Potential Distribution of the Alien Invasive Brown Tree Snake, Boiga Irregularis (Reptilia: Colubridae)1

Potential Distribution of the Alien Invasive Brown Tree Snake, Boiga irregularis (Reptilia: Colubridae)1

Dennis Ro¨dder2,3,4 and Stefan Lo¨tters3

Abstract: The Brown Tree Snake (Boiga irregularis) is native to Southeast Asia and Australia and has been introduced to Guam. There it causes major ecological and socioeconomic problems and is considered to belong to the 100 worst alien invasive species worldwide. We used a maximum entropy-based Climate Envelope Model to identify worldwide areas outside the species’ known range that are potentially suitable under current climatic conditions. Projections revealed that this invasive alien species potentially occurs in tropical and some subtropical regions. In the closer vicinity of the snake’s known distribution, highest suitability was found for the Northern Mariana Islands, Hawaiian Islands, Madagascar, New Caledonia, and Fiji Islands. If predictions are inter- preted as depicting invasiveness potential of B. irregularis, strategies to prevent invasion should focus on these regions. An analysis of potential distributions under different future anthropogenic climate-change scenarios showed that the Fiji Islands, Hawaiian Islands, and Northern Mariana Islands will remain overall most suitable habitat for the Brown Tree Snake. In addition, we noted an increase of suitability in New Zealand.

Alien invasive species are a concern in Its status in Sulawesi (Indonesia) is uncertain. nature conservation because they may have Rodda et al. (1999) listed populations from negative impact on native biodiversity and Sulawesi as ‘‘native,’’ but Iskandar and Tjan can have major socioeconomic impacts. A re- (1996) suggested that B. irregularis may have markable example is the Brown Tree Snake, reached the island only via trade. Ecologi- Boiga irregularis (Merrem, 1802) (Savidge cally, this snake can be regarded a ‘‘general- 1987, Rodda and Fritts 1992, Fritts and ist’’ concerning habitat requirements and Rodda 1998, Wiles et al. 2003). This venom- prey selectivity. Boiga irregularis is known ous, 1- to 3-m-long arboreal colubrid snake is from natural forest and grassland as well as native to Papua New Guinea, the Solomon agricultural and urban areas and does not Islands, and the northern and eastern coasts show particular food preferences (Rodda et of Australia (Figure 1) (Rodda et al. 1999). al. 1999). The Brown Tree Snake was brought accidentally to several islands in the Pacific 1 This work benefited from a grant by the Graduier- shortly after World War II where it success- tenfo¨rderung des Landes Nordrhein-Westfalen to D.R. Manuscript accepted 1 February 2009. fully established local populations due to ob- 2 Herpetology Department, Zoologisches For- viously unoccupied niches and plentiful naive schungsmuseum Alexander Koenig, Adenauerallee 160, prey (Rodda et al. 1992). As a result, B. irreg- 53113 Bonn, Germany. ularis today is listed among the 100 worst 3 Biogeography Department, Trier University, 54286 alien invasive species worldwide (Lowe et al. Trier, Germany. 4 Corresponding author: (e-mail: roedder@uni-trier 2000). The most devastating consequences .de). of its introduction are known from Guam, where it led to decimation of vertebrate species including flying foxes, several small terrestrial mammals, and lizards as well as the Pacific Science (2010), vol. 64, no. 1:11–22 doi: 10.2984/64.1.011 extinction of eight of the 11 endemic bird : 2010 by University of Hawai‘i Press species in the 1980s (Savidge 1987, Wiles All rights reserved et al. 2003). It is suggested that B. irregularis

11 12 PACIFIC SCIENCE . January 2010

Figure 1. Native distribution of Boiga irregularis. (Adapted from Rodda et al. [1999].)

has substantially altered food web structures, Micronesia (Pohnpei), the Ryukyu Islands enhancing its effect on native biota and ex- (Okinawa), Spain (Rota), Taiwan, the United plaining its extraordinarily devastating effect States (Texas, Corpus Christi), and Wake Is- (D’Antonio and Dudley 1995, Fritts and land (Mito and Uesugi 2004, IUCN Invasive Rodda 1998). Species Specialist Group 2007). Although it One of the major means of increase of B. apparently has so far not become residential irregularis distribution is as a stowaway in mil- at any of these places, this list of geographic itary equipment and cargo (Fritts 1987, Kraus sites and their distances from the native oc- 2007). Currently, anthropogenic dispersal is currence demonstrate the snake’s ‘‘efﬁciency’’ most associated with large-scale exercises in dispersal. A result is that this clearly de- and personal cargo of people moving from mands special concern with regard to conser- Guam to other duty posts. Also sea freight vation (Buden et al. 2001, Gill et al. 2001, (container/bulk) may play a role in dispersal Kraus and Carvalho 2001, Rodda et al. (Kraus 2007), and several tree snakes trans- 2002). In the United States, incursions by B. located in aircraft cargoes have been detected irregularis have been repeatedly intercepted at Honolulu International Airport in recent (McCoid et al. 1994), but eradication of es- years (e.g., Atkinson and Atkinson 2000). tablished populations is expensive and difﬁ- During the last decades, the Brown Tree cult (Rodda et al. 2002). Snake arrived in the Cocos Islands, Diego It appears to be a question of time until Garcia, the Hawaiian Islands (Honolulu, the Brown Tree Snake will establish addi- O‘ahu), Indonesia ( Java), Japan, New Zea- tional invasive populations. Fritts and Rodda land, the Northern Mariana Islands (Saipan, (1998) studied the risk of invasion by the spe- Tinian), Malaysia (Singapore), the Marshall cies using biological and natural history infor- Islands (Kwajalein), the Federated States of mation. They concluded that the Mariana, Potential Distribution of Boiga irregularis . Ro¨dder and Lo¨tters 13

Hawaiian, and Caroline islands are most at the altitude of the locality in a digital eleva- risk because a native snake fauna is absent and tion model, using the Extract Values by potential prey density is high. Invasion risk Points function of DIVA-GIS. Of the total will depend on the arrival of founder individ- records available, 337 were situated in unique uals over time and also on environmental grid cells within the species’ native range, conditions making more areas available for classified as being accurate and hence suitable the ectothermic species. for model building (see later section). We see an urgent need to identify those Information on current climate was ob- regions in which this species finds suitable tained from the WorldClim database, version conditions for successful establishment of in- 1.4 (www.worldclim.org), which is based on vasive populations. Climate Envelope Models weather conditions recorded between 1950 are a useful tool to study potential distribu- and 2000 with a grid cell resolution of 2.5 tions of species under past, current, and fu- min (Hijmans et al. 2005). It was created by ture climatic scenarios (e.g., Hijmans and interpolation using a thin-plate smoothing Graham 2006, Malcom et al. 2006, Peterson spline of observed climate at weather sta- and Nya´ri 2007, Carnaval and Moritz 2008, tions, with latitude, longitude, and elevation Ro¨dder et al. 2008, Ro¨dder 2009). The scope as independent variables (Hutchinson 1995, of the study reported here is to assess the 2004). worldwide potential distribution of B. irregularis under current climatic conditions to Variable Selection identify regions with high potential for invasions. Rodda et al. (2007) investigated the climate envelope of the Brown Tree Snake in its na- materials and methods tive geographic range regarding annual mean monthly temperature and precipitation. They Species Records and Climate Data found that the amount of precipitation was a A total of 2,679 Boiga irregularis records good predictor for its distribution. Such a re- was available through the Global Biodiversity lationship appears to be further supported by Information Facility (GBIF, www.gbif.org) natural history observations because the sea- and HerpNet databases (www.herpnet.org). sonal activity of the Brown Tree Snake is In addition, records of invasive populations limited to the warmer and wetter summer were obtained from the IUCN Invasive Spe- months in its Australian range (Covacbvich cies Specialist Group (www.issg.org), where- and Limpus 1973, Shine 1991a). It was also by only records within areas with confirmed observed that the snake is unable to shed reproduction were included. For georeferenc- properly when the relative humidity is lower ing, the Alexandria Digital Library Gazetteer than 60% (Rodda et al. 1999). Within the na- Server Client (www.middleware.alexandria tive distribution of the target species, this de- .ucsb.edu/client/gaz/adl/index.jsp) was used. pendency is reflected in the geographic range We used DIVA-GIS 5.4 (Hijmans et al. in Australia, where B. irregularis inhabits 2001) to test the accuracy of coordinates mainly humid areas close to the sea. There- (Check Coordinates tool) by comparing in- fore, next to the mean annual values, mini- formation accompanying the species records mum and maximum annual values may be and locality data extracted from an adminis- important, suggesting that ‘‘annual precipita- trative boundaries database at the smallest tion,’’ ‘‘precipitation of the wettest month,’’ possible level (country/state/city). This in- and ‘‘precipitation of the driest month’’ are formation should be the same, and any mis- suitable predictor variables. matches may reflect errors (see Hijmans et al. Temperature is a key factor influencing 1999). In addition, we used altitudinal infor- ectothermic species; therefore the ‘‘annual mation to spot likely errors in the coordinate mean temperature’’ is an important variable data when this information was provided with related to energetic balances and digestive the records used. Altitude was compared with turnover rates. Mathies and Miller (2002) 14 PACIFIC SCIENCE . January 2010 showed that temperature seasonality is physi- known species (Pearson et al. 2007) and the ologically important in influencing reproduc- outcome of introductions of alien invasive tion. Those authors investigated the effect of species outside the native distribution (e.g., two different temperature regimes for elicit- Peterson and Vieglais 2001, Ficetola et al. ing reproduction in male and female B. irreg- 2007, Jeschke and Strayer 2008, Ro¨dder et ularis. They found that specimens maintained al. 2008, Ro¨dder 2009). at 24C followed by a 60-day cool period at Runs used herein were conducted using 19C exhibited substantial reproductive ac- the default values for all program settings, tivity. Clutches were produced by females whereby randomly chosen background points shortly after returning to 24C. In contrast, were restricted to an area defined by a mini- individuals maintained at 28C followed by mum convex polygon including all native rec- an identical 19C cooling period exhibited ords. Herein, the logistic output format with relatively little reproductive activity, and suitability values ranging from 0 (unsuitable) none of the females produced eggs. Con- to 1 (optimal) was used, and areas where sidering these effects, we added the ‘‘maxi- ‘‘clamping’’ (i.e., nonanalogous climatic con- mum temperature of the warmest month’’ dition present in the training area) occurred and ‘‘minimum temperature of the coldest during projections were subsequently ex- month’’ as predictor variables for model com- cluded (Philips and Dudı´k 2008). putation. Maxent allows for model testing by calcu- lation of the Area Under the Curve (AUC), referring to the ROC (Receiver Operation Climate Envelope Models Characteristic) curve; herein we assessed the Maxent 3.2.1 (www.cs.princeton.edu/ ability of the model to distinguish back- ~shapire/maxent) (Phillips et al. 2006) was ground points from training points (Hanley applied for Climate Envelope Model cal- and McNeil 1982, Phillips et al. 2006). This culation to assess the potential distribution method is recommended for ecological appli- of the Brown Tree Snake and to map it into cations because it is nonparametric (Pearce geographic space. Maxent is a grid-based and Ferrier 2000, but see Lobo et al. 2008). machine-learning algorithm following the Values of AUC range from 0.5 (random) for principles of maximum entropy ( Jaynes models with no predictive ability to 1.0 for 1957). The general concept is to find a prob- models giving perfect predictions. According ability distribution covering the study area to the classification of Swets (1988), AUC that satisfies a set of constraints derived from values >0.9 describe ‘‘very good,’’ >0.8 occurrence data. Each constraint requires that ‘‘good,’’ and >0.7 ‘‘useful’’ discrimination the expected value of an environmental vari- abilities. Maxent allows for an assessment of able or a function thereof must be within a the relative contribution of variables included confidence interval of its empirical mean using a jackknifing approach. over the presence records. The program chooses the distribution that is closest to uni- form and therefore maximizes entropy within results all distributions that satisfy the constraints Current Potential Distribution because any other choice would represent constraints on the distribution that are not We obtained ‘‘very good’’ AUC values in the justified by the data. Maxent is able to incor- model (AUC ¼ 0.971), and the known inva- porate complex dependencies between pre- sive range of Boiga irregularis on Guam is sit- dictor variables and has been shown to reveal uated within higher Maxent classes (>0.780), better results than other comparable methods also confirming the predictive power of the (e.g., Elith et al. 2006, Wisz et al. 2008). The model. The lowest observed Maxent value at reliability of the results obtained from Max- the presence point used for model training ent models has been confirmed by its capacity was 0.201. Analyses of variable contributions to predict novel presence localities for poorly in the model revealed that ‘‘annual precipita- Potential Distribution of Boiga irregularis . Ro¨dder and Lo¨tters 15

Figure 2. Potential distribution of Boiga irregularis under current climatic conditions. Any area with a Maxent value above 0.201 may be climatically suitable for B. irregularis, although higher Maxent values suggest higher climatic suitability. tion’’ with 33.3% has the highest explanative (e.g., Congo and Amazon basins) and mon- power, followed by ‘‘maximum temperature tane (e.g., southern coastal Brazil, Ethiopian of the warmest month’’ (28.5%), ‘‘annual highlands) regions. In addition, high suit- mean temperature’’ (13.8%), ‘‘precipitation ability was found in the Hawaiian Islands of the driest month’’ (11.7%), ‘‘precipitation (lower elevations) (Figure 4) and Madagas- of the wettest month’’ (7.7%), and the ‘‘min- car (mainly east coast and higher elevations) imum temperature of the coldest month’’ (Figure 5). (4.9%). The relative variable importance obtained via jackknifing from the Maxent model discussion appears to be consistent with natural history observations (see earlier). ‘‘Clamping’’ oc- When interpreting results and assessing an curred at very few sites, such as at a very small invasion risk, it is important to evaluate possi- ridge on the west coast of India. Those areas ble discrepancies between the realized and were excluded from further analyses. fundamental climatic niche of B. irregularis Under current climatic conditions, the and the relative contribution of ecological Brown Tree Snake has a geographically wide factors other than climate, which may limit potential distribution throughout almost all the species’ current distribution. In addition the tropics and adjacent subtropical regions, to a species’ climate envelope, these are acces- especially widespread within the Southern sibility limitations and/or biotic interactions Hemisphere but also extending to North such as competition or predation (see also So- America including Florida and coastal areas beroń and Peterson 2005). of the Gulf of Mexico (Figure 2). Regarding the general region in which the Brown Tree Biotic Interactions and/or Accessibility Snake occurs (i.e., Southeast Asia and Austra- lia), major parts of New Caledonia, New Although the modeled potential distribution Zealand, the Fiji Islands, and Vanuatu were in the eastern portion of the native range identified as highly suitable for B. irregularis of Boiga irregularis is coincident with the by the Climate Envelope Model (Figure 3). reported distribution (i.e., current species rec- Guam, the Caroline Islands, and the North- ords), that of the western portion is not (Fig- ern Mariana Islands including Saipan are all ure 3). Climatically, the Brown Tree Snake suitable for B. irregularis (Figure 3). In addi- can find suitable areas on Borneo, Java, and tion to regions in the vicinity of the species’ Sumatra west of Wallace’s line, but none of native geographic range in Southeast Asia, its these islands has been successfully invaded, climate envelope is mirrored in both lowland as far as is known (Rodda et al. 1999, IUCN Figure 3. Potential distribution of Boiga irregularis under current climatic conditions within Southeast Asia and Aus- tralia. Higher Maxent values suggest higher climatic suitability. Native-distribution records are indicated as open circles and invasive ones as open triangles.

Figure 4. Potential distribution of Boiga irregularis under current climatic conditions in the Hawaiian Islands. Higher Maxent values suggest higher climatic suitability. Potential Distribution of Boiga irregularis . Ro¨dder and Lo¨tters 17

century and is generally high today (Espada and Kumazawa 2005). Therefore, the chance for accidental introduction of specimens is expected to be high considering that cargo is one of the major means of spreading this species (Kraus 2007). In the case of the Brown Tree Snake, presence of competitors perhaps best explains the observed pattern. Shine (1991a,b) suggested that colubrid snakes such as B. irregularis are not in noteworthy competition with the endemic elapid snakes and pythons in Australia because they have invaded the con- tinent with ecological specializations that are rare among the endemics. Differences in foraging habitats and in the preferred prey spectrum in comparison with the endemic Australian snakes especially may have enhanced their success (Shine 1991b). This might not be true for the snake fauna west of Wallace’s line. The genus Boiga currently comprises 34 species; 11 of them inhabit allo- patric ranges adjacent to the native range of B. irregularis (B. angulata, B. bengkuluensis, B. Figure 5. Potential distribution of Boiga irregularis cynodon, B. dendrophila, B. drapiezii, B. jaspidea, under current climatic conditions in Madagascar. Higher B. multimaculata, B. nigriceps, B. philippina, Maxent values suggest higher climatic suitability. B. schultzei, B. tanahjampeana) (Uetz et al. 2007), but none of them is known to be invasive (IUCN Invasive Species Specialist Group Invasive Species Specialist Group 2007). Pos- 2007). It is remarkable that although other sible explanations could be a nonequilibrium snakes occur in sympatry with the Brown of the actual range with climate due to Tree Snake, other Boiga species commonly limited accessibility or ongoing spreading, do not. Only at Sulawesi B. dendrophila and limited prey availability, or the presence of B. multimaculata occur together with the predators or competitors (e.g., Rodda et al. Brown Tree Snake, but there B. irregularis 1999, Arau´jo and Pearson 2005, Ro¨dder et al. was only reported from the immediate neigh- 2008). borhood of seaports, where it might have Accessibility highly inﬂuences the number been introduced (Iskandar and Tjan 1996). It of jurisdictions where a species can be intro- is unclear if Sulawesi inland populations actu- duced, which is an important predictor of the ally do exist (Inger and Voris 2001). probability that a species can establish inva- It is interesting that the Brown Tree Snake sive populations (Bomford et al. 2008). Lim- is the only member of the genus known to ited accessibility appears to be an unlikely have established invasive populations. Differ- explanation for the absence of B. irregularis ences in behavioral traits, which might pro- west of Wallace’s line. Although faunal ex- vide superior abilities to establish invasive change between the westernmost populations populations, are not known. In our opinion, in Sulawesi and New Guinea and those of ad- the most likely explanation is that the native jacent Indonesian islands in past geological range of the Brown Tree Snake is much times was rather restricted (e.g., Inger and closer to areas lacking a native terrestrial Voris 2001), cargo trafﬁc within Indonesia snake fauna, enhancing the chance of translo- and adjacent islands increased during the last cation. 18 PACIFIC SCIENCE . January 2010

Climate logical and natural history information. They concluded that the Northern Mariana, Ha- Mathies and Miller (2002) showed that a pe- waiian, and Caroline islands are most at risk riod of cool temperatures elicits reproductive because a native (competing) snake fauna is activity in both sexes. Those authors pointed absent and prey is available. Because the Mar- out that temperatures experienced during the iana and Caroline islands are comparable in artificial hibernation were much lower than terms of faunal composition with Guam, an the snakes would experience on Guam, where invasion in those islands may have similarly temperature seasonality is relatively invari- dramatic ecological consequences. The faunal ant throughout the year (G1C [Rodda et al. composition of the Hawaiian Islands might 1999]). Comparing native populations in Aus- also generally support the establishment of tralia and invasive populations on Guam, Boiga irregularis (Kraus and Carvalho 2001). Moore et al. (2005) found that snake speci- However, predictions based on natural his- mens from Guam exhibited substantially tory and biological information appear to reduced body conditions compared with indi- bear a higher degree of uncertainty there be- viduals from Australia. Those authors sug- cause the faunal composition of the Hawaiian gested that Brown Tree Snakes on Guam Islands is distinctly different from that found were living under stressful conditions, possi- on Guam (Fritts and Rodda 1998). bly due to overcrowding and overexploitation Climate Envelope Model predictions ob- of food resources, resulting in decreased adult tained in this paper confirm the expected cli- size/weight and suppressed reproduction. matic suitability of the Northern Mariana, The findings of Mathies and Miller (2002) Hawaiian, and Caroline islands. In addition, might provide an additional explanation for our models highlight major parts of central their finding because environmental stress Africa, Central America, South America, a may cause absence of hibernation and dis- broad swath of the southeastern United turbed seasonal reproductive cycles. How- States coastal plain (coincident with the pre- ever, although reproductive activity on diction presented by Rodda et al. [2007]), Guam may be suppressed and desynchro- New Caledonia, New Zealand, and Madagas- nized, only minor enhanced fluctuations in car as having high climatic suitability for B. temperature throughout the year may be suf- irregularis. Records of Brown Tree Snakes ficient to elicit reproduction; therefore fluctu- that were accidentally imported to Hawai‘i ations and/or extreme weather events due to and New Zealand already exist (Gill et al. anthropogenic climate change might enhance 2001, IUCN Invasive Species Specialist the snake problem. Furthermore, the obser- Group 2007), and it may be feared that the vation that moderate hibernation enhances Brown Tree Snake will establish there. Im- the reproductive cycle allows the hypothesis pacts may be enhanced due to high availabil- that may also find climati- Boiga irregularis ity of prey affecting population dynamics of cally suitable habitats in more temperate re- the invader. Establishment of B. irregularis in gions with a higher degree of seasonality, New Zealand, in addition, may be facilitated such as the northern parts of New Zealand, due to the absence of competitors, because Japan, or Taiwan. Generally, climatic similar- only sea snakes occur there (Hydrophiidae). ities of a novel environment compared with Similarly, for New Caledonia and the Fiji Is- that of the native range enhance the probabil- lands only 19 and seven snake species are ity of successful establishment of alien inva- known, respectively (Uetz et al. 2007). Six- sive species (Bomford et al. 2008). teen of the New Caledonian species also are hydrophiids, two are fossorial Thyphlopida, but (Boidae) could be a poten- Which Regions Are at High Risk of Invasion Candoia bibroni tial competitor to the Brown Tree Snake. At by the Brown Tree Snake? Fiji, there are three species each of hydro- Fritts and Rodda (1998) studied the risk of phiids and thyphlopids plus Candoia bibroni. invasion by the Brown Tree Snake using bio- It must be noted that possible discrepan- Potential Distribution of Boiga irregularis . Ro¨dder and Lo¨tters 19 cies between the realized climatic niche (cli- tions via aircraft cargoes (Atkinson and Atkin- matic conditions within realized distribution) son 2000). Preventing further spreading to and the fundamental climatic niche of B. ir- the areas highlighted may depend on main- regularis may lead to an underestimation of taining strict cargo checks of boats and threat in some regions. Unfortunately, the re- searches at airports. lationship between realized and fundamental niches can be addressed only experimentally but not with climate envelope model ap- acknowledgments proaches, leaving some degree of uncertainty. We are grateful to Marcelo R. Duarte, Tom However, it is interesting that the potential C. Mathies, and Mirco Sole´, who helped distribution of B. irregularis derived from our with literature used herein. Three anony- model is remarkably coincident with the real- mous reviewers helped improve this paper. ized distributions of all other members of the genus Boiga, implying that climatic niches are rather conservative within the genus (at Literature Cited least regarding the variables chosen herein) and that we captured a great part of its funda- Arau´jo, M. B., and R. G. Pearson. 2005. mental niche with our model. However, areas Equilibrium of species’ distribution with outside the proposed potential distribution climate. Ecography 28:693–695. may exhibit climatic conditions different Atkinson, I. A. E., and T. J. Atkinson. 2000. from those present within the realized distri- Land vertebrates as invasive species on the bution of B. irregularis, but they may not nec- islands of the South Pacific Regional Envi- essarily be unsuitable. ronment Programme. Pages 19–84 in G. Sherely, ed. Invasive species in the Pacific: conclusions A technical review and draft regional strat- egy. South Pacific Regional Environment We conclude that Boiga irregularis is a species Program, Samoa. causing a high extinction risk to endemic Bomford, M., F. Kraus, S. C. Barry, and E. faunas in regions where specimens are fre- Lawrence. 2008. Predicting establishment quently translocated via military shipment success for alien reptiles and amphibians: and cargo (Fritts 1987, Kraus 2007, Bomford The role of climate matching. Biol. Inva- et al. 2008) and that are suitable under cur- sions 11:713–724. rent climatic conditions and lack competitors, Buden, D. W., D. B. Lynch, and G. R. Zug. namely the Northern Mariana and Hawaiian 2001. Recent records of exotic reptiles on islands, New Caledonia, and the Fiji Islands. Pohnpei, eastern Caroline Islands, Micro- Climatically suitable regions harboring mini- nesia. Pac. Sci. 55:65–70. mal or lacking native snake species are likely Carnaval, A. C., and C. Moritz. 2008. Histor- to be most impacted. There, availability of ical climate modelling predicts patterns of resources may be enhanced compared with current biodiversity in the Brazilian Atlan- an ecosystem already harboring competing tic forest. J. Biogeogr. 25:1187–1201. snakes. Covacbvich, J., and C. Limpus. 1973. Two Under future anthropogenic climate large winter aggregations of three species change, the situation will not change essen- of tree-climbing snakes in south-eastern tially, but climatic suitability for B. irregularis Queensland. Herpetofauna 6:16–21. will generally remain high in Madagascar and D’Antonio, C. M., and T. L. Dudley. 1995. New Caledonia and even increase in New Biological invasions as agents of change Zealand (D.R., unpubl. data). Possible routes on islands versus mainlands. Pages 103– for spreading of the species to Madagascar, 121 in P. M. Vitousek, L. L. Loope, and New Caledonia, and the Fiji Islands include H. Andersen, eds. Islands: Biological di- sea cargo, military equipment (Fritts 1987, versity and ecosystem function. Ecological Kraus 2007), and also accidental transloca- Studies. Springer, Berlin. 20 PACIFIC SCIENCE . January 2010

Elith, J., C. H. Graham, R. P. Anderson, M. species distributions. Global Change Biol. Dudik, S. Ferrier, A. Guisan, R. J. Hij- 12:2272–2281. mans, F. Huettmann, J. R. Leathwick, A. Hijmans, R. J., M. Schreuder, J. De la Cruz, Lehmann, J. Li, L. G. Lohmann, B. A. and L. Guarino. 1999. Using GIS to Loiselle, G. Manion, C. Moritz, M. Naka- check coordinates of genebank accessions. mura, Y. Nakazawa, J. M. M. Overton, Genet. Resour. Crop Evol. 46:291–296. A. T. Perterson, S. J. Phillips, K. Richard- Hutchinson, M. F. 1995. Interpolating mean son, R. Scachetti-Pereira, R. E. Shapire, rainfall using thin plate smoothing splines. J. Soberoń, S. Williams, M. S. Wisz, and Int. J. Geogr. Inf. Sys. 9:385–403. N. E. Zimmermann. 2006. Novel methods ———. 2004. Anusplin, version 4.3. Centre improve prediction of species’ distribu- for Resource and Environment Studies, tions from occurrence data. Ecography The Australian National University, Can- 29:129–151. berra. Espada, I. C., and K. Kumazawa. 2005. O-D Inger, R. F., and H. K. Voris. 2001. The bio- structure of domestic maritime traffic in geographical relations of the frogs and Indonesia. Proc. East. Asia Soc. Transport. snakes of Sundaland. J. Biogeogr. 28:863– Stud. 5:647–660. 891. Ficetola, G. F., W. Thuiller, and C. Miaud. Iskandar, D. T., and K.-N. Tjan. 1996. The 2007. Prediction and validation of the po- amphibians and reptiles of Sulawesi, with tential global distribution of a problematic notes on the distribution and chromosome alien species: The American Bullfrog. number of frogs. Pages 39–46 in D. J. Divers. Distrib. 13:476–485. Kitchener and A. Suyanto, eds. Proceed- Fritts, T. H. 1987. Movements of snakes ings of the 1st International Conference via cargo in the Pacific region. Elepaio on Eastern Indonesian-Australian Verte- 47:17–18. brate Fauna, Mandao. Western Australian Fritts, T. H., and G. H. Rodda. 1998. The Museum for lembaga Ilmu Penetahuan In- role of introduced species in the degrada- donesia, Perth, Australia. tion of island ecosystems. Annu. Rev. IUCN Invasive Species Specialist Group. Ecol. Syst. 29:113–140. 2007. www.issg.org. Gill, B. J., D. Bejakovich, and A. H. Whi- Jaynes, E. T. 1957. Information theory and taker. 2001. Records of foreign reptiles statistical mechanics. Phys. Rev. 106:620– and amphibians accidentally imported to 630. New Zealand. N. Z. J. Zool. 28:351–359. Jeschke, J. M., and D. L. Strayer. 2008. Use- Hanley, J., and B. McNeil. 1982. The mean- fulness of bioclimatic models for studying ing of the use of the area under a receiver climate change and invasive species. Ann. operating characteristic (ROC) curve. Ra- N. Y. Acad. Sci. 1134:1–24. diology 143:29–36. Kraus, F. 2007. Using pathway analysis to in- Hijmans, R. J., S. E. Cameron, J. L. Parra, form prevention strategies for alien rep- P. G. Jones, and A. Jarvis. 2005. Very tiles and amphibians. Pages 94–103 in high resolution interpolated climate sur- G. W. Witmer, W. C. Pitt, and K. A. faces for global land areas. Int. J. Climatol. Fagerstone, eds. Managing vertebrate in- 25:1965–1978. vasive species: Proceedings of an interna- Hijmans, R. J., J. M. Cruz, R. Rojas, and L. tional symposium. USDA/APHIS/WS, Guarino. 2001. DIVA-GIS, version 1.4. National Wildlife Research Center, Fort A geographic information system for the Collins, Colorado. management and analysis of genetic re- Kraus, F., and D. Carvalho. 2001. The risk to sources data. Manual. International Potato Hawai‘i from snakes. Pac. Sci. 55:409–417. Center and International Plant Genetic Lobo, J. M., A. Jimeńez-Valverde, and R. Resources Institute, Lima, Peru. Real. 2008. AUC: A misleading measure Hijmans, R. J., and C. H. Graham. 2006. of the performance of predictive distri- The ability of climate envelope models to bution models. Global Ecol. Biogeogr. predict the effect of climate change on 17:145–151. Potential Distribution of Boiga irregularis . Ro¨dder and Lo¨tters 21

Lowe, S., M. Browne, S. Boudjelas, and M. ling of species geographic distributions. De Poorter. 2000. 100 of the world’s worst Ecol. Modell. 190:231–259. invasive alien species: A selection from the Phillips, S. J., and M. Dudı´k. 2008. Modeling Global Invasive Species Database. The In- of species distributions with Maxent: New vasive Species Specialist Group (ISSG), extensions and comprehensive evaluation. Auckland. Ecography 31:161–175. Malcom, J. R., D. Liu, R. P. Neilson, L. Han- Rodda, G. H., and T. H. Fritts. 1992. The sen, and L. Hannah. 2006. Global warm- impact of the introduction of the colubrid ing and extinction of endemic species snake Boiga irregularis on Guam’s lizards. from biodiversity hotspots. Conserv. Biol. J. Herpetol. 26:166–174. 20:538–548. Rodda, G. H., T. H. Fritts, E. W. Campbell, Mathies, T., and L. A. Miller. 2002. Cool K. Dean-Bradley, G. Perry, and C. P. temperatures elicit reproduction in a bio- Qualls. 2002. Practical concerns in the logically invasive predator, the Brown eradication of island snakes. Pages 260– treesnake. Zoo Biol. 22:227–238. 265 in C. R. Veitch and M. N. Clout, eds. McCoid, M. J., T. H. Fritts, and E. W. Turning the tide: The eradication of in- Campbell. 1994. A Brown tree snake vasive species. Proceedings of the Inter- (Colubridae: Boiga irregularis) sighting in national Conference on Eradication of Texas. Tex. J. Sci. 46:365–368. Island Invasives. Occasional Paper of the Mito, T., and T. Uesugi. 2004. Invasive alien IUCN Species Survival Commission, species in Japan: The status quo and new Gland, Switzerland, and Cambridge, regulation for prevention of their adverse United Kingdom. effects. Global Environ. Res. 8:171–191. Rodda, G. H., T. H. Fritts, and P. J. Conry. Moore, I. T., M. J. Greene, A. T. Lerner, 1992. Origin and population growth of C. E. Asher, R. W. Krohmer, D. L. Hess, the Brown tree snake, Boiga irregularis, on J. Whittier, and R. T. Mason. 2005. Phys- Guam. Pac. Sci. 46:46–57. iological evidence for reproductive sup- Rodda, G. H., T. H. Fritts, M. J. McCoid, pression in the introduced population of and E. W. Campbell. 1999. An overview brown tree snakes (Boiga irregularis)on of the biology of the Brown tree snake, Guam. Biol. Conserv. 121:91–98. Boiga irregularis, a costly introduced pest Pearce, J., and S. Ferrier. 2000. An evaluation on Paciﬁc islands. Pages 44–80 in G. H. of alternative algorithms for ﬁtting species Rodda, Y. Sawai, D. Chiszar, and H. Ta- distribution models using logistic regres- naka, eds. Problem snake management: sion. Ecol. Modell. 128:128–147. The Habu and the Brown Tree Snake. Pearson, R. G., C. J. Raxworthy, M. Na- Cornell University Press, Ithaca, New kamura, and A. T. Peterson. 2007. Pre- York. dicting species distributions from small Rodda, G. H., R. N. Reed, and C. S. Jarne- numbers of occurrence records: A test vich. 2007. Climate matching as a tool for case using cryptic geckos in Madagascar. predicting potential North American J. Biogeogr. 34:102–117. spread of Brown treesnakes. Pages 138– Peterson, A. T., and A` . S. Nya´ri. 2007. Eco- 145 in G. W. Witmer, W. C. Pitt, and logical niche conservatism and Pleistocene K. A. Fangerstone, eds. Managing verte- refugia in the Thrush-like Mourner, Shif- brate invasive species: Proceedings of an forinis sp., in the Neotropics. Evolution international symposium. USDA/APHIS/ 62:173–183. WS, National Wildlife Research Center, Peterson, A. T., and D. A. Vieglais. 2001. Fort Collins, Colorado. Predicting species invasions using ecologi- Ro¨dder, D. 2009. ‘Sleepless in Hawaii’: Does cal niche modeling: New approaches from anthropogenic climate change enhance bioinformatics attack a pressing problem. ecological and socioeconomic impacts of BioScience 51:363–371. the alien invasive Eleutherodactylus coqui Phillips, S. J., R. P. Anderson, and R. E. Thomas, 1966 (Anura: Eleutherodactyli- Schapire. 2006. Maximum entropy model- dae)? North-West. J. Zool. 5:16–25. 22 PACIFIC SCIENCE . January 2010

Ro¨dder, D., M. Sole´, and W. Bo¨hme. 2008. Swets, K. 1988. Measuring the accuracy of Predicting the potential distribution of diagnostic systems. Science (Washington, two alien invasive housegeckos (Gekkoni- D.C.) 240:1285–1293. dae: Hemidactylus frenatus, Hemidactylus Uetz, P., J. Goll, and J. Hallermann. 2007. mabouia). North-West. J. Zool. 4:236–246. Die TIGR-Reptiliendatenbank. Elaphe Savidge, J. A. 1987. Extinction of an island 15:22–25. forest avifauna by an introduced snake. Wiles, G. J., J. Bart, R. E. Beck, and C. F. Ecology 68:660–668. Aguon. 2003. Impacts of the Brown tree Shine, R. 1991a. Australian snakes: A natural snake: Patterns of decline and species per- history. Cornell University Press, Ithaca, sistence in Guam’s avifauna. Conserv. Biol. New York. 17:1350–1360. ———. 1991b. Strangers in a strange land: Wisz, M. S., R. J. Hijmans, A. T. Peterson, Ecology of the Australian colubrid snakes. C. H. Graham, A. Guisan, and NPSDW Copeia 1991:120–131. Group. 2008. Effects of sample size on Sobero´n, J., and A. T. Peterson. 2005. Inter- the performance of species distribution pretation of models of fundamental eco- models. Divers. Distrib. 14:763–773. logical niches and species’ distributional areas. Biodiv. Inf. 2:1–10.