The urban fire ant paradox: native fire ants persist in an urban refuge while invasive fire ants dominate natural habitats
- First Online:
- Cite this article as:
- Plowes, R.M., Dunn, J.G. & Gilbert, L.E. Biol Invasions (2007) 9: 825. doi:10.1007/s10530-006-9084-7
- 224 Views
In contrast to the widespread extirpation of native fire ants (Solenopsis geminata) across southern US following the invasion by imported red fire ants (S. invicta), some residential areas of Austin form unexpected refuges for native fire ants. Ironically, these urban environments provide refuges for the native fire ants while adjacent natural habitats have been overrun by invasive fire ants. Resistance to invasive fire ants in these urban areas occurs mainly in older residential properties constructed prior to the S. invicta invasion, while more recent construction has allowed establishment by S. invicta. The invasive ability of S. invicta is often attributed to escape from parasitoids and efficient dispersal of polygyne multiple queen colonies. Here we also show the importance of landscape parameters in the invasion process, where low levels of disturbance and continuous plant cover in older residential areas form possible barriers to colonization. Dense leaf cover (high NDVI) was also found to be associated with native ant refuges. Long term residential land ownership may have resulted in lower recent disturbance levels and increased plant cover that support refuges of native fire ants.
KeywordsSolenopsisinvictaSolenopsis geminataFire antInvasive speciesInvasibilityNDVIDisturbancePolygyneUrban landscape
Contrary to the expected associations of invasive species with anthropogenically disturbed habitats and native species with natural habitats, a striking contrast was observed in Austin, Texas, where the invasive fire ant, Solenopsis invicta, has become established in natural habitats while nearby residential areas form refuges for the native fire ant, S. geminata. The red imported fire ant, S. invicta, negatively impacts the US economy by around $ 6 billion per year (Drees and Lard 2006), while also disrupting native biodiversity and ecosystem services (Vinson 1997). Like many invasive species, S. invicta has leveraged its escape from top-down controls to spread region-wide (Porter et al. 1992, 1997), but the pathway and outcome of its impact at local scales may depend on a variety of local biotic and abiotic factors and history of anthropogenic disturbance (Holway et al. 2002, DiGirolamo and Fox 2006). Almost everywhere it has spread, S. invicta has altered the native arthropod community and decimated populations of its congener, S. geminata, (Roe 1973, Porter et al. 1988, Porter and Savignano 1990, Wojcik 1994). It is informative to consider under what circumstances native species can persist as this may allow insights for biocontrol and management of invasives, and a better understanding of which habitats and communities are at risk.
Initial surveys indicated that S. geminata was associated with an older neighborhood with extensive oak tree cover and well-tended yards. In immediately adjacent schools, parkland and commercial areas, the typical pattern of extremely dominant S. invicta was found. In contrast to Tarrytown, a similarly aged residential area in east Austin was dominated by S. invicta, which begged the questions why Tarrytown remained as a S. geminata holdout, and whether other neighborhoods were also relatively free from S. invicta. The reasons for the existence of this refuge may be a combination of biotic, environmental and landscape factors (Crawley 1987; With 2002), and this study examines some differences between the invaded and non-invaded areas based on landuse, soil types and vegetative cover. We interpret these habitat factors with respect to biological differences between S. geminata and S. invicta, in particular the preponderance of polygyny in S. invicta in this community (Porter et al. 1991).
Polygyne colonies occupy multiple mounds within their habitats and expand spatially through a budding clonal-growth process leading to sub-colonies consisting of several queens and many workers (Porter et al. 1988; Tsutsui and Suarez 2003). In central Texas, S. invicta colonies are usually polygyne (Porter et al. 1991), while S. geminata colonies are more frequently monogyne. Networks of polygyne colonies show low intraspecific aggression and increased interspecific competitive dominance (Morel et al. 1990, Vander Meer et al. 1990) allowing large interconnected colonies to form across wide fronts (Porter et al. 1988; Macom and Porter 1996). The clonal spread of polygyne colonies is expected to be limited by resource gradients or significant landscape barriers such as roads or lot boundaries in residential areas, while alates of either polygyne or monogyne forms could disperse to remote sites beyond such boundaries. However polygyne queens are smaller than monogyne counterparts and may have less success in claustrally founding colonies (Vargo and Fletcher 1989). Therefore polygyne colonies are expected to dominate large contiguous areas but are less likely to occur in smaller remote sites such as residential lots.
A clear example of clonal spread in polygyne S. invicta occurred at Brackenridge Field Laboratory in Austin during the 1980’s (Porter et al. 1988; Vargo and Porter 1989) with a gradual advance across a broad front (Fig. 1), while there were few cases of founding ahead of this front by alate queens despite continual propagule pressure. A similar case arises in south Texas where S. invicta failed to establish across the Rio Grande river barrier for many years, and only recently has penetrated to the south of the river (Sanchez-Pena et al. 2005). Such barriers also affect other polygyne ant species that rely on budding for colony reproduction, and depend on human-mediated dispersal to colonize new habitats (Suarez et al. 2001).
Elevated disturbance levels are often associated with ant invasibility (Bolger et al. 2000; Holway et al. 2002) and S. invicta has strong associations with anthropogenic disturbance and open habitats (Tschinkel 1988). Several mechanisms may contribute to these associations including disruption of native ant communities, directed dispersal by founding alates (DeHeer et al. 1999), the use of open sites for brood thermo-regulation (Porter 1988) and higher foraging rates in open areas (Porter and Tschinkel 1987). We hypothesize that in the Austin study area, house construction in new neighborhoods has provided opportunities for invasion by S. invicta through soil disturbance and loss of plant cover while older neighborhoods constructed prior to the invasion have high levels of resistance to S. invicta and form refuges for S. geminata.
Disturbance intensity and timing can affect environmental correlates of insect distributions such as microclimates, soil types and vegetation composition and structure (With 2002). Causal relationships between vegetative cover and insect communities may be uncertain, but differences have been shown between S. invicta and S. geminata in response to habitat structure (Tschinkel 1988). Plant cover may be quantified over large landscapes and time periods using remote sensing methods, and we propose to use the Normalized Difference Vegetation Index (NDVI) to provide repeatable, standardized measures of plant cover rather than broad categorical descriptors.
Another key environmental factor for many arthropod communities is soil type. Soils may indirectly affect these ants through associations of plants and arthropods that specialize on certain soil types. For example, Tschinkel (1988) proposed that the hypogeic Solenopsis molesta may impact S. invicta by predation of brood. Additionally, soil properties may have a direct role in niche determination based on structural requirements for mound construction and foraging. Both Solenopsis species construct mounds, which in the case of S. invicta may be sizeable structures, while S. geminata more often has small mounds adjacent to rocks or trees. Excess clay may hinder mound construction and durability since clay soils are often expansive and poorly drained, although clay may improve the integrity of shallow tunnels used for foraging. It is not clear a priori which soils are most likely to be associated with invasibility and the situation may be confounded by neighborhood effects of landuse and period of construction.
The aim of this study was to determine the landuse and environmental correlates of S. geminata residential refuges as potential explanatory variables for resistance to S. invicta invasion. Specifically, we examined landuse differences in (1) year of construction, (2) lot size and (3) lot valuation while environmental differences were estimated by (4) plant cover (NDVI) and (5) soil properties. These landscape properties of refuge areas were interpreted in light of known ecological differences between S. invicta and S. geminata, in particular the prevalence of polygyny. Since both S. invicta and S. geminata have become invasive pests in many countries, this study contributes to understanding the mechanisms of invasion and persistence by these species through the use of remote-sensing and GIS applications.
Site and survey method
Landuse classes and attributes were determined from the City of Austin GIS database, TCAD2000. The classes used in this study were Residential, Open Space (greenbelts and parks), Undeveloped (vacant lots), Civic (schools, institutes, hospitals), Apartments, Roads and Commercial (including offices and industrial). Each lot had attributes of year of construction, lot size, and 2000 market value. Since samples were collected along road edges, the samples were assigned attributes of the closest lot. No data were available concerning which lots had recent building reconstruction activity.
We estimated vegetative cover from Landsat images using the NDVI which provides a relative measure of photosynthetically active plant cover for each pixel. NDVI can be readily estimated at landscape scales using historic chronosequences to allow averaging across several scenes for different seasons. Similar NDVI values do not necessarily reflect the same underlying causes since plant cover at any site depends on combinations of factors such as soil type, nutrients, soil moisture and plant species. High NDVI values may occur in cooler, moister and more shady sites as found in well-established and watered properties with extensive tree cover and low rates of disturbance. Some habitats show strong seasonal differences in NDVI, depending on plant types and water availability. Urban landscapes include several components that were expected to have distinct NDVI signatures such as watered lawns, undisturbed green belts, commercial areas and irrigated parklands.
NDVI utilizes the difference in reflectance properties of vegetation in the red and near infrared bands of satellite images to create an image showing vegetative vigor, where NDVI = (NIR − Red)/(NIR + Red) = (band3 − band4)/(band3 + band4) for Landsat images and pixel values range from 0 to 256. We used an alternate formulation of NDVI = 100*(Band3 –Band4)/(Band3 + Band4 + 1), so that NDVI ranged between ±100, and we added 1 to the denominator to avoid divide-by-zero errors.
To account for seasonality, we calculated NDVI during leaf-on and leaf-off periods, and took averages across several dates to minimize within-day sources of error. NDVI (leaf-on) was the average of days 189, 202 and 248 (7/8/02, 7/21/01 and 9/4/00), while NDVI (leaf-off) was the average of days 74 and 80 (3/15/03 and 3/21/03). Each ant record was assigned NDVI (leaf-on) and (leaf-off) values. Later during analysis we found that NDVI (leaf-off) values provided similar results to NDVI (leaf-on), and so we only report NDVI (leaf-on) results.
To assess NDVI patterns for each landuse class within each neighborhood, the NDVI and landuse layers were intersected to provide new composite layers. Each composite layer was clipped to the extents of the neighborhood blocks. This provided an image containing numerous smaller polygon elements each with attributes of neighborhood, NDVI, and landuse parameters that were exported to a database table to explore correlations between NDVI and landuse.
Soil clay content
Soil maps were imported from the SSURGO survey of Travis County published by USDA Natural Resources Conservation Service, 2006. Average clay content percentage was estimated from ranges given in the SSURGO database and these attributes were added to all ant records. We also generated a composite image of soil types by landuse categories for correlation analysis of soil and landuse.
For each neighborhood we used a logistic regression test to examine the relationship between the presence of S. geminata and construction period, market value, lot size, NDVI (leaf-on) and clay content. The construction period was coded as a binary categorical variable based on construction before or after 1980. We used manual stepwise regression, where we successively added the next variable at each step if it yielded a significant improvement to the model using the G-statistic for likelihood ratios (Hillborn and Mangel 1997).
We tested correlations between year of construction and NDVI or clay content using data from the composite images clipped to each neighborhood. We expected some degree of correlation between plant cover (NDVI) and construction period, but NDVI could also have a strong component driven by soils or native plant communities. In addition, we used a Pearson test to describe correlations between market value and year of construction, and between market value and lot size.
The association between soils and Solenopsis species was tested by comparing their relative abundances in each soil type. To avoid pseudo-replication, we only generated one data point for each soil type in each neighborhood, and estimated the percentage of samples with S. geminata for each soil type. We only included soil types for which there were more than 10 ant records. We used Pearson’s correlation test to see how S. geminata abundance varied with clay content.
Solenopsis geminata associations with landuse and environmental variables
Period of construction
Plant cover (NDVI)
During winter (leaf-off), all sites are drier and there is less difference between S. geminata and S. invicta sites (Fig. 5). Areas that remain greener in winter (such as Civic and Open Space) may include watered lawns, winter grassland and riparian sites with evergreen tree cover, or deciduous woodlands with evergreen understories. The high average NDVI values in Open Space areas may be consequence of extensive areas with continuous plant cover containing few roads or buildings, rather than an indication of higher plant densities within each pixel.
Distribution of S. geminata and S. invicta in non-residential properties
Open Space and undeveloped
Commercial and industrial
% with invicta
We also considered other soil properties in the SSURGO dataset, but found no association between Solenopsis and soil pH or calcium carbonate content. Clay content was correlated with cation exchange capacity, organic content and water retention, such that soils with high clay content could potentially support dense plant cover (high NDVI). However there was no consistent relationship between NDVI and clay content (Pearson r = 0.70 East, −0.38 Northwest, 0.15 South, 0.08 Tarrytown).
All non-residential areas have been thoroughly invaded by S. invicta (Table 2) with particularly high rates of infection in Civic areas. These properties are much larger than residential lots with few roads and other features that could form barriers to spread by polygyne S. invicta. Sites in the Open Space, Undeveloped and Civic landuse categories often show increases in winter plant cover (NDVI leaf-off), in contrast to other landuse categories (Fig. 5). Civic areas include parks and school grounds that may be irrigated through winter, and these sites show particularly high levels of infection by S. invicta. Open Space and Undeveloped areas may support growth of winter and early spring grass during the leaf-off period, resulting in higher NDVI.
Older residential properties in Austin provide rare and unexpected refuges for native fire ants, compared to the pervasive spread of S. invicta across the southern United States. Refuges of native S. geminata were found mainly in residential properties constructed prior to the S. invicta invasion around 1980, particularly in sites with abundant plant cover (high NDVI), while adjacent natural habitats were overrun by S. invicta. The explanation for these urban refuges appears to be the low disturbance rates in older residential areas that impedes colonization by the common polygyne form of S. invicta, together with possible human intervention of new S. invicta mounds. Old neighborhoods occupied predominately by monogyne S. geminata have undergone low levels of disturbance for many decades, allowing S. geminata to persist at high densities and recolonize most new sites within the neighborhood. Conversely, we see a failure by either monogyne or polygyne forms of S. invicta to colonize older residential sites by either winged queens or clonal spread.
Despite continuous pressure from clonal spread by many adjacent populations of polygyne S. invicta, older neighborhoods have resisted invasion along several contact zones. Furthermore, although dispersal by winged S. invicta queens must provide intense propagule pressure, they too have had low colonization success. We suggest that S. invicta cannot colonize these older neighborhoods either along contact zones or by dispersal to new interior sites because of several potential reasons: (1) a lack of suitable sites due to low disturbance levels, (2) barriers to clonal expansion in small sized plots, (3) mortality following founding through resource mismatch, (4) competition with native ants or (5) landowner action.
The low rate of recent disturbance in older residential sites may account for the extremely strong effect of construction period on invasibility. Construction activities have several possible impacts on ant communities that may render them more invasible by S. invicta including soil disturbance, disruption of native arthropod fauna and decrease in plant cover. Although S. geminata still dominates properties constructed prior to 1980, the East block has high levels of S. invicta across all construction ages. We suggest that since the East block comprises a low-cost neighborhood, these yards may have sparse canopies (low NDVI) with pronounced seasonal changes, disturbances may be more frequent and landowner response to S. invicta may be lower.
In non-residential areas dominated by S. invicta, the larger lot sizes of schools, parks and Open Space areas allow establishment of polygyne networks that can track resource fluctuations within a patch without encountering barriers or intra-specific conflict. Polygyne clonal expansion and seasonal movement in small residential lots could be limited by landscape barriers such as roads or property boundaries. But since S. invicta is found in low-cost neighborhoods such as East Austin which have similar road and plot configurations these barriers may not be effective against clonal spread when there is low density plant cover, exposed ground, or less garden care (Feagan and Ripmeester 1999).
In rural areas, there are many sites that have low disturbance histories with dense plant cover, yet these have been invaded by S. invicta. Rural and urban habitats differ in respect of plot sizes, watering and pest management, all of which could inhibit S. invicta colonization. Chemicals applied to lawns and gardens could both encourage S. invicta takeover by removing native ants, and conversely prevent its spread when new S. invicta mounds are treated. Chemical pest-control was not assessed in this study, but we assume that the extent of treatment is related to neighborhood wealth (Robbins et al. 2001; Steer and Grey 2006).
Dense plant cover, as measured by NDVI, was associated with S. geminata refuges in residential sites and the effect occurred independently of construction period. NDVI is expected to be higher at sites with low levels of disturbance and with continuous tree canopy cover, offering few opportunities for colonization by S. invicta and poor conditions for brood production (Porter 1988). In Florida S. invicta was found in open canopy, heavily disturbed and pond-side sites, while S. geminata was common in mesic forested sites and where the water table was deep (Tschinkel 1988). Similarly, in rural areas of central Texas, S. invicta occurs most frequently in disturbed areas, around pastures, riparian zones and at the edges of woodland. S. geminata is now rare in rural Travis County, and little is known about its present habitat associations.
Areas with high clay contents in soils were generally dominated by S. geminata but clay content did not have a significant additional effect in logistic regression models that first considered lot age and NDVI. Outside the study area S. invicta occurs in all soil types, and the apparent correlation seen in residential areas may be an artefact of neighborhood development patterns.
This study suggests that invasibility depends on local combinations of biotic and abiotic factors, whereby refuges of native species may occur despite persistent pressure from invasive species. Even when an invasive species appears to have escaped its native biotic controls on a region-wide scale, it may encounter local conditions that inhibit its spread. These conditions are generally expected to occur near the edges of its expansion front where environmental stresses limit further expansion but here we find an exception to this pattern with a refuge of uninvaded territory embedded in the invasion zone. Residential areas in Austin contain otherwise suitable conditions and resources for S. invicta invasion, but anthropogenic controls on disturbance and landscape structure appear to have slowed the invasion.
Thus, we propose that native fire ants persist in some urban zones of Austin because of a stable mosaic of landscape structure and management regimes. If our view is correct, then this urban refuge for native ants will continue to exist as long as residents of the area continue their present gardening methods and new construction activities are limited in size and frequency. Furthermore, assuming similar circumstances occur elsewhere, any neighborhoods of cities in southern United States developed prior to invasion by imported fire ants should show a similar pattern if the prevalent form of S. invicta is polygyne.
Solenopsis invicta populations are mainly polygyne, S. geminata are usually monogyne in the study area.
Polygyne S. invicta colonies spread mainly by clonal expansion into large and disturbed areas with open canopies.
Non-residential sites and low-cost neighborhoods typically have low plant cover (low NDVI) arising from open canopies and high disturbance levels. These sites have high infection rates by S. invicta.
Solenopsis geminata refuges occur on older residential lots with low levels of recent disturbance.
The refuge areas contain many small residential lots where roads and boundaries may form barriers to the spread of polygyne S. invicta.
Refuge areas have dense plant cover (high NDVI) in well-established yards and gardens resulting in moderated microclimates.
Solenopsis geminata colonies are less conspicuous and unlikely to be treated by landowners while incipient populations of S. invicta are probably subject to pest management in wealthier neighborhoods.
This study has described the paradox of human-settled areas providing a refuge for a native species while nearby natural habitats have been overwhelmed by an invasive congener. Often we consider anthropogenic disturbances to be linked to invasibility, and generally S. invicta follows this pattern of association with human landuse. Here we see a converse pattern where long term residential land ownership may have resulted in lower recent disturbance levels and increased plant cover that support refuges of native fire ants.
This study’s conclusions were based on an accumulation of knowledge about the fire ant invasion at Brackenridge Field Laboratory and in Texas which, in turn was made possible by many years of support from various agencies and foundations as acknowledged in the papers referenced herein. Recent funding was provided by the State of Texas fire ant initiative (FARMAAC), Helen C. Kleberg and Robert J. Kleberg Foundation, USDA and the Lee & Ramona Bass Foundation. E. Le Brun and N. Plowes provided useful comments on the manuscript. Brackenridge Field Laboratory technicians assisted with surveys, data collection and identification work, especially H. Allard. O.M. Gilbert provided pilot study information that stimulated this research.