Dynamic expansion in recently introduced populations of fire ant parasitoids (Diptera: Phoridae)
- First Online:
- Cite this article as:
- LeBrun, E.G., Plowes, R.M. & Gilbert, L.E. Biol Invasions (2008) 10: 989. doi:10.1007/s10530-007-9177-y
Combating invasive species requires a detailed, mechanistic understanding of the manner and speed with which organisms expand their ranges. Biological control efforts provide an opportunity to study the process of species invasions and range expansions under known initial conditions. This study examines the rate, pattern and mechanisms of spread for two populations of the biological control agent Pseudacteon tricuspis, phorid-fly parasitoids of imported fire ants. We employ a trap-based survey method that detects phorid flies in low-density populations, and provides data on abundance. This technique allows us to differentiate between continuous population spread and effective long-distance dispersal and to examine density gradients of phorid flies across the expanding population front. We find that occupied sites in front of the leading edge of continuous populations were common; forming small populations we refer to as satellite populations. Satellite populations are tens of kilometers from the nearest possible source. Wind governs the dynamics of spread in these two central Texas populations. Population edges expanding with the wind exhibited a higher frequency of effective long-distance dispersal than did populations expanding into the wind. This enhanced effective long-distance dispersal rate translated into a five times faster rate of spread for population edges traveling with the wind. This planned invasion shares many characteristics in common with unplanned species invasions including: protracted establishment phase during which densities were below detection thresholds, and slow initial spread immediately after establishment followed by rapid, accelerating spread rates as population sizes grew.
KeywordsBiological controlInvasive speciesLong-distance dispersalPopulation spreadPseudacteon tricuspisRange expansionSolenopsis invictaWind dispersal
The rate and mechanisms by which populations spread is fundamental to their ability to persist under changing environmental conditions, evolve, and invade new environments (Suarez et al. 2001; Travis and Dytham 2002; Parmesan 2006). The process of population spread results from the interaction of several factors including individual characteristics such as dispersal capabilities, life history parameters such as generation time, and population level effects like the frequency of long-distance dispersal events, and intrinsic growth rates (Hastings et al. 2005) and abiotic factors such as environmental gradients and landscape features. A detailed and predictive understanding of the mechanics of species spread is central to efforts such as controlling invasive species (Johnson et al. 2006). A necessary component of this effort is empirical case studies of newly established and spreading organisms from a variety of taxa and life histories. Biological control agents are particularly well suited to this effort because we typically know critical information such as their introduction date and location information that must be inferred in most other invading organisms (Ehler 1998).
This study examines the rate, pattern and mechanisms of spread for two central Texas populations of phorid fly parasitoids, Pseudacteon tricuspis. These flies were introduced as biological control agents for the red imported fire ant, Solenopsis invicta. There is much interest in characterizing the spread of invasive species (Bohrer et al. 2005; Hastings et al. 2005), and this “planned invasion” of phorid flies provides a unique opportunity to track and quantify their spread, and to look at associated phenomena such as long-distance dispersal, Allee effects and responses to regional environmental gradients.
These parasitoids, which are about 1.0 mm in length, attack adult worker ants from the air, injecting an egg into the thorax of the ant. Their larvae then develop and pupate in the head capsules of the infected worker (Consoli et al. 2001). Adult ant-parasitizing phorids apparently use olfaction to locate their hosts (Feener et al. 1996). Pseudacteon tricuspis is primarily attracted to disturbed ant nests and presumably use the large amounts of alarm pheromones released during nest defense as a host location cue (Orr et al. 1997). Both males and females of P. tricuspis are attracted to disturbed nests and mating occurs at these sites (Porter 1998).
The difficulty in detecting and measuring individual dispersal constitutes a central limitation in the study of expanding populations. This problem is particularly acute in the case of rare long-distance dispersal events, and these events are even more difficult to measure in small bodied, flying insects, like phorids, where marking individuals is not feasible. However, despite its rarity, long-distance dispersal can often be the primary determinant of population spread rates and has been implicated in the rapid and extensive spread of many invasive species (Clark et al. 2001; Neubert and Parker 2004; Nathan 2005).
Because of the difficulty of studying individual dispersal in phorid flies, investigators have measured the effective dispersal of Pseudacteon populations. A common theme has been an accelerating rate of spread in the first few years after release and establishment. In Florida, P. tricuspis populations spread, on average, 1.5 km in the first year after release, 15 km in the second year and up to 30 km/year in the third year (Porter et al. 2004). In comparison, populations of P. curvatus in Mississippi, Alabama, and Florida spread from 0.8 to 1.6 km the first year post-release (Graham et al. 2003a; Vazquez et al. 2006), and exhibited 5–44 km total spread by the end of the second year (Graham et al. 2003a; Thead et al. 2005). Given these large distances covered in short periods, it appears likely that long-distance dispersal is occurring in these systems and, given that these are small bodied, flying insects, we hypothesize that it is a wind-assisted phenomenon.
Previous studies of population spread in Pseudacteon have all relied on the observer intensive technique of disturbing nest mounds some distance from the release site and waiting a set time interval to see if phorid flies appear. Unless substantial effort is made to standardize sampling effort across sites, this technique provides largely presence–absence data. It is also labor intensive, subject to observer bias, and may not detect low-density populations (LeBrun et al. in prep.). Because of these limitations, no attempt has been made to distinguish between continuous population spread and effective long-distance dispersal or examine density gradients of phorid flies across the expanding population front.
In this study, we employ an improved, trap-based survey method that allows for the detection of Pseudacteon in very low-density populations, is less subject to observer bias, and provides quantitative data on abundance. We evaluate the current distribution of P. tricuspis in two central Texas populations. Using trap lines, we characterize the wave front of the expanding P. tricuspis populations. We assess the frequency, and magnitude of effective long-distance dispersal, and we test the role of wind in mediating these events and determining rates of population spread.
The limits, temporal, and spatial dynamics of phorid populations were characterized using sticky traps baited with live ants and freshly frozen ant colony fragments, hereafter referred to as midden. The traps were modified 14 × 14 × 5 cm food storage containers. Commercial sticky fly paper was hung 2 cm in from the side of the container by inserting 00 test tube stoppers into the corners and attaching the fly paper with push pins. The inside walls of the container were coated with Fluon™ to prevent ants from being able to climb out of the trap (design described in LeBrun et al. in prep.). Traps were placed in small holes dug into S. invicta mounds. Alarmed workers stream up the outside walls of the trap, fall inside, and cannot climb out, serving as a persistent signal for phorids attempting to find a host. To supplement these live workers and to ensure that all traps were sufficiently attractive to searching phorids, we added 2 tablespoons of freshly frozen midden to each trap. This frozen material provides a long-term attractive signal to phorids (Smith and Gilbert 2003) and simulates cues associated with natural fire ant colony midden. Traps were placed in the midmorning and collected the following day, for a total of 24 h of exposure. After being retrieved, the fly-paper strips were scored under a microscope for the number, species identity, and sex of all Pseudacteon caught. Occasionally traps were destroyed, damaged, or disappeared before being picked-up. These were excluded from all analyses.
To evaluate the role of the prevailing wind in determining the frequency of long-distance dispersal events, we sampled additional transects in August 2006 that ran into and with the prevailing wind, for a total of 5 transects in the central zone running with the wind and 5 running into the wind, and 3 in the Gulf Region running into the wind and 3 running with the wind. From March to September in both the central Texas and the gulf coast region the wind consistently comes from the southeast (Fig. 1). Following Porter et al. 2004, we calculated wind drift data during the period sampled, April–October, 2006 as the sum of the wind speeds for all phorid active hours that the wind blew towards each of 36 compass bearings. Phorid active hours were defined as daylight hours when the air temperature was above 21°C. We divided all transects running up to 45° from the axes of the prevailing wind direction as running with or into the wind. Transects with headings ranging from 270° to 360° were considered to be running with the wind and transects heading from 180° to 90° to be running into the wind. Trap sampling followed the same protocol as above.
To evaluate the dependence of long-distance dispersal events on wind direction, we compared the number of traps in front of the expanding edge of the Pseudacteon population that captured phorid flies for with the wind and into the wind transects. We defined the edge of the continuous Pseudacteon population to be the last point on a transect where the traps consistently caught phorids. Operationally, this edge was located at the trap preceding the first section with at least two empty traps in a row. An effort was made to trap for 30 km in front of the population edge. However, because this location was not known prior to sampling, and because in the Gulf Region the population edge was located from 14 to 28 km from the ocean, it was not possible to standardize, in advance, the amount of trapping in front of the population edge for the two directions. To account for this, transects running into and with the wind for the central population and gulf population were paired within region by the amount of distance sampled beyond the population edge. The data from the longer of the two transects was then truncated to the length of the shorter transect. This ensured equal sampling effort within regions for transects going into versus with the wind, and resulted in 136 total km sampled in front of the population edge both into and with the wind. For comparisons between the central versus Gulf Region, the same approach was applied. However, transects were paired across regions by bearing, and sampling effort was then equalized across region.
To examine how wind direction and frequency of long-distance dispersal effect rate of population spread, we resampled the transects running into and with the wind in the central zone during the month of November 2006. Amount of spread between the two samples was calculated as the straight-line distance between the edge of the continuous population front at time period 1 and 2. This was transformed into a per generation rate of spread. Generation time, the time interval for a fly to go from egg to reproductive adult, in P. tricuspis is approximately 35 days (Consoli 2001; Morrison et al. 1997; Porter et al. 1997).
In order to gain a more dynamic view of population expansion in this system, two transects running with the wind on the Western edge of the central Texas population near Fredricksburg, Texas were repeatedly sampled during August and September 2006. Transects were run using the methods described above once every 2 weeks. Two weeks represents half a generation for these parasitoids, so the interval ensured that we would detect any variance associated with a particular cohort. These data also provide a method for assessing the frequency of false negatives in transect samples. False negatives were evaluated using sites on transects where previous sampling revealed the presence of an established population. Subsequent samples of these points were scored as either positives, P. tricuspis detected, or false negatives, no P. tricuspis detected. A breeding population was considered to be established at a site if a single sample captured 3 or more phorids, or if phorid flies were captured at a site on two successive trapping intervals. In addition these data allow for an assessment of how often P. tricuspis captured on traps beyond the edge were or were not from a locally established population. Phorid flies were ascertained to be products of an established population if previous samples revealed an established population at the site or if the subsequent sample contained phorids. Where the previous and subsequent traps from that site did not capture P. tricuspis, phorids were deemed to not be the product of a locally established population.
All data associated with the occurrence of satellite populations and the density of P. tricuspis both behind the fronts and within the satellite populations were non-normally distributed. These data were analyzed with non-parametric statistics (Conover 1971). Pseudacteon tricuspis densities vary seasonally and with respect to distance from edge, so to compare phorid densities across ecoregions, while controlling for seasonal and distance effects, transects were paired by month and then the number of sites sampled behind the front was equalized between transects. For each transect, the number of P. tricuspis per trap for sites at least 6 km behind the front was averaged. These averages were compared across ecoregions using a Mann-Whitney test. To compare the density of P. tricuspis in satellite populations to the density of phorid flies at the edge of the continuous population, the average number of phorids per trap was calculated for each satellite population. These values were then averaged for all satellites that occurred on a transect. This per transect average of P. tricuspis density beyond the edge was then compared to the average number of P. tricuspis per trap for the edge and the first positive sample behind the edge. The average of these two samples was chosen because satellites were on average 2 sites long and, by definition, contained no samples without phorid flies. These measures were paired by transect and compared with a Wilcoxon Signed-Rank test. All data were analyzed with the statistical package JMP IN® (JMP®2000).
Satellite populations have lower average densities of P. tricuspis (2.6 ± 2.8 phorids per trap) than paired sites at and immediately behind the edge of the continuous population fronts (8.2 ± 10.0 phorids per trap) (Wilcoxon Sign-Rank Test: W+ = 21.5, n = 10, P = 0.03). As an indication of our ability to detect P. tricuspis in these low-density populations, the data from the repetitively sampled transects revealed approximately 15% (n = 29) of traps at sites known to contain established populations at or beyond the continuous population edge failed to capture P. tricuspis. In addition at sites beyond the continuous population front, phorid flies from approximately 9% (n = 34) of positive samples were not from locally, established populations.
There was no difference between the central and Gulf Regions in the frequency (Mann–Whitney Test: U = 37.5, n = 12, P = 0.85) or the average distance (Mann–Whitney Test: U = 123.5, n = 25, P = 0.30) that P. tricuspis was detected in front of the population edge, so data from both regions was combined. There was also no difference within the central Texas region in the average fly densities behind the front between with the wind and into the wind transects (Mann–Whitney Test: U = 43, n = 6, P = 0.58).
In both regions sampled, the into the wind population front exhibited a slow decline in P. tricuspis density from well behind the population edge towards the edge. This slow decline was followed by a precipitous decline to 0 at the population edge, with almost no P. tricuspis detected beyond this edge (Fig. 4). However, the population edges moving with the wind had a very different character. They exhibited a decline to zero at the edge, but beyond that point sites, even distant sites, commonly contained P. tricuspis (Fig. 4).
In the central Texas population, population edges traveling with the wind spread at a faster rate compared to edges traveling into the wind. In the central Texas population, the average radial distance that the population expanded was 2.4 ± 1.0 km per generation (n = 3) for population edges expanding into the prevailing winds and 11.1 ± 5.8 km per generation (n = 3) for population edges expanding in a direction with the prevailing winds.
Not surprisingly, given the difficulty of monitoring the movements of small insects across large areas, few attempts have been made to measure the frequency and magnitude of individual, long-distance dispersal for Pseudacteon species (but see: Morrison et al. 1999). This study provides the first demonstration of effective long-distance dispersal in an expanding phorid population and the first characterization of the satellite populations that result from this process. These populations were common in both of the regional populations sampled. On average, effective long-distance dispersal, the distance between a satellite population and the nearest upwind population, was 10 km, while the most isolated satellite was 29 km from the nearest upwind source. Successful efforts to establish populations of P. tricuspis have invariably required inputs of large numbers of individuals (Graham et al. 2003b; Porter et al. 2004; Plowes et al. in prep.), indicating that single gravid females are insufficient to establish populations. Remarkably, these tiny flies must disperse on a scale of tens of kilometers and do so in sufficient numbers to establish populations. Members of the extremely diverse and varied family, Phoridae, are common constituents of aerial plankton surveys (reviewed in Porter 2004), demonstrating the potential at the family level for long-distance dispersal. These surveys do not report the genera of phorids caught, so it is unclear how commonly Pseudacteon get into the upper atmosphere.
Data from transects sampled repetitively indicate that the vast majority of these satellites represent locally established, breeding populations and not dispersing individuals. Satellite populations are less dense than populations immediately behind the expanding edge. This likely arises from a combination of the lower propagule pressure these distant populations receive and, potentially, a more recent average establishment than edge populations. Some edge populations were once satellites that are in the process of coalescing with the expanding front, and are thus not recent colonization events. A result of the low density of satellite populations is that 15% of repeated samples taken at sites in front of the edge that were known to contain established populations were false negatives. Because densities rise rapidly behind the population front (Fig. 4), and two sites with no P. tricuspis in a row were required to define a population edge, these false negatives should not substantively alter our determination of the location of the population fronts. However, we likely overlooked nascent satellite populations.
Wind plays a dramatic role in governing the dynamics of spread in these two central Texas populations. During the time of year that P. tricuspis is active, wind drift is consistently from the SE towards the NW both in the central and in the Gulf Regions (Fig. 1). Correspondingly, population edges expanding with this prevailing wind, from between 270° to 360°, exhibited a much higher frequency of effective long-distance dispersal than did populations expanding into this wind, from between 90° and 180°. In the central Texas population, this enhanced effective long-distance dispersal rate translated into per generation rate of spread, five times faster than for population edges traveling with the wind. Within the central Texas population, into the wind transects lie within the Blackland Prairie Ecoregion and with the wind transects lie in the Edwards Plateau Ecoregion. However, there is no difference in phorid densities within the continuous population between with the wind and into the wind transects, indicating that these higher rates of spread are not the result of habitat differences influencing spread by enhancing local population densities. In contrast to this study, Porter et al. (2004) found no influence of prevailing wind on the spread of a P. tricuspis population in north-central Florida. This may result from the wind being much less directional in that system. In north-central Florida, the wind comes out of the ENE or WSW with about equal frequency and intensity, yielding no direction that would qualify as into the wind in the current study. In addition, because of the highly directional nature of the prevailing wind in Central Texas, the magnitude of annual wind drift (Fig. 1) is approximately double that of north-central Florida.
Short-range, directed dispersal of phorid flies occurs commonly as adults seek out hosts. Morrison et al. (1999) measured individual dispersal distances in a mixed population of four species of native Pseudacteon exploiting S. geminata. They found a maximum individual dispersal distance of 650 m (Morrison et al. 1999). However, this was close to the largest distance sampled. This measurement of short-distance dispersal provides an estimated maximum population spread rate of about 650 m per generation from short-distance dispersal alone. The into-the-wind rate of spread of 2.5 km per generation suggests that either P. tricuspis individuals are commonly moving much farther than 650 m, or both into the wind and with the wind rates of spread are dominated by longer-distance dispersal events, but with markedly different underlying dispersal kernels.
Flight activity in Pseudacteon is temperature dependent with phorid flies being largely inactive below 21°C (Porter et al. 2004). Thus most dispersal of individuals and in turn population spread occurs during times when daytime highs are above this temperature. In Central Texas, depending on elevation, from 6 to 7 months a year (March–November) have daily highs above this activity threshold. In addition, generation time in Pseudacteon is temperature dependent with egg to adult developmental intervals for P. tricuspis ranging from 33 to 42 days between 24 and 30°C (Morrison et al. 1997; Porter et al. 1997; Consoli et al. 2001; Folgarait et al. 2002). Given these ranges in activity windows and developmental rates, 4–6 P. tricuspis generations elapse during the active period of the year. This provides an estimate of current annual population spread rate in Central Texas of about 45–70 km per year with the wind, to the north and west, and 10–14 km per year into the wind, to the south and east. Supporting this estimate, the one with the wind population edge in Central Texas tracked throughout this time period in 2006 expanded 68 km. At this rate, P. tricuspis should reach the western edge of S. invicta’s range in Texas in 2–4 years and the Northern edge in 3–6 years (range data from: http://www.aphis.usda.gov/ppq/maps/fireant.pdf), provided increasingly cold (north) and dry (west) abiotic conditions do not slow or stop its spread.
This with the wind rate of spread is faster than rates reported in other studies examining recently established, introduced Pseudacteon populations. In Florida, 4 years post release, Porter et al. (2004) found the average dispersal rate for a P. tricuspis population to be 23 km per year. While studies of P. curvatus populations 2 years post release in Alabama and Mississippi, show spread rates of approximately 20 km per year (Graham et al. 2003a; Thead et al. 2005). However, a follow up study on the Florida P. tricuspis population 8 years after release reported spread rates up to 57 km per year (Pereira and Porter 2006), a rate that falls within the range estimated for the central Texas population reported here.
In multiple Pseudacteon populations during the first 1–3 years post establishment, investigators have reported an accelerating population expansion rate (Graham et al. 2003a; Porter et al. 2004). Given the existence of long-distance dispersal, as small populations grow, the number of propagules arriving at and beyond the edge should increase as a function of the population density at the core and the area it covers leading to acceleration in the rate of radial spread. This early acceleration will be amplified by any Allee effects that might be dampening the spread of very small populations (Taylor and Hastings 2005). Given a strongly leptokurtic propagule dispersal kernel, this acceleration phase can be prolonged (Kot et al. 1996), or indefinite if the tail of the dispersal kernel is exponentially unbounded (Clark et al. 2001). This counter intuitive prediction is realized in species invasions with relatively small potential ranges, such as the rice water weevil invasion of Japan, or in species invasions dominated by human-mediated, long-distance dispersal, such as the cheat grass invasion of the western USA (Shigesada et al. 1995). There is no available data on the length or shape of the tail of the dispersal kernel for Pseudacteon individuals. However, the distribution of effective, long-distance dispersal documented in this study (Fig. 3) suggests that the length of this tail is at least 50 km. In both the central Texas population reported here, and the Florida population (Pereira and Porter 2006), the rate at which populations have spread has continued to accelerate beyond this point.
Alternatively, this prolonged acceleration could be due to non-demographic processes. Perhaps, both the Florida and the central Texas population have expanded into more favorable habitats, leading to higher local densities and faster spread. However, in the central Texas population, the phorid population densities relatively near the release site on the south and east population edge are not different from densities on the rapidly expanding north and western edges. Or, these populations may have evolved enhanced dispersal abilities. Enhanced dispersal is predicted to be selected for in expanding populations (Travis and Dytham 2002), and its evolution has been documented in species including cane toads (Phillips et al. 2006) and two bush crickets (Simmons and Thomas 2004).
Despite the dominant role of prevailing wind in determining population spread, a comparison of the distribution maps of P. tricuspis in the central and Gulf Regions (Fig. 2) with the pattern of prevailing wind drift (Fig. 1), reveals that the prevailing wind does not explain all aspects of population spread. In the Gulf Region the shape of the polygon circumscribing the current P. tricuspis distribution reflects the pattern of wind drift, suggesting that in this coastal plain environment, wind is sufficient to explain the overall pattern of spread. In the central Texas population, the South-eastern sections of the population polygons, also in an environment with little topography, reflect the expectation arising from large amounts of passive wind transport and smaller magnitude active dispersal. However, the North-western section of the main central Texas polygon does not support simple wind-directed drift. There is too much spread to the south-west, a direction perpendicular to the prevailing wind. This divergence from the expectations based on wind drift is coincident with the transition from the flat, open environments of the Blackland Prairie ecoregion to the very complex landscape of the Balcones Canyonlands and Edwards Plateau (Fig. 2). This hilly region of limestone is characterized by sinuous ridges, bisected by steep sided canyons. In environments, such as this, with complex topography, the prevailing wind direction is less predictive of the wind conditions at ground level (Nathan et al. 2005).
From the data presented here and data on release efforts, a consistent picture is emerging of the growth and spread of invading P. tricuspis populations. This picture reflects several general observations of species invasions. A lag-time, commonly observed in species invasions (Liebhold and Tobin 2006), of between 1 and 3 years occurred between release and establishment in all attempts to establish populations of P. tricuspis in the central and gulf coast region. The density of flies remained below the threshold for detection using observer intensive mound-disturbance sampling during this period despite intensive attempts to recover wild-bred flies focused on the exact location of release. This lag occurred regardless of the number of original flies released. Following establishment, initial spread rates were slow and likely the result of local diffusion processes, arising from flies dispersing by actively seeking out mounds. After this period, the populations entered a phase of rapid and accelerating spread. This change appears to result from the transition from purely diffusive spread to spread by stratified diffusion (Shigesada et al. 1995), in which spread by diffusion is combined with effective long-distance dispersal, and subsequent infilling of the gaps between the main and satellite populations.
We thank John Dunn, Naomi Gebo, Jerod Romine, John Sprague and Phebe van der Meer for technical assistance in the field. Steven Gibson and Tonya Simmons provided laboratory support. Funding was provided by the State of Texas Fire Ant Initiative (FARMAAC), the Helen C. Kleberg and Robert J. Kleberg Foundation, and the Lee and Ramona Bass Foundation. We thank Richard Patrock and Patricia Folgarait for comments on the manuscript.