Open Access
Stainbrook, David Philip
Graduate Program:
Wildlife and Fisheries Science
Master of Science
Document Type:
Master Thesis
Date of Defense:
July 28, 2011
Committee Members:
  • Duane R Diefenbach, Thesis Advisor/Co-Advisor
  • Global Positioning System (GPS)
  • Lincoln-Petersen
  • mark-resight
  • Bowden
  • abundance
  • density
  • bias
  • non-uniform transects
  • roads
  • distance sampling
  • Odocoileus virginianus
  • white-tailed deer
  • resource selection
  • habitat use
  • Gettysburg
  • Pennsylvania
The mission at Gettysburg National Military Park and Eisenhower National Historic Site (GNMP-ENHS) is to preserve the historic character of the parks to enable current and future generations to understand and interpret the events that took place at each park. Management objectives include maintaining the landscape as it existed during the historic 1863 Civil War battle (e.g., dense understory in woodlots) in GNMP, and as it existed during the Eisenhowers’ occupancy (e.g., patchwork of cropfields). Browsing by white-tailed deer (Odocoileus virginianus) diminished regeneration of native trees in woodlots and prevented crops from reaching maturity. Thus, to increase regeneration in woodlots and reduce crop damage, the National Park Service (NPS) began culling deer in 1995 to reach a density of 10 deer per square km of forest. However, park managers required an accurate population estimate to determine if this management goal has been met. Therefore, I captured and fitted adult and juvenile male and female deer with Global Positioning System (GPS) collars and performed surveys at dusk and at night, from April 2009 to November 2010, to estimate abundance using mark-resight methods. I found that the average detection probability during the April 2010 dusk count was 0.25, compared to 0.54 from research conducted over 20 years prior. Previous research used only marked female deer, and a number of factors that influence detectability of deer likely changed over time. Park managers can use my updated as their sighting index to estimate deer abundance during future deer counts, which may provide more accurate estimates of abundance. However, factors that influence detection probability can change over time; thus, accuracy of future estimates of abundance could change. Additionally, I conducted distance sampling surveys from roads during mark-resight surveys at night to test assumptions when roads are used as transects with distance sampling. A critical requirement of distance sampling is that transects are placed randomly on the landscape to obtain a representative sampling of the study area and to meet the assumption that the distribution of deer is uniform with respect to perpendicular distances to transects. Roads have been used as transects for distance sampling and provide logistical advantages, but roads may be correlated with habitat characteristics and the distribution of the animals. Distance sampling can be a useful estimator for monitoring abundance; however, if roads are used as transects, the magnitude and direction of the bias are unknown unless information on the distribution of the animal is available. Therefore, I used GPS locations from GPS-collared deer to model the distribution of deer relative to roads using a resource selection function (RSF). During the hours when I conducted spotlight surveys, the distribution of deer was not uniform with respect to the location of roads in both forested and non-forested areas. Deer avoided areas close to roads, were more likely to be found near the park boundary, and selected for forested areas and open areas near forest edges. The estimator of detection probability was positively biased when deer avoided roads; thus, estimates of density in the sampled area were negatively biased. Although I failed to reject the null hypothesis that using roads as transects provided a representative sampling of the study area, extrapolating biased density estimates from the sample to the study area likely resulted in biased estimates of abundance in the study area. Further, estimates of abundance from distance sampling, using roads as transects, were lower than estimates of abundance from mark-resight estimators. Additionally, I demonstrated how the RSF can be used to account for non-random placement of transects to obtain more accurate estimates of abundance. Both mark-resight and distance sampling estimators provided density estimates approximately 4 times greater than the park’s goal of 10 deer per square km of forested land. I estimated abundance for April 2010 as 368 (95% CI = 322 – 421) and density as 43 deer/km2 of forest during dusk mark-resight surveys; 403 (95% CI = 297 – 546; density = 48 deer per square km of forest) during spotlight mark-resight surveys; and 381 (95% CI = 238 – 607; density = 45 deer per square km of forest) during distance sampling surveys. I estimated abundance for November 2010 as 425 (95% CI = 196 – 921; density = 50 deer per square km of forest) during dusk mark-resight surveys; 598 (95% CI = 420 – 852; density = 71 deer per square km of forest) during spotlight mark-resight surveys; and 366 (95% CI = 255 – 525; density = 43 deer per square km of forest) during distance sampling surveys. However, density within the entire study area was not homogenous. I observed more deer on private lands in the study area and fewer deer on NPS owned property. Park staff have observed increased tree regeneration and reduced crop damage since culling was initiated, even though current deer densities are approximately 4 times greater than the goal specified in the GNMP-ENHS deer management plan. Consequently, the NPS may want to consider re-evaluating deer density goals.