Associations of Fire Ant Phorids and Microhabitats

Abstract

We examined flight activity patterns for a guild of fire ant parasitoids in western Argentina in relationship to their host’s location (mound/foraging trail) and light condition (full sun/partial sun/full shade) at different scales, from the individually sampled mound to the full day’s summation for each species. We asked first whether taxa showed preferences among these conditions, and second, whether certain species and sexes might be found together more frequently than expected to by chance. All species, except the P. obtusus species complex, were significantly more likely to be found attacking ants at disturbed mounds than at paired foraging trails. The P. nocens complex and P. litoralis were more likely to be in the shade when temperatures were above the overall mean of the study (28.3°C), whereas others, such as the P. obtusus complex and P. tricuspis, were more likely to be in full sun under these same conditions. Our analyses indicated that a limited set of species, particularly P. nocens with P. litoralis, and males with female P. obtusus and P. tricuspis, were more likely to be found together than expected. We also found decreasing proportions of males with increasing time of analysis. We discuss the implications of host location, metereological conditions, and sex ratios in relationship to ongoing classical biological control efforts using species of these phorids.

biological control, host location, light treatments, parasitoids, sex ratios

Phorid flies in the genus Pseudacteon are internal parasitoids of adult ants, with the majority of taxa specializing on Solenopsis fire ants (Disney 1994). The adult female flies are attracted for considerable distances (Morrison et al. 1999b) to exposed hosts. Once near ant colonies, the flies hunt individual victims (Orr et al. 1995, Wuellner et al. 2002) and attempt internal oviposition, using piercing-like ovipositors (Disney 1994). After flies initiate attacks, worker ants will often attempt to escape underground or other shelters or they maintain stereotypic postures (Feener and Brown 1992, Wuellner et al. 2002) that most likely reduces oviposition attempts. The responses of the fire ants to these parasitoids are at the heart of biological control efforts for the imported fire ants, S. invicta Buren and S. richteri, in the southern United States (Gilbert and Patrock 2002, Porter and Gilbert 2004, Porter et al. 2004). Ant community structure is largely considered to be regulated by competitive interactions among the ants themselves and possibly mediated by phorid flies (Feener 1981,2000, LeBrun 2005) that modify the behavior of their hosts as described here, potentially leading to a reduction in competitive ability of the parasitized ants relative to their neighbors.

Sets of 2-14 species of phorid parasitoid species will be found attacking fire ants in a single locality, independent of whether the hosts are in the related S. geminata complex (Trager 1991) in North America (Morrison et al. 2000, Wuellner and Saunders 2003) or the imported fire ants and their sister taxa in the S. saevissima complex of South America (Porter et al. 1995, Orr et al. 1997, Folgarait et al. 2003,2007). The structure of such guilds is thought to derive from the ways in which taxa partition their use of the common host resource. Time is an important niche axis that separates species with time of day or season (Pesquero et al. 1996, Folgarait et al. 2003,2007). Parasitism of different host sizes and attraction to different host locations may also reduce encounters or indirect interactions among members of the guild (Porter et al. 1995, Orr et al. 1997, Morrison et al. 2000). Division of resources is not absolute, however, and multiple species may commonly be found together attacking ants at the same nest (Williams et al. 1973, Folgarait et al. 2007).

The extent to which temporal-spatial overlap occurs among Pseudacteon species is not well documented (but see Bruzzone 2004) but has potentially important consequences ecologically and economically. One of the strategies for the classical biological control of fire ants using these flies is to use multiple species (Porter and Gilbert 2004, Gilbert and Patrock 2002). This strategy attempts first to mirror the diversity and structure of fire ant parasitoid guilds observed in the native range of Saevissima group fire ants (Orr et al. 1997, Folgarait et al. 2003,2007) and second to seek ecotypes of phorids that might accommodate for the large climatic and host variability (Folgarait et al. 2005a) documented across the vast and diverse introduced range of Solenopsis invicta and S. richteri. If there is substantial overlap among species in their timing of attacks on fire ants, such synchrony may facilitate host regulation, such as is found in the predator-aphid system described by Losey and Denno (1999). However, if there is too much overlap, interference or exploitative competition might lead to competitive exclusion of one or another of the species.

This study was undertaken to quantify interactions and particular abiotic preferences among Pseudacteon species in western Argentina that may act as a source of flies for our classical biological control efforts (Gilbert and Patrock 2002). Some of our antecedent work (Folgarait et al. 2007) showed an extremely high diversity in the parasitoid guild and that there was substantial overlap in terms of the season and time of day when various parasitoid species might be found attacking Solenopsis ants. We followed up this examination by making additional observations along two niche dimensions, host location and light conditions, that we considered important to the flies. As a way to promote their coexistence, we expected to find species partitioning among the mentioned niche axes, at least for the most common species that overlap in their phenological and circadian patterns.

Orr et al. (1997) showed, in their Brazilian localities, that certain taxa most commonly frequented the disturbed colony nests (i.e., mounds), whereas others were disproportionately found along foraging trails. Mounds and trails represent two substantially different prospects for finding hosts for the flies in terms of patch quantity and quality, as well as probabilistic occurrences of available hosts (Table 1). In general, disturbed mounds make for the best opportunities for host location, despite their unpredictable nature of occurrence, because of their sessile nature and rich concentration of all available ages and sizes of hosts. The challenge for a phorid is being available during that indeterminate time when a colony′s workers are exposed.

Table 1.

Listing of qualities that can differentiate between mounds and trails with respect to possible effects on Pseudacteon foraging

However, fire ants may forage during all except the extreme hot or cold periods of the day (Porter and Tschinkel 1987), so ants may be found almost continually on above ground trails. These locations are constantly changing, however, and the age and quality and quantity of the ants are poorer or more restrictive than at mounds (Table 1). A mound-trail dichotomy was not observed in Buenos Aires by Bruzzone (2004), which suggests that niche overlap within a guild may depend on local conditions and/or the combination of species present. From the biological control perspective, flies that are found attacking ants on foraging trails rather than mounds are more likely to impact the competitive position of the fire ants in their community hierarchy (Folgarait and Gilbert 1999, Orr et al. 2003).

Imported fire ants colonies are largely found in open areas, such as lawns, pastures, and annual crops, although foragers can also be prevalent in shady conditions, such as in pecan plantations (Tedders et al. 1990). Their parasitoids might also be more likely to search for the ants in well-lit areas where the hosts may be more prevalent. The dynamics of competition between fire ants and other sympatric ants also varies with light conditions, either as a result of the corresponding vegetation and its communities (Ali and Reagan 1985) or different temperature conditions or time of day (Wuellner and Saunders 2003). We might expect different levels of behaviorally mediated competition by the flies, according to the light condition, as well as conditional responses by the ants (Morrison et al. 2000).

Finally, we were also interested in relative attraction of flying males and female phorids to the ants. Mate seeking by males at the ovipositional site is found in a few Pseudacteon species associated with Solenopsis (Morrison et al. 1999b, Calcaterra et al. 2005). Field sex ratio observations have been highly variable, with frequent strongly biased male ratios (Morrison et al. 1999a, personal observations). Laboratory rearing of all species suggest that one component of these field observations is the emergence patterns of the flies, in that sex ratios vary enormously among cohorts. One factor underlying this variation for some species is the relationship between host size and sex of the emerging fly (Morrison et al. 1999a, but see Folgarait et al. 2002a), with males developing in smaller ants and females in larger ants.

Ant size also differs by colony and age but also varies strongly with the social form of the ant (Greenberg et al. 1985). Typically, colonies with multiple queens (polygyne) have significantly smaller ants than do single queen (monogyne) colonies. In the areas of Texas where our classical biological control efforts are directed, the polygyne form of fire ants is prevalent, and we have suggested that the male-biased sex ratios observed for introduced P. tricuspis (Gilbert and Patrock 2002) results from the high proportions of smaller ants. Field sex ratios in the native range have not been well documented, however. In one study (Calcaterra et al. 2005), the data indicated relatively even or female-biased sex ratios for individual taxa, presumably because polygyny may not be common in South America fire ants (Jouvenaz et al. 1989).

Therefore, in this study, we compared sympatric parasitoid species with respect to degree of preference for light conditions (and correlated temperature and vapor pressure deficit) and host location (mound versus trail). Our primary aims were (1) to better understand the coexistence of species that overlap in many dimensions of their niches and thus (2) to optimally select suites of complementary species to introduce for the biological control of fire ants. Secondarily, we were interested in quantifying sex ratios in the field across temporal scales with the aim of developing expectations for the already introduced or future introductions of parasitoids in the United States.

Materials and Methods

Our field sites were located near Brea Pozo (28.25° S, 63.95° W), which lies ~50 km south of the city of Santiago del Estero in the province of Santiago del Estero, Argentina. Climatically, the area is designated as Tropical-Xeric (Rivas-Martinez et al. 1999; http://www.globalbioclimatics.org/report/BIOC056.htm) and lies in the phytogeographical region of the dry western Chaco (Cabrera and Willink 1980). Characteristic vegetation includes xerophyllic trees such as the quebrachos Schinopsis and Aspidosperma, along with Brea (Cercidium praecox (R. et P.) Harms) and various Prosopis and Acacia. We had five sites altogether, with four being within 300 m of each other and the last separated from the others by 1.3 km. Each differed qualitatively from the others with respect to differences in vegetation cover, proximity to irrigation, and host densities. Two host species, S. interrupta Santschi and S. invicta were found existing in mosaics within each of these sites. S. interrupta was the more common host found in all except site 4, where S. invicta was almost the exclusive Solenopsis fire ant and the only one sampled. This difference in host relative abundance did not seem to affect species richness because a similar number of species was found in site 4 and the other sites (Folgarait et al. 2007), and no particular set of species seemed to occur at different mounds (personal observations). Specimens were identified using J. Trager's professional services in addition to using the diagnostic keys in Pitts (2002) and Trager (1991).

From November 2003 to August 2004, we monitored each of these sites on a monthly basis. Observations began in the first or second hour after sunrise and continued at least until sunset. We established two sets of treatment conditions for observing the flies, a mound-foraging trail dichotomy and a light-shade trichotomy, with a Sun, Partial Sun-Shade, and Shade set of treatments, explained below. Our mound-foraging trail treatments were set up as follows. We established a transect of six mounds, at each site. Mounds were spaced ≈10-15 m apart and paired with baited foraging trails ≈1.5 m from each mound. Mounds were disturbed initially by digging a small hole in their sides until ants streamed from below. The foraging trails were started by placing balls of tuna fish (≈2 cm diameter) on the ground. To maintain high numbers of exposed ants in the mound, we also used tuna fish plugs. The underlying difference in treatment between mound and foraging trail was distance to an active nest. Mounds, however, were redisturbed as needed, and the presence of fire ants on the foraging trail was required for an observation.

Establishment of observation points for the light conditions was largely opportunistic in that these points were naturally exposed to full sun (Sun), shaded by the trunks of trees (Shade), or received dappled light under tree branches (Partial Sun/Shade). We did use tiles to act as full shade covers in the more open areas. We were familiar with patterns of light and shade because we had been working in this area for 10 mo preceding the initiation of this study and had mapped the positions of each of the mounds and trees. The positions of many of the observation points were changed according to solar angle during the day. In the other instances, however, the observation point could have remained fixed as a single treatment through the day. Our mound and trail treatments were nested in these light conditions. We refer to the set of mounds and trails in all light conditions, at one observation time, as a transect.

Our sample unit was a 10-min observation period where we stopped at each mound-foraging trail pair along a transect, looking for phorids around the ants. We alternated at each pair as to whether the mound or trail was observed first. Our alternation of observations among light treatments and among mound-trail pairs was haphazard. When found, the flies were collected by aspirating into a labeled vial. The labels were recorded as to time and treatment and the vials placed in a cooler for later counting and identification of the flies to sex or species. We could normally identify taxa for females, but not males, in the field using a ×15-20 magnifying lens. If field identification was problematic, we would bring it back to the laboratory in Buenos Aries for confirmation using the key in Porter and Pesquero (2001).

Field metereological conditions were recorded in the full sun, full shade, and partial sun-shade conditions with (1) HOBO dataloggers (ONSET; Pocasset, MA), set off the ground surface and under shade covers, as well as (2) probes measuring soil temperatures at 10 cm depth. HOBO data (temperature and humidity) were logged on a 10 min basis, and soil temperatures were collected three times an hour. Relative humidity data were used to calculate vapor pressure deficit.

Treatment Analysis.

We asked to what extent, host location, and light conditions, in conjunction with a few correlated meteorological conditions, affected species richness and individual taxa occurrences and abundances. Our data were complicated by being serially replicated on a daily basis. To account for this way of sampling, we structured our analysis using daily contrasts among treatments as replicates. To examine frequencies, we first found the number of signed differences between treatments for each day. For instance, if the abundance value found at an observation for a mound was higher than at a paired trail, this would be given a score of 1; if tied, it was given a score of 0; and if lower, it was given a score of -1. Because we had ties, we found daily proportions for each sign. Light conditions were compared using abundances, averaged by day.

Correlated to light, metereological variables were categorized as follows. Meteorological conditions were measured on different scales so we standardized these variables using standardized deviates (Sokal and Rohlf 1969) or Z-scores of the mean hourly values for soil temperature, air temperature, and vapor pressure deficits. Z-scores are expressed in the units of the variable's standard distribution and provide both the distance and direction of each data point from the mean of the distribution. We found the mean rank of the different conditions and separated these into two metereological treatment categories, corresponding to (1) above mean (ABOVE) or (2) below mean values (BELOW). Tallies for each category were adjusted for their proportion of total daily observations. Mean values for the study were 28.3°C, 17.9°C and 3.9 kPa for ambient and soil temperatures and vapor pressure deficits, respectively.

For analyses involving contrasts among light treatments, we took the following two steps. First, observations around solar noon were taken only when shadows were sufficiently long to cover observation points. We considered some recorded data problematic when predicted shadow lengths based on solar angle at this location by date and time (Terwilliger and Sawyer 1996) were <20% of a standard height (gnomon) and excluded this data. We determined sunrise and sunset times for the geographical coordinates of Brea Pozo using the U.S. Naval Observatory converter at http://aa.usno.navy.mil/data/docs/RS_OneYear.html. Data where no flies were seen in a specific treatment comparison were excluded because we were interested in explaining their presence and not their absence.

For both frequencies and abundances, we compared treatments using repeated-measures ANOVAs. Depending on the analyses, site was considered a fixed effect and metereological treatment category and host location were considered random factors.

Species Associations.

One potential consequence of our classical biological strategy of establishing a suite of species is the potential of negative interactions among the taxa, leading to a displacement of taxa or their effects. We studied two components of this problem: dispersion of taxa and associations in occurrence among the common taxa. The examination of dispersion was to determine the extent of temporal clustering of flies. To do this, we used the nonparametric Wald-Wolfowitz tests for each uninterrupted set of samples within each day. In our data, a positive run would be the sequence of 10-min samples where flies were present and negative, when absent. A significant result indicates that the flies were appearing in a nonrandom order.

We examined species associations in two ways. First, we asked the explicit question of whether relative frequencies of co-occurrences between taxa were related to the overall frequencies of the individual species. Our expectation was that all species would co-occur most frequently (rank first) with the most common species and then in descending order according to the relative abundance of each taxa. To examine this question, we found the rank order of abundance for all the taxa and the rank frequency of co-occurrences for each of the taxa relative to all the others. We tested whether these co-occurrence ranks differed significantly from the abundance ranks of the taxa using Kendall's coefficient of concordance. We looked at this question at the hourly and 10-min sample level. Our null hypothesis was that species should rank according to their overall frequency in the population. Significant differences in rankings would suggest that other factors might be affecting the temporal distribution of the taxa, especially where relative differences in frequencies are large. For example, if the two most common species were found at entirely different times of the day, we would expect significantly different rankings (disconcordances) between their occurrences and co-occurrences. They would be ranked first and second in the community but they would rank each other at the bottom of their species associations. Because males of P. tricuspis and P. obtusus were found in the field on ant mounds (Calcaterra et al. 2005), presumably looking for females, we asked specifically whether the co-occurrences of males (overall, without distinguishing species) with P. obtusus and/or P. tricuspis were concordant with their frequency rankings.

We also asked whether species associations were significantly recurrent, that is whether there were sets of taxa that were more likely to occur together than by chance. To address this question, we used Fager's nonhierarchical complete linkage clustering method (NHCLC) (Fager 1957, Hayes 1978, Legendre and Legendre 1998). Although this method is the oldest of its class specifically designed to uncover biological associations, it gives meaningful as well as reproducible species groupings (Legendre and Legendre 1998). A major limitation is that is probably only useful for identifying positive associations (Fager 1957), but we addressed possible negative associations by our concordance analysis above. An alternative newer method to NHCLC that might be considered is path analysis, which is also based on decomposition of correlation coefficients. However, much of our data are presence-absence, which would be analyzed using phi-coefficients (Chen and Popovich 2002) being equivalent to using χ² analyses for species associations. Fager (1957) showed that χ² is extremely poor in determining the correct type of association. The NHCLC is therefore superior to χ² analysis for finding associations based on presence-absence data.

An outline of the NHCLC procedure is as follows. Recurrent species pairs are first identified according to whether their joint occurrences are significant. We used the index of Fager (1957) with P < 0.05 after adjusting for multiple correlations. This index is similar to a correlation coefficient and takes on a t-distribution but differs in its weighting of co-occurrences. Hayes (1978) explored other statistical properties of this index, such as the useful ability to correct for sample sizes. Once pairs are identified, higher-level groupings can be formed where there are significant co-occurrences within and among the group's species pairs. We formed these groupings following the explicit decision rules set out in Fager (1957) and Krylov (1968) and updated in Legendre and Legendre (1998). Our time frames of analysis were the 10-min sample and hour. Graphs are typically used to represent results of this analysis with significant associations among species or groups represented by common lines, whereas other associations are null or disconnected.

Sex Ratios.

Although we were not able to identify males to species, we asked how sex ratios (considering all males together) varied according to the scale of observation. However, data from Calcaterra et al. (2005) indicated that, of the species we found at our sites, males of P. obtusus, in addition to those of P. tricuspis (Porter et al. 1995), might be commonly attracted to females around ants. They found males of two other species, P. curvatus and P. nr. obtusus, on a rare basis (<0.5-1.6% of their respective taxa). Therefore, we further explore sex ratios in two ways: (1) that of males to the composite numbers of female P. obtusus Borgmeier and P. tricuspis Borgmeier and (2) males to female P. tricuspis only. In addition to asking whether the sex ratios differed according to light conditions, we were concerned about the relative arrival time of males to females of these species among samples and across days. On a daily basis, we determined first arrival times for the different species and sexes with the specific hypothesis that males would arrive before females.

Transformation of data before analysis included square-root (abundances and times) and arcsine-square-root transforms (proportions). If significant differences were found among treatments, Scheffé post hoc tests were run to examine for differences among groups within treatments. Analyses were done using JMP 6.0 (Lehamn et al. 2005) or Statview 5.1 for the Macintosh (SAS Institute 1998).

Voucher specimens for the ants and flies are kept in the laboratory at the Universidad Nacional de Quilmes.

Results

Light Conditions and Correlated Metereologial Variables.

We made 29,460 observations of possible host ant-parasitoid fly interactions equally distributed on mounds and trails over the course of 9 mo (38 d, 477 h, 2,455 transects). Further details of presence of flies according to location and light condition are found in Tables 2-4. Overall, female phorids were found at 45.3% of observation times, with the percentage of observations of phorids increasing only slightly to 46.7% when males were included (Table 2). We also found more female flies at mounds than on trails on a daily-site basis (Table 2; analysis of variance [ANOVA], F_1,33 = 23.23, P < 0.0001).

Table 2.

Percentage of 10-min samples when flies were found (averaged within hours and with 1 SD) according to light condition and location

Table 3.

Numbers of transect observations at mounds (M) and foraging trails (T) in the different light conditions for each taxon

a

Taxa are ordered according to the overall no. transects on which they were found.

b

Asterisks in the overall columns represent the location (M or T) with significantly higher frequencies of those taxa, according to ANOVAs (all P< 0.05, see text). Only species (ormales) with >100 observations were tested.

c

It should be noted that taxa may have been found in more than one light condition per transect so that overall numbers will be different than the sums for mounds or trails by light condition.

d

The Any Pseudacteon row shows total counts for each light and mound or trail conditions.

 

NS, not significant at P = 0.05.

Table 4.

Mean no. flies per hour (±SD) for the conunon species of Pseudacteon in the different light conditions

a

Different letters within rows imply significantly different mean no. flies among light conditions at P = 0.05. Only taxa with >100 observations were tested.

 

NS, not significant.

We found 13 species of flies during the study, of which 8 species were found on at least 100 transects (≈4% of the total; Tables 3 and 4). The two most common species in any treatment were P. nocens, followed by P. cultellatus, with males nested in frequency between these two (Tables 3 and 4). Of the common species, all had significantly higher numbers at mounds than on trails except the two members of the P. obtusus complex (Table 3). The mean proportional difference in occurrences for any species (excluding the P. obtusus complex) was ≈27.6 ± 10% higher for mounds than on trails.

With respect to light treatment, there was a mixed pattern of differences in the response of individual species (Table 3). Summing mound and trails, five of the species and males had their maximum number of occurrences in the sun, whereas the maxima was in the shade for seven of the other species. For eight species and males, the minima were under partial shade conditions. Significant differences among light conditions on a daily basis were found for some of these taxa. For the P. nocens complex (P. nocens and P. nr. nocens (Table 3) and P. litoralis, significantly more flies were found in the shade than in either the sun or partial light conditions, whereas for P. obtusus and males, significantly more flies were found in the sun than partial light or shade conditions (Table 4). The opposing tendencies among the taxa with respect to light condition are seen in the nonsignificant differences among light conditions for species richness (F_2,64 = 1.02, P = 0.37).

We found significant differences with respect to light treatment when metereological condition (air and soil temperature, vapor pressure deficit) was taken into account in our ANOVAs for the five most common species plus males (F_5,70 = 5.94, P = 0.009, P. nocens; F_5,45 = 4.6, P = 0.003, males; F_5,35 = 3.947, P = 0.006, P. cultellatus; F_5,35 = 4.6, P = 0.0018, P. litoralis, F_5,20 = 3.28, P = 0.025, P. tricuspis; and F_5,25 = 5.94, P = 0.0009, P. obtusus). For P. nocens, more flies were found in the Shade-Above than for all other light-metereological conditions (all P < 0.0066, Scheffé post hoc tests; Fig. 1). More P. litoralis were also found in the Shade-Above than other conditions, but this value was only significantly different than Sun-Above (P = 0.034, Sheffe's post hoc test). The mean number of flies in Sun-Below was significantly higher than Shade Below for P. cultellatus (P = 0.041, Sheffe's post hoc test) and P. obtusus (P = 0.031, Sheffe's post hoc test), as well as being significantly higher than Partial Light-Above for males (P = 0.039, Sheffe's post hoc test) and P. obtusus (P = 0.024, Sheffe's post hoc test). The mean number of P. obtusus in the Sun-Below was significantly higher than in Partial Light-Below (P = 0.017, Sheffe's post hoc test). We did not find any significant differences for contrasts within ambient conditions for P. tricuspis.

Fig. 1.

Mean number of flies (P. nocens, P. litoralis, P. cultellatus, males, P. tricuspis, and P. obtusus) per observation averaged by day under the three light conditions. Observations are grouped by metereological condition according to their Z-scores. Letters associated with symbols represent contrasts where significant differences between light treatments and metereological condition were found. Symbols with no letters associated indicates that the mean was not significantly different from any other treatment mean.

Within each light regimen (Fig. 1), we also found the following patterns. Within Sun, Sun-Below has a higher value for all taxa except P. tricuspis, indicating that all the taxa have some tendency to be in the sun when it is cooler than average. In the Shade treatment, all species were found more commonly in the Shade-Above condition than in the Shade-Below. This suggests that there may have been a tendency for all taxa to be in the shade when metereological conditions were hotter and drier than the mean of the study. There was no trend within Partial Shade, with three means higher for Partial-Shade Above and three means higher for Partial Shade-Below.

Species Associations.

Species associations were found in 61.5% of the 1,136 observation times when flies were present with two (34.7%), three (18.0%), or greater than three species (8.9%) per sample. We found the time scale to be fundamentally important for understanding these associations. For instance, any individual taxon could have been found with at least one other in 36.8 ± 23.8% of cases at the individual mound or foraging trail level within the 10-min sample, but this value increased to 86.7 ± 11.1% of cases when adjusted to an hourly basis. The maximum number of species found increased less rapidly with scale, with 7 (50%), 9 (64%), 11 (79%), and 13 species per sample, hour, day, and overall, respectively. The mean total number of flies found per sample (when found) was relatively low on a per sample basis, 4.6 ± 3.9 (4.0 ± 3.6 females) and merely tripled to 15.5 ± 14.7 SD (13.2 ± 13.0 SD females) at the hour scale.

Temporal dispersion of arrival times was significantly nonrandom (overall Wald-Wolfowitz Z = -26.28, 48 of 66 daily sequences with P < 0.05, all with >10 samples). The distribution of sightings was highly aggregated, with 68.4% of these samples coupled with a sequential sighting and an additional 20.8% of sightings interspersed with less than half an hour between observing flies. The median proportion of time that flies could have been seen on a daily basis, that is from the first to the last sighting, was 88.2% of the day (mean = 72.5 ± 28.2%). These observations suggest that the arrival times of the flies were found in lengthy pulses with the smallest breaks within the pulses.

The rank order for overall frequencies on an hourly basis was (1) P. nocens (20.5% of total sightings), (2) males (14.2%), (3) P. litoralis (12.1%), (4) P. cultellatus (11.3%), (5) P. obtusus (10.5%), (6) P. tricuspis (9.9%), (7) P. nr. nocens (7.8%), (8) P. nr. obtusus (7.4%), and (9) P. nudicornis (6.3%). P. nocens co-occurred most frequently with all taxa and males (it was ranked first), with the exceptions of P. obtusus and P. tricuspis. These latter two species were more frequently associated with males. Males, however, co-occurred most frequently with P. nocens but ranked P. tricuspis second, followed by P. obtusus. The species occurrence rankings for P. litoralis had a strong concordance with their frequencies rankings, as did those for P. cultellatus, with the exception of the lowest ranked species P. nr. nocens, P. nr. obtusus, and P. nudicornis (Kendall W, a posteriori tests, all P > 0.05).

At the 10-min sample level, the mean frequency rankings of the species were the same as those in the hourly examination. In all light and mound-trail treatments, P. nocens was ranked first, whereas P. nudicornis was ranked last, as seen in Table 3. Kendall's Ws for all light and mound-trail treatments were significant (all P < 0.0001, Friedman's χ²), indicating that, in all treatments, species joint association rankings were not concordant with the individual taxa's frequency in the population. We found significant a posteriori tests in all treatment conditions, as well with all but P. nr. nocens, P. obtusus, and males showing at least one significant difference across treatments.

Changes in expected joint associated ranks among specific taxa across treatments exhibit the following general patterns: (1) P. tricuspis ranked males higher than P. nocens, in all conditions except Shade; (2) P. litoralis ranked P. cultellatus lower than expected in all treatments except Shade; (3) P. cultellatus ranked P. litoralis lower than expected in all except Mound and Sun treatments; (4) P. nocens ranked P. litoralis higher than expected and P. cultellatus lower in all treatments; (5) P. obtusus ranked males higher than expected in all treatments; and (6) P. nr. nocens and P. nr. obtusus ranked each other first or second in all treatment levels except Partial Shade.

Results of the NHCLC recurrent group analysis also differed by time scale. At the 10-min sample level, we found two significant joint associations: one between P. nocens and P. litoralis and the second between P. tricuspis and males (both P < 0.0001). At the hour scale, we found 7 of the possible 36 associations to be significant. These associations are diagramed in Fig. 2. Males were significantly associated with P. obtusus and P. tricuspis, as well as with P. cultellatus and P. nocens. P. nocens was also significantly associated with P. litoralis, P. cultellatus, and P. nudicornis. No higher-level groupings were possible on either time scale.

Fig. 2.

Line diagram of recurrent species associations using nonhierarchical complete linkage clustering for the common Pseudacteon species at the hour scale. Lines connecting taxa denote significant associations between them. Numbers next to lines represent the observed significance levels. If not specified, the level is P < 0.0001, after adjustment for multiple comparisons. Note that P. nr. obtusus and P. nr. nocens are not associated significantly with any species.

Sex Ratios.

The relative proportion of males to females was another statistic that was highly dependent on the time scale of the tally. On a location basis (a mound or trail at the 10′ scale), the sex ratios were largely bimodal, with only one sex being present almost three fourths of the time (adding the percentages of males only and females only = 71.3%; Fig. 3). On the daily scale, however, the sex ratios were more equitable, with a mean of ≈48.8 ± 23.6% females and with only 14.3% of the days finding only one sex (11.4% of days with females only). The percentage of samples with males only, accounting for just female P. tricuspis, decreased from 52.1% to 8% when going from the 10-min sample to a daily sample basis. The proportions of male:female P. tricuspis above 3:2, however, increase when summing the total collections for the day (Fig. 3). This indicates that, although females may be present, they are still in lower relative numbers.

Fig. 3.

Sex ratios on 10-min sample and daily bases. The dark portion of the bars represents the ratios of all males to females of both P. trictispis and P. obtusus, whereas the clear portions represent samples where males are compared only with P trictispis females. Values for all bars are read directly from the y axis.

Light conditions were also a significant factor in sex ratios. We found that the mean sex ratio was significantly lower in the shade than either in the sun or partial conditions at the day scale (F_2,68 = 4.28, P = 0.018) and at the hour scale but only for below metereological condition for the latter scale (F_2,34 = 7.60, P = 0.019 and F_2,32 = 0.206, P = 0.81 for Below and Above, respectively).

There was extensive temporal overlap between males and females of both P. obtusus and P. tricuspis. Males were absent on 3.2% of days when females of either species were present, and males were present on 3.2% (N = 1) of days when both species were absent and 6.4% (N = 2) when females of either P. obtusus or P. tricuspis were absent. We found a significant difference between arrival times of males relative to P. obtusus and P. tricuspis. The first males arrived before P. tricuspis females on 86.2% of cases but before P. obtusus on only 44.8% of days surveyed. Males first arrived before females of either of these species on 45.7% of days. The difference between taxa resulted largely from the fact that P. obtusus arrived before P. tricuspis on 75% of the 28 d when both were present.

Discussion

We found species richness and the presence of most species to be significantly higher at mounds than on nearby foraging trails. We had expected this finding because "trail specialists" are typically less abundant than those categorized as "mound species" (Orr et al. 1997). Two findings dealing with this contrast in location are notable. The first is that we did not find significant differences between mound and trails for the P. obtusus complex, although the power of these analyses was low (0.34 and 0.27 for P. obtusus and P. nr. obtusus, respectively). These values indicate that if this study were to be repeated, we would find this same set of differences 34 and 27% of the time. Nonetheless, these findings suggest that the P. obtusus complex behave differently from the other species with respect to factors that attract or retain them along foraging trails. This trait should be considered to consistently add this species (Folgarait et al. 2005b) within the suite of candidates for the biological control of fire ants, because P. obtusus will spend a great deal of the time attacking at foraging trails, therefore complementing the activity of other Pseudacteon species.

The second notable finding with respect to the mound-trail dichotomy is that flies were ≈30% more likely, on average, to be on a disturbed mound than on a foraging trail (Table 2; divide trail overall by mound overall). This result might suggest that there is less difference between the flies with respect to being found at mounds or trails than previously suspected. However, the distance between disturbed mounds and the experimentally created foraging trails was relatively short (≈1.5 m). Given that Pseudacteon flies are quite vagile, once attracted to a location by the larger numbers of ants at a mound, the short distance to the paired foraging trail could increase the encounter rate with the foraging trails while searching for ants in the vicinity. Indeed in our release experiments with P. tricuspis in Texas, female P. tricuspis typically begin attacks in the center of the disturbed mound but soon can be observed pursuing ants in the immediate area (e.g., up to a meter from the mound edge). Two factors probably encourage females to search more widely: (1) tendency of target hosts to flee underground in the zone of attack and (2) harassment by males in lekking in the mound disturbance (L.E.G., personal observations) Encounter rates might also have been high away from the mound because most areas were relatively barren of vegetation, which might have increased the flies searching ability. Although the foraging trails were maintained throughout the day, we removed flies as they arrived so all flies could be considered naïve with respect to the trail location. This conclusion cannot be made for the experience of the flies as to mound locations, which were fixed in terms of the study, and their positions may have been learned by the flies on previous days. If these points are valid, we might expect lower numbers of flies attracted to foraging trails that are further removed in space or time from a disturbed mound. Solenopsis fire ants typically have subterranean foraging trails that are below ground and only emerge several meters away from the colony. This foraging pattern might offer a reduction in parasitoid load for the ants by increasing independence of foraging trails from attacks at the mound.

We found differences among the taxa with respect to their occurrences in the different light conditions, with approximately one half the species being found most commonly in full sun and the other half in full shade. It is tempting to dichotomize the guild according to this finding. This treatment, however, was influenced by both available light and surrounding temperatures. Along with differences in light, mean vapor pressure deficits and air and soil temperatures were significantly higher in full sun than those in partial and shade conditions (repeated-measures ANOVA, P < 0.0001 in all cases, independent of time scale). Our results suggest that P. nocens in particular, plus other species (Table 4; Fig. 1), were accommodating their temperature preferences by moving in and out of the light (Fig. 1). When temperatures were cooler, higher numbers were found in the sun, whereas when it was warmer, higher numbers were associated with shade.

The species that we found to be in significantly higher numbers in full shade was classified as a "more temperate" species (Folgarait et al. 2007), whereas those in higher numbers in the sun were classified as "hotter" species based on significant differences in the mean temperature each was found. Given this temperature-humidity dimension, the relative differences in abundances with shade are most likely related to the overall effect solar radiation has on the microclimate rather than to possible phorid preferences to light qualities.

From a practical perspective, such as in planning a release protocol for individual species, attention to both light conditions and temperature would need to be considered. For instance, releases for all of the taxa might be most efficiently done in full sun or shade conditions while the coolest or hottest conditions prevailed, respectively. For the more frequently encountered intermediate ranges of temperature conditions, a combination of light and shade might be appropriate, according to the species. In Texas, of the species examined, we have only released P. tricuspis, which we found in this study is most frequently found under sunny conditions both Above and Below metereological conditions (Table 2). Therefore, we had not considered light conditions during our planning, but the data presented here indicate that, in any case, shading would not have been a useful factor to include because P. tricuspis prefer sunny conditions (Fig. 1).

An additional consideration relates to examining the behavioral modification by the flies in reducing foraging by the ants. At high temperatures, many fewer flies were seen on the trails compared with mounds in the Sun and Partial Shade conditions, but the difference was marginal in the Shade condition (Table 1). We suspect that some of this difference might be explained by our ability to see the flies. At a mound, our observation was confined to a cavity that contained the ants and flies, whereas at a foraging trail, the observation point was more diffuse, particularly when the ants could find external shelter from both the sun and the flies. Protocols using artificial shading, such as Folgarait and Gilbert (1999) or a combination of artificial shade and full sun (R.J.W.P. and L.E.G., unpublished data) for sampling foraging trails might offer different results depending on the prevailing temperature regimens.

One potential consequence of the strategy of establishing a suite of species is the potential of competitive or negative interactions among the taxa, leading to a displacement of taxa or a reduction in their modification of the ants' behavior. We studied two components of this problem: dispersion of taxa and associations in occurrence among the common taxa. Temporal dispersion of the guild was clumped, and this result suggested there might be strong positive joint occurrences among the taxa. We found several significant positive associations in both the Kendall's test of concordance and Fager's NHCLC (Fig. 2). A positive association denotes that the taxa are found together significantly more often than would be expected by chance, but the connotation of this is dependent on the interacting taxa.

For instance, both of our association tests found significant associations between males and P. obtusus (hour) and P. tricuspis (sample and hour scale). This association should be related to mating and is considered a set of positive interaction. Although we could not identify the taxa for the males, this association fits neatly with our expectation for these species mating at the oviposition resource site (Calcaterra et al. 2005, Folgarait et al. 2007). Given that we found significant associations between males and P. tricuspis at both scales of analysis but only at the hour scale for P. obtusus, we are likely to have had higher relative numbers of P. tricuspis males than P. obtusus males. This difference in relative numbers is also suggested by the fact that males arrived before P. tricuspis significantly earlier than before P. obtusus.

We consider the finding of an association between males and P. nocens and P. cultellatus to result from our inability to discriminate among males as to taxa. The following supports this conjecture. When we grouped P. tricuspis and P. obtusus females as an entity (because males can be either P. tricuspis or P. obtusus), we found significant correlations between both P. cultellatus and P. nocens and this pair at the hour level (both P < 0.05), although not at the sample level (both P > 0.4). These correlations are similar to those we found with the males and these taxa.

The positive associations among the other taxa might represent the possibility for interference competition resulting from higher than expected joint occurrences, as well as possible additive effects in host behavioral modification. The strongest overlap was found between P. litoralis and P. nocens at both the sample and hour level with the NHCLC. Both species are larger than the average taxon in the guild and can use larger ants (Folgarait et al. 2002b,2005b), suggesting an overlap of both temporal and resource use for these two species. P. nocens, however, has a greater size range than P. litoralis, which might represent more flexibility in its host size use, depending on the female and narrowing their resource overlap.

There were also significant positive associations between P. nocens and both P. cultellatus and P. nudicornis at the hour level with the NHCLC. In addition, the Kendall's test of concordance found an increase in rank for P. nudicornis with P. litoralis, which is in line with the significant positive association found for both of these two species with P. nocens. While the concordance analysis found a reciprocal lowering of rank of P. cultellatus with both P. nocens and P. litoralis, this change in each species' relative position can be seen as resulting from the stronger association of P. nocens with P. litoralis at the sample and hour level of analysis, according to the NHCLC.

Pseudacteon cultellatus and P. nudicornis are smaller on average than P. nocens or P. litoralis (Folgarait et al. 2002a,b,2007) so that any overlap in host use among these flies would be most pronounced in smaller ants. This size class is much more common than that of larger ants in most colonies, potentially leading to less overlap in resource use than that between P. nocens and P. litoralis. Although we found a positive relationship in our concordance analysis between two of the other small taxa, P. nr. obtusus and P. nr. nocens, this relationship was not supported in the NHCLC analysis.

In this study, we found that light conditions affected the sex ratios. Another perspective is obtained by examining these sex ratios on different time scales. We found that at the finest scale (10-min sample), the sex ratios were largely bimodal, but at the daily scale, there was greater equity in the numbers of the different sexes. This suggests that daily cumulative numbers might offer a better understanding of effective population sizes than hourly samples. Temporal pulses of one sex or the other seem to be a common feature of Pseudacteon, but each may be corrected by pulses of the other sex emerging over the course of the few days of their lives. We consider the ant populations studied here to be largely monogyne, and the distributions of sex ratios found in this study could be used to compare with resulting sex ratios for P. tricuspis (and perhaps later for P. obtusus for which releases have now been initiated in Texas) where by default they have been intentionally established on polygyne populations. Potential differences within the appropriate time frame could be useful for better understanding the influence of social form on the population dynamics of these species.

From a biological control perspective, our results are informative and encouraging. The surprisingly high overlap in main niche dimensions among the most common species in a species rich local community of Pseudacteon reinforces the idea of a multispecies biocontrol strategy as previously suggested (Gilbert and Patrock 2002, Folgarait et al. 2003,2005b,2006, Porter and Gilbert 2004). Indeed, the system studied includes the greatest number of co-existing phorid species attacking Solenopsis recorded to date (Folgarait et al. 2007). Our results point to certain ideal species combinations based on ecological and behavioral traits. Thus, P. tricuspis has been consistently found as a heat-tolerant species (Folgarait et al.,2005a,2007), and our results confirmed this finding for P. tricuspis and shows P. obtusus to have similar abiotic requirements, tolerances, and preferences. Therefore, ecotypes of these two species from Santiago del Estero should be excellent candidates for establishment in similar hot and arid areas such as south Texas, especially considering that P. obtusus complements P. tricuspis in its greater preference for attacking at foraging trails (Folgarait et al. 2005b,2007). Other set of species that should be considered in the suite of candidates are P. nocens, P. litoralis, P. cultellatus, and P. nudicornis, which we found positively associated, and they should complement the previous ones either because they prefer more shady (less hot) sites such as P. nocens and P. litoralis or because they will attack smaller size hosts such as P. cultellatus and P. nudicornis. Our results also showed that P. nocens, P. litoralis, and P. cultellatus from Santiago del Estero are the most abundant species, which reflects their ability to deal with extremely harsh dry conditions.

We thank R. Plowes for careful and thoughtful comments on this manuscript. G. Azzimonti did much of the fieldwork and was assisted by R. Carrara on occasion. We are indebted to the staff of FAUNA in Santiago del Estero who provided permits and logistic support to perform this research. P.J.F. thanks the Universidad Nacional de Quilmes for support and CONICET for oversight. This research was supported by the Helen C. Kleberg and Robert J. Kleberg Foundation and The State Of Texas Fire Ant Project.