Predation is the largest source of mortality among avian populations (Ricklefs 1969, Martin 1993, Thompson 2007, Conner et al. 2010). Early avian life stages are particularly vulnerable to predation mortality (O'Conner 1991, Coté and Sutherland 1997). High levels of nest predation can limit avian recruitment and reduce population growth (Cowardin et al. 1985). Because ground-nesting birds are particularly susceptible to nest predation by meso-mammals (Rogers and Heard 2000, Jimenez and Conover 2001), managers are often interested in manipulating this predator community to enhance avian reproduction and population densities (Coté and Sutherland 1997, Ellis-Felege et al. 2012). Manipulation could include numerical changes to predator population size (e.g. removal via trapping) or behavioral changes to modify the functional relationship between prey and predator (Holling 1959). However, the efficacy of predator manipulation is predicated on predation being a limiting factor for the population, that is, the predation limitation hypothesis.

Under the predation limitation hypothesis, a reduction in predation should increase population rate of growth or population metrics such as survival or reproduction. Managers have implemented both indirect and direct techniques, such as habitat alteration and meso-mammal trap and removal (MMTR), to reduce predator numbers and increase avian recruitment (Lokemoen 1984, Jimenez and Conover 2001). Although MMTR is often considered a controversial management tool (Messmer and Rohwer 1996, Ellis-Felege et al. 2010), it is commonly used in gamebird and waterfowl management (Potts 1986, Coté and Sutherland 1997). Seasonal predator removal has been shown to increase nest success above levels necessary for population maintenance in upland nesting ducks (Cowardin et al. 1985, Garrettson and Rohwer 2001, Pieron et al. 2013). Seasonal predation control has also proven effective at increasing brood size, production of young, and subsequent breeding populations of grey partridge Perdix perdix and ground-nesting moorland birds (Tapper et al. 1996, Fletcher et al. 2010).

However, uncertainty surrounds the existence of relationships between meso-mammal activity and northern bobwhite Colinus virginianus reproductive success, as well as the efficacy of MMTR as a management tool for improving reproductive success. There is a consensus in the literature regarding the efficacy of predator removal for enhancing bird populations (Coté and Sutherland 1997, Smith et al. 2010); however, the literature is scant for bobwhite in particular — only two studies included in Smith et al. (2010). Predator trapping for bobwhite is a unique case of predator removal because of the diverse predator community that depredate their nests with snakes contributing a significant portion (∼29%; Stallet et al. 2005). Generally, snakes cannot legally be removed as part of a trapping program potentially reducing or negating its effectiveness through compensation on some sites (Ellis-Felege et al. 2012). However, Palmer et al. (2005) did find a positive effect of meso-mammal predator removal on bobwhite nest success.

Our objectives were to determine the efficacy of MMTR in the context of typical management on properties managed for bobwhite. We hypothesized support for the predation limitation hypothesis as tested through predator removal. We predicted that removal of meso-mammals would increase bobwhite reproductive output. This hypothesis is plausible on these sites because habitat is managed intensively and resources (i.e. food and cover) are not likely limiting bobwhite reproduction (Stribling and Sisson 2009, Burger et al. 1998, Palmer and Wellendorf 2007, Terhune et al. 2007), leaving the possibility of predation as a limiting factor more likely. Furthermore, we were interested in understanding how the residual meso-mammal community (i.e. post-trapping) may impact bobwhites. We hypothesized that MMTR effectively eliminates the numerical relationship between predator abundance and bobwhite reproduction metrics. Under this hypothesis we predict no negative relationship between bobwhite reproduction metrics and predator abundance for trapped sites would exist. This hypothesis is not only novel in the predator trapping literature — as no other authors investigated the effects of residual predators — it has management implications because, if supported, no additional predator removal in these systems is warranted that year as predation is no longer a dominant limiting factor. Lastly, we were interested in evaluating the utility of a commonly used scent-station survey, known as the predator index, as a tool to determine predation pressure of a site.

Material and methods

Study area

We studied the relationships between the predator index, MMTR, and bobwhite reproductive demographics at 11 sites in Florida, Georgia and Alabama during 1999–2006. Sites ranged from 1012 ha to 10 927 ha and averaged 4541 ha (Table 1). All study sites were located in the southeastern Coastal Plain and were privately owned with a long history of bobwhite management. Study sites practiced a consistent management program that included maintaining low basal area timber (e.g. 3–9 m² ha^-1), two year prescribed fire interval, and other mechanical vegetation treatments including mowing, roller-chopping, seasonal disking, and herbicide application. Supplemental feed was broadcast at a rate of 0.06–0.13 metric tons ha^-1 on all sites. Tall Timbers and the Red Hills plantations were comprised of 80% loblolly Pinus taeda, and shortleaf pine P. echinata with associated early-successional ground cover vegetation and longleaf pine P. palustris with associated wiregrass Aristida stricta ground cover. Harwood hammocks were interspersed throughout the Red Hills study sites and comprised 10% of the area. Annually disked fallow fields ranging from 0.4 to 1.2 ha in size covered the remaining 10% of the study sites (Staller et al. 2005, Ellis-Felege et al. 2012). Georgia sites were dominated by upland pine forests (80%) consisting of loblolly, longleaf, and slash Pinus elliotti pine. Ground cover vegetation consisted of bunch grasses (Andropogon spp.), wiregrass, blackberry Rubus spp., goldenrod Solidago spp., partridge pea Chamaecrista fasciculata and ragweed Ambrosia artemisiifolia. Rotational fallow fields ranging from 0.41– 2.43 ha in size were interspersed throughout the study sites and covered the remaining 20% of the area (Terhune et al. 2006, 2010, Sisson et al. 2009). Alabama sites consisted of low density pine forests comprised of shortleaf, longleaf, loblolly and slash pine with associated early successional plant communities, covering approximately 70% of the area. Annually disked fallow fields were scattered throughout the sites and ranged from 0.5–3.0 ha in size covering approximately 16% of the area. The remaining 14% of the area was comprised of thinned hardwood forests (Crouch 2010). Meso-mammal trap and removal using box and leg-hold traps occurred throughout the bobwhite breeding season every year during the study on five sites. Six sites had some or all years during the study when meso-mammal trap and removal did not take place. Combined, there were 37 site—year combinations monitored when MMTR occurred and 20 combinations when it did not occur (Table 1). Typical trap density ranged from one trap per 8 hectares to one trap per 20 hectares. The number of meso-mammals removed was recorded for 12 site—year combinations and ranged from 243–737 and averaged 430 meso-mammals removed annually per site.

Table 1.

Study sites where meso-mammal activity indices, bobwhite reproductive demographics, and trap years were monitored in Georgia, Florida and Alabama, USA, 1999–2006.

Results

We collected bobwhite reproductive information from a total of 3935 radio-tagged bobwhites ( =69 site^-1, SD = 31.37) across 8 years (1999–2006), resulting in 2499 nests ( =44 site^-1, SD = 24.11). Nest success ranged from 0.29 to 0.72 among sites and averaged 0.52 (SD = 0.11). The number of nests per hen ranged from 0.22 to 1.92 and averaged 0.71 (SD = 0.32). Broods per hen averaged 0.37 (SD = 0.21) with a range of 0.06 to 1.02 and the number of chicks per hen averaged 4.09 (SD = 2.42) and ranged from 0.42 to 11.52 (Table 2). The mean predator index was 0.13 (range = 0.03-0.38, SD = 0.08) with a species indices proportion of 0.061 raccoons (range = 0-0.27, SD = 0.06), 0.024 nine-banded armadillos (range = 0-0.09, SD = 0.02), 0.017 Virginia opossums (range = 0-0.11, SD = 0.02), 0.021 bobcats (range = 0-0.08, SD = 0.02), and 0.008 foxes (range = 0-0.03, SD = 0.008; Table 3).

Trapping was effective at reducing predator activity. The model containing the trap effect was supported by the LRT (χ² = 34.95, df = 1, p< 0.001). The model predicted mean predator index for non-trapped sites was 0.21 (95% CLs: 0.18, 0.24) compared to 0.10 (95% CL: 0.07, 0.13) for trapped sites. The lack of variability surrounding the predator index across trapped sites (residual variance = 0.0023) suggests that trap effort had similar effects on predator activity across sites.

We found support for the predation limitation hypothesis among all bobwhite reproduction metrics. The nest success model containing the trap effect was supported by the LRT (χ² = 7.59, df = 1, p = 0.006). Nests were 1.33 times (odds ratio, 95% CL: 1.09, 1.62, β = 0.29 ± 0.10) more likely to be successful on trapped sites than non-trapped sites (Fig. 1). Mean nest success on trapped sites was 0.56 (95% CL: 0.52, 0.59) and mean nest success on non-trapped sites was 0.48 (95% CI: 0.43, 0.53, Fig. 1). The nests hen^-1 model containing the trap effect was also supported by the LRT (χ²= 19.45, df=1, p< 0.001) — for every 100 hens they produced 14 more nests on trapped sites (β = 0.25, 95% CL: 0.13, 0.37, Fig. 2). The broods hen^-1 model containing the trap effect was supported by the LRT (χ² = 24.81, df = 1, p < 0.001) — for every 100 hens they produced 12 more broods (β = 0.37, 95% CL: 0.21, 0.53, Fig. 3). Lastly, the chicks hen^-1 model containing the trap effect was supported by the LRT (χ² = 4.16, df= 1, p = 0.041) — for every 100 hens they produced 109 more chicks on trapped sites (β = 1.09, 95% CL: 0.07, 2.11, Fig. 4).

We found support for the residual predator hypothesis using the trapped site—year data subset, i.e. our predictions were supported in each case. We found no relationship between residual predators and nest success (β = -0.36, 95% CL: -3.14, 2.42), the hypothesis was not supported over the null model (χ² = 0.065, df= 1, p = 0.798; Fig. 5). The nests hen^-1 model supported our hypothesis even though the null was rejected (χ² = 16.04, df = 1, p< 0.001; Fig. 6), because we did not find a negative effect between residual predators and nests hen^-1 (β = 2.92, 95% CL: 1.52, 4.32). Concomitantly, with broods hen^-1 our hypothesis was supported (χ² = 7.37, df= 1, p = 0.007; Fig. 7), because we did not find a negative effect between residual predators and broods hen^-1 (β = 2.72, 95% CL: 0.79, 4.64). Furthermore, we did not find a negative effect between residual predators and chicks hen^-1 among trapped sites (β = 5.56, 95% CL: -8.77, 19.89) and this model was not selected over the null, (χ² = 0.605, df= 1, p = 0.437; Fig. 8).

Table 2.

Summary statistics of monitored bobwhite metrics among study sites in Georgia, Florida and Alabama, USA, 1999–2006.

Table 3.

Summary statistics of scent-station surveys among study sites in Georgia, Florida and Alabama, USA, 1999–2006.

Figure 1.

Predicted bobwhite nest success on trapped study sites versus non-trapped study sites in Georgia, Florida and Alabama, USA, 1999–2006. Median nest success value depicted by the center lines of each box plot, 75% upper quartiles represented by the upper bounds of each box, 25% lower quartiles represented by the lower bounds of each box,and associated outliers.

Figure 2.

Predicted number of nests per 100 hens on trapped study sites versus non-trapped study sites in Georgia, Florida and Alabama, USA, 1999–2006. Median number of nests per 100 hens depicted by the center lines of each box plot, 75% upper quartiles represented by the upper bounds of each box, 25% lower quartiles represented by the lower bounds of each box, and associated outliers.

Discussion

We found support that predation limits bobwhite reproduction by reducing nest success, nests hen^-1, broods hen^-1 and chicks hen^-1. After controlling for predator removal (i.e. sub-setting analysis to only trapped sites), we did not detect a negative relationship between residual predator activity and reproduction suggesting that trapping was effective at reducing predation. We found trapped sites to have a lower average predator index than non-trapped sites, suggesting that MMTR reduces predator activity. Collectively, our results suggest management of meso-predators to improve reproduction is feasible when habitat resources are not deficient, but methods of measuring predator activity need further study.

The predation limitation hypothesis characterizes predation as a limiting factor for prey (e.g. northern bobwhite) populations. Our findings support this hypothesis and are similar to previous studies (Baiser et al. 1968, Chesness et al. 1968, Duebbert and Lokemoen 1980, Garrettson and Rohwer 2001), further strengthening the justification and utility of MMTR as a management tool. While nest success alone is a tenuous metric for determining population growth in bobwhite (Etterson et al. 2011, Ellis-Felege et al. 2012), nests hen^-1, broods hen^-1 and chicks hen^-1 are stronger predictors of reproductive success and population trajectories. Bobwhite population growth is most sensitive to the variation in reproductive demographics (Sandercock 2006, Stahl and Oli 2006); thus, it is reasonable to believe that the increases in reproduction, especially nests hen^-1, broods hen^-1 and chicks hen^-1, we observed would lead to population growth. Given that bobwhite populations experience dramatic annual mortality (Stoddard 1931, Rosene 1969, Klimstra and Roseberry 1975, Roseberry and Klimstra 1984, Brennan 1991), increased chick and brood production can help stabilize populations by offsetting local declines due to adult mortality (Sandercock et al. 2008). Increased fecundity (i.e. chicks hen^-1) can lead to higher recruitment to the fall, higher fall abundance, and increased breeding densities the following year when overwinter survival remains good (Tapper et al. 1996). In areas where habitat and resources are not limiting, managers can use MMTR to reduce predation pressure and benefit bobwhite productivity.

Figure 3.

Predicted number of broods per 100 hens on trapped study sites versus non-trapped study sites in Georgia, Florida and Alabama, USA, 1999–2006. Median number of broods per 100 hens is depicted by the center lines of each box plot, 75% upper quartiles represented by the upper bounds of each box, 25% lower quartiles represented by the lower bounds of each box, and associated outliers.

While MMTR can reduce predation pressure on bobwhite populations, bobwhites and meso-mammals coevolved such that bobwhites themselves, have strategies to mitigate the risk of predation (e.g. cryptic coloration, renesting). However, the loss of apex predators and landscape changes can release and elevate meso-mammal populations to unprecedented levels (Soulé et al. 1988, Crooks and Soulé 1999, Conner and Morris 2015), potentially destabilizing the predator—prey relationship between bobwhites and meso-mammals (Rogers and Heard 2000, Prugh et al. 2009, Conner et al. 2010). Managers focused on promoting bobwhite populations often utilize MMTR to combat perceived negative impacts of elevated mesomammal populations. However, the goal of MMTR is not to eliminate meso-mammals from a particular site, but rather to reduce them to a level where predation no longer limits reproduction (Ellis-Felege et al. 2010). Meso-mammal trap and removal is effective at reducing meso-mammal populations within a year, but annual implementation is likely required to impede their reestablishment upon cessation of trapping (Tapper et al. 1996, Ellis-Felege et al. 2010). We demonstrate that MMTR is an effective tool for bobwhite managers, however, it should be used under appropriate conditions within a holistic management regime. That is, MMTR should complement other management actions, as habitat management is fundamental to bobwhite management (Stoddard 1931, Rosene 1969, Brennan 1991).

Figure 4.

Predicted number of chicks per 100 hens on trapped study sites versus non-trapped study sites in Georgia, Florida and Alabama, USA, 1999–2006. Median number of chicks per 100 hens depicted by the center lines of each box plot, 75% upper quartiles represented by the upper bounds of each box, 25% lower quartiles represented by the lower bounds of each box, and associated outliers.

Figure 5.

Relationship between nest success and residual predators on trapped study sites in Georgia, Florida and Alabama, USA, 1999–2006.

Figure 6.

Relationship between nests hen^-1 and residual predators on trapped study sites in Georgia, Florida and Alabama, USA, 1999–2006.

We found support that trapping reduced meso-mammals to a level that residual predators were not limiting bobwhite reproduction. Thus, the intensity of MMTR in this study was effective at reducing predation pressure within a given year. But, we found positive effects on trapped sites between residual predators and nests hen^-1, broods hen^-1 and chicks hen^-1. We postulate that on trapped sites with few residual predators, alternative mechanisms may be driving bobwhite reproductive success. Residual predators may consist primarily of bobcats and foxes, as they may be more wary of novel things in their environment (e.g. traps; Hernandez et al. 1997). Therefore, the positive effects we found between residual predators and bobwhite reproduction may be a result of the predator species remaining on a site post-trapping. Therefore, the effects predicted by our residual predator models could be spurious relationships between residual predators and bobwhite reproduction.

The predator index is an effective tool for monitoring predator activity on a site (Best and Whiting 1990, Diefenbach et al. 1994, Ellis-Felege et al. 2010). We detected differences in the predator index between trapped and non-trapped sites, suggesting it provides value for managers to evaluate the efficacy of trapping programs or determine if predator activity levels warrant trapping. However, the predator index surveys were conducted post-bobwhite breeding season and may not accurately depict actual predator abundance or even predator activity during peak breeding season (Diefenbach et al. 1994). Future studies conducting surveys during the bobwhite breeding season and incorporating detection of meso-predators into the predator index may yield improved estimates and result in stronger predictive power.

Figure 7.

Relationship between broods hen^-1 and residual predators on trapped study sites in Georgia, Florida and Alabama, USA, 1999–2006.

Figure 8.

Relationship between chicks hen^-1 and residual predators on trapped study sites in Georgia, Florida and Alabama, USA, 1999–2006.

While meso-mammal populations impact bobwhite reproduction, the predator index does not account for factors known to impact reproduction. Adult bobwhite survival, for instance, is a critical driver of population performance (Sandercock et al. 2008), and while meso-mammals prey on adults, raptors pose a more significant threat to adult survival (Burger et al. 1998, Rollins and Carroll 2001, Sisson et al. 2009). Similarly, the predator index does not provide a measure of snake activity, which are significant nest predators and known to prey on chicks and adult bobwhites (Stoddard 1931, Staller et al. 2005, Ellis-Felege et al. 2012). Furthermore, heterogeneity in individual fitness, habitat quality, weather, and supplemental feeding can also impact production (Brennan 1991, Doerr and Silvy 2006, Rolland et al. 2011, Tri et al. 2013, Buckley et al. 2015). Therefore, the predator index should be used as a crude predictor of mesomammal activity and the associated effects differing levels may have on bobwhite reproduction, with the knowledge that additional factors affect overall bobwhite reproductive output and population trajectories.

Our study was conducted on sites with a long history of bobwhite management and maintaining consistently high abundance compared to other areas throughout the range of bobwhites (Stribling and Sisson 2009, Terhune et al. 2007). Therefore, where habitat is limited, meso-mammal populations could be different and the effects of trapping on bobwhite reproduction could be different as well. Furthermore, we only studied the effectiveness of MMTR for a single biological objective — bobwhite reproduction. The impact of MMTR to overall recruitment, population growth rate, and ultimately abundance needs more study to further justify the use and costs of MMTR as a management tool as well as the potential impact on ecosystem health.