Wildfire Impacts on Red Imported Fire Ant Captures Around Forest Ponds in the Lost Pines Ecoregion of Texas

The red imported fire ant (RIFA) Solenopsis invicta is an invasive species in the southern and southeastern United States. It first appeared in Alabama in the 1930s and has since expanded east to North Carolina and west to central Texas (Callcott and Collins 1996); it is also present in California (Ward 2005). Negative ecological impacts of RIFA invasion have affected a wide range of vertebrates and invertebrates (Porter and Savignano 1990; Allen et al. 1994, 2004; Stuble et al. 2009; Diffie et al. 2010; Epperson and Allen 2010). Among vertebrates, juveniles tend to be particularly vulnerable to RIFA predation (Landers et al. 1980; Freed and Neitman 1988; Pedersen et al. 1996).

The red imported fire ant appeared in Bastrop County, central Texas, between 1973 and 1977, and has since become well established throughout the Lost Pines ecoregion (Cokendolpher and Phillips 1989; Taber and Fleenor 2003), which has historically been the last remaining stronghold for the federally endangered Houston toad Bufo [Anaxyrus] houstonensis (U.S. Endangered Species Act [ESA] 1973, as amended). This species is a known predator of juvenile Houston toads (Freed and Neitman 1988), and thus there is interest in understanding the potential and realized impacts of RIFA on Houston toad survivorship.

Brown et al. (2012) recently conducted a study in the Lost Pines aimed at delineating annual RIFA activity patterns, elucidating the relationship between overstory canopy cover and RIFA captures, and evaluating the effects of low-intensity prescribed fire around pond edges (where juvenile Houston toads are vulnerable to predation after emergence) on RIFA captures. Brown et al. (2012) found that annual RIFA activity was highest between May and October, which includes the period in which juvenile Houston toads disperse from ponds. They also found a strong inverse relationship between RIFA captures and canopy cover near pond edges. It is possible that greater sunlight exposure, and thus greater heat, in open-canopied environments results in more suitable habitat for RIFA (Vogt et al. 2003). Lastly, Brown et al. (2012) detected no effects of low-intensity prescribed fire on RIFA captures, indicating no sustained direct impact of fire on the species. However, the authors noted that research on high-intensity fire impacts was warranted, given the greater potential for habitat alteration consequent of hotter fires.

At 1400 hours on 4 September 2011 a high-intensity wildfire began in the Lost Pines. The fire was unstoppable due to wind gusts in excess of 58 kph caused by Tropical Storm Lee, coupled with extreme drought conditions in central Texas. After 18 d the fire was 95% contained, with the total burn area encompassing 13,406 ha (ca. 39% of the Lost Pines ecoregion). Seven of the 10 ponds used in the Brown et al. (2012) study burned in the wildfire, allowing us the opportunity to assess the impacts of a high-intensity wildfire on RIFA in the Lost Pines ecoregion using seven burned and three unburned ponds. Given the inverse relationship between canopy cover and RIFA captures found in the previous study, we hypothesized that RIFA captures would increase at ponds with substantial tree mortality caused by the wildfire.

We conducted this study on the 1,948 ha Griffith League Ranch (GLR). Vegetation on the GLR is typical of the Lost Pines ecoregion, with an overstory dominated by loblolly pine Pinus taeda, post oak Quercus stellata, blackjack oak Quercus marilandica, and eastern red cedar Juniperus virginiana, and an understory dominated by yaupon holly Ilex vomitoria, American beautyberry Callicarpa americana, and farkleberry Vaccinium arboreum. The GLR contains three permanent ponds (i.e., ponds have not dried in at least 12 y), 10 semipermanent ponds (i.e., ponds typically dry several times per decade), and dozens of ephemeral ponds that hold water for days to months annually depending on rainfall. For this study we used the seven burned and three unburned ponds with RIFA sampling data prior to the wildfire, all of which are known Houston toad breeding ponds. Three of the ponds were ephemeral, six were semipermanent, and one was permanent.

We used prewildfire RIFA capture data from 26 September 2009 to 10 October 2010, which included 27 samplings for the area around each of the 10 ponds (Brown et al. 2012). We collected postwildfire RIFA capture data for this study between 25 September 2011 and 4 August 2012, sampling the area around each of the 10 ponds a total of 11 times. To maximize comparability we used a sampling design identical to that of Brown et al. (2012). Days between sampling ranged from 15 to 78, with a mean of 31 days between samples. We used a standard bait cup method for sampling ants (Porter and Tschinkel 1987; Mueller et al. 1999), which consisted of half a Vienna sausage placed in a 9-cm-diameter round plastic dish. We placed between 4 and 14 dishes around the perimeter of each pond, depending on pond size (i.e., perimeter length). During each sampling event, we placed dishes 1–2 m from the pond edge at 3–4 m intervals, and allowed them to attract ants for 30 min. After 30 min we collected ants from each dish and euthanized them by freezing. We removed native ants from samples and counted the number of RIFA individuals collected in each bait cup at each pond on each sampling date. Because the number of bait cups varied by pond and sampling date, we calculated captures-per-unit-effort (CPUE; defined as the number of captures per number of bait cups), and used CPUE as our response variable.

The prewildfire canopy cover data that we used in this study were from 20 July 2008 to 16 April 2010, with 4–24 estimates of percent canopy cover around each pond depending on hydroperiod status at the time of sampling (i.e., no estimates when ponds did not contain water; Brown et al. 2012). We estimated percent canopy cover around each pond four times between 18 April 2012 and 27 June 2012. For both studies, we estimated percent canopy cover using a spherical densiometer (Forestry Suppliers Inc., Jackson, MS) at two to six randomly selected points at the pond edge, with higher numbers of estimation points corresponding to larger ponds. We then averaged the estimates at each pond per sampling date. For this study we used the difference between the mean prewildfire and postwildfire canopy cover estimates as a predictor of postwildfire RIFA CPUE.

Our data included both spatial (multiple samples were taken from around the same pond) and temporal (the same ponds were sampled repeatedly) nonindependence. To account for this we used a mixed effects modeling approach for our analyses, treating within-day sample nested within pond as a random effect. Preliminary analyses indicated a strong positive relationship between air temperature and RIFA captures, and thus we included mean daily temperature as a covariate in our final analyses. We obtained temperature data from the Austin-Bergstrom International Airport weather station, located approximately 36 km from the GLR (station number: 410429).

We initially tested interactions among and between treatment (control/wildfire), time (preburn/postburn), and canopy cover change (change in mean percent canopy cover between the prewildfire and postwildfire sampling periods). This analysis indicated an interaction between canopy cover change and time, and changes in canopy cover were not consistent among ponds due to variability in tree mortality caused by the wildfire, as well as pond size (canopy cover at the edge of ponds typically increases as pond size increases). Thus, we analyzed each pond individually using mixed effects models to determine if RIFA CPUE differed by time. We investigated assumptions of normality and homoscedasticity using graphical diagnostics plots (Zuur et al. 2009). Because the data appeared to violate the assumption of normality, we transformed RIFA CPUE using the arcsinh (i.e., inverse hyperbolic sine) transformation (Fowler et al. 1998). We performed these analyses with the program R (R Version 2.10.1, www.r-project.org) using the nlme (Pinheiro et al. 2009) package. We considered effects to be significant at α  =  0.05. Finally, we tested the relationship between changes in RIFA CPUE between sampling periods and changes in mean percent canopy cover between sampling periods using ordinary least squares (OLS) regression (Sokal and Rohlf 1995). For this analysis we used RIFA captures between May and early August (i.e., the peak RIFA activity months common to both the prewildfire and postwildfire sampling periods). We considered effects to be significant at α  =  0.05.

The initial analysis testing interactions among and between treatment, time, and canopy cover change indicated no treatment × time × canopy cover change interaction effect (F_1,365  =  0.03, P  =  0.853), no treatment × canopy cover change interaction effect (F_1,6  =  1.72, P  =  0.237), and no treatment × time interaction effect (F_1,365  =  0.02, P  =  0.902). However, we detected a time × canopy cover change interaction effect (F_1,365  =  27.52; P < 0.001). These results indicate changes in CPUE between sampling periods (i.e., time) were not directly associated with the wildfire. Rather, CPUE differed between sampling periods, and changes in percent canopy cover influenced the effect. The pond-specific tests indicated a time effect on RIFA CPUE at 5 of the 10 ponds, including 1 of 3 control ponds, and 4 of 7 wildfire ponds (Table 1; Table S1, Supplemental Material). The time coefficients indicated that postburn CPUE decreased at a control pond (pond 6), decreased at two of the wildfire ponds (ponds 10 and 11), and increased at two of the wildfire ponds (ponds 13 and 14). The directional change in RIFA captures for all of these ponds was inversely related to canopy cover change (pond 7 [+5.14%]; pond 10 [+5.01%]; pond 11 [+17.52%]; pond 13 [−11.30%]; pond 14 [−41.86%]). The OLS regression indicated that RIFA CPUE decreased with increased mean percent canopy cover (F_1,7  =  13.17; P  =  0.008; r²  =  0.65; Figure 1).

We found that a high-intensity wildfire did not have a direct impact on RIFA CPUE, which was the same conclusion as for low-intensity prescribed fires in this system (Brown et al. 2012). Rather, changes in canopy cover drove CPUE differences between the two sampling periods. Our results indicated the wildfire indirectly benefited RIFA through overstory tree mortality (and thus canopy cover reduction), supporting our hypothesis based on previous research (Brown et al. 2012). We speculate that the two wildfire ponds with significant postwildfire decreases in RIFA CPUE actually positively impacted RIFA, but an increase in canopy cover at these ponds masked the effect due to heavy rainfall in fall 2011 that resulted in the pond edges reaching into the adjacent upland forest canopy.

Our primary motivation for conducting research on RIFA was to assist with management activities that benefit recovery of the Houston toad. The results of this study are cause for concern in that the wildfire burned a significant portion of the Lost Pines ecoregion (ca. 39%), with an estimated overstory tree loss of 78% across the burned area (Lost Pines Recovery Team 2011). Based on 13 randomly placed 20 m × 50 m vegetation plots on the GLR within the wildfire zone, the mean overstory tree loss was 81.9%, with a corresponding 30.2% decrease in canopy cover (D.J. Brown, unpublished data). Further, the Federal Emergency Management Agency, Bastrop County, and local landowners are currently removing a substantial amount of dead trees due to conflicts with human-based recovery needs. Red imported fire ants are known to prefer forest gaps (Colby and Prowell 2006), and to use openings such as roads and powerline cuts as expansion corridors (Tschinkel 1988; Stiles and Jones 1998; Todd et al. 2008). Thus, the mortality and subsequent removal of trees in the Lost Pines has the potential to dramatically increase both area occupied by RIFA and site-specific RIFA densities in the Lost Pines.

The research in this study and Brown et al. (2012) has increased our understanding of the spatial and temporal habitat use of RIFA in the Lost Pines ecoregion. It is clear that these patterns overlap with those of the Houston toad (Brown et al. 2012), and thus RIFA has the potential to be a serious threat to the persistence of the Houston toad. However, although we know that RIFA prey upon juvenile Houston toads (Freed and Neitman 1988), and we have evidence that they also prey upon adult Houston toads (M. C. Jones, Texas State University, personal communication), we do not have a good understanding of their impacts on Houston toad survivorship rates. Thus, it is possible that RIFA predation is partially or entirely compensatory (i.e., some or all of the mortality caused by RIFA could be substituting for mortality that would have occurred due to other factors). However, we believe the increased prevalance of RIFA in the Lost Pines as a result of postwildfire canopy cover loss will likely add additional stress to Houston toad persistance in the ecoregion. Fortunately, RIFA control through insecticides, if warranted, is a viable option for Houston toad recovery efforts.

Please note: The Journal of Fish and Wildlife Management is not responsible for the content or functionality of any supplemental material. Queries should be directed to the corresponding author.

Table S1. Red imported fire ant (RIFA) captures around ponds on the Griffith League Ranch, Bastrop County, Texas between 26 September 2009 and 4 August 2012 at control and wildfire treatment ponds, mean daily temperature for each sampling date, and difference in mean percent canopy cover at each pond between the prewildfire and postwildfire sampling periods.

Found at DOI: http://dx.doi.org/10.3996/012013-JFWM-004.S1 (375 KB XLS).

This study was supported by the Texas Parks and Wildlife Department and U.S. Fish and Wildlife Service through a traditional Section 6 grant, and relied upon methods developed with funding from the Texas Department of Transportation. The Capitol Area Council of the Boy Scouts of America provided access to the Griffith League Ranch, and we appreciate their continuing support of our research. We thank B. DeVolld, W. DeVolld, and L. Eddins for their assistance with data collection for the prewildfire data set. Two anonymous reviewers and the Journal Subject Editor provided comments that improved the quality of this manuscript.

Any use of trade, product, or firm names is for descriptive purposes only and does not imply endorsement by the U.S. Government.