OUP user menu

A Climate Model of the Red Imported Fire Ant, Solenopsis invicta Buren (Hymenoptera: Formicidae): Implications for Invasion of New Regions, Particularly Oceania

Robert W. Sutherst, Gunter Maywald
DOI: http://dx.doi.org/10.1603/0046-225X-34.2.317 317-335 First published online: 1 April 2005


The paucity of empirical data on processes in species life cycles demands tools to extract insight from field observations. Such insights help inform policy on invasive species and on impacts of climate change at regional and local scales. We used the CLIMEX model to infer the response of the red imported fire ant, Solenopsis invicta Buren (Hymenoptera: Formicidae), to temperature and moisture from its range in the United States. We tested hypotheses on the mechanisms that limit the distribution of the ant and estimated the potential global area at risk from invasion. The ant can spread further in the United States, including north along the west coast, where patterns of infestation will differ from those in the east. We analyzed the risk of colonization in Australia and New Zealand, where the ant was recently discovered. The patterns of infestation of the ant in Oceania will differ from those in the eastern United States, with slower growth and less winter mortality. This study adds to earlier temperature-based models by incorporating a moisture response; by replacing arbitrary categories of colony size to predict overwintering success with a site-specific model based on the balance between annual growth and survival; and by comparing different hypotheses on low temperature-related mechanisms that limit the geographical distribution. It shows how the response of a species to climate can be synthesized from field observations to provide useful insights into its population dynamics. Such analyses provide a basis for making decisions on regional management of invasive species and an informative context for local studies.

  • species mapping
  • invasion
  • biogeography
  • climate

Risks to natural and managed ecosystems from invasive species and environmental change are creating demand for tools to assess regional risks in a data-poor environment. Specifically, ecologists need to estimate the likelihood of establishment of species in new regions and their impacts on biodiversity, agriculture, and built environments under current and possible future climates. The need for simple tools arises because there is rarely enough empirical data available on biological processes to build rigorous, predictive population models for even the most intensely studied species (Sutherst and Maywald 1985, Korzukhin et al. 2001). As climate is such a basic driver of biological processes, tools that capture the response of species to climate can make a contribution. Phenological models (Regniere and Nealis 2002) and simple process-based models (Sykes and Prentice 2004) have a role to play if tolerances to extreme climatic conditions can be defined. There are numerous descriptive, statistical, or rule-based approaches available, but their success in making predictions of changes in geographical distributions has been questioned repeatedly (Sutherst 1998, Kriticos and Randall 2001, Thuiller 2003). Such static models suffer from the inherent deficiency that the more they are parameterized to describe a given geographical distribution, the less they are able to accommodate different climatic patterns. They also provide minimal insight into the biology of the species, and the use of sophisticated statistical tests of goodness-of-fit (Cumming 2000, Anderson and Lew 2003) gives false impressions of precision because it ignores systemic errors in the model paradigms. Furthermore, the sophistication of any tool used to make global extrapolations is ultimately limited by the availability of environmental databases with global coverage on one hand and the quality of the distribution records on the other. Thus, we need pragmatic and parsimonious approaches to make the most use of the global climatic data and field observations that are available.

An alternative approach is inverse or inferential simulation modeling. It infers the growth and stress responses of a species to climate from its geographical distribution, relative abundance, and seasonal phenology. The model is parameterized using field observations, supported where possible with empirical data. This approach forms the basis of the CLIMEX model (Sutherst and Maywald 1985, Sutherst 2003). Given the richness of spatial, climatic data compared with experimental or site-specific field data, the resulting models can be quite robust (Sutherst 1998). They can be used to test hypotheses on climatic factors that limit a species' distribution and to estimate the climatic suitability of other regions for that species (Sutherst 2003).

In this paper, we use CLIMEX to infer the climatic responses of the red imported fire ant, Solenopsis invicta Buren (Hymenoptera: Formicidae), referred to hereafter in this study as the fire ant, from its geographical distribution and seasonal phenology in the United States. The result is an overview that synthesizes the species relationship with climate and provides an informative context within which to conduct plot-scaled studies. As such, CLIMEX analyses complement process-based modeling and are a useful first step in any ecological study.

The fire ant originates from the Pantanal region in Brazil, Paraguay, and northern Argentina (Buren et al. 1974, Shoemaker et al. 1996, Ross and Shoemaker 2005). It has a broad range of negative impacts on biodiversity (Allen et al. 1994,2001, Gotelli and Arnett 2000) and urban environments (Lard et al. 2002), but mixed effects on agriculture (Davidson and Stone 1989). Four attempts have been made to predict the limits to the ant's spread in North America using statistical or simulation models (Pimm and Bartell 1980, Stoker et al. 1994, Killion and Grant 1995, Korzukhin et al. 2001). Each model made a useful contribution, and the latter authors discussed the strengths and weaknesses of each approach. All have been hindered by the fact that the fire ant is a ground nesting, social insect, which moves its brood up and down in the subterranean nest to maintain it at the optimal available temperature (Morrill et al. 1978). In addition, its range is still expanding. Anonymous (2001), in a preliminary analysis, used the CLIMEX model to assess the risks of establishment in Oceania. That report concluded that the fire ant could establish in much cooler climates than those classified as "possible" by Korzukhin (2001), with minimum alate production of >2100. Morrison et al. (2004) then liberalized this threshold to 1,500 "… to reduce the chances of underestimating future range limits … " and applied the model to selected global locations.

The geographical distribution of the fire ant covers a wide range of climatic conditions in the United States (Fig. 1; http://www.aphis.usda.gov/ppq/maps/fireant.pdf, updated April 2004 and also showing historical infestations in northwest Texas). In the southern United States, the current quarantine area includes the area from Mexica to Virginia and inland to western Texas and from southern Oklahoma to North Carolina. It persists in irrigated habitats in southern California (Callcott and Collins 1996, Anonymous 2000), west Texas, and New Mexico (MacKay and Fagerlund 1997). The fire ant was eradicated from a horticultural nursery in Arizona (Frank 1988), and a large infestation was the target for eradication in Yuma, AZ (Thaxton 1999). The current quarantine area does not include historical urban infestations in Lubbock lasting 7 yr (Thorvilson et al. 1992) and Sherman county in Texas close to Oklahoma (Stoker et al. 1994, http://fireant.tamu.edu/materials/newsletters/fatrails_s.html).

Fig. 1.

Geographical distribution of the fire ant in the United States as indicated by the quarantine zone (after APHIS updated 2004) and including locations of two historical infestations in northwest Texas.

In February 2001, the fire ant was found to have established in the subtropical, east coast city of Brisbane (27.5° S) in Australia. In the same month, a single nest, including three alate queens, was detected at the Auckland airport in New Zealand (Anonymous 2002). A further nest was found at the Port of Napier, New Zealand, in February 2004 (A. Flynn, personal communication). These events raise the question of how far the ant could spread in Oceania and how much damage it could cause. A primary aim of this study was therefore to define the risks that are posed to Australia and New Zealand by the fire ant. We now finalize the preliminary risk analyses of Anonymous (2001) and Sutherst (2002) and assess the risks to the rest of the world from this severe pest species.

Materials and Methods

The CLIMEX software (Sutherst and Maywald 2004) contains the CLIMEX model, which is a dynamic model, of intermediate complexity, that describes the response of an organism to climate in different geographical locations (the Compare Locations function). It is therefore an explanatory model, albeit with limitations in scope and precision that could be achieved with a comprehensive population model.

Persistence of populations in a given habitat depends on the balance between growth during the favorable season and mortality during the unfavorable season. Therefore, CLIMEX builds a composite picture of the growth and stress responses and any constraints imposed by daylength or the length of the growing season. It incorporates a hydrological model that simulates available soil moisture. CLIMEX integrates the weekly effects of temperature (TI) and moisture (MI) to calculate weekly and annual hydrothermal growth indices, GIW and GIA, respectively. The GIW gives weekly snapshots of the suitability of the climate for growth and estimates instantaneous rates of growth throughout the year. A series of stress indices estimate the response of the species to prolonged or intense periods of adversely cold, hot, dry, or wet conditions. If the annual value of the total stresses reaches 100, the species is deemed not to be able to survive at that location. The GIA and stress indices are combined into an ecoclimatic index, EI, scaled from 0 to 100, to represent the overall favorableness of the given geographical location for persistence of that species. On past experience, EI values of >0-10 indicate marginal habitats, 10-20 will support substantial populations, and >20 are highly favorable. A value of zero indicates that the species is unable to persist at that location.

CLIMEX models have been used for many species (http://www.ento.csiro.au/climex/bibliography.htm) and validated on occasions by a priori predictions being fulfilled (Bruce and Wilson 1998, Hall and Wall 1995). They have also provided new insights into the biogeography of species, such as the livestock tick, Boophilus microplus Canestrini, even when they have been studied extensively (Sutherst 1987, Sutherst et al. 1995a). In addition, a lack of internal consistency during parameter fitting has helped to detect nonclimatic factors limiting species' distributions (Sutherst and Maywald 1985, Vera et al. 2002).


The principle assumption when using CLIMEX is that the geographical distribution of the target species is ultimately limited by climate (Sutherst and Maywald 1985). This carries the implicit assumption that a positive record represents the ability of the species to persist in the given location with the inherent climatic variability that is associated with the average climatic data.

Other variables may prevent a species from occupying all climatically favorable habitats, so it is necessary to identify potential nonclimatic limits as part of any analysis (Sutherst 2003). These may include one or more of the following factors that could affect S. invicta and so constrain the interpretation of the role of climate.

Nonclimatic Constraints

Physical Barriers.  

Habitat types (dry cerrado versus flooding pantanal) and soil disturbance have large effects on the occurrence of the fire ant in South America (Allen et al. 1974). There were no reports of similar soil-related barriers in the United States.

Hosts or Food Sources.  

The fire ant is a generalist feeder with plant and animal food sources, so food is unlikely to limit the distribution except in arid zones.


Alate queens disperse locally, so no vectors are required for dispersal of the fire ant. Movement of horticultural material is a particularly efficient long-distance vector (Stoker et al. 1994).

Other Species.  

There are reports of hybridization with Solenopsis richter in the United States (Ross and Robertson 1990, Shoemaker et al. 1994). While survival of the hybrids over winter was not different to that of the fire ant (Diffie et al. 1997), the hybrids were less fit than the parental genotypes, leading to the establishment of mosaic hybrid zones in which both genetic fitness and environmental stressors contribute to the outcomes (Shoemaker et al. 1996). A consequence of this phenomenon is that the inferred climatic tolerance of the fire ant is likely to be underestimated in the United States. There are interactions with other ant fauna (Banks and Williams 1989) but little evidence that these latter species block the spread of the fire ant in the United States (Gotelli and Arnett 2000). The fire ant's distribution in South America (Mescher et al. 2003) may be limited by interactions with other ant species such as Pheidole spp or other species of Solenopsis . There is also exchange of genetic material between Solenopsis spp. in the region (Buren 1972, Buren et al. 1974, Ross and Trager 1990, Ross and Shoemaker 2005). Furthermore, the incidence of Wolbachia varies across the ant's range in South America (Shoemaker et al. 2003), with unknown implications for the geographical distribution. These interactions precluded the use of the native range in estimating the model parameter values.

Artificial Environments.  

Built infrastructure can provide favorable habitats in arid regions with irrigation as well as artificial heat-sinks that facilitate overwintering (Thorvilson et al. 1992, Stoker et al. 1994). Nurseries provide artificially benign habitats in otherwise hostile environments (Frank 1988, Anonymous 1999).


As with any process that attempts to parameterize models or set thresholds based on geographical distributions, there is an implicit assumption that the species' range has reached a stable state. That is not true in the case of the fire ant in North America (Callcott and Collins 1996), so prediction of the potential range was constrained to giving minimum estimates based on extrapolation from the most adverse climatic conditions that the ants have so far been shown to tolerate.


Some species have specific life cycle attributes that require certain conditions to be met before they can persist in a given habitat. The first is obligate diapause, but that does not occur in the fire ant. The second is the amount of annual heat accumulation that is needed to complete a generation. This does not apply to multivoltine species or social insects with indeterminate generation times, but accumulated degree-days have been used to infer the annual alate production of colonies (Korzukhin et al. 2001) and their related ability to persist in a habitat.

CLIMEX requires monthly average maximum and minimum temperatures, rainfall, and relative humidity data. The Intergovernmental Panel on Climate Change global splined grid of mean monthly surface climate data with a resolution of 10′ (New et al. 2002) from the Climate Research Unit (CRU) at Norwich, United Kingdom, was used in the analyses. In addition, comparisons were made between climatic data from specific locations using the internal CLIMEX database.

Rural and urban irrigation practices were simulated by assuming that up to ≈30 mm of water was applied to farms and urban gardens weekly (4.3 mm/d) throughout the year to top up the rainfall until there was 30 mm of moisture being input each week. Summer drought in the United States was simulated using the CLIMEX climate change scenario feature to change the rainfall and temperature as required.

Model Fitting

A visual, iterative process is used to fit CLIMEX parameter values using the boundaries of the geographical distribution to estimate values for each stress variable in turn. This process has considerable heuristic value and acknowledges the modest quality of observations on geographical distributions. An automated parameter fitting procedure is under development for use in routine risk assessments. Observations on seasonal phenology are used to infer the suitability of conditions for colony growth. The model is then run to compare the ecoclimatic indices with the geographical distribution. Finally, the model estimates are examined for consistency with empirical data if they are available. The geographical distribution of the fire ant quarantine area (Callcott and Collins 1996), updated to 2003 (http://www.aphis.usda.gov/ppq/maps/fireant.pdf), was used to estimate the parameter values for the stress parameters (Table 1). All of the current simulations were run with the parameter values in Table 1, unless stated otherwise. The model fitting procedures involved testing a number of hypotheses related to overwintering in particular.

View this table:
Table 1.


Heat Limits.  

The fire ant can tolerate extreme heat, judging by the temperatures in the infested areas in the western United States. The highest monthly average maximum temperature associated with recorded infestations of the fire ant was 41.6°C in July in Yuma, AZ (Thaxton 1999). Because this is one of the hottest places in North America, there was no evidence that heat per se excluded the ant from anywhere on the continent.

Cold Limits.  

Interpretation of the limiting effects of cold was the most demanding of the parameter-fitting processes, because this soil-nesting ant is still extending its northern range, and a hybrid zone exists with S. richteri along the boundary. Also, some records of infestations in cold climates in western and northwestern Texas (Thorvilson et al. 1992, Stoker et al. 1994) and an unconfirmed case at Knox County, TN were outside the range of minimum temperature of any rural infestations in the infested zone. Artificial heat sinks increased the overwintering survival of the fire ants at Lubbock from 40 to ≈80%, and irrigation promoted growth in summer. Model parameters were fitted to simulate cold as a limiting factor in rural environments.

Two cold stress (CS) hypotheses were tested, based on (1) minimum temperatures and (2) average daily temperatures, to take account of the buffering effects of soil on diurnal temperature cycles with different amplitudes. With both models, CS was calibrated to equal 100 at Knoxville, adjacent to the northern boundary of the fire ant quarantine zone in eastern Tennessee, where mean daily temperatures were the lowest on the boundary.

In view of the uncertainties surrounding the accuracy of parameterization because of the continuing expansion of the range in North America and the uncertainty of the significance of the reported overwintering success at Sherman County, a second approach was taken to check the results. In this approach, the winter temperatures for one of the coldest cities in Australia, Canberra, were compared with locations in North America using the CLIMEX Match Climates function. It compares climatic variables from different places without reference to any given species' response. The index of similarity, called a match index, CMI is scaled between 0 and 1.

Population Growth

Colony growth is sensitive to temperature, with a reported lower threshold temperature for colony growth of 24°C (Porter 1988). This value is much higher than the theoretical threshold value of 17°C for development of most stages derived by Porter. A range of estimates of developmental thresholds of 13-17°C for the different instars has been reported (O'Neal and Markin 1975). A threshold value of 21°C was adopted for modeling colony growth (Korzukhin et al. 2001), but temperatures are not comparable with climatic data because these authors used soil temperatures that were several degrees higher than air temperatures. Because development of some instars proceeded below 21°C and our aim was to simulate growth of the colony as a whole, a threshold air temperature of 17°C was used, and a sensitivity analysis was conducted to examine the effects of using higher values.

An optimum average temperature of 28°C (Cokendolpher and Francke 1985) was below Porter's minimum optimum of 32°C and uncomfortably close to the latter's minimum of 24°C. An upper threshold of 36°C was reported (Porter 1988), with colony growth at temperatures above 32°C being curtailed by high mortality rates of workers (O'Neal and Markin 1975). Nanitic brood (first generation from a new founding queen) developed ≈35% faster than did minor worker brood (O'Neal and Markin 1975, Porter 1988).

The parameter values for the temperature response (Table 1) were fitted, using field observations on seasonal variation of brood production at six locations in the southeastern United States (Markin and Dillier 1971, Markin et al. 1973). The seasonal phenology of the ant indicated that temperatures in the southern United States are usually suitable for population growth from spring to autumn, but production of larvae usually ceased in winter and was reduced in midsummer to autumn in most places. Markin's data represented colony size unadjusted for emigration of alates, and they measured relative population size of larvae and workers not colony growth rates, so the opportunity to optimize the growth parameters was constrained. Our principal aim was to simulate the seasonal start and end to fire ant colony development and broad seasonal pattern of colony development. Meteorological data for each location was obtained from the corresponding cell in the climate grid.


Dryness Limits.  

Direct physiological effects of aridity on fire ants are difficult to separate from the lack of a source of prey in arid zones. Reports of moisture limitation in western Texas and New Mexico (Thorvilson et al. 1992, Allen et al. 1993, Stoker et al. 1994, MacKay and Fagerlund 1997) were used to set the soil moisture threshold that prevented populations of S. invicta from persisting. The fire ant thrives in irrigated lands (Callcott and Collins 1996), so simulations were run with irrigation to examine its effect on the potential range.

Wetness Limits.  

Because the fire ant is most abundant in the high rainfall areas in the southeast United States and is present in arid areas when associated with irrigation, it is evident that high rainfall is favorable for S. invicta . Judging by its success in Florida and the coastal States around the Gulf of Mexico, S. invicta is highly adapted to wet conditions including flooding. This is consistent with its origin in annually inundated habitat in the Pantanal in South America (Allen et al. 1974, Buren et al. 1974). Parameter values for wet stress were set high enough to exclude any limiting effect of high rainfall on the geographical distribution in the United States.

Population Growth

Observations on seasonal phenology in the eastern United States (Markin and Dillier 1971) were used to estimate the moisture-related growth parameters.


The results provide a composite description of the suitability of each component of the climate around the United States for fire ants. This is shown with results from representative locations: Tampa, Gulfport, Lubbock, Chattanooga, Knoxville, and St. Louis in the east, and Yuma, Fresno, and Sacramento in the west.

Risks in the United States

Seasonal Population Growth.  

The trends of high summer temperatures reducing population growth rates and winter temperatures stopping colony growth in all but the most southerly habitats were captured in the seasonal CLIMEX temperature index, TI, values. The results are compared with observations at six sites in the southeast United States (Markin and Dillier 1971) (Fig. 2).

Fig. 2.

Mean monthly percentages of colony biomass represented by total larvae (- - -) at six sites (a, Auburn, AL; b, Clay, MS; c, Conway, SC; d, Gulfport, MS; e, Monroe, LA; f, Tampa, FL) in the southeast United States (Markin and Dillier 1971) compared with the CLIMEX weekly growth (▪▪▪▪), temperature (●●●●), and moisture (——) indices.

The values of the seasonal TI in Gadsden County, FL, were compared with data on colony development including the larval and pupal stages of colony development (Morrill 1974) (Fig. 3). The results using a threshold temperature of 17°C corresponded better with the data on spring activity, including the larval and pupal stages of colony development, than either 21 and 24°C (Porter 1988, Korzukhin et al. 2001). They also compared well with the data on seasonal production by colonies of the fire ants at Tallahassee, near Quincy (Tschinkel 1993).

Fig. 3.

(Top) Average maximum (——) and minimum (●●●) temperatures and weekly rainfall (histograms) at Quincy, FL. (Bottom) Development of alate-producing larvae (—) and pupae (—) of S. invicta in Gadsden County, north Florida (Morrill 1974) compared with the CLIMEX growth index with threshold temperature (DVO) of 17 (heavy solid line), 21 (—●—), and 24°C (●●●).

The relationship between soil moisture, relative humidity, and colony size and reproductive status was examined in Louisiana (Pranschke and Hooper-Bui 2003). Differences in Harlan ratings were attributed largely to vertical movements of the ants within the nests. Suppressed colony growth in dry summers was reported in Georgia (Haney et al. 1996) and Mississippi (Callcott et al. 2000), but the effects of low soil moisture and high temperatures were not separated. The simulated population growth was reduced during periods of moisture deficits in summer and autumn (Fig. 2), suggesting that average seasonal soil moisture constrains population growth of the fire ant under natural precipitation conditions in those states, with a greater effect in the drier areas.

To study the sensitivity of the model to low soil moisture levels, simulations were run with values of the threshold soil moisture parameter SM0 = 0.20 and 0.10 compared with the fitted value of 0.15. The effect was to move the boundary of the potential range in western Texas by 100 km east or west, respectively. The seasonal growth results show that low moisture became progressively more limiting when moving west, especially west of a north-south transect through central Texas. The drier trend also reduced the potential for population growth of the fire ant toward the western end of the northern boundary of the quarantine zone in Oklahoma.

Integration of the effects of temperature and moisture into the GIW described the potential for growth of the populations as shown in Fig. 2. The results showed the underlying seasonal cycle being driven by temperature, with restriction of population growth in summer and autumn by suboptimal soil moisture at some sites.

The sensitivity of the growth-related parameters to summer conditions was tested further by reducing rainfall by 50% and simultaneously increasing temperature by 2°C to simulate the harsh summer in southern Mississippi in 1994 (Callcott et al. 2000) and at Tifton, GA (Haney et al. 1996), in which fire ant colonies decreased in size. The results (Fig. 4) showed a substantial depression of growth rates in the hotter, drier scenario, with the GIA values being reduced by about two-thirds compared with results in Fig. 2. This effect could be attributed to both reduced moisture and unfavorably temperatures in midsummer.

Fig. 4.

CLIMEX weekly growth (—), moisture (●●●●), and temperature (- - -) indices for (A) Gulfport +2°C, −50% rainfall for comparison with Fig. 1, and Tifton, with (B) average temperature and rainfall and (C) +2°C, −50% rainfall.

Simulations were also run to examine the effect of irrigation on population growth. It made the arid zones very suitable, consistent with the infestations in urban and horticultural habitats.

Annual Population Growth Around the United States.  

Annual CLIMEX Growth Index, GI A. Inadequate opportunity for a colony to reach a critical size to enable it to overwinter could also limit its geographical distribution. The number of annual degree-days is therefore one measure of the suitability of a habitat for growth of fire ant colonies (Korzukhin et al. 2001). The CLIMEX indicator of the potential size of colonies available to overwinter is the GIA. The above simulations of the seasonal growth patterns in relation to temperature and moisture provided the basis for extrapolation of the results spatially to examine the likely geographical patterns of annual population growth around the United States.


The annual CLIMEX temperature index, TIA, is designed to provide a more realistic indicator of potential size of a colony than degree-days because it simulates the nonlinear changes in population growth rates rather than just linear increases in development rates. It therefore accounts more accurately for suboptimal conditions at high and low temperatures, which reduce development and reproduction and increase mortality rates. The results, shown in Fig. 5a, indicate that current infestations of the fire ant occur in regions with an annual TIA greater than ≈25. The values of TIA showed a gradient from 48 at Tampa to 32 at Gulfport and 25-28 at Lubbock, Chattanooga, Knoxville, and St. Louis. In the west, the values for Fresno, Yuma, and Sacramento were 16, 17, and 21, respectively. The trend of predicted suitability based on the TIA followed that based on the degree-days in the eastern states but differed significantly in the west, where much of the excessive heat in Yuma was not exploitable by the ant.

Fig. 5.

Geographical distribution of annual suitability of the growing season for fire ant in the United States as estimated (a) by the annual CLIMEX temperature index, TIA, and (b) the annual CLIMEX moisture index, MIA, under natural rainfall conditions.


The effect of moisture on potential annual colony growth in the United States was studied (Fig. 5b). The values of the annual CLIMEX moisture index, MIA (Table 2), were very high at Gulfport and Tennessee, dropped at St. Louis and Tampa, and was only five at Lubbock. On the west coast, the MIA values for Yuma, Fresno, and Sacramento were 0, 15, and 41, respectively. Year-round irrigation increased the MIA values greatly at Lubbock and on the west coast (Table 2).

View this table:
Table 2.

By combining the effects of temperature and moisture, we derived the weekly and annual values of the GIA in habitats with natural rainfall across the United States. Simulations were run to compare potential colony growth in the United States (Fig. 6a and b; Table 2). The values of GIA showed gradients from east to west and from south to north. They were high at Gulfport and Tampa, dropped to Tennessee, and then dropped to 16 at St. Louis and 2 at Lubbock. Much of the far west was too dry. Values were just above zero in the California valleys and southern coast. They were low because of the lack of synchrony between the favorable summer temperatures and the winter rainfall. Year-round irrigation expanded the area in which fire ant colonies are able to grow (Fig. 6b). There was potential for enhanced colony growth in a belt of country on both sides of the Mexican border.

Fig. 6.

Geographical distribution of annual suitability of the growing season for fire ant in the United States as estimated by annual CLIMEX growth index, GIA, under (a) natural rainfall and (b) irrigation conditions,


There was no evidence that high temperatures limited the geographical distribution of the fire ant, given the infestation in Yuma. The most contentious question about the fire ant relates to the effect of low temperatures on the potential for northern spread of the ant. Two hypotheses were tested on the effect of the severity and duration of the winter cold stress in different habitats, with the models calibrated to give a value of CS of 100 for Knoxville, as explained above, to enable a model comparison.

Model 1: Minimum Temperature.  

A threshold minimum temperature of 0°C and a CS accumulation rate of -0.00447 increased CS rapidly approaching the current boundary (Fig. 7a). There was a gradient of values of CS from 0 in Tampa and Gulfport to 41 at Chattanooga, 74 at Lubbock, 100 at Knoxville, and 192 at St. Louis in the eastern states, and 0 at Yuma, Fresno, and Sacramento in the west.

Fig. 7.

Distribution of cold stress CLIMEX (CS) experienced by the fire ant in the United States. Cold stress based on (a) minimum temperature of zero, below which stress is accumulated, and (b) average winter temperatures. See text for parameter values.

Model 2: Mean Daily Temperature.  

The warmer days in Lubbock than at Knoxville, TN, with the same minimum temperatures could have raised average mound temperatures sufficiently to enable higher survival at Lubbock (Thorvilson et al. 1992). A better indicator of winter cold, as experienced by the ant colonies, may therefore be the mean daily temperature. Simulations were run to examine the effects of limiting the range based on average winter temperatures. An average daily temperature of 6°C, corresponding to Lubbock February temperatures, and a stress accumulation rate of -0.0135 gave an annual CS of ≈60 at Lubbock. Note that CS and mortality rates are not necessarily linearly related but are used here for simplicity, given the difficulty in defining the relationship precisely from the available data. Such a model excluded the ants entirely from Tennessee and adjoining areas to the south, and restricted persistence in Oklahoma to its southern border. This clearly underestimated their potential range. The rate of accumulation of cold stress was therefore reduced to -0.0041 to allow the value of the CS index in Knoxville to equal 100. The potential CS then closely matched that from model 1 in the eastern states, but there were significant differences in the west (Fig. 7b), with reductions in the northwest and in New Mexico (in the absence of moisture limitations to growth). The model therefore favored the high country with its warm days compared with that based on minimum temperature above (Fig. 7a). It gave a lower CS value of 18 for Lubbock.

Despite the fact that there were significant differences between the results from the two models, model 1 was used in further analyses for brevity and because the western areas favored by model 2 are unsuitable because of aridity. Further field data are needed to discriminate between the two models.


The analysis did not identify any limiting effects of high soil moisture on the potential geographical distribution of the fire ant. However, extreme aridity was estimated to not only reduce population growth (Fig. 4), but also to prevent the colonies from surviving, thus excluding ants from arid zones (Fig. 8). Irrigation removed the dryness limits in western Texas and in southern and central California.

Fig. 8.

Distribution of dry stress (CLIMEX DS) experienced by the fire ant in the United States.


By combining the GIA with the estimated reduction in relative abundance in the unfavorable season, as indicated by the total stress, CLIMEX produces an overall measure of likely persistence and relative abundance in the form of the ecoclimatic index, EI. The results for the fire ant in the United States, with and without irrigation, are shown in Fig. 9a and b.

Fig. 9.

Geographical distribution of suitability of habitats for permanent persistence of the fire ant in the United States, as estimated by the CLIMEX ecoclimatic index, EI, under (a) natural rainfall and (b) irrigation.

There was a gradient of values of EI from 39 at Tampa and 37 at Gulfport to low values in Tennessee and 0 at Lubbock and St. Louis. Thus, Lubbock, Knoxville, and St. Louis are estimated to be unable to sustain fire ant populations with natural rainfall. The EI values were very low (1-2) in the northwest because of low temperatures in the Sacramento, San Joaquin, and Willamette valleys, and further south in the coastal region around Los Angeles because of low rainfall. The values of the EI in the south increased from ≈2 up to 10-20 with irrigation across the United States. There was a shortage of thermal accumulation in the north that will result in very slow colony growth coupled with low winterkill in the west.

We examined the sensitivity of the projected boundary to changes in winter temperatures, reflecting either future spread of the ant into colder habitats than it currently occupies in the United States (and hence current underestimation of the fire ant's tolerance of cold), difference between soil, air, and laboratory temperatures, or global warming. Increasing or decreasing the winter (1 October to 1 March inclusive), temperatures moved the receptive zone north or south by ≈100 km/°C, with variations depending on local topography (Fig. 10). Increasing temperature alone in summer (2 March to 30 September) resulted in a retraction of the range from the western limits in Texas because of greater moisture deficits with the associated increase in evaporation. A 10% increase in rainfall offset this effect.

Fig. 10.

Sensitivity of the potential distribution of the fire ant in the United States to winter temperatures (a) +2°C, (b) + 1°C, (c) default, (d) −1°C, and (e) −2°C.

Risks in Oceania

The global areas at risk of invasion by the fire ant, in the absence of nonclimatic barriers, were explored. CLIMEX was run, with parameters set for rural areas using the CRU global climatic database without irrigation. The results are shown in Fig. 11. Climate per se will not constrain the ant from colonizing all moist tropical and subtropical zones and countries bordering the Mediterranean and western France. Irrigation would allow it to establish in arid zones and increase colony growth in Mediterranean climates. Particular attention was given to the risks to Australia and New Zealand, but the receptivity of the climate of other regions was noted as follows. In contrast to the observed distribution, in South America, almost the whole continent except the Andes and southern tip is climatically very receptive to establishment of the fire ant. In Africa, there are few climatic barriers apart from the Sahara and Namib deserts, with the humid tropics of west and central Africa being particularly favorable. The whole of tropical and subtropical Asia is climatically suitable with its combination of high rainfall and high temperatures.

Fig. 11.

Potential global distribution of the fire ant under natural rainfall as estimated by the CLIMEX ecoclimatic index, EL


The potential for seasonal population growth around Australia was studied, and representative results for Brisbane and Sydney on the east coast, for Melbourne in the south, and for Perth on the west coast, with and without summer irrigation, are shown in Fig. 12. All sites except Melbourne had the potential to support substantial populations if moisture deficits were overcome by irrigation. Low temperatures in Melbourne and Canberra (Table 2) limited growth potential.

Fig. 12.

CLIMEX growth index, GI, for S. invicta in Australia with (a) natural rainfall and (b) irrigation.

Scenarios were used to examine the potential for population growth in rural environments with natural rainfall and irrigation. Australia is the driest inhabited continent, and the ant was restricted to coastal areas with summer rainfall (Fig. 13a). Low summer temperatures restricted the potential to marginal relative abundance south of the New South Wales-Victorian border. The addition of irrigation enabled populations of the ant to grow over most of the continent (Fig. 13b).

Fig. 13.

(Bottom) CLIMEX growth index (——), temperature index (- - -), and moisture index (····) for S. invicta at (a) Brisbane, (b) Sydney, (c) Perth, and (d) Perth with irrigation. (Top) Rainfall is shown by histograms and maximum and minimum temperature by dotted and solid lines, respectively.

The potential barriers posed to the ant's establishment by the length of the growing season in southern areas were studied. The ants could produce a generation of alates in the first year (PDD = 840) everywhere in Australia except the extreme southeast and high altitude areas. Milder summer temperatures reduced the potential for population growth and production of alates in much of the temperate areas compared with the southeast United States. However, as in California, there is little indication of mortality in winter, so it is likely that colonies will grow slowly but persist in southern areas.

Extreme low winter temperatures were only limiting in the high altitude parts of the Snowy Mountains and in Tasmania, reflecting the much milder winters in Australia than in the United States. The daily average winter temperature cold stress model 2 increased the severity of cold stress in these areas and extended the area affected slightly.

To remove any doubt about the ability of the fire ant to overwinter in southern Australia, a direct comparison was made between the winter temperatures in Canberra with locations in the United States. The best match (CMI = 0.92) was found to be with Tupelo, in northern Mississippi, which is renown for its fire ant infestations within the current quarantine zone. This confirmed that winter temperatures are only likely to exclude the fire ant from the coldest habitats in the high altitude or high latitude areas in Australia. There were 475 DD for development, so there was limited growth potential in summer months.

Finally, the continental risk from permanent establishment of S. invicta is indicated by the EI, with and without irrigation. The potential distribution closely matched that of the area that was suitable for population growth, excluding only the small areas limited by cold stress in the presence of irrigation. A sensitivity analysis of the model to temperature was carried out. With the assumption that the length of the growing season would not constrain persistence, the effect of changes of ±2°C was to increase or decrease the TIA coupled with changes in the MIA in the opposite direction because of increased evaporation. An increase in temperature increased the values of the EI in southern regions slightly: Melbourne EI increased from 4 to 6, and Sydney (Bankstown) EI increased from 17 to 22, whereas in Brisbane (Archerfield), the extrathermal accumulation was offset by increased evaporation. Reduced temperature was offset by increased available moisture and so had little effect in most areas in Australia.

New Zealand.  

The likely limits of the distribution in New Zealand were then studied. The results show the extent of the potential geographical distribution and relative abundance of the fire ant in relation to climate, as indicated by the EI values (Fig. 14a).

Fig. 14.

CLIMEX (a) ecoclimatic index and (b) cold stress for S. invicta in New Zealand.

Excessive moisture, aridity, and high temperatures were not apparent in New Zealand, so it is likely that a combination of mild summer temperatures and cold stress from the low winter temperatures in the highlands of the South Island (Fig. 14b) will determine the potential success of the ant. The potential for colony growth in summer was limited by the low maximum temperatures, with the GI only reaching ≈10 and ≈5 in the northern parts of the North and South Islands, respectively, associated with up to annual totals of 400 and 200 DD, respectively. Auckland had a GI of 8 and 270 DD above 17°C. Degree-day accumulation was very sensitive to the threshold temperature, with only 27 DD above 21°C and 383 DD above 16°C. Allowing for a 2°C increase in temperature associated with urban infrastructure or natural thermal sinks such as rocks, there was still only 526 DD above 17°C, but the EI increased to 16.

Results from representative sites in the United States, Australia, and New Zealand are compared in Table 2. They emphasis the large regional differences to be expected in the patterns of infestation of the fire ant.


The current analyses project the potential ranges of S. invicta in the United States, based on fitting of CLIMEX parameters to accommodate populations at the most extreme rural and urban conditions so far reported for the ant in the United States. Projections of areas at risk are conservative for the reasons given above. Simulated trends in seasonal population growth rates were broadly consistent with the field data (Markin and Dillier 1971, Markin et al. 1973, Morrill 1974, Callcott et al. 2000). They are sufficient for present purposes, given that the available field observations were approximate, long-term average meteorological data were used, and the objective was a regional-scale risk assessment. In the northeast of the United States, limited heat accumulation in summer prevented exploitation of the favorable moisture regimen. Our model restricted the growth of fire ant populations until it reached zero when the annual precipitation fell below ≈400 mm. Irrigation greatly increased the area at risk of colonization in the west. Low heat accumulation in summer combined with low winter mortality in the northwest indicated a likely pattern of slow colony growth and high winter survival.

CLIMEX contributed to the interpretation of the limiting effects of low temperature mortality by allowing alternative hypotheses on the effects of cold on fire ants to be tested. We were able to discriminate between the effects of minimum and average daily temperatures in winter. CS model 2 seemed, on basic principles, to be more realistic than model 1, but the results agreed less with the current northwest portion of the quarantine boundary, probably because a lack of moisture for population growth (Fig. 6b) prevents the ants from displaying their ability to withstand the cold in that region. Each of the potential CS mechanisms is likely to be triggered in some region of the world. Specific threshold temperatures per se have little direct biological meaning because cold has its effect by combining a given temperature with a corresponding duration of exposure. Therefore any particular CS threshold is simply an indicator above and below which stress accumulates at slower or faster rates, respectively. The results of the CS analyses indicate that there is the potential for northern expansion of the range in the United States, consistent with the more liberal projections of (Korzukhin et al. 2001). There are a number of uninfested areas abutting the boundary, which are as warm in winter as areas that are currently infested. These include southern and eastern Oklahoma and the northern half of Arkansas.

Rigorous validation of the CS models was not possible because comparison of the results with that portion of the northern border of the quarantine zone (i.e., non-Tennessee) that was not used in the fitting process (Korzukhin et al. 2001) has three weaknesses. First, the final limits to colonization have not yet been reached. Second, such data are not independent and therefore provide a weak test comprised of local interpolation and extrapolation to correlated habitats. Third, aridity restricts the range substantially in the northwest. The results of the two CS hypotheses have implications for future projections of the geographical range of the fire ant. Differences between model projections were up to 200 km.

In this study, it was immediately evident from the riverine distribution of S. invicta in South America (Mescher et al. 2003) that the ant is not limited by climate on that continent. We were unable to define a CLIMEX model that would explain the distribution, reinforcing the statement (Sutherst and Maywald 1985) that internal inconsistencies in CLIMEX model parameter values can flag when other factors limit the range of a species. The roles of soil disturbance and annual inundation, and interactions with other species of ants, described above, need further clarification.

The desirability of using a simplified, process-based model, given the deficiencies in available life cycle data, was recognized by (Korzukhin et al. 2001). Despite using such a mechanistic model, those authors were forced to infer threshold colony alate production rates from selected areas on the invasion front of the ant "to adjust the calculated range to the furthermost points in the present distribution." This inferential process is similar to that used in CLIMEX, but it overrides empirically determined parameter values in a life cycle model. The Korzukhin et al. (2001) threshold alate production figures for persistence are specific to the northern boundary of the United States quarantine area. The liberalization of the possible risk category by Morrison et al. (2004) brings their revised projections more in line with those above for Australia (Anonymous 2001), New Zealand (Sutherst 2002), and the current analyses. In contrast with their arbitrary and subjective risk categories, the CLIMEX EI is a site-specific indicator of likely persistence and relative abundance, which must exceed zero to indicate persistence.

CLIMEX extends the findings from other modeling studies in its treatment of moisture as a determinant of colony growth. A threshold of 510 mm of annual rainfall was used (Korzukhin et al. 2001) to delineate the range, but that makes no provision for the effects of suboptimal seasonal moisture values on the performance of colonies within the infested area. Thus their thermal model overestimated production of alates in the western part of the quarantine zone. In this analysis, we included weekly soil moisture rather than a threshold annual amount of rainfall, so the seasonal availability of moisture for growth was accommodated.

The use of long-term, average climatic data implicitly incorporates the biogeographical effects of annual variation in climate. Use of such data for global comparisons also carries the underlying assumption that variances at each location are equal in different regions of the world. That is not true for either temperature (Sutherst and Maywald 2005) or rainfall (Peel et al. 2004). This problem is pertinent to the current comparison of the United States with Oceania because there is much higher interannual variation in extreme winter temperatures in the southeast United States than occurs in Australia. The result is that the absolute minimum temperature recorded at a site in the United States, with the same average minimum temperature as a site in Australia, will be much lower than in Australia. These Australian locations have a less restricting climate than their United States counterparts (Sutherst et al. 1995b). We await inclusion of measures of variation in global databases of climatic averages.

Data on geographical distributions are qualitative and opportunistic, being compiled over years with variable climates from biased sampling efforts. In attempting to understand the climatic factors limiting geographical distributions, it would be helpful to have access to better quality data. This will justify a more rigorous statistical fitting process for CLIMEX parameter values, at the expense of the current heuristic and visual iterative fitting process. This process has nevertheless been found to be robust because each variable is fitted individually and differences are usually readily apparent because of geographical heterogeneity. Such errors of approximation are much less serious that the systemic errors associated with statistical and rule-based models. Fortunately, gradients of abundance along climatic gradients are fairly obvious from anecdotal observations. This makes it possible to parameterize coarse-scale models such as CLIMEX. Comparative studies of seasonal phenology and relative abundance, along transects up to the edge of the range (Markin and Dillier 1971, Bourne et al. 1988, Nealis et al. 1999) are particularly useful.

This analysis highlights the large areas of the world where the climate is suitable for establishment of the fire ant. The results agree with those from the liberalized criteria of Morrison et al. (2004). The extent of the area suggests that the species should be given the highest possible weighting when evaluating risks to other countries. Much of Africa and Asia are at risk, and it would be prudent to study the ability of the local fauna to prevent establishment of the fire ant. Similarly, in South America, there is a large area at risk if the ant is introduced into areas that are not protected by other fauna or unsuitable habitat. Acceleration of movements of goods contaminated with fire ants may overcome the biotic resistance currently containing them. If other ant species are indeed constraining the ant's geographical distribution, such species may pose as much of a risk to other continents as that posed by the fire ant. They should be included in any risk assessments of exotic invasive ant species.

The projected distribution of S. invicta in Australia indicates that the ant has the potential to impact widely on agriculture and the native fauna of Australia. The ecology of the ant is likely to be quite different in Oceania and North America, with the potential for small, slow-growing colonies to persist in Oceania and even to gradually develop into large infestations. Consider the small colony of fire ants with some alates that was reported in the edge of a lawn in Auckland, New Zealand. With only 270 DD above 17°C each year, CLIMEX identifies the area as being at risk (EI = 8) from slow-growing but persistent populations. Because the nest had three alate queens in February, the colony must have overwintered the previous year. These observations also lend credence to the lower temperature threshold used in this analysis, as well as the absence of any fixed annual degree-day threshold for persistence in this environment. They highlight the great difference in the population dynamics of the ant in different regions.

It will be useful to calibrate the human health, economic, and environmental damage caused by different densities of the fire ant in different habitats with the CLIMEX EI values. This will enable injury and economic thresholds to be estimated for the species in relation to the suitability of the climate in different regions, as has been done for another pest species, the Queensland fruit fly, Bactrocera (Dacus) tryoni (Froggatt) (Sutherst et al. 2000).

In summary, CLIMEX extends the earlier modeling of the potential geographical distribution of the fire ant by incorporating a hydrological response, by comparing the effects of different hypotheses on the projected effects of low temperatures, by computing site-specific combinations of growth and stress to estimate limits to persistence, and by providing a vehicle for synthesizing a species response to climate and making global projections of areas at risk of invasion. This analysis shows how information on the geographical distribution of a species can be used to infer its seasonal and spatial responses to climate and to make useful statements about its potential to occupy other environments in the absence of nonclimatic barriers. In this case, the fire ant has been shown to pose a threat to vast areas of the world, and it warrants high priority by biosecurity agencies responsible for prevention of nonindigenous invasions.


The Standing Committee on Agriculture and Resource Management (SCARM) contributed funding for this study. W. Bottomley assisted with the illustrations, and J. Goolsby gave helpful advice on the biology of the fire ant. R. Thornton and A. Flynn of MAF New Zealand provided valuable guidance on the New Zealand analysis and information on the Auckland infestation, respectively. Plant Health Australia and the National Centre for Disease Investigation, MAF, New Zealand, contributed funding for the analysis. The Climate Research Unit at Norwich University, UK, made the global climate surfaces available through the collaborative global change research program of the IGBP. NIWA provided the climatic data for the original analysis for New Zealand.


View Abstract