Malabar Savage JM (2002) The amphibians and reptiles of . Mehrtens JM (1987) Living snakes of the world. Sterling University of Chicago Press, Chicago Publishing, New York Schwartz A, Henderson RW (1991) Amphibians and reptiles of Meshaka WE Jr, Butterfield NP, Hauge JB (2004) The exotic the West Indies. University of Florida Press, Gainesville amphibians and reptiles of Florida. Krieger Publishing Spawls S, Howell K, Drewes R, Ashe J (2002) A field guide to Company, Malabar the reptiles of east Africa. Academic Press, London Pheloung P (1995) Determining the weed potential of new Stafford PJ, Meyer JR (2000) A guide to the reptiles of . plant introductions to Australia. Report. Australia Weeds Natural History Museum, London Committee and the Plant Industries Committee Stebbins RC (1985) A field guide to western reptiles and Pheloung PC, Williams PA, Halloy SR (1999) A weed risk amphibians. Houghton Mifflin Company, Boston assessment model for use as a biosecurity tool evaluating Sutherst RW, Maywald GF, Kriticos D (2007) CLIMEX ver- plant introductions. J Environ Manag 57:239–251 sion 3.0 user’s guide. Hearne Scientific Software Pty Ltd, Pope CH (1935) Natural history of central Asia. Vol 10, the Melbourne reptiles of . American Museum of Natural History, Venables WN, Ripley BD (2002) Modern applied statistics New York with S, 4th edn. Springer, New York Pritchard PCH, Trebbau P (1984) The turtles of Venezuela. Williamson M, Fitter A (1996) The varying success of invad- Society for the study of amphibians and reptiles, Grand ers. Ecology 77:1661–1666 Rapids

123