SEARCH

SEARCH BY CITATION

Keywords:

  • Bayesian analysis;
  • Colinus virginianus;
  • compensatory mortality;
  • ground-nesting birds;
  • infrared nest camera;
  • mesomammal;
  • northern bobwhite;
  • predator control

Summary

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusions and management implications
  8. Acknowledgements
  9. References

1. Nesting birds can be vulnerable to predation. Wildlife managers sometimes manipulate predator communities to enhance avian productivity and abundance. Managers need to know the predation risk from different predator species responsible for nest failures to maximize success. This issue is especially important when considering reductions in only a part of the predator community in complex ecosystems.

2. We conducted a 7-year crossover experiment at four study sites to examine the effect of mesomammalian predator control on nest success of northern bobwhite Colinus virginianus in the southeastern USA. Nests were monitored using 24-h near-infrared video. We hypothesized that nest failures caused by different predator guilds may not be independent and may lead to compensation by other predators as one predator guild was reduced.

3. We compared levels of bobwhite nest predation by mesomammals, snakes and other predators in years with and without mesomammal control.

4. Control of mesomammal predators reduced the levels of mesomammal nest predation, but predation levels by snakes and other predators increased such that total nest mortality was not reduced. Nest mortality among predator groups was best described as compensatory, and total nest mortality differed among sites.

5.Synthesis and applications. Our findings suggest that reductions in predation risk from one predator guild can be compensated by an increased risk from other predators in complex ecosystems. Predator removal within one group may not translate to additive increases in overall nest success, but rather results in shifts in the identity of predators responsible for nest failures. Management efforts focused on manipulating predator communities to enhance avian reproduction are encouraged to examine cause-specific nest fates to determine the effectiveness of predator reduction programmes.


Introduction

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusions and management implications
  8. Acknowledgements
  9. References

Knowledge of the key factors leading to avian nest failure is pivotal for understanding and managing annual recruitment. Many potential causes of avian nest failure exist, including parasitism, adverse weather, abandonment, egg failure and a variety of predators (Etterson, Nagy & Robinson 2007). These failures may not be independent of one another posing interesting applied and theoretical questions for quantifying cause-specific nest failures (Etterson, Nagy & Robinson 2007). For wildlife managers, partitioning failure among predator guilds provides insight into predator–prey dynamics. It also permits estimation of shifts in predation risk as a result of annual changes or manipulation of predator abundance. This information can inform management decisions directed at increasing avian reproduction through the use of predator reduction programmes.

The concept of non-independence among mortality factors, or compensatory mortality, is not new to the wildlife field (Errington 1946, 1956, 1967). Traditionally, the focus among mortality factors is on the compensatory relationship between hunting and natural mortality factors (Allen 1954; Anderson & Burnham 1976; Clark 1987; Servanty et al. 2010). Specifically, density-dependent mechanisms are thought to cause reduced per-capita mortality risk when populations are reduced through harvest, and vice versa, typically up to a threshold limit set by mortality in the absence of harvest (Anderson & Burnham 1976). Thus, interactions of multiple mortality sources may fall along a continuum of completely compensatory to completely additive relationships between the failure causes, including complex combinations of compensation up to a point after which the interaction of mortality factors becomes additive (i.e. partial compensation).

Compensatory interactions associated with nest failures have been anecdotally discussed but not experimentally evaluated. For example, Errington (1967) removed specific predators to enhance waterfowl nest success and noted that total nest mortality appeared to remain constant each year. However, the specific predators responsible for failures appeared to shift. The fundamental idea is a nest at risk to predator one and two will probably succumb to predation by predator two if predator one is removed. These results, if general, have ramifications for the management of predation for nesting birds in complex ecosystems where specific predator species are reduced in an attempt to increase productivity.

A model system for exploring such interactions in predation resulting from predator management programmes is the northern bobwhite Colinus virginianus Linnaeus in the southeastern USA. Only middle-sized mammalian (mesomammalian) predators (e.g. raccoon Procyon lotor Linnaeus, Virginia opossum Didelphis virginiana Kerr, bobcat Lynx rufus Schreber, nine-banded armadillo Dasypus novemcinctus Linnaeus, red fox Vulpes vulpes Linnaeus, gray fox Urocyon cinereoargenteus Schreber, and coyotes Canis latrans Say) can legally and logistically be removed during the bobwhite breeding season in an attempt to increase bobwhite production (Felege 2010). However, several other predator species exist within the community, such as snakes (particularly rat snakes Pantherophis alleghaniensis and P. guttatus Say and kingsnakes Lampropeltis getula Linnaeus), fire ants Solenopsis spp. Westwood and several incidental predator species (Staller et al. 2005; Ellis-Felege et al. 2008; Terhune et al. 2008). Some of these species, such as snakes, serve roles as both bobwhite predators and as prey items for the mesomammals being reduced (Lang 2008; Howze 2009). Therefore, reduction in certain predator species may release other predator groups or affect interactions related to foraging efficiency, with unintended consequences on predation events at nests (Henke & Bryant 1999; Tewes, Mock & Young 2002). This suggests that cause-specific nest predation rates may shift as a result of changes in predator community dynamics, resulting in an overall compensatory rather than additive effects of predator reductions in nest success.

Although there has been interest in the quantification of cause-specific mortality, challenges exist in the best analytical approaches to evaluate non-independence among nest failure causes (Etterson, Nagy & Robinson 2007). Much of the work conducted on cause-specific mortality has been related to marked individuals (Servanty et al. 2010) and frequently relative to harvest and natural mortality of game species (Anderson & Burnham 1976; Burnham, White & Anderson 1984; Boyce, Sinclair & White 1999). Previous nesting studies have estimated cause-specific mortality by assuming mortality sources are independent of one another because analytical challenges exist in creating a model with dependence among failure causes when incorporating discovery bias and uncertainty in nest fates (Etterson, Nagy & Robinson 2007; Etterson et al. 2011).

Managers often are interested in manipulating predation rates by reducing predators to enhance avian reproduction; however, studies suggest variable success (Côté & Sutherland 1997). In diverse ecosystems, the response in nest mortality of the target species to predator reduction can range from completely additive, where noticeable increases in nest success could be observed, to completely compensatory depending on the predators reduced and the complexity of the ecosystem. In this study, we hypothesized that nest failures caused by the different predator guilds may not be independent and may lead to compensation by other predator species as one predator guild was reduced. Using an experimental approach, we examined potential shifts in nest predation by reducing one predator guild, mesomammals, in a complex predator community. To avoid traditional problems of nest discovery bias (Etterson et al. 2011) and the inability to identify predators with certainty (Staller et al. 2005; Rader et al. 2007), we used radio-telemetry and 24-h nest cameras to monitor incubation periods of a ground-nesting species to observe whether predator reduction efforts impacted total nest mortality and cause-specific predation rates.

Materials and methods

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusions and management implications
  8. Acknowledgements
  9. References

We studied components of bobwhite nest predation at three properties in southern Georgia and northern Florida during 2000–2006. Tall Timbers Research Station (Leon County, FL; 30°39′39″N, 84°13′35″W; hereafter TTRS) and Pebble Hill Plantation (Thomas and Grady County, GA; 30°46′13″N, 84°5′48″W; hereafter PH) were located in the Red Hills physiographic region. Pinebloom Plantation (Baker County, GA; 31°24′42″N, 84°22′45″W) is located in the Upper Coastal Plain physiographic region near Albany, Georgia. Pinebloom was divided into two 1400-ha study plots (hereafter PB East and PB West) with a bald cypress Taxodium distichum Rich. swamp buffer approximately 607 ha in size between the two sections. Predator management was undertaken on each of the 1400 ha areas, and the core bobwhite data were collected on a 400-ha area in the centre of each plot. We did this to minimize edge effects of predator removal and effects of neighbouring properties. Detailed site descriptions can be found in Staller et al. (2005) and Sisson et al. (2009). All three sites are managed for bobwhites with frequent fire, disking, roller-chopping, and mowing to maintain an open, low-density pine forest structure. Sites are dominated by loblolly pine Pinus taeda Linnaeus and shortleaf pine P. echinata Miller with associated ‘old-field’ ground cover vegetation and areas of longleaf pine P. palustris Miller with associated wiregrass Aristida stricta Michx. ground cover. Hardwood drains, hammocks and fallow fields are interspersed across the landscape.

Each year between January and April, we captured approximately 100 bobwhites on each site using baited funnel traps (Stoddard 1931) and fitted them with 6·5 g (∼4% body-weight) collar-style radio-transmitters equipped with mortality switches (American Wildlife Enterprises, Tallahassee, FL, USA) or motion-sensitive switches (Holohil Systems Ltd., Ontario, Canada). Trapping, handling and marking followed University of Georgia procedures (Institutional Animal Care and Use Committee permit #A2004-10109-c1 and A3437-0). We used radio signals to locate bobwhites ≥5 days per week from 15 April to 1 October each year to monitor nesting behaviour and determine nest fates. Bobwhites found in the same location on two consecutive days were assumed to be incubating. Flagging was placed near the nest site to mark its location and relocate the nest when the incubating bobwhite was away foraging. We were able to find nests and begin monitoring within the first few days of incubation. Thus, nests were discovered at approximately the same stage and only included incubation, not the egg-laying phase of nesting (Taylor et al. 1999).

While the incubating bobwhite was absent from the nest, we installed small continuously recording cameras (Furhman Diversified, Seabrook, TX, USA) 1–1·5 m from the nest (Staller et al. 2005). A near-infrared lighting source (950 nm) was attached with the camera to provide lighting for night-time recording. The camera and light source were supported on an articulating arm camouflaged with surrounding vegetation. A 25-m cord linked the camera and lighting unit to a time-lapse VHS recorder and 225-ampere reserve capacity deep cycle battery. Camera batteries and VHS tapes were changed every 24 h. Unlike many nesting studies, we checked nests daily until failure or hatch via telemetry, minimizing errors in failure dates common to nesting studies. Additionally, we viewed videos to confirm nest fate and identify nest predators in the event of a depredation. Failures were partitioned into successful hatch and three main failure categories including mesomammals (species described above), snakes, and other (e.g. fire ants, incubating adult killed, and incidental predators such as white-tailed deer Odocoileus virginianus Zimmerman, great horned owl Bubo virginianus Gmelin, barred owl Strix varia Barton and fox squirrels Sciurus niger Linnaeus).

Predators were removed from two pairs of study sites, the Red Hills Region and the Albany Region. In each study region, we established a treatment and a control plot of approximately 1300–1400 ha in size. In 2000, a year of baseline data was collected on bobwhite nesting activities. Prior to this, no predator control occurred on any of the plots. From 1 March to 30 September in 2001 to 2003, one plot in the Red Hills Region (PH) and one in Albany (PB East) received intensive predator reduction by personnel from Georgia USDA-Wildlife Services, while predators were not removed at the other plot. During 2004–2006, the treatments were reversed (i.e. TTRS and PB West trapped). The experiment follows a blocked, repeated measures crossover design. Mesomammalian predators targeted in the study included raccoon, Virginia opossum, bobcat, nine-banded armadillo, red and gray fox, coyotes and feral animals such as pigs Sus scrofa Linnaeus (Table 1).

Table 1.   Number of mesomammalian predators removed by species for each of the four study sites in southern Georgia and northern Florida, 2001–2006
PredatorPB EastPB WestPHTTRSTotal
200120022003200420052006200120022003200420052006
  1. PB, Pinebloom; PH, Pebble Hill Plantation; TTRS, Tall Timbers Research Station.

  2. *Feral animals included normally domestic animals breeding in the wild, such as pigs Sus scrofa.

Raccoon19323610417514016990475710487951497
Opossum601792764061641741061861402461532032293
Armadillo4345511196147614431150149153954
Bobcat20402226981459201316202
Coyote9146471413134121212120
Fox48712360106543
Feral*86863262104652
Total3375284747373864172962972435324244905161

Statistical analysis

We developed two models of compensation in a predation context following the framework developed by Anderson & Burnham (1976). We generalized this model to reflect compensations that could occur among nest mortality because of different predator guilds. First, we explored three separate causes of bobwhite nest failure (mesomammals, snakes, other failures), as we had specific interests in these predator groups. We then simplified the models to explore mesomammalian predators and all other predators into a single category to address the more general management question of shifts in nest failure between those manipulated with predator control and those not manipulated.

3-failure cause model

In our three-cause model, we constructed a multinomial logit random effects model within a Bayesian framework using Markov chain Monte Carlo (MCMC) algorithms to examine predator-specific mortality rates during incubation. We built and ran the model in Python (version 2.5) and the module PyMC (version 2.0; http://code.google.com/p/pymc). The model was developed in a hierarchical fashion conditioned on nest failure. The number of nest failures because of cause (xi) in each year resulted in a multinomial distribution (eqn 1).

  • image(eqn 1)

where P3 = 1−P1P2. The probability of each cause of failure was modelled as multinomial logit mixed model:

  • image(eqn 2)

where the inline image and third source of mortality serves as the baseline (in our case, mesomammals acted as the baseline for comparisons). β0 is the intercept, β1 is the effect of the binary covariate for predator reduction (w), uj is the random effect for site (= 1, 2, 3, 4), zk is the random effect for year (= 1, 2, …,7) and i represents the cause of nest failure (1–3). We modelled random effects of site and year for the predator probabilities of the multinomial model using standard errors that were drawn from a half-normal distribution with hyperparameter τ = 0·001. The multinomial model included an intercept (β0) and trap effect for predator control treatments (β1) with prior distributions specified as a normal distribution (μ = 0, τ = 0·001), where τ represents the squared inverse of the standard error.

We specified a diffuse prior distribution as a beta with parameters of value 1 = 1, value 2 = 1 for three distinct models: (i) full compensation model where total nest mortality, or mortality from all causes of nest failure, was a single constant across all sites and years, (ii) site-specific mortality where total nest mortality was constant across years, but different among sites and (iii) no compensation where total nest failure varied over all years and sites. Each of these models was considered with and without the predator control covariate to create six competing models. We modified the harvest models from Anderson & Burnham (1976) and Burnham & Anderson (1984) to reflect relationships between nest failures for completely additive and completely compensatory models. Thus, absolute mortality for each nest failure caused by predators (Mi,t where = predator-specific failure cause, = year) in each year was derived from total mortality (Mtot) and the predator-specific causes of failure. The compensation models (i.e. models 1 and 2 above) were constrained such that the third source of mortality (M3) was found from subtracting the first two causes of failure (M1 and M2, respectively) from total nest mortality (Mtot); therefore, modelling dependence among the failure causes:

  • image(eqn 3)

In models without compensation, each mortality factor was independent of the others, and total nest mortality was based simply upon the sum of the three causes of nest failure each year:

  • image(eqn 4)
2-failure cause model

We simplified the above model to evaluate the relationship between managed (i.e. mesomammals) and unmanaged causes of nest failure (i.e. snakes, ants and incidental causes) by adjusting the previously described multinomial logit mixed model to only two causes of nest failure (i.e. multinomial model collapses to a binary logistic model when only two alternative options are considered, eqn 5).

  • image(eqn 5)

where the inline image. These included mesomammals and all other causes of nest failure.

For both the 2-failure and 3-failure cause models, six models (described above) were developed based upon the three causes of nest failure and the inclusion of the covariate for mesomammal predator control. We ran MCMC algorithms for 1 000 000 iterations with a burn-in of 800 000 and thinned every 50 to ensure convergence and minimize autocorrelation. Models were evaluated using deviance information criteria (DIC), analogous to Akaike information criterion (AIC), to find the most parsimonious model explaining bobwhite nest failure (Spiegelhatler et al. 2002). Thus, smaller DIC values indicate the best approximating yet simplest model. Bayesian goodness-of-fit (GOF) was assessed by comparison of deviance from simulated data to that of the observed data (Gelman 2004). A model with a perfect fit would have a GOF = 0·5, which indicates that half of the simulated data deviances exceeded the deviances of the observed data. Much debate exists about the appropriate methods for model-averaging for comparison of Bayesian models (Link & Barker 2010). Rather than model-averaging, we chose to focus inference on the best DIC-supported models and reported the median and 95% credibility intervals (95% CI) of the posterior distribution.

Results

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusions and management implications
  8. Acknowledgements
  9. References

During 6 years across all study areas combined, we removed a total of 5161 mesomammalian predators (Table 1). Examination of the predator community using occupancy modelling from scent station data found mesomammal predator use of the study sites to be 2·5 times less likely in years with predator control efforts than in years without control efforts (Felege 2010). We monitored 746 bobwhite nests with cameras. Of these, the fate of 30 (4·0%) was unknown because of thick vegetation, 8 (1·1%) were subject to camera failure and 32 (4·3%) were abandoned because of research activities (primarily as a result of camera installation), leaving 676 (90·6%) nests with known causes of failure. Of these, 429 (63·5%) nests hatched at least one egg and 247 (36·5%) nests failed during incubation. When incorporating nest failures because of unknown causes, a total of 706 nests were monitored, and 277 nests did not hatch at least one egg. Causes of failure included 98 (35·4% of total depredated nests with known and unknown causes) predation events by mesomamals, 79 (28·5%) snake depredations, 70 (25·3%) failed because of other causes [fire ants (11·9%), incidental predator species (1·4%), or mortality of the incubating bobwhite (11·9%)] and 30 (10·8%) where the cause of failure was unknown. Failures classified as unknown were not a result of mesomammal predators because these predators were apparent even in thick vegetation. Therefore, failures because of unknown causes were incorporated into the 2-failure model, because our best evidence suggested that failure was caused by snakes (Staller et al. 2005); we could not distinguish snakes or smaller incidental predators from camera footage, so these 30 nests were not included in the 3-failure cause model.

3-failure cause model

We first examined the role of compensation with respect to 3-failure causes (i.e. mesomammals, snakes and other causes of nest failure) and obtained posterior statistics, goodness-of-fit and information theoretical statistics for the six alternative models. Goodness-of-fit was adequate for all models, but lower for the two full compensation models (Table 2a). The best approximating model, based upon DIC values, for these data was the site-specific compensation model with the mesomammal predator control effect covariate included (Table 2a). In other words, the best model from our candidate set demonstrated compensation. Total mortality because of complete nest failure was different at each site but similar from year to year at a site with and without predator control. Additionally, individual probabilities of the causes of failure were dependent on the covariate of predator control. This model had 77·2% of the model weight and was 3·9 times more likely than the next best-fitting model, which described site-specific compensation without the trap effect included.

Table 2.   Goodness-of-fit (GOF) statistics and model selection statistics using DIC (Deviance Information Criterion) to assess best approximating model for (a) 3-nest failure causes (mesomammals, snakes, other causes) and (b) 2-nest failure causes (mesomammals and all other causes) at northern bobwhite nests on four study sites over 7 years in northern Florida and southern Georgia, USA
ModelGOFDICΔDICWi
  1. *Site-specific compensation refers to total mortality being different by sites.

  2. †Pred. control refers to mesomammal predator control.

  3. ‡Not compensatory refers to independence among causes of nest failure.

  4. §Full compensation refers to total mortality being constant across all sites, rather than site-specific.

(a)
Site-specific compensation*, Pred. control0·64332·9 00·772
Site-specific compensation*0·61335·62·70·200
Not compensatory, Pred. control0·64341·28·30·012
Full compensation§, Pred. control0·71341·38·40·012
Not compensatory0·61343·510·60·004
Full compensation§0·71353·320·40·000
(b)
Site-specific compensation*, Pred. control0·53236·7 00·929
Full compensation§, Pred. control0·61242·45·70·054
Site-specific compensation*0·54244·88·10·016
Full compensation§0·61250·613·90·001
Not compensatory0·68280·7440·000
Not compensatory, Pred. control0·68281·444·70·000

The credibility intervals (CI) of the probability of total nest mortality overlapped for three of the four sites, suggesting similar mortality among all sites except PB West (Fig. 1). The probability of mesomammal predation events decreased when predator control occurred, whereas the probability of snakes and other causes of failure increased during this period (Fig. 2). Based upon the top model, the effect of predator control made the probability of snake mortality 2·80 times more likely (95% CI: 1·44–5·61) and other failures 3·58 times more likely (95% CI: 1·84–7·24) than mesomammal predation events.

image

Figure 1.  Posterior predicted total annual bobwhite nest mortality (median ± 95% Bayesian credibility interval) between 2000 and 2006 for four sites, each of which had mesomammalian predator control in 3 of the 7 years of the study.

Download figure to PowerPoint

image

Figure 2.  Posterior predicted probability of predator-specific nest mortality (median ± 95% Bayesian credibility interval) when examining 3-failure causes from our top model with site-specific compensation and a covariate for predator control at bobwhite nests in the southeastern USA. Years of mesomammal trapping for each site are highlighted by the box.

Download figure to PowerPoint

2-failure cause model

The 2-failure cause model allowed us to incorporate the 30 nest failures that were known not to be mesomammals but could not be positively identified as snakes. This increased the sample size to 706 total nests, of which 277 failed (39·9%). As for the 3-failure cause models, we obtained posterior estimates, goodness-of-fit and information theoretical statistics for the six alternative models. All models exhibited adequate fit (Table 2b). The best approximating model for the 2-failure cause data was also the site-specific compensation model with the mesomammal predator control effect included as a covariate (Table 2b). This model had 92·9% of the model weight and was >17 times more likely than the second best-fitting model describing full compensation with the trap effect covariate included. The top model for the 2-failure causes had a similar pattern in total mortality with PB West exhibiting greater estimated annual total mortality (0·53) than other sites (PB East: 0·40, PH: 0·37, and TTRS: 0·35); however, the total annual mortality across years was still similar at a site with and without predator control. Mesomammal depredations were 2·73 times less likely (95% CI: 1·51–4·86) than other types of nest failure when predator control occurred (Fig. 3).

image

Figure 3.  Posterior predicted probability of predator-specific mortality (median ± 95% Bayesian credibility interval) when examining 2-failure causes from our top model with site-specific compensation and a covariate for predator control at bobwhite nests in the southeastern USA. Years of mesomammal trapping for each site are highlighted by the box.

Download figure to PowerPoint

Discussion

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusions and management implications
  8. Acknowledgements
  9. References

Our study provided evidence that predator reduction in one major group of nest predators (mesomammals) reduced levels of mammalian predation at bobwhite nests; however, increased predation by snakes and other uncontrolled predators resulted in no overall reduction in total annual nest mortality. Thus, management actions to reduce specific nest predators in complex ecosystems may not result in an overall reduction in nest mortality. Additionally, we found evidence that compensation dependency was site specific. Therefore, total annual mortality was affected by other local factors that probably influenced predator–prey relationships.

Mechanisms of compensation

We observed decreased depredations by mesomammals as a result of predator control whereas other uncontrolled predator guild depredations increased. Although there is clear dependence among the causes of nest failure, the mechanism for this relationship is unclear. One explanation is that mesomammal manipulations in the predator community resulted in compensatory predator releases of other predators within the community (e.g. snakes increased because of a reduction in the number of predators that may prey upon them; see summary in Elmhagen & Rushton 2007). Another reason maybe that by removing the more efficient mesomammal nest predators, other less efficient predators or predators with different foraging behaviours (e.g. ants depredating nests at hatching) were given the opportunity to prey upon the nest.

Dietary studies of mesomammal predators, such as bobcats (Neale & Sacks 2001; Godbois, Conner & Warren 2003; Schoch 2003; Lang 2008; Howze 2009), coyotes (Litvaitis 1981; Neale & Sacks 2001), foxes (Neale & Sacks 2001) and opossum (Reynolds 1945; Llewellyn & Uhler 1952) are known to eat snakes. However, given the relatively short time horizon, we observed in response to shifts in causes of mortality relative to predator control, we believe it is unlikely that this was the case. Our initial year of trapping at each site resulted in compensation immediately. If snake predation by mesomammals was reduced then we would expect a time lag in response, unless this was simply a behavioural response to reduced predator numbers.

Interactions among the predators include predators preying upon other predators and also competition among species with different foraging behaviour. For example, predators such as ants generally do not depredate nests until eggs are hatching because they are unable to directly access eggs that were intact (Staller et al. 2005). If nests are available in the landscape longer as the result of reduced mesomammal depredations, then these nests become susceptible to depredations by other predators, which have different foraging techniques. Alternatively, foraging behaviour might change as a result of reduced competition. Therefore, reduction in one species or guild in a complex predator community may have unpredictable or indirect effects on the population size, distribution or availability of resources from other predators (Henke & Bryant 1999; Tewes, Mock & Young 2002), some of which may be more serious threats to ground-nesting species.

Manipulation of predator community

Manipulation of predators to increase reproduction of nesting birds has had highly variable results (Côté & Sutherland 1997; Newton 1998). Most of these studies have examined nest success. However, this commonly used metric is considered to be only an index to annual fecundity and may not necessarily reflect production accurately (Etterson et al. 2011). Therefore, increases in reproduction as the result of partial manipulation of the predator community may not be observed as straightforward increases in nesting success. For example, W. E. Palmer & J. P. Carroll (unpublished data) found increased bobwhite productivity, measured as an increase in chicks/hen, following mesomammalian predator control, despite also finding that nest success remained relatively constant from year to year regardless of predator reductions. As our study demonstrates, reductions in some members of the predator community may result in reduced predation rates by the species targeted, but this may not translate into increased nest success for ground-nesting birds.

In ecosystems with few nest predators and relatively simple community dynamics, such as those of northern boreal forest and agricultural ecosystems, predator reduction efforts have led to increased nesting success of some ground-nesting birds (Chesness, Nelson & Longley 1968; Schranck 1972; Duebbert & Lokemoen 1980; Sargeant, Sovada & Shaffer 1995; Tapper, Potts & Brockless 1996). These simple ecosystems with fewer predators may not exhibit compensations as strongly as complex systems possessing many prey and predator species, such as those in the southeastern USA. The complexity is further complicated by the tremendous number of links within the system when generalist predators, such as we have in our study and commonly found elsewhere, are incorporated in community dynamics models (Closs, Balcombe & Shirley 1999). Although we only examined nests once incubation had begun, our results suggest that nest survival may not be sensitive enough to detect the underlying processes occurring, and therefore this approach is unable to detect the actual effect predator control efforts have upon reproduction and community dynamics.

Furthermore, we focused on the direct effects of predator manipulations on predation rates of a target ground-nesting bird species. We did not address longer-term issues such as cascade events related to disease or behavioural changes (e.g. Packer et al. 2003; Fortin et al. 2005). However, Felege (2010) observed predator occupancy increased to pre-control levels in the following year after predator reductions ceased, suggesting little carry-over of predator control.

Impacts of camera monitoring

Our study will have been susceptible to the impact of camera equipment on nest survival and predator behaviour. Few studies have examined this issue, although Conner, Rutledge & Smith (2010) observed higher nest survival in songbird nests monitored with cameras while Staller et al. (2005) and Coates, Connelly & Delehanty (2008) did not observe any impact of cameras on ground-nesting birds. Nest predators are frequently misidentified when cameras are not employed (Staller et al. 2005; Rader et al. 2007), so it is difficult to quantify the impact of this methodology by comparisons between camera monitored and not monitored nests. However, because both predator reduction sites and control sites had camera monitoring conducted similarly and the study was conducted with a crossover design, the technique is unlikely to have biased our results.

Previous attempts at understanding different causes of nest failure have frequently suffered from an inability to accurately identify predators (King et al. 2001), issues of discovery bias (Etterson, Nagy & Robinson 2007) and inaccurate times of nest failure (Etterson, Nagy & Robinson 2007). Many studies have examined cause-specific failures as independent risks, but without considering potential dependence among failure causes when accounting for discovery bias, suggesting another potential topic for future research (Etterson, Nagy & Robinson 2007). Through the use of radio-telemetry, we were able to monitor almost the entire 23-day incubation period for our nests to reduce discovery bias. However, the expense of video equipment, as well as coupling nest monitoring with radio-telemetry makes comprehensive studies with large sample sizes, such as our study, very rare.

Conclusions and management implications

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusions and management implications
  8. Acknowledgements
  9. References

Predator populations are sometimes reduced to protect nests of ground-nesting birds with the aim of increasing populations. Our results demonstrate that in ecosystems with large and diverse predator communities, the effect of reducing part of the nest predator community results in increased levels of nest failure because of other causes, including other predators. Specifically, total nest failure was insensitive to partial predator removal. In most cases, it is logistically (e.g. fire ants) or legally (e.g. snakes or raptors) impossible to manipulate all predator guilds in complex ecosystems. Each site will differ depending on the predator–prey community structure and the number of potential predators being manipulated. Predicting those cases where predator shifts would be likely to nullify predator control efforts requires a thorough understanding of the predator community, their dynamics and interactions.

In addition, measuring nesting success alone may not be sensitive enough to detect the underlying biological processes occurring with respect to predation and the potential impacts of management actions. Other researchers have found a lack of correlation between nesting success and other reproductive metrics (Murray 2000). Results of our study suggest that these reproductive measures could be decoupled. Therefore, when managing ground-nesting birds using predator control, it is imperative that biologists evaluate cause-specific nest mortality in addition to nest success to determine the impacts of predator reductions on management objectives.

Acknowledgements

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusions and management implications
  8. Acknowledgements
  9. References

We would like to thank the many technicians and interns that radio-tracked bobwhites, maintained cameras and watched hours of video footage. In addition, Shane Wellendorf and D. Clay Sisson provided valuable logistical assistance in data collection. Margaret A. Voss provided comments on earlier drafts. Funding was provided by a Direct Congressional Appropriation to USDA-Wildlife Services, The University of Georgia Graduate School, the Warnell School of Forestry and Natural Resources, McIntire-Stennis Projects GEO 100 and 136, the Albany Quail Project, and Tall Timbers Research Station and Land Conservancy, Inc. Additional funding was provided by the Northeast Georgia Chapter of Quail Unlimited and the American Association of University Women for S.N.E.

References

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusions and management implications
  8. Acknowledgements
  9. References
  • Allen, D.L. (1954) Our Wildlife Legacy. Funk & Wagnalls, New York, NY.
  • Anderson, D.R. & Burnham, K.P. (1976) Population Ecology of the Mallard: VI. The Effect of Exploitation on Survival. Resourse Publication 128. U.S. Fish and Wildlife Service, Washington, D.C.
  • Boyce, M.S., Sinclair, A.R.E. & White, G.C. (1999) Seasonal compensation of predation and harvesting. Oikos, 87, 419426.
  • Burnham, K.P. & Anderson, D.R. (1984) Tests of compensatory vs. additive hypotheses of mortality in Mallards. Ecology, 65, 105112.
  • Burnham, K.P., White, G.C. & Anderson, D.R. (1984) Estimating the effect of hunting on annual survival rates of adult Mallards. Journal of Wildlife Management, 48, 350361.
  • Chesness, R.A., Nelson, M.M. & Longley, W.H. (1968) Effect of predator removal on pheasant reproductive success. Journal of Wildlife Management, 32, 683697.
  • Clark, W.R. (1987) Effects of harvest on annual survival of muskrats. Journal of Wildlife Management, 51, 265272.
  • Closs, G.P., Balcombe, S.R. & Shirley, M.J. (1999) Generalist predators, interaction strength and food-web stability. Advances in Ecological Research, Vol 28 (eds A. H. Fitter & D. Raffaelli), pp. 93126. Academic Press Ltd, London.
  • Coates, P.S., Connelly, J.W. & Delehanty, D.J. (2008) Predators of Greater Sage-Grouse nests identified by video monitoring. Journal of Field Ornithology, 79, 421428.
  • Conner, L.M., Rutledge, J.C. & Smith, L.L. (2010) Effects of mesopredators on nest survival of shrub-nesting songbirds. Journal of Wildlife Management, 74, 7380.
  • Côté, I.M. & Sutherland, W.J. (1997) The effectiveness of removing predators to protect bird populations. Conservation Biology, 11, 395405.
  • Duebbert, H.F. & Lokemoen, J.T. (1980) High duck nesting success in a predator-reduced environment. Journal of Wildlife Management, 44, 428437.
  • Ellis-Felege, S.N., Burnam, J.S., Palmer, W.E., Sisson, D.C., Wellendorf, S.D., Thornton, R.P., Stribling, H.L. & Carroll, J.P. (2008) Cameras identify white-tailed deer depredating Northern Bobwhite nests. Southeastern Naturalist, 7, 562564.
  • Elmhagen, B. & Rushton, S.P. (2007) Trophic control of mesopredators in terrestrial ecosystems; top-down or bottom-up? Ecology Letters, 10, 197206.
  • Errington, P.L. (1946) Predation and vertebrate populations. Quarterly Review of Biology, 21, 144177.
  • Errington, P.L. (1956) Factors limiting vertebrate populations. Science, 124, 304307.
  • Errington, P.L. (1967) Of Predation and Life, 1st edn. Iowa State University Press, Ames, IA.
  • Etterson, M.A., Nagy, L.R. & Robinson, T.R. (2007) Partitioning risk among different causes of nest failure. Auk, 124, 432443.
  • Etterson, M.A., Ellis-Felege, S.N., Evers, D., Gauthier, G., Grzybowski, J.A., Mattson, B.J., Nagy, L.R., Olsen, B.J., Pease, C.M., Post van der Burg, M. & Potvien, A. (2011) Modeling fecundity in birds: conceptual overview, current models and considerations for future developments. Ecological Modelling, 222, 21782190.
  • Felege, S.N.E. (2010) Nest predation ecology of the Northern Bobwhite in the southeastern USA. PhD thesis. The University of Georgia, Athens, GA.
  • Fortin, D., Beyer, H.L., Boyces, M.S., Smith, D.W., Duchesne, T. & Mao, J.S. (2005) Wolves influence elk movements: behavior shapes a trophic cascade in Yellowstone National Park. Ecology, 86, 13201330.
  • Gelman, A. (2004) Exploratory data analysis for complex models. Journal of Computational and Graphical Statistics, 13, 755779.
  • Godbois, I.A., Conner, L.M. & Warren, R.J. (2003) Bobcat diet on an area managed for Northern Bobwhite. Proceeding of the Annual Conference of Southeastern Associations of Fish and Wildlife Agencies, 57, 222227.
  • Henke, S.E. & Bryant, F.C. (1999) Effect of coyote removal on the faunal community in western Texas. Journal of Wildlife Management, 63, 10661081.
  • Howze, M.B. (2009) The effect of predation on white-tailed deer recruitment at the Joseph W. Jones Ecological Research Center. MS Thesis, The University of Georgia, Athens, GA.
  • King, D.I., DeGraaf, R.M., Champlin, P.J. & Champlin, T.B. (2001) A new method for wireless video monitoring of bird nests. Wildlife Society Bulletin, 29, 349353.
  • Lang, M.C. (2008) The effects of intensive predator harvest during the quail nesting season on diet, age, and reproduction of mesomammalian predators. MS Thesis, The University of Georgia, Athens, GA.
  • Link, W.A. & Barker, R.J. (2010) Bayesian Inference with Ecological Applications. Academic Press, Boston.
  • Litvaitis, J.A. (1981) A comparison of coyote and bobcat food habits in the Wichita Mountains, Oklahoma. Proceedings of the Oklahoma Academy of Science, 61, 8182.
  • Llewellyn, L.M. & Uhler, F.M. (1952) The foods of fur animals of the Patuxent Research Refuge, Maryland. American Midland Naturalist, 48, 193203.
  • Murray Jr, B.G. (2000) Measuring annual reproductive success in birds. Condor, 102, 470473.
  • Neale, J.C.C. & Sacks, B.N. (2001) Food habits and space use of gray foxes in relation to sympatric coyotes and bobcats. Canadian Journal of Zoology, 79, 17941800.
  • Newton, I. (1998) Population Limitation in Birds. Academic Press, San Diego.
  • Packer, C., Holt, R.D., Hudson, P.J., Lafferty, K.D. & Dobson, A.P. (2003) Keeping the herds healthy an alert: implications of predator control for infectious disease. Ecology Letters, 6, 797802.
  • Rader, M.J., Teinert, T.W., Brennan, L.A., Hernandez, F., Silvy, N.J. & Wu, X.B. (2007) Identifying predators and nest fates of bobwhites in southern Texas. Journal of Wildlife Management, 71, 16261630.
  • Reynolds, H.C. (1945) Some aspects of the life history and ecology of the opossum in central Missouri. Journal of Mammalogy, 26, 361379.
  • Sargeant, A.B., Sovada, M.A. & Shaffer, T.L. (1995) Seasonal predator removal relative to hatch rate of duck nests in waterfowl populations areas. Wildlife Society Bulletin, 23, 507513.
  • Schoch, B.N. (2003) Diet, age, and reproduction of mesomammalian predators in response to intensive removal during the quail nesting season. MS Thesis, The University of Georgia, Athens, GA.
  • Schranck, B.W. (1972) Waterfowl nest cover and some predation relationships. Journal of Wildlife Management, 36, 182186.
  • Servanty, S., Choquet, R., Baubet, E., Brandt, S., Gaillard, J., Schaub, M., Toïgo, C., Lebreton, J., Buoro, M. & Gimenez, O. (2010) Assessing whether mortality is additive using marked animals: a Bayesian state-space modeling approach. Ecology, 91, 19161923.
  • Sisson, D.C., Terhune, T.M., Stribling, H.L., Sholar, J. & Mitchell, S. (2009) Survival and cause of mortality of Northern Bobwhite (Colinus virginianus) in the Southeastern USA. Gamebird 2006: Quail VI and Perdix XII (eds S.B. Cederbaum, B.C. Faircloth, T.M. Terhune, J.J. Thompson & J.P. Carroll), pp. 467478. Warnell School of Forestry and Natural Resources, Athens, GA.
  • Spiegelhatler, D.J., Best, N.G., Carline, B.P. & van der Linde, A. (2002) Bayesian measures of model complexity and fit. Journal of the Royal Statistical Society, 64, 583639.
  • Staller, E.L., Palmer, W.E., Carroll, J.P., Thornton, R.P. & Sisson, D.C. (2005) Identifying predators at Northern Bobwhite nests. Journal of Wildlife Management, 69, 124132.
  • Stoddard, H.L. (1931) The Bobwhite Quail: Its Habits, Preservation and Increase. C. Scribner’s Sons, New York.
  • Tapper, S.C., Potts, G.R. & Brockless, M.H. (1996) The effect of an experimental reduction in predation pressure on the breeding success and population density of Grey Partridges Perdix perdix. Journal of Applied Ecology, 33, 965978.
  • Taylor, J.S., Church, K.E., Rusch, D.H. & Cary, J.R. (1999) Macrohabitat effects on summer survival, movements, and clutch success of Northern Bobwhite in Kansas. Journal of Wildlife Management, 63, 675685.
  • Terhune, T.M., Sisson, D.C., Palmer, W.E., Stribling, H.L. & Carroll, J.P. (2008) Raptor predation of Northern Bobwhite nests. Journal of Raptor Research, 42, 148151.
  • Tewes, M.E., Mock, J.M. & Young, J.H. (2002) Bobcat predation on quail, birds, and mesomammals. Proceedings of Fifth National Quail Symposium, 5, 6570.