Original Article White-Tailed Deer Fawn Recruitment Before and After Experimental Coyote Removals in Central Georgia WILLIAM D. GULSBY, 1 Warnell School of Forestry and Natural Resources, University of Georgia, Athens, GA 30602, USA CHARLIE H. KILLMASTER, Georgia Department of Natural Resources, Wildlife Resources Division, 2070 United States Highway 278, SE, Social Circle, GA 30025, USA JOHN W. BOWERS, Georgia Department of Natural Resources, Wildlife Resources Division, 2070 United States Highway 278, SE, Social Circle, GA 30025, USA JAMES D. KELLY, New York State Department of Environmental Conservation, Division of Fish, Wildlife, and Marine Resources, 625 Broadway, 5th floor, Albany, NY 12233, USA BENJAMIN N. SACKS, Mammalian Ecology and Conservation Unit, Veterinary Genetics Laboratory, Department of Population Health and Reproduction, University of California, Davis, One Shields Avenue/Old Davis Road, Davis, CA 95616, USA MARK J. STATHAM, Mammalian Ecology and Conservation Unit, Veterinary Genetics Laboratory, University of California, Davis, One Shields Avenue/Old Davis Road, Davis, CA 95616, USA KARL V. MILLER, Warnell School of Forestry and Natural Resources, University of Georgia, Athens, GA 30602, USA ABSTRACT Recent evidence from across the southeastern United States indicating high predation rates by coyotes ( Canis latrans ) on white-tailed deer ( Odocoileus virginianus ) fawns has led some managers to implement coyote control. Although some evidence suggests coyote control can improve recruitment, success appears to be site dependent. Therefore, we designed an experiment to assess feasibility of coyote control as a management action to increase recruitment on B.F. Grant and Cedar Creek Wildlife Management Areas (WMA) in central Georgia, USA. We estimated annual coyote abundance during 2010–2012 using a noninvasive mark–recapture design and fawn recruitment using infrared-triggered camera surveys. During March–June 2011 and March–April 2012, trappers removed coyotes from both sites. Estimates of coyote abundance on B.F. Grant WMA after trapping were 81% (2011) and 24% (2012) lower than during preremoval. Coyote abundance estimates were similar among years on Cedar Creek WMA. Fawn recruitment on B.F. Grant WMA averaged 0.65 fawns/adult female prior to removal and 1.01 fawns/adult female during the 2 years following the removals. Fawn recruitment on Cedar Creek WMA did not differ among years during the study, and was similar to that prior to coyote arrival. The differential coyote impacts and variable effectiveness of trapping we observed on nearby sites suggest coyote control may not achieve management objectives in some areas. Furthermore, transient behavior and the potential for coyotes to adapt to control efforts likely reduce efficacy of this management action. However, we observed an increase in recruitment on B.F. Grant WMA during one year, and others have seen similar responses. Therefore if lowered fawn recruitment is hindering achievement of management objectives, we recommend managers who opt to control coyotes continuously monitor recruitment to determine whether a response occurs. Ó 2015 The Wildlife Society. KEY WORDS Canis latrans , coyote, fawn, fecal genotyping, Odocoileus virginianus , predation, recruitment, trapping, white-tailed deer. Recent observations of declining white-tailed deer ( Odocoileus virginianus ) fawn recruitment coincident with increasing coyote ( Canis latrans ) abundance in some parts of the southeastern United States has prompted a series of investigations aimed at evaluating the impacts of this relatively novel predator on fawns in the Southeast. Researchers have analyzed coyote scat contents (Schrecengost et al. 2008, Kelly 2012), quantified coyote-induced fawn mortality (Cook et al. 1971, Saalfeld and Ditchkoff 2007, Kilgo et al. 2012, McCoy et al. 2013), and examined recruitment before and after coyote removals (Howze et al. 2009, VanGilder et al. 2009) to improve understanding of this predator–prey dynamic. In general, findings indicate that coyotes can be significant fawn predators in the Southeast, although some variation exists among studies. For example, in South Carolina, coyotes depredated as much as 62% of radiocollared fawns (Kilgo et al. 2012), whereas 28% of fawns in Alabama (Saalfeld and Ditchkoff 2007) were depredated. On one central Georgia site, 62% of coyote scats contained fawn remains during the Received: 3 May 2014; Accepted: 13 November 2014 Published: 18 March 2015 1 E-mail: gulsbyw@warnell.uga.edu Wildlife Society Bulletin 39(2):248–255; 2015; DOI: 10.1002/wsb.534 248 Wildlife Society Bulletin 39(2) fawning season versus 27% of scats on a site 8 km away (Kelly 2012). In response to recent evidence indicating high predation rates by coyotes on fawns, managers have become interested in the use of coyote control to increase deer recruitment (Kilgo et al. 2014). Some evidence indicates coyote control can significantly improve fawn recruitment. In southwestern Georgia, the fawn:adult female ratio on an area where coyotes and bobcats ( Lynx rufus ) had been removed was more than twice that observed on an untreated control site during the autumn following the removal (Howze et al. 2009). In northeastern Alabama, the fawn:adult female ratio increased by 200% following removal of coyotes and bobcats from an 800-ha study site (VanGilder et al. 2009). Similarly, net productivity of deer in predator-removal areas was 74% greater than in untreated areas in Texas (Beasom 1974). However, these studies were conducted on single, relatively small study areas, and usually for only 1 year. In contrast, Kilgo et al. (2014) monitored fawn survival for 4 years prior to and 3 years during intensive coyote removal on 3 32-km 2 areas on a study site in western South Carolina. Although fawn survival increased and indices of coyote abundance decreased following trapping, fawn survival differed among years during the removal period and the overall effect of coyote removal on survival was modest. Despite the significant contribution of this work, further experimenta- tion under a variety of conditions is warranted. For example, the authors highlighted the need for additional information on sites with greater deer population density and cover conditions. Furthermore, coyote density in that study was quite high; the average removal rate (1.6 coyotes/km 2 ) was significantly greater than average range-wide density (0.2– 0.4 coyotes/km 2 ; Knowlton 1972). Although interest in coyote control across the Southeast appears to be increasing, the efficacy of this management action is relatively unknown and success appears to be site dependent. Furthermore, the immense resource require- ments associated with intensive and extensive coyote removal and deer monitoring in Southeastern systems makes replication difficult within a single research effort. Therefore, we designed an experiment to assess feasibility of coyote control as a management action to increase fawn recruitment on 2 sites in central Georgia with differing deer densities and landscape features. We monitored relative abundance of coyotes and fawn recruitment on both sites for 1 year before, and each year after, 2 successive years of coyote removal. Our primary objective was to gain additional insight on expected outcomes of coyote removal programs in the Southeast, thereby increasing the ability of managers to make recommendations regarding this issue. STUDY AREAS Research was conducted on portions of B.F. Grant and Cedar Creek Wildlife Management Areas (WMA), which were separated by approximately 8 km and located in the Piedmont physiographic region of Georgia, USA (Fig. 1). Elevation ranges from approximately 120 m to 180 m, and the terrain is gently rolling. Wildlife management on B.F. Grant WMA (owned by the University of Georgia) and Cedar Creek WMA (part of the Oconee National Forest) is overseen by the Georgia Department of Natural Resources through cooperative agreements for hunting, fishing, and outdoor recreation. Although B.F. Grant and Cedar Creek WMA cover approximately 50 km 2 and 160 km 2 , respec- tively, research activities were primarily limited to a 2,000-ha block lying in the interior of each area. The majority of B.F. Grant’s vegetation cover consisted of loblolly pine ( Pinus taeda ) plantations managed on approxi- mately 30-year rotations. As a result, approximately 14% (700 ha) of B.F. Grant’s forested area consisted of early successional forest, generally lasting from the first growing season following timber harvest until canopy closure at approximately 7 years. In addition, the University of Georgia maintained an agricultural research station within the property that consisted primarily of fescue ( Schedonorus arundinaceus ) and Bermuda grass ( Cynodon dactylon ) pastures and hayfields, which also contained a variety of forbs. These pastures and hayfields accounted for an additional 14% (700 ha) of the area. From 1974 until the conclusion of research activities, the Georgia Department of Natural Resources managed B.F. Grant WMA for quality, male white-tailed deer by limiting hunter access through a quota system, and by restricting the harvest of adult males to only Figure 1. B.F. Grant (BFG) and Cedar Creek (CC) Wildlife Management Areas (WMA), Putnam County, Georgia, USA. Research activities involved estimating annual coyote abundance during 2010–2012 using a noninvasive mark–recapture design and fawn recruitment during 2010–2013 using infrared-triggered camera surveys, and were limited to the area north of Highway 212 (area shown) on Cedar Creek WMA. Early successional habitat covered approximately 28% of the study area on B.F. Grant WMA and 7% on Cedar Creek WMA. Gulsby et al. Fawn Recruitment Before and After Coyote Removal 249 19385463a, 2015, 2, Downloaded from https://wildlife.onlinelibrary.wiley.com/doi/10.1002/wsb.534 by University Of Florida, Wiley Online Library on [11/09/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License those with antlers meeting minimal measurement criteria. The combination of productive early successional and field habitats and tightly regulated deer harvest allowed for a moderate-to-high deer density ranging from approximately 19 deer/km 2 to 23 deer/km 2 (C. H. Killmaster, unpublished data). Research on Cedar Creek WMA was limited to areas north of Georgia State Highway 212. Timber management on Cedar Creek WMA was primarily limited to salvage operations (E. Caldwell, United States Forest Service, personal communication). As a result, much of Cedar Creek WMA consisted of mature, closed-canopy forest, and only approximately 7% (420 ha) of the study area consisted of early successional habitat. The Georgia Department of Natural Resources managed the deer population to afford maximum hunter opportunity by allowing increased harvest of juvenile ( 1.5-year old) males and open hunter access during specified dates. The combination of mature forest and increased deer hunting opportunity on Cedar Creek WMA resulted in a deer density ranging from approximately 8 deer/ km 2 to 12 deer/km 2 , roughly half of that on B.F. Grant WMA (C. H. Killmaster, unpublished data). Although the historical record of coyote range expansion into Georgia is coarse, coyotes first occupied the southwest- ern portion of the state during the early 1980s (Hill et al. 1987). Approximately 10 years later, northern and southern coyote populations converged along the Appalachian Mountains and Mid-Atlantic region (Parker 1995). There- fore initial occupation of our study sites by coyotes likely occurred during the late 1980s, with animals becoming relatively abundant sometime during the mid-to-late 1990s. METHODS Estimation of Fawn Recruitment We used data collected by the Georgia Department of Natural Resources at mandatory check stations on each site to calculate indices of fawn recruitment prior to the study period. These data were intended to provide a historical context of fawn recruitment prior to coyote arrival on our sites (1977–1986), during initial occupation (1987–1996), and during the period when coyotes likely became relatively abundant (1997–2008). Specifically, we calculated the ratio of fawns to adult females in the hunter harvest during each 10-year period. Because the only spatial data collected for each deer was the site of harvest (Cedar Creek or B.F. Grant WMA), these indices represented recruitment across the entirety of each site. Therefore, we do not present these data for the study period because research activities were limited to 12.5% and 40% of Cedar Creek and B.F. Grant WMA, respectively. From January 2010 to February 2013, we monitored fawn recruitment (fawns/adult F) on both sites using infrared- triggered cameras positioned over bait, as described by Jacobson et al. (1997). Surveys were conducted during January and February both to avoid the deer hunting season and because accuracy of fawn crop estimates is increased during this period (McKinley 2002). We arranged Cuddeback Capture 1 (Non Typical, Inc., Green Bay, WI) digital trail cameras to cover a 2,000-ha grid on each site at a density of approximately 1 camera/65 ha. We positioned cameras in areas with abundant deer sign (e.g., tracks, trails, feces) near the center of each grid cell. Sites were prebaited with shelled corn for 1 week prior to each survey. After the prebaiting period, cameras were positioned over bait and set on a 15-min delay between photographs. Surveys ran for a 10-day period. We analyzed camera-survey data to estimate fawn recruitment using 2 methods. We used the Jacobson method (Jacobson et al. 1997) of dividing the total number of fawn pictures by the total number of adult female pictures, because this technique was frequently used in prior studies similar to ours. However, the Jacobson method provides no measure of uncertainty for recruitment estimates. Therefore, we also estimated recruitment according to Weckel et al. (2011). This method compares the number of raw photographic occurrences of each sex–age class to their probability of being photographed (i.e., trap success) using linear regression and generates a standardized photographic occurrence for each group. Then, standardized photographic occurrence esti- mates are used to generate standardized demographic ratios. Uncertainty of standardized photographic occurrence and demographic ratio estimates are estimated using 1,000 nonparametric bootstraps of camera stations, and the distribution of standardized photographic occurrence in each demographic group, respectively. The primary objec- tives of this analysis are to account for differences in trap success among sex–age classes and to provide a measure of uncertainty for demographic ratio estimates. Coyote Removal Professional trappers removed coyotes from the area covered by camera surveys within each study site from March to June 2011. During 2012 the trapping period only included March and April because nearly 90% of animals were captured during this period in 2011. Trapping occurred just prior to fawning season, which typically occurs during May and June on our study sites (C. H. Killmaster, unpublished data). Traps were located in areas frequented by coyotes (e.g., along dirt roads, intersections, trails, and firebreaks) and all nontarget species were released unharmed. Coyotes were euthanized via gunshot. Capture and euthanasia procedures were approved by the University of Georgia Institutional Animal Care and Use Committee (A2009 09-157-Y3-A0). We recorded the date of capture and sex of each coyote trapped. We collected 3–5 g of tongue tissue postmortem for genetic analysis and placed them into a Fisherbrand 1 15-mL polystyrene centrifuge tube (Fisher Scientific, Pittsburgh, PA) filled with 9 mL of 95% EtOH. We extracted the left mandibular canine tooth from each coyote and submitted them to Matson’s Laboratory LLC (Missoula, MT) for age determination via cementum annuli analysis as previously described (Linhart and Knowlton 1967). Estimation of Coyote Abundance Beginning in January 2010, we collected putative coyote scats on a series of permanently identified transects along unpaved 250 Wildlife Society Bulletin 39(2) 19385463a, 2015, 2, Downloaded from https://wildlife.onlinelibrary.wiley.com/doi/10.1002/wsb.534 by University Of Florida, Wiley Online Library on [11/09/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License roads and trails in each area covered by camera surveys. We drove along each transect approximately weekly and collected all scats detected. We evaluated species of origin in the field through evaluation of scat size, shape, and contents. For each scat we recorded the date of collection and estimated age, based on moisture content and degree of decomposition. For each scat estimated to be 3 days old we preserved a sample for genetic confirmation of species of origin and genotyping to the individual level. We placed approximately 3 mL of fecal matter, collected from the outside edges of each scat, into a Fisherbrand 1 15-mL polystyrene centrifuge tube filled with 9 mL of 95% EtOH. DNA was extracted from tissue using the Qiagen DNeasy blood and tissue kit and from feces using the Qiagen QiaAmp Stool Kit (Qiagen, Germantown, MD). Specifically, we PCR-amplified a portion of the cytochrome b using the primer pair RF14724 and RF15149, PCR chemistry, and thermocycle conditions of Perrine et al. (2007), followed by BLAST search in Genbank. Individual genotypes and genetic sex were determined based on 12 microsatellites (AHT137, AHT142, AHTh171, CPH18, CXX-279, CXX-374, CXX-468, CXX-602, INU055, REN162C04, REN169O18, REN54P11) and a sex marker (based on X and Y chromosome paralogs of the amelogenin gene; Moore et al. 2010). All fecal genotypes were replicated twice. We calculated the allelic dropout rate by dividing the number of dropouts by the number of successful replicates for each heterozygous locus (with heterozygosity determined from consensus genotypes). Finally, we calculated the probability of identity for each locus and for increasing combinations of the 12 loci using GenAlEx 6.5 (Peakall and Smouse 2006, 2012). It was impossible to collect enough scats in a sufficiently short time to assume a closed population. Even open population models require periods of closed-population sampling. Therefore, we used recapture data as though it were obtained from a closed population to calculate (presumably biased) estimates of abundance, but used them strictly to index the relative abundances of coyotes among time periods and study areas (for which deviations from population closure were presumably similar). We used 2 approaches. First, we applied a simple closed-capture design in Program MARK (White and Burnham 1999). We constructed encounter histories for each coyote identified on each site. No individual was encountered on both sites. Although scats were collected weekly to minimize degradation of fecal DNA and increase genotyp- ing success, we divided each calendar year into 3 biologically relevant seasons, with each season considered a sampling occasion. Seasons were defined as breeding (January–April), denning and pup-rearing (May–July), and dispersal (August–December). For each site, we computed these estimates during 3 time periods: January 2010 to Febru- ary 2011 (prior to coyote removal), June 2011 to February 2012 (year following first coyote removal), and during April 2012 (year following second coyote removal). Although some individuals were encountered more than once during a given sampling occasion, we treated these instances as a single encounter. Second, we analyzed encounter histories during the previously described periods using a rarefaction approach in CAPWIRE (Miller and Waits 2005). This method of population estimation takes advantage of multiple captures of an individual within a session and provides estimates with small bias and good coverage, along with high accuracy and precision, even when the data contain capture heterogeneity. We used estimates of coyote abundance as an index to the approximate number of coyotes using each site during a given time period, rather than as a density estimate. Second, we analyzed encounter histories during the previously described periods using a rarefaction approach in CAPWIRE (Miller and Waits 2005). This method of population estimation takes advantage of multiple captures of an individual within a session and provides estimates with small bias and good coverage, along with high accuracy and precision, even when the data contain capture heterogeneity. We used estimates of coyote abundance as an index to the approximate number of coyotes using each site during a given time period, rather than as a density estimate. RESULTS Fawn Recruitment Estimates Check-station data indicated a hunter harvest of 0.83 fawns/ adult female on Cedar Creek WMA between 1977 and 1986 (Fig. 2a). Although this ratio appeared to decline to 0.65 fawns/adult female between 1997 and 2008, confidence limits overlapped. Jacobson estimates of fawn recruitment on Cedar Creek WMA averaged 0.83 during the 2 years prior to coyote removal and 0.77 fawns/adult female following removal. Weckel estimates similarly averaged 0.84 fawns/ adult female before removal and 0.85 fawns/adult female after removal. Both estimators indicated that throughout the study, recruitment on Cedar Creek WMA was similar to the average ratio of fawns to adult females in the harvest for each 10-year period from 1977 to 2008 (Fig. 2a). Between 1977 and 1996, check-station data on B.F. Grant WMA indicated an average hunter harvest of 0.87–0.89 fawns/adult female (Fig. 2b). However, from 1997 to 2008, concurrent with coyote expansion in the region, the ratio of fawns to adult females in the harvest trended downward to an average of 0.63 fawns/adult female. Jacobson estimates of fawn recruitment on B.F. Grant WMA averaged 0.54 fawns/ adult female during the 2 years before coyote removal and 0.85 fawns/adult female after removal (Fig. 2b). Weckel estimates averaged 0.65 fawns/adult female before removal and 1.01 fawns/adult female after coyote removal. Recruit- ment on B.F. Grant WMA during 2012 (year following first removal) appeared to be greater than during preremoval (Weckel ¼ 1.07 fawns/ad F; Jacobson ¼ 0.93 fawns/ad F), but confidence limits of the Weckel estimate slightly overlapped those of 2010, making this conclusion somewhat uncertain. Recruitment declined during 2013, and did not differ from pretreatment. Coyote Removal Trappers removed 9 coyotes from Cedar Creek WMA in 2011. Seven (78%) were captured during March, 1 in April, Gulsby et al. Fawn Recruitment Before and After Coyote Removal 251 19385463a, 2015, 2, Downloaded from https://wildlife.onlinelibrary.wiley.com/doi/10.1002/wsb.534 by University Of Florida, Wiley Online Library on [11/09/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License and 1 in May. Six (67%) were male and the remaining 3 (33%) were female. Eight of the animals were successfully aged and ranged from < 1 year old to 6 years old. Four (44%) of the captured coyotes had been previously detected via fecal genotyping. The number of previous encounters of these 4 coyotes ranged from 1 to 4. Only 1 coyote, a yearling female, was captured on Cedar Creek WMA during 2012. She was captured during March and had never been previously encountered. Trappers removed 15 coyotes from B.F. Grant WMA during 2011 and 6 during 2012. During 2011, 12 of the 15 (80%) were captured during March. Seven (47%) of the coyotes were female and 8 (53%) were male. Ages of 14 of the animals ranged from < 1 year old to 5 years old. Ten (67%) of the coyotes had not been detected via fecal genotyping prior to the removal. Of the 5 (33%) that were previously detected, the number of previous detections ranged from 1 to 7. All 6 coyotes removed from B.F. Grant WMA in 2012 were captured during March. Three were male, 3 were female, and all 6 were 1 year old. One of the coyotes was previously detected via fecal genotyping on 2 occasions prior to capture; the others were never previously encountered. Coyote Abundance Estimates We collected 340 scats on B.F. Grant WMA over the course of the study. Of the scats collected, 238 (70%) were 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 a Weckel Jacobson Harvest rao 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 77–86 87–96 97–08 2010 2011 2012 2013 b Fawn recruitment Figure 2. Estimates of fawn recruitment on (a) Cedar Creek and (b) B.F. Grant Wildlife Management Areas in Putnam County, Georgia, USA. Estimates were based on the ratio of fawns to adult females in the hunter harvest from 1977 to 2008 and camera surveys from 2010 to 2013. Camera surveys were conducted according to Jacobson et al. (1997) and results were analyzed according to Jacobson et al. (1997) and Weckel et al. (2011). Error bars represent 95% confidence limits and arrows indicate timing of coyote removals conducted by professional trappers. 252 Wildlife Society Bulletin 39(2) 19385463a, 2015, 2, Downloaded from https://wildlife.onlinelibrary.wiley.com/doi/10.1002/wsb.534 by University Of Florida, Wiley Online Library on [11/09/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License sufficiently fresh for genetic analysis. The species of origin could not be determined for 72 (30%) of those scats. Of the remaining 166 scats, 51 (31%) were from bobcats, 13 (8%) were from gray foxes ( Urocyon cinereoargenteus ), and 102 (61%) were from coyotes. Genetic analysis of these scats yielded 68 useable coyote genotypes representing 38 unique individuals. Thus, 29% of scats collected on B.F. Grant WMA and analyzed were confirmed as coyote scats and successfully genotyped to the individual level. Of the 235 scats collected on Cedar Creek WMA, 196 (83%) were used for genetic analysis. The species of origin could not be determined for 46 (23%) of those scats. Of the remaining 150 scats, 3 (2%) were from bobcats, 30 (20%) were from gray foxes, and 117 (78%) were from coyotes. Genetic analysis of these scats yielded 68 useable coyote genotypes representing 32 unique individuals. Thus, 35% of scats collected on Cedar Creek WMA and analyzed were confirmed as coyote scats and genotyped to the individual level. The average allelic dropout rate across loci was 0.28 alleles/ replicate. On average, individual genotypes were constructed from genotyping results at 11.24 microsatellite loci (range ¼ 7–12). Using all 12 loci, probability of identity was 5.8 10 15 and probability of identity for siblings was 6.4 10 6 . Probability of identity at 7 loci (min. no. of loci at which an individual was genotyped) was 5.0 10 8 overall and 9.4 10 4 for siblings. Estimates of preremoval coyote abundance were similar between sites (Table 1). Both MARK and CAPWIRE population estimates indicated that the coyote population on B.F. Grant WMA declined from 2010 to 2011 (based on nonoverlapping confidence limits), following the initial trapping period. Although confidence intervals were wide in 2012 on account of small sample size, both estimators suggested that coyote numbers increased from 2011 to 2012. Based on point estimates from MARK, coyote numbers on B.F. Grant WMA decreased approximately 80% from 2010 to 2011, then increased by 75% from 2011 to 2012 to levels similar to those observed prior to coyote removal (Table 1). Both abundance estimators indicated a moderate decline in coyote numbers on Cedar Creek WMA following both trapping periods, although confidence limits overlapped across all years and thus abundance was likely constant (Table 1). Thus, it appeared that trapping efforts were more successful on B.F. Grant than Cedar Creek, resulting in a greater reduction in coyote numbers, particularly during 2011. DISCUSSION Coyotes were likely never an important source of fawn mortality on Cedar Creek WMA. Preremoval estimates of fawn recruitment (2010–2011) were similar to the ratio of fawns to adult females in the annual harvest during 1977–1986—prior to suspected coyote occupation of the site. In contrast, we suspect that coyotes may have been limiting recruitment on B.F. Grant WMA based on 3 lines of evidence. First, the ratio of fawns to adult females in the annual harvest declined from around 0.9 fawns/adult female to approximately 0.65 fawns/adult female during 1997–2008, when coyotes likely became relatively abundant on the site based on the best available evidence. Second, this lowered recruitment rate remained similar throughout the first 2 years of our study (2010–2011), only increasing after an 80% reduction in relative coyote abundance during 2012. Third, recruitment trended back down following a subsequent increase in relative coyote abundance during 2013. Because we did not monitor fawn survival or mortality causes, we cannot conclude with absolute certainty that the increase in recruitment on B.F. Grant WMA during 2012 was a direct result of coyote removal. However, recent literature does support this assertion. Coyotes were the leading cause of fawn mortality on study sites in Alabama (Saalfeld and Ditchkoff 2007, Jackson 2011), South Carolina (Kilgo et al. 2012, McCoy et al. 2013), North Carolina (Chitwood 2014), and Georgia (Nelson 2013). When coyotes are the primary predator on white-tailed deer populations below carrying capacity, predation is typically additive to other sources of mortality (Ballard et al. 2001). Furthermore, noncoyote-related fawn mortality did not compensate for a reduction in coyote predation on 1 South Carolina site (Kilgo et al. 2014), and fawn survival was inversely related to indices of predator abundance on another South Carolina site (McCoy et al. 2013). Our finding that recruitment tended to be greatest on B.F. Grant WMA during 2012 when coyote relative abundance was lowest is, therefore, consistent with these results. We believe the lack of a treatment response (i.e., increased recruitment) on B.F. Grant WMA during 2013 resulted from the rebound in coyote abundance to nearly pretreat- ment levels. Approximately 30%–50% of Southeastern Table 1. MARK and CAPWIRE estimates of coyote abundance based on noninvasive mark–recapture using fecal genotyping on B.F. Grant and Cedar Creek Wildlife Management Areas in Putnam County, Georgia, USA, from 2010 to 2012. Closed model CAPWIRE N Lower 95% Upper 95% N Lower 95% Upper 95% B.F. Grant 2010 21 19 33 24 18 32 2011 4 4 17 8 4 14 2012 16 7 82 18 6 50 Cedar Creek 2010 16 15 28 22 15 30 2011 9 5 40 16 6 36 2012 9 7 22 14 8 22 Gulsby et al. Fawn Recruitment Before and After Coyote Removal 253 19385463a, 2015, 2, Downloaded from https://wildlife.onlinelibrary.wiley.com/doi/10.1002/wsb.534 by University Of Florida, Wiley Online Library on [11/09/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License coyotes are transients or nomads (Chamberlain et al. 2000, Hinton et al. 2012, Hickman 2014), and likely serve as population founders in areas vacated by coyotes following removal efforts (Hickman 2014). Given our estimated 80% reduction in coyote abundance during 2012, any transient animals in the surrounding area likely detected the vacancy and recolonized the site quickly. Although the effects of recolonization should be offset by additional removal, we removed less than half the number of animals during 2013 as in 2012 despite consistent trapping effort and personnel. This reduction in trapping efficacy may have resulted from learned avoidance behavior by coyotes exposed to trapping but not captured the previous year; however, we are uncertain of its cause. Therefore, we present this finding as an important consideration for managers that efficacy of coyote control may vary among years, regardless of trapping effort. Our finding that recruitment remained relatively constant on Cedar Creek WMA is not surprising given our inability to alter coyote abundance on the site. However, we did not expect to observe recruitment rates similar to those prior to coyote arrival, or greater recruitment than on B.F. Grant WMA despite similar preremoval coyote abundance. We hypothesize that relatively low coyote predation on Cedar Creek WMA was perhaps linked to the site’s deer density and landscape characteristics. Optimal foraging theory suggests that animals select food items according to their profitability, and lower ranking food items are consumed in order of decreasing profitability as higher ranking food items fall below a threshold abundance (Charnov 1976, Stephens and Krebs 1986). Because deer density on Cedar Creek WMA was approximately half that of B.F. Grant WMA, profitability of fawns as a coyote food item could have been lower there, leading to decreased targeting of fawns by coyotes. Perhaps coyotes also used Cedar Creek WMA less intensively because of the lack of habitat diversity, particularly early successional vegetation types, which are preferred both by coyotes (Holzman et al. 1992, Chamber- lain et al. 2000, Kays et al. 2008, Schrecengost et al. 2009) and many small mammal species important in their diet (Atkeson and Johnson 1979). Our findings on B.F. Grant WMA are similar to those of previous coyote removal studies in Georgia (Howze et al. 2009), Texas (Beasom 1974), and Alabama (VanGilder et al. 2009), which reported an increase in fawn recruitment following intensive coyote removal. However, despite intensive trapping efforts, we were not able to significantly decrease coyote abundance on Cedar Creek WMA throughout the study period or on B.F. Grant WMA during 2012. This apparently led to unequal responses in recruitment rates between sites and among years, a finding which is not unprecedented. In South Carolina, the effect of 3 consecutive years of coyote removal on fawn survival varied among years (Kilgo et al. 2014). In fact, fawn survival was similar to the pretreatment level during the second year of coyote removal. In that study, the authors cited immigration of dispersing coyotes and low success at removing the coyotes responsible for the majority of predation (alphas) as potential causes. We acknowledge the possibility that our treatment effects were confounded by annual variation in fawn recruitment unrelated to coyote predation. However, we believe this was unlikely because recruitment on B.F. Grant WMA appeared to closely track our estimates of relative coyote abundance and recruitment remained constant among years on Cedar Creek WMA where coyote abundance did not differ. Future research assessing the effect of coyote removal on deer populations would benefit from quantification of mortality attributable to coyotes as in Kilgo et al. (2014), as well as replication of treatment and control sites. However, the immense costs and other resources associated with intensive and extensive coyote control, deer capture, and deer monitoring often prohibit such endeavors. Therefore, interpretation of results from this study and others should be made carefully and within the context of the entire body of literature. MANAGEMENT IMPLICATIONS Our results indicate that coyote predation may vary across relatively small spatial scales, limiting some deer populations and not others. Thus, monitoring population parameters using camera surveys or other techniques is requisite prior to consideration or implementation of a coyote control program. However, the success of trapping programs at sufficiently reducing coyote abundance to increase fawn recruitment may be inconsistent among sites and years. These considerations, combined with the significant cost of large-scale, intensive trapping programs, suggest coyote control may not be prudent in some scenarios and other options, such as reduced antlerless deer harvest, should be considered. However, if coyote control is implemented, our data indicate that trapping should be intense, focused on removal of a significant percentage of the coyote population, and conducted annually to maintain levels of fawn recruitment observed prior to the establishment of abundant coyote populations. ACKNOWLEDGMENTS Funding for this project was provided by the Georgia Wildlife Resources Division through the Wildlife Restora- tion Program, which derives monies through an excise tax on sporting arms and ammunition paid by hunters and recreational shooters. Furbearers Unlimited, Inc. provided additional financial assistance. The Georgia Trapper’s Association, including R. Johnson, T. Key, and J. Lee, provided trapping services and consultation throughout the study. Additional funding was contributed by the UC Davis, Mammalian Ecology and Conservation Unit. Field support was provided by J. Hickman, C. Paschal, F. Hays, F. Mahone, D. Thompson, and E. Caldwell. We also thank M. Chamberlain, J. Kilgo, and J. Nairn for providing prelimi- nary reviews of this manuscript. Finally, we thank R. Boertje and 2 anonymous reviewers for their thoughtful reviews and improvements to earlier drafts. 254 Wildlife Society Bulletin 39(2) 19385463a, 2015, 2, Downloaded from https://wildlife.onlinelibrary.wiley.com/doi/10.1002/wsb.534 by University Of Florida, Wiley Online Library on [11/09/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License LITERATURE CITED Atkeson, T. D., and A. S. Johnson. 1979. Succession of small mammals on pine plantations in the Georgia Piedmont. American Midland Naturalist 101:385–392. Ballard, W. B., D. Lutz, T. W. Keegan, L. H. Carpenter, and J. C. deVos, Jr. 2001. Deer-predator relationships: a review of recent North American studies with emphasis on mule and black-tailed deer. Wildlife Society Bulletin 29:99–115. Beasom, S. L. 1974. Intensive short-term predator removal as a game management tool. Transactions of the North American Wildlife and Natural Resources Conference 39:230–240. Chamberlain, M. J., C. D. Lovell, and B. D. Leopold. 2000. Spatial-use patterns, movements, and interactions among adult coyotes in central Mississippi. Canadian Journal of Zoology 78:2087–2095. Charnov, E. L. 1976. Optimal foraging: attack strategy of a mantid. American Midland Naturalist 110:127–154. Chitwood, M. C. 2014. White-tailed deer population dynamics in the presence of a novel predator. Dissertation, North Carolina State University, Raleigh, USA. Cook, R. S., M. White, D. O. Trainer, and W. C. Glazener. 1971. Mortality of young white-tailed deer fawns in South Texas. Journal of Wildlife Management 35:47–56. Hickman, J. E. 2014. Home range, habitat use, and movement patterns of female coyotes in the Piedmont region of Georgia: implications for fawn predation. Thesis, University of Georgia, Athens, USA. Hill, E. P., P. W. Sumner, and J. B. Wooding. 1987. Human influences on range expansion of coyotes in the Southeast. Wildlife Society Bulletin 15:521–524. Hinton, J. W., M. J. Chamberlain, and F. T. Van Manen. 2012. Long- distance movements of transient coyotes in eastern North Carolina. American Midland Naturalist 168:281–288. Holzman, S., M. J. Conroy, and J. Pickering. 1992. Home range, movements, and habitat use of coyote