Large Mammal Use of Seminatural Grasslands and Implications for Aviation Strike Risk

Wildlife collisions with civil aircraft cost an estimated US$718 million annually in the United States (U.S.) and up to US$2 billion worldwide (Allan 2013; Dolbeer 2013). Deer Odocoileus spp. and coyote Canis latrans accounted for 83.5% of reported damaging wildlife strikes by terrestrial mammals during 1990–2014 among continental U.S. airports (Dolbeer et al. 2015). White-tailed deer O. virginianus are one of the most-hazardous wildlife species for aircraft, with 84% of deer strikes causing damage (839 of 1,001 deer strikes; Wright et al. 1998; Biondi et al. 2011; DeVault et al. 2011). Coyote strikes are less likely to cause damage (42 of 469 coyote strikes [9%] caused damage from 1990 to 2014; Dolbeer et al. 2015). Biologists and airport personnel continue to mitigate wildlife strike hazards using lethal and nonlethal techniques, reducing the total number of strikes within airport environments and total damaging strikes (Dolbeer 2011; Dolbeer et al. 2015). Although removal methods (lethal and nonlethal) and fences are available mitigation methods (DeVault et al. 2008; Biondi et al. 2014), new types of habitat manipulation could expand the mitigation toolbox while offering a long-term solution when other methods are impractical.

Switchgrass monocultures and native warm-season grass polycultures could be unattractive for coyotes and deer (DeVault et al. 2012; Martin et al. 2013; Schmidt et al. 2013). Common airport landscapes of intermittent grasslands, wooded areas, and sometimes row crops, do not deter deer and coyote use (Wright et al. 1998; Blackwell et al. 2009; DeVault et al. 2009). Existing data on mammalian use of native warm-season grass polycultures and switchgrass monocultures are restricted predominately to small mammals (Kamler et al. 2005; Schmidt et al. 2013), despite aircraft strike risk concerns of deer and coyotes (DeVault et al. 2011). Greater small mammal diversity and abundance in native warm-season grasses could suggest greater prey availability for predators like coyotes. Coyotes are omnivorous, so animal prey availability is but one of many factors that influence their use of an area (Chamberlain and Leopold 1999; Fedriani et al. 2001; Hidalgo-Mihart et al. 2004; Randa and Yunger 2006; Randa et al. 2009).

We compared deer and coyote use of experimental switchgrass monoculture fields to that of native warm-season grass polyculture fields. We determined potential hazards to aircraft and whether either treatment demonstrated promise as an alternative land cover for airports according to deer and coyote use. We predicted less use of switchgrass fields than native warm-season grass fields because greater plant species diversity in polycultures would complement deer and coyote diets, such as greater forage quality and increased small mammal prey, respectively (Wright et al. 1998; Blackwell et al. 2009; DeVault et al. 2009; Schmidt et al. 2013).

We conducted fieldwork on B. Bryan Farms, West Point, Mississippi, USA, within the Blackland Prairie geological province (from 33.65° to 33.66°N, 88.56° to 88.57°W; Barone 2005). The site was previously soybean fields surrounded by wooded fencerows, pasture, riparian areas, and forest patches; soil types include Sessum, Leeper, Ozan, and Kipling series (Murphree and Miller 1976). We delineated 16 plots of 5.0–8.4 ha (x̄ = 7.55 ha, σ = 1.08 ha) using a paired design. Although plot sizes were less than home ranges of focus species, plot sizes and distributions represented a typical airport where portions of the airport property could be converted to native grasses (e.g., outside of the airport fence or controlled area) with habitat corridors connecting patches. Similar to airport properties, animals were free to move among plots. We grouped plots by soil type and adjacent habitat type (n = 4 plots/block), and randomly assigned a grass type (native warm-season grass polyculture or switchgrass monoculture) resulting in six plots of a switchgrass Panicum virgatum monoculture and eight plots of native warm-season grass polyculture(e.g., big bluestem Andropogon gerardii, little bluestem Schizachyrium scoparium, Indian grass Sorghastrum nutans, roundheaded lespedeza capitata, greyheaded coneflower Ratibida pinnata, Canada tick trefoil Desmodium canadense, tickseed sunflower Bidens aristosa, Illinois bundleflower Desmanthus illinoensis, and wild blue lupine Lupinus perennis). We originally had eight plots per grass type (eight pairs), but two replicates of switchgrass plots from two different blocks failed to establish after planting in 2011 resulting in fourteen total plots.

We used a photographic index to estimate the rate of occurrence for deer and coyotes for each treatment. We placed two motion-sensor cameras (Reconyx PC800 Hyperfire Professional Semi-covert IR, Holmen, WI) on each plot for 14 consecutive d each month from June to August and November, 2011 and 2012. We chose these sampling months based on data characteristics such as months during which maximum detections among plots occurred, months sampled both study years, perceived maximum population densities in summer after breeding (June to August), and frequent deer movements in Mississippi (November) similar to frequencies of national deer strike incidents (Biondi et al. 2011). We did not have enough cameras to sample all plots, so we randomly selected eight plots to sample first and then rotated cameras monthly to the remaining six plots. We assumed less variability in movements within months than across months. We placed cameras 120 m apart in each plot facing each other and secured them to a steel post approximately 30 cm above ground. We positioned cameras ≥55 m from the plot edge to maximize distance from edge and distance from cameras on adjacent plots. We programmed cameras to take three images when triggered, and set picture intervals to 3 s, with no delay between triggers. We maintained grass height at ≤10 cm between cameras and scent stations to facilitate identification of deer and coyotes. We used five drops of commercial skunk Mephitis mephitis scent on cotton balls placed in a closed canister with air holes attached to a stake 5 m in front of each camera as a weak attractant.

We reviewed images monthly and identified individuals to species. Individual detections were defined as the camera being triggered by ≥1 deer or coyote. If >1 individual was observed in images from a single interval, then the number of detections equaled the number of individuals observed during that interval. We deemed consecutive pictures of the same species independent when either a ≥15-min time interval passed or the individual(s) was identifiably different from the previous photograph (e.g., branch-antlered deer). We assumed minimum effects of detection error by species among plots due to standardized camera stations (skunk scent and cleared vegetation), plot distribution, equal access to plots by animals, and how we determined a detection event (time-interval).

We used a multivariate generalized linear mixed model using package MCMCglmm in Program R (version 2.13.1, The R Foundation for Statistical Computing, Vienna, Austria) to compare coyote and deer use between treatments (Hadfield 2010; Data S1, Supplemental Material). Using MCMCglmm, we assessed responses of both species combined while accounting for temporal autocorrelation using random factors in a Bayesian model-selection framework. This method avoids issues associated with multiple hypothesis testing such as inflated error rate. We specified vague, proper priors for unknown parameters with minimal degree of belief (v = 0.002) emphasizing data-driven parameter estimation instead of strong influence of prior assumptions (Spiegelhalter et al. 2002). We used deviance information criterion to select the best random structure of our treatment model with random effects (e.g., block, plot, year, and month) associated with each observation (idh variance structure) or among all observations (Spiegelhalter et al. 2002). Using the selected random structure, we ran three models (treatment × species, species, and null) for 100,000 iterations, with a burn-in of 50,000 and a thinning interval of 10. We ran each model three times to assess error and convergence (Hadfield 2010). We selected the model with the least average deviance information criterion value. In addition to summary statistics of the posterior distribution, we calculated the proportion of posterior distribution values >0 as a metric of the strength of the directional response when 95% credible intervals overlapped 0. The proportion of posterior distribution values >0 provided an index of the probability of a parameter estimate being greater than or less than 0 (positive or negative response, respectively). We also calculated monthly cumulative hazard score per visit per plot as the sum of each species' occurrence multiplied by their respective hazard score. We used hazard scores assigned to deer and coyotes by Biondi et al. (2014) based on log body mass (g).

We observed 248 unique detections of coyotes and deer among 3,136 camera-days (Table S1, Supplemental Material). Coyote and deer use differed between treatments according to the top model including species-specific responses to treatments (Table S2, Supplemental Material), but uncertainty in the magnitude of effect existed with all posterior distributions overlapping 0 (Table S3, Supplemental Material). Most (58–65%) parameter estimates in posterior distributions were <0, suggesting both species used switchgrass monocultures less than native warm-season grass polycultures. However, coyotes demonstrated a stronger directional response to treatments with more parameter estimates <0 (Table S3, Supplemental Material). Use patterns across months resulted in greater cumulative hazard scores within polycultures, suggesting greater aviation wildlife strike risk than monocultures (Table S1, Supplemental Material). November had substantially greater cumulative hazard in polycultures than monocultures likely due to >3 times more deer occurrences in polycultures than monocultures. Overall, November demonstrated greater variability in deer and coyote use between treatments than summer months (Table S1, Supplemental Material).

Deer and coyotes used native warm-season grass polycultures more than switchgrass monocultures, especially during November, but substantial differences between treatments were lacking. The realized danger that coyotes and deer pose to aircraft prioritizes controlling their use of airport grounds, which may include decreasing the attractiveness of land covers within and surrounding an airport (Biondi et al. 2011). Switchgrass monocultures may be less attractive than native warm-season grass polycultures to deer and coyotes, but differences may only be noticeable in high-animal-use areas or during high-animal-use months (Biondi et al. 2014). Monitoring wildlife year round could indicate peak use times during which targeted mitigation efforts would be most effective (Biondi et al. 2011; Schwarz et al. 2014).

Airports host abundant open areas within and outside air operations areas with turf grasses dominating most actively managed areas (Cleary and Dolbeer 2005; Blackwell et al. 2013; Washburn and Seamans 2013). Within such grassland-dominated landscapes, deer often select wooded areas for bedding and limit use of open areas to nocturnal foraging (Deperno et al. 2002; Rouleau et al. 2002; Volk et al. 2007). Reducing forage availability within open areas by perpetuating monocultures such as switchgrass could differentiate deer and coyote use of switchgrass monocultures versus typical airfield grasslands, which often include multiple forbs and legumes. Typical airport grass management also includes frequent cutting to maintain 15–35-cm (6–14-in) height (Civil Aviation Authority 2008; U. S. Air Force instruction 91-202, 7.11.2.3). In contrast, switchgrass monoculture and native warm-season grass polycultures would introduce tall grass management that may augment deer and coyote movements. Vegetation height management alone may not deter deer and coyote use, but changes in vegetation composition or prey availability may affect deer and coyote use, respectively (Andelt and Andelt 1981; Person and Hirth 1991; Kamler et al. 2005; Walter et al. 2009; Washburn and Begier 2011).

Establishing land covers that deter wildlife on airport grounds could complement other deterrence methods such as fencing and wildlife control to reduce aircraft–wildlife strikes (Martin et al. 2013). Switchgrass monocultures demonstrate potential for reducing hazardous wildlife use compared with native warm-season grass polycultures, but our research only provides a first step toward understanding deer and coyote responses to alternative airport land covers. Additional investigations of deer and coyote use among similar land covers across new landscapes would inform managers of the feasibility and effectiveness of this mitigation approach.

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

Table S1. Average monthly detections of coyotes Canis latrans and white-tailed deer Odocoileus virginianus as detected by 14-d camera-trapping surveys in native warm-season grass polyculture and switchgrass Panicum virgatum monoculture fields near West Point, Mississippi, from June to August and November 2011 and 2012. Cumulative hazard scores weight detections by aviation strike risk. Scores were calculated as the sum of animal occurrence multiplied by their respective aviation hazard scores (Coyote = 62, White-tailed Deer = 94) from Biondi et al. (2014).

Found at DOI: http://dx.doi.org/10.3996/022017-JFWM-019.S1 (12 KB DOCX).

Table S2. Markov chain Monte Carlo multivariate generalized linear mixed-model results comparing coyote Canis latrans and white-tailed deer Odocoileus virginianus use of native warm-season grass polyculture and switchgrass monoculture fields near West Point, Mississippi, sampled with 14-d camera-trapping surveys from June to August and November 2011 and 2012.

Found at DOI: http://dx.doi.org/10.3996/022017-JFWM-019.S2 (12 KB DOCX).

Table S3. Summary statistics of posterior distributions from Markov Chain Monte Carlo multivariate generalized linear mixed model comparing coyote Canis latrans and white-tailed deer Odocoileus virginianus use between switchgrass monocultures and native warm-season grass polycultures collected using 14-d camera-trapping surveys near West Point, Mississippi, June to August and November 2011 and 2012.

Found at DOI: http://dx.doi.org/10.3996/022017-JFWM-019.S3 (13 KB DOCX).

Data S1. Coyote Canis latrans and white-tailed deer Odocoileus virginianus frequency of occurrences and associated aviation hazard scores used for comparing mammal use between switchgrass monocultures and native warm-season grass polycultures collected using 14-d camera-trapping surveys near West Point, Mississippi, June to August and November 2011 and 2012. Archival data includes associated manuscript information, paired study design designations of plot grouping (block) and plot (experimental unit), sampling year, sampling month, associated treatment (switchgrass Panicum virgatum monoculture or native warm-season grass polyculture), coyote monthly frequency of occurrence (COY), white-tailed deer monthly frequency of occurrence (WTD) and associated hazard score (HScoywtd, sum of animal occurrences multiplied by their respective aviation hazard scores [Coyote = 62, White-tailed Deer = 94] from Biondi et al. [2014]).

Found at DOI: http://dx.doi.org/10.3996/022017-JFWM-019.S4 (62 KB DOCX).

Reference S1. Civil Aviation Authority. 2008. CAP 772: birdstrike risk management for aerodromes. Sussex, United Kingdom.

Found at DOI: http://dx.doi.org/10.3996/022017-JFWM-019.S5 (364 KB PDF); also available at <http://www.birdstrike.it/birdstrike/file/images/file/CAP772.pdf>.

Reference S2. Cleary EC., Dolbeer RA. 2005. Wildlife hazard management at airports: a manual for airport personnel. U.S. Department of Agriculture National Wildlife Research Center - Staff Publications, Paper 133.

Found at DOI: http://dx.doi.org/10.3996/022017-JFWM-019.S6 (2968 KB PDF); also available at http://digitalcommons.unl.edu/cgi/viewcontent.cgi?article=1127&context=icwdm_usdanwrc.

Reference S3. DeVault TL, Kubel JE, Rhodes JOE, Dolbeer RA. 2009. Habitat and bird communities at small airports in the midwestern U.S.A. Proceedings of the Wildlife Damage Management Conference 13:137–145.

Found at DOI: http://dx.doi.org/10.3996/022017-JFWM-019.S7 (168 KB PDF); also available at http://digitalcommons.unl.edu/cgi/viewcontent.cgi?article=1134&context=icwdm_wdmconfproc.

Reference S4. Dolbeer RA, Wright SE, Weller JR, Anderson AL, Begier MJ. 2015. Wildlife strikes to Civil Aircraft in the United States, 1990–2014. Washington, D.C.: U. S. Department of Transportation, Federal Aviation Administration.

Found at DOI: http://dx.doi.org/10.3996/022017-JFWM-019.S8 (2676 KB PDF).

Reference S5. Volk MD, Kaufman DW, Kaufman GA. 2007. Diurnal activity and habitat associations of white-tailed deer in tallgrass prairie of eastern Kansas. Transactions of the Kansas Academy of Science 110:145–154.

Found at DOI: http://dx.doi.org/10.3996/022017-JFWM-019.S9 (478 KB PDF).

We thank the U.S. Department of Agriculture Animal and Plant Health Inspection Service Wildlife Services National Wildlife Research Center for their support of this project and cooperative agreement with Mississippi State University and interagency agreement with the U.S. Department of Transportation, Federal Aviation Administration. We are also thankful for assistance in the field and data organization by Matthew Thorton and Tara Conkling and a review by Dr. Bradley Blackwell and three anonymous reviewers. This manuscript is WFA 431 of the Forest and Wildlife Research Center at Mississippi State University.

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