(a) Increasing invasibility of extreme regions
Our results suggest an expansion of suitable habitat area for invasive cane toads in Australia. We predict that cane toads now have the potential to inhabit over 2
of the continent. This estimate includes three-quarters of Australia's coast, a region where most of the continent's human population and biological diversity are concentrated. If the cane toad's advance continues, this prolific and problematic species is likely to cause further harm to Australia's unique wildlife and economy.
Our model differed from prior work in a variety of ways, including its higher spatial resolution, statistical derivation and the inclusion of anthropogenic variables. However, the predicted distribution of cane toads and the total area of suitable habitat were relatively insensitive to the removal of anthropogenic factors (see electronic supplementary material). The most important variables in our model, including maximum and minimum temperatures, precipitation, moisture index and evaporation, were similar to those used to build a predictive model of Australian toad range based on the species' native range (Sutherst et al. 1995
). We also applied approximately the same criterion (greater than or equal to 50% probability of colonization) to determine the eventual range size predicted by native and introduced populations. Hence, we believe that the major difference between our model and past work, and the primary reason for the observed increase in the predicted area of suitable toad habitat, is that our model does not depend on data from the cane toad's native range or its native physiology in that range. Instead, the invasive range model incorporates changes in the toad's bioclimatic niche that have arisen in the exotic habitat.
Our invasive range model predicts a broader distribution of toads in both colder and hotter climates in Australia when compared with those based on native range (Sutherst et al. 1995
). Australian cane toads now occupy regions where the minimum monthly temperature falls below 5.0°C and the maximum monthly temperature climbs above 37.0°C. Hence, Australian cane toads now are predicted to be able to inhabit a broad region of cooler climates in southern Australia (b
). At the same time, the toads' recent and successful invasion into high-temperature habitats of the Northern Territory suggests that cane toads are capable of surviving in extensive areas of northern and western Australia. Since we based our climate data on monthly means, daily thermal fluctuations will probably be more extreme and might expose toads to their lower or upper physiological temperature limits in these regions (Floyd 1983
). In addition, in hot regions, increases in potential water loss correlated with high temperatures are likely to elevate the potential for lethal desiccation stress (Zug & Zug 1979
). Although broad-scale climatic variation will not always equate with the range of available thermal microhabitats on the ground, our study suggests that toads increasingly have colonized areas where thermal and desiccation stresses had been expected to prevent their successful establishment.
(b) Invasion dynamics
By updating our model with time-limited data subsets, we constructed a time-series of invasive range predictions. From this series, we discovered a remarkable increase in the predicted area of suitable habitat that has occurred over the last 10 years (). This expansion in suitable habitat followed a longer-term rise in the 95th percentile maximum temperature of colonized toad habitat from 36.1 to 38.3°C. Moreover, colonized regions after 1994 were more likely to occur where maximum annual temperatures rose above 37.0°C, according to a classification tree analysis. This pattern cannot be explained by a simple time lag following introduction before cane toads encountered extreme temperatures, because comparably hot and dry regions found in interior Australia halted westward invasions prior to their expansion into the Northern Territory. In contrast to maximum temperatures, 95th percentile minimum annual temperatures did not change significantly over the same period.
These results suggest two points. First, the dynamics of invasion into hot and cold regions of Australia appears to be quite different. Although toads inhabit cooler regions of Australia than expected based on their native range, this cold tolerance has not expanded greatly over the last 30 years. In contrast, cane toads rapidly expanded their range into hot regions in the last decade. This rapid expansion was preceded by the colonization of climatic regions close to the toad's known upper physiological limits of maximum temperature. This pattern is in agreement with the empirical data demonstrating that cane toads from the warmer northern invasion front are expanding their range at increasing rates (Phillips et al. 2006
). Moreover, these recent toad expansions suggest that model predictions may be conservative for regions characterized by high maximum annual temperatures.
Clearly, the breadth of predictors associated with an expanding invasion range can be expected to increase until the invasive taxon reaches the edge of its bioclimatic envelope in a new region (Wiens & Graham 2005
; Facon et al. 2006
). However, several observations suggest that the cane toad's niche has expanded in Australia. (i) Range expansions have been restricted to specific geographical regions (i.e. northern expansion front) rather than to broad regions with similar bioclimatic conditions. (ii) Australian toad range boundaries now lie outside those predicted by their native range. (iii) Toad range expansions have accelerated rather than decelerated with time.
Both ecological and evolutionary explanations can be offered to explain niche expansions in introduced species (Lee 2002
; Torchin et al. 2003
; Holt et al. 2005
). Ecologists have argued that a species' native niche, as defined by its ancestral range climate envelope, may not equate with its realized niche in an exotic habitat because the negative demographic effects of native species are absent (Holt et al. 2005
). The successful establishment of introduced species has often been attributed to their ecological release from native enemies and competitors (Diamond 1970
; Keane & Crawley 2002
; Torchin et al. 2003
). Ecological release could result in a higher tolerance of extreme abiotic conditions in novel habitats if both biotic interactions and abiotic conditions combine to limit fitness in the ancestral habitat (Holt et al. 2005
). For example, abundant ticks are thought to increase cane toad susceptibility to dehydration and starvation in their native range (Zug & Zug 1979
). The cane toad left most of its parasites, pathogens and predators behind, when it was introduced to exotic habitats (Speare 1990
). Hence, ecological release then may explain part of the cane toad's success in hot and cold areas of Australia. However, evidence for ecological release remains equivocal (Keane & Crawley 2002
; Colautti et al. 2004
) and little is known about how biotic interactions and abiotic conditions jointly affect cane toad demography in Australia.
We suggest that an evolutionary, in addition to an ecological, explanation may be necessary to account for the changing distribution of Australian toads. Cane toads increasingly occupy regions in Australia with extreme temperatures, with daily temperatures potentially falling outside experimentally determined physiological limits (Floyd 1983
). The number of cane toads in Australia probably has surpassed the population genetic threshold at which mutation rate constrains emergence of new genetic variation (Butin et al. 2005
). The observed lag in range expansion followed by explosive growth (Phillips et al. 2006
) corresponds with predicted patterns of expansion following niche evolution (Holt et al. 2005
). Most importantly, recent evidence suggests that toads from the invasion front in the Northern Territory have evolved enhanced dispersal ability (Phillips et al. 2006
). Adequate genetic variation also may be expected for other traits, including those involved in overcoming physiological limits. Toads may be evolving to move not only faster, but also further into extreme abiotic habitats in Australia.
Therefore, invasion dynamics may include a fourth stage (after arrival, establishment and spread): once adaptation and niche expansion are sufficient, adaptive genetic variation accrues in an expanding invasive population (Lee 2002
). Ecological considerations are expected to dominate early on in a successful invasion as the newly introduced species becomes established in a habitat that matches its ancestral niche. Following establishment of a stable population source—a beachhead of sorts—an increasing number of propagules can disperse into suboptimal habitats with little impact on the demography of the source population (Holt et al. 2005
). In the meantime, the rapidly growing core population may restore the genetic variation lost through the initial bottleneck, via recombination and mutation (Butin et al. 2005
). Together, these factors can increase the probability that advantageous adaptations arise in marginal habitats such that range expansion can occur through niche evolution. Demographic or genetic traits that allow species to reach the evolutionary invasion stage may explain why some taxa become globally invasive while others remain infrequent intruders. For example, cane toads are explosive breeders and often are found at extremely high densities in their introduced habitats (Lever 2001
). These demographic characteristics both support high rates of population increase and influence the rate at which successful mutations arise in new habitats.
From a practical standpoint, forecasts of species distributions, such as those applied to invasive species or to species distributions following climate change, should incorporate the possibility that the niches of invasive species may not be conserved in novel habitats. This may occur due to either ecological release or evolutionary dynamics. More accurate forecasts can be accomplished using flexible models that are based on the introduced, in addition to the native, range and that are dynamically updated during a species' range expansion. Unlike the standard approach of using native range to predict the potential distribution of invasive species in novel habitats, models based on the invasive range include the potential for niche expansion through either ecological or evolutionary means (Lee 2002
; Kolbe et al. 2004
; Butin et al. 2005
; Holt et al. 2005
; Facon et al. 2006
). Range predictions through time can indicate areas of model uncertainty, highlight specific variables to which organisms may be evolving higher tolerances and delineate particular geographical regions where trait divergence is suspected.
Controlling the further expansion of cane toads in Australia remains a daunting task. Future range expansions could devastate Australia's endemic species. However, effective quarantine might still limit the range of this invasive taxon. Model predictions can facilitate attempts to curb further expansion, by focusing monitoring and eradication in cities, at habitat bottlenecks, and between isolated regions of suitable habitat in western parts of Australia. However, if the continuing acceleration of toad expansion reflects continuing adaptation to a new environment (Phillips et al. 2006
), then even our current forecast may prove conservative.