A Rapid Habitat Assessment Sampling Design and Protocol to Monitor Forest Vegetation - Diefenbach and Fritsky
Effects of Demographic Manipulation on Dispersal and Breeding Ecology of Male White-tailed Deer in Pennsylvania - Long and Diefenbach
Use of Reclaimed Surface Mines by Grassland Birds - Mattice and Diefenbach
Distribution and Coarse-Scale Habitat Association of Snowshoe Hares in Pennsylvania - Diefenbach, Rathbun, and Vreeland
The Ability of Aerial Surveys using Thermal Infrared Imagery to detect changes in abundance of white-tailed deer - Diefenbach
Investigation of the Use of Catch-Effort Models to Estimate Abundance of White-tailed Deer at Fort Indiantown Gap National Guard Training Center - Diefenbach and Vreeland
Hunter movement activities and opinions on public lands in Pennsylvania - Diefenbach, Finley, Luloff, San Julian, Stedman, Zinn, and Swope
The Implications of Inter-annual Movements of a Migratory Songbird on Annual Survival Estimation - Marshall and Diefenbach
Survival Rates, Cause-specific Mortality, and Landscape Influence on Survival of White-tailed Deer Fawns in Northcentral Pennsylvania - Vreeland, Diefenbach, and Walllingford
Population Demographics of a Suppressed Wild Turkey Population in Pennsylvania - Lowles, Diefenbach, and Casalena
Dr. Duane R. Diefenbach, Principal Investigator
Richard S. Fritsky, Research Associate
The final report is available here in pdf format (Appendix A and Appendix B).
Eleven DMAP areas were selected for this study that were located in the Moshannon, Susquehannock, Tioga, Elk, Bald Eagle, Tuscarora, Delaware, Loyalsock, Micheaux, Sproul, and Gallitzin state forests and encompassed 311 square miles. The sampling design was a 2-stage design. First, we delineated square-mile blocks across each study area and 54-100% of blocks were selected to be sampled. Second, within each square-mile block we visited 10 sample sites. Thus, there were two sources of variability that needed to be accounted for in the estimation of variances of parameters: variability among blocks and variability among sample points (within blocks).
At each sample point we collected data on tree basal area and diameter at breast height (dbh) by species (via a prism plot), stem densities of shrubs and saplings by species (>1.5 m tall and <10 cm dbh; 1.5 m × 40 m plot), stem densities of tree seedlings by species (30–150 cm tall; 1.5-m radius plot), whether each tree seedlings had been browsed by deer, counts of Indian cucumber, trillium, Canada mayflower, and Jack-in-the-pulpit (1.5-m radius plot), heights of the tallest individual of each of the four flower species (if present), and percent cover of Rubus, grasses, ferns, and forbs (3.5-m radius plot).
During summer 2006 we sampled 234 blocks (square miles) across the 11 DMAP areas using three 2-person teams. Within each DMAP area we sampled 90–100% of the blocks on smaller areas (5–20 square miles) and >54% of the largest areas (<116 square miles). Vegetation data were collected at more than 2,000 sample points. We intentionally over-sampled blocks to obtain sufficient data to evaluate the statistical precision of results and improve the efficiency of the sampling design.
The precision of estimates ranged from good to poor depending on the vegetation characteristic being measured. The coefficient of variation (CV = SE/mean×100%) is a measure of precision, in which a CV = 20–25% is considered sufficient for management decisions. We formally evaluated the statistical power of this sampling design to detect changes in tree seedling stem densities and heights of Indian cucumber.
For tree seedlings, the precision of stem density estimates was poor (CV = 43–95%). However, we found that one had a >80% chance (α = 0.05) of detecting increases of >800 stems/acre if current stem densities were <400 stems/acre. On sites with greater seedling stem densities one is unlikely to be able to detect even large changes in stem density, but sites with >1,000 stems/acre already are likely to have good advanced tree regeneration.
The precision of counts of Indian cucumber (mean = <0.1–5.4 plants/plot), the most abundant and widespread flower in this study, were poor (CV = 60–223%) but the precision of percent of plots occupied by this species were better (plots occupied = 2–43%; CV = 34–224%). Mean heights of the tallest Indian cucumber plant had the best precision (CV = 28–63%); however, mean heights were small (7.0 to 21.3 cm) compared to the reported height for typical specimens of this species (20–90 cm). We estimated that this sampling design could have a >80% chance (statistical power) of detecting height increases of 8–30 cm depending on the DMAP area.
The precision of percent of plots adequately stocked with advanced tree regeneration was poor (mean adequately stocked = 10–72%; CV = 26–107%), but most sites had <20% of plots adequately stocked, which explains the large CVs (e.g., Susquehannock SF had a CV = 26% and 72% of plots adequately stocked) and suggests that substantial changes in the amount of advanced regeneration could be detected. Similarly, the precision of counts of Canada mayflower had poor precision (CV = 52–522%), but given that few plants were encountered that were flowering we believe substantial changes in the ratio of flowering to non-flowering plants may be possible to detect and we believe this might be a suitable indicator of forest vegetation conditions.
It is possible that if a paired difference or repeated measures statistical analysis were conducted on data that represent forest conditions at two or more points in time that this sampling design would have greater statistical power to detect differences. Because this study only had data from one point in time, however, we could not evaluate the statistical power of such analyses. We believe that our power analyses were conservative and that the true ability of this sampling design to detect changes in forest vegetation may be equal to, or better, than what is presented in this report.
A consistent pattern among all measurements was that variability among blocks was almost inconsequential compared to the variability among sample points (block:plot variance ratios <<1). This means that sampling plots across each square mile block captured much of the heterogeneity in the landscape (which occurred at a fairly local scale), such that the vegetation characteristics averaged across square-mile blocks was similar among blocks. Thus, our recommendation for more efficient sampling is to reduce the number of blocks visited and to increase the number of sample points within each block. For DMAP areas of <20 square miles, we recommend visiting five blocks (sampling fraction >25%). For larger DMAP areas, visit an additional block for every additional 10 square miles of area above 20 square miles (sampling fraction >20% for 30 mi2, >16% for 50 mi2, etc.). Also, we recommend each block contain 20 sample points (instead of 10). These changes to the sampling design greatly reduce the number of blocks that need to be visited but result in equivalent precision of estimates at reduced cost.
Under the proposed sampling design, we believe a trained, 2-person crew could sample about five blocks per week. Thus, on smaller DMAP areas (<20 square miles) two people could sample five blocks in less than eight 10-hour days. To minimize the effects of phenological changes on vegetation measurements, we recommend surveys be conducted during June-August and that when an area is re-sampled that the re-visit be conducted within two weeks of the date it was previously sampled. Surveys could probably be conducted every 3–4 years, but costs, management or research objectives, and logistical issues greatly affect the optimal choice for time intervals between samples and we cannot provide specific guidelines based solely on the results of this study.
We estimate it would cost about $15,000–$20,000 each summer data were collected, which would include a two-person crew for about 800–1,000 hours and 5,000 vehicle miles. This crew could likely sample 50–60 square-mile blocks during a summer. Additional expenses would involve database management and data analysis but possibly could be performed by existing staff if an operational database management system were developed. Also, the agency could consider incorporating some of these measurements in their stand analysis protocols.
We recommend retaining the following data collection in the protocols:
· Tree (>10 cm dbh) basal area and dbh to be able to calculate overstory stocking and assess understory light conditions;
· Stem density, by species, of shrubs and saplings >1.5 m tall and <10 cm dbh to assess advanced tree regeneration and identify sites with problems with interference vegetation;
· Percent cover of Rubus, ferns, grasses, and forbs primarily to identify sites with >25% fern cover and potential tree regeneration problems;
· Stem density of tree seedlings (30–150 cm tall), by species, to assess advanced tree regeneration;
· Counts of Indian cucumber and Canada mayflower, and to record the number of flowering Canada mayflower; and
· Height of the tallest Indian cucumber on each plot.
Under the proposed sampling protocols, the following forest vegetation indicators could be monitored:
· Percent of plots adequately stocked with advanced tree regeneration on plots with <25% fern cover, <1,000 stems/acre of interference shrubs and saplings, and <75% overstory stocking;
· Stem density of tree seedlings 30–150 cm tall, which could also account plots with interference vegetation and inadequate overstory conditions;
· Counts of Indian cucumber and Canada mayflower;
· Percent of Canada mayflower plants that are flowering; and
· Height of Indian cucumber.
This study by itself does not provide any direct information on the effects of deer browsing on forest vegetation conditions. Furthermore, we do not know by how much the measures that were chosen for this study will actually respond to changes in deer browsing as influenced by changes in deer density. For example, will percent of flowering Canada mayflower increase by 10% or 50% for a given reduction in deer density?
To further refine a vegetation monitoring program based on the recommendations presented in this report, changes in deer density are required during which repeated vegetation measurements are collected. We believe DCNR lands enrolled in DMAP are large enough for such an endeavor. However, there are some challenges. First, such an undertaking requires a long-term perspective and commitment because vegetation responses may require >10 years, although responses by some species of forest herbs may occur sooner. Second, hunter harvest is the single greatest mortality factor for deer in Pennsylvania, and an accurate accounting of hunter harvest would permit stronger inferences about changes in deer densities. Third, it may be necessary to install deer exclosures on the study area to make sure that reduced deer densities should result in a detectable change in vegetation and to identify what type of changes should be expected to occur.
The vegetation monitoring protocol proposed in this report would be fundamental to any attempt to perform forest restoration in a management-research (i.e., adaptive resource management) context. That is, deer and forest land management decisions would be accompanied by a monitoring program so that outcomes could be assessed in a quantitative, objective manner. As monitoring proceeds new data are collected to evaluate and help refine management decisions as well as improve our understanding of how the ecosystem being managed functions. In this context, deer management, forest vegetation monitoring, and land management decisions are all integrated along with a research component.
Effects
of demographic manipulation on dispersal and breeding ecology of male
white-tailed deer in PennsylvaniaEric S. Long, Graduate Research Assistant
Dr. Duane R. Diefenbach, Advisor
Information on this project on the Pennsylvania Game Commission website can be found here
Dissertation based on this project is available here
In Pennsylvania, as in many other states, historically intense hunting pressure has focused on adult male white-tailed deer (Odocoileus virginianus), and disproportionate harvest has resulted in populations demonstrating skewed sex- and age-structures with abundant does and relatively few adult bucks. In October 2002, the Pennsylvania Game Commission instituted state-wide antler-point restrictions that reduced annual huntingrelated mortality rates of yearling (i.e., 18-month-old) male white-tailed deer from approximately 80% to 32%. In subsequent years, this management change doubled the number of 2.5-year-old bucks in the population and decreased the ratio of yearling bucks to older bucks from approximately 4:1 to 2:1. Concurrently, increased hunting opportunities for antlerless deer reduced abundance of adult does and decreased density of deer populations across the state. Together, these large-scale demographic manipulations likely affected many aspects of deer sociobiology. From 2002 – 2004, I investigated the effects of demographic change on dispersal, which is influenced by social mechanisms and has important implications for ecological processes such as gene flow, population dynamics, and disease spread. In two study areas in Pennsylvania, Armstrong County in western Pennsylvania and Centre County in Central Pennsylvania, I captured and radio-tracked 454 juvenile male white-tailed deer to estimate dispersal parameters during this time of large-scale population change.
I found that dispersal rates varied between areas, among years, and within years. In both areas, throughout the entire study, dispersal during spring fawning (mid-April to early June) and immediately prior to fall breeding season (mid-September to early iv November) accounted for 98% of all observed dispersal. In Armstrong County, dispersal rates remained relatively constant, ranging from 71.0% (95% CI = 60.0 – 79.9%) in 2003 to 78.2% (62.5 – 88.6%) in 2002. In Centre County, however, dispersal rates increased throughout the study, from 31.5% (17.8 – 49.4%) in 2002 to 73.5% (59.5 – 83.9%) in 2004. Reasons for these differences between study areas and the regularly increasing trend observed in Centre County are unclear. In both study areas in 2002, prior to management changes, most dispersal was observed during spring fawning (73% of all dispersal in Centre County and 50% of all dispersal in Armstrong County). In 2003 and 2004, after management changes increased density of adult bucks and decreased density of adult does, the majority of dispersal occurred in fall, prior to breeding season (69 – 79% in Centre County, 63 – 70% in Armstrong County). These results are consistent with hypotheses that maternal influences cue spring dispersal and fraternal social pressure elicits fall dispersal of juvenile male white-tailed deer.
From a meta-analysis of 14 North American populations of white-tailed deer, I found that dispersal rate does not relate to population density; however, average dispersal distance is predicated by percent forest cover (r2 = 0.92, P < 0.001), such that white-tailed deer disperse farther in habitats with less forest cover. To my knowledge, these results represent the first study to document landscape-related plasticity in dispersal of a large mammal and the most complete effort to relate landscape patterns to dispersal distance for a single vertebrate species.
Based on juvenile male deer equipped with radio-collars using global position systems, I found that dispersal paths were generally straight (mean straightness = 0.81 ± 0.07), and dispersal durations were short (median = 12 h). In Armstrong County, v distribution of dispersal directions did not differ from uniformity (URao = 122.7, P > 0.90); but in eastern Centre County regularly trending topography comprised of numerous parallel ridges and valleys, tended to direct dispersal along these features ( x = 75° and 255°, 95% CI = 51 – 108, 231 – 287, respectively; UGD = 2.18, P < 0.05). Further, major roads and rivers were semi-permeable barriers to dispersal of juvenile male white-tailed deer, such that dispersal paths in Armstrong and eastern Centre counties were 34% (Z = -5.17, P < 0.001) and 51% (Z = -4.46, P < 0.001) less likely than simulated paths to intersect these barriers.
In this way, dispersal patterns of juvenile male
white-tailed deer are influenced by both social processes and landscape
interactions. Sociobiological cues are important for the initiation of dispersal
(e.g., dispersal probability and timing), and landscape influences affect
dispersal transition and termination (e.g., dispersal distance, path, and
direction). Understanding dispersal mechanisms and processes for this species
will improve management strategies for white-tailed deer; for example, these
data may provide useful information for modeling spread of epizootics such as
chronic wasting disease and for developing appropriate management efforts to
control disease spread. Further, these findings improve understanding of
mammalian dispersal, especially those of large mammals, whose dispersal patterns
generally remain poorly understood.
Use of Reclaimed
Surface Mines by Grassland BirdsJennifer A. Mattice, Graduate Research
Assistant
Dr.
Duane R. Diefenbach, Advisor
Daniel W. Brauning, Cooperator, Pennsylvania Game Commission
Loss of prairies and grassland habitats has been substantial in North America, and grassland songbirds have exhibited rangewide population declines. In western Pennsylvania, coal mining and subsequent reclamation have created >35,000 ha of grasslands used by grassland-obligate songbirds. However, the largest reclaimed areas are an order of magnitude smaller than reclaimed mine areas studied in the midwestern United States and, unlike Midwestern grasslands, reclaimed surface mines in Pennsylvania are embedded in a forest-dominated landscape. In 2001 and 2002, we estimated densities of grasshopper (Ammodramus savannarum), Henslow’s (A. henslowii), and savannah (Passerculus sandwichensis) sparrows on reclaimed surface mines to assess area sensitivity and identify characteristics of reclaimed surface mines more likely to be occupied by these species. All 3 species exhibited area sensitivity, although for only Henslow’s sparrows could we detect a critical reclaimed mine size (~60 ha) at which density reached an asymptote (15–23 singing males/100 ha). Savannah sparrows exhibited lower densities of singing males ( = 2–3 singing males/100 ha) than Henslow’s and grasshopper sparrows ( = 20–21 singing males/100 ha). We identified vegetative, physical, and landscape characteristics related to occupancy of reclaimed surface mines. Henslow’s sparrows were positively associated with increasing litter depth, increasing number of standing dead stems >50 cm tall, decreasing tree and shrub density, and increasing percent non-forested area within 500 m of the reclaimed mine area. Grasshopper sparrows were positively associated with decreasing tree and shrub density, decreasing litter depth, and increasing number of standing dead stems >50 cm tall. Both grasshopper and savannah sparrows were less likely to occupy sites with an obstructed view of the horizon, and savannah sparrows were negatively associated with increasing areal coverage of grass. We found densities of grasshopper and Henslow’s sparrows were similar to those in traditional grassland habitats or larger reclaimed surface mines. Moreover, minimum patch sizes required for maximum density, or specified occupancy rates, were much smaller than reported in the literature.
Additional information
Diefenbach, D. R., J. T. McQuaide, and J. A. Mattice. 2002. Using PDAs to collect geo-referenced data. Bulletin of the Ecological Society of America 83:256-259.
Diefenbach, D. R., D. M. Brauning, and J. A. Mattice. 2003. Effects of observer variability and species detection on estimates of abundance of grassland songbirds. Auk 120:1168-1179.
Mattice, J. A., D. W. Brauning, and D. R. Diefenbach. 2005. Abundance of grassland songbirds on reclaimed surface mines in western Pennsylvania. in C. J. Ralph and T. D. Rich, editors. Bird Conservation Implementation and Integration in the Americas: Proceedings of the Third International Partners in Flight Conference. USDA Forest Service General Technical Report PSW-GTR-191.
Duane R. Diefenbach, Co-Principal Investigator
Stephen Rathbun, Co-Principal Investigator
Justin K. Vreeland, Research Associate
Click here for a pdf of the Final Report (417 KB)
Executive Summary
The snowshoe hare (Lepus americanus) is a charismatic species of interest
to hunters and nonhunting
wildlife enthusiasts. The hare is a lagomorph named for its disproportionately
large
hind feet (11–14 cm), which with dense fur and stiff hairs form “snowshoes”
well adapted for
locomotion in deep, powdery snow. It is also called varying hare because it has
a pure white
pelage in winter, except for black eyelids and ear tips, which changes to a
black-peppered rusty
brown or grayish summer pelage. In Pennsylvania, hares likely are distributed
patchily and
associated with specific habitat types. However, Pennsylvania is within the
southern periphery
of the hare’s range, and habitat use by this species in Pennsylvania is not
well understood.
Important hare habitat in northern portions of its range is dense, young,
regenerating stands of
hardwoods and conifers, as well as scrub-shrub wetlands. In contrast, conifer
cover is scarce in
the unglaciated regions of Pennsylvania, which may comprise the largest portion
of the range of
hares in the Commonwealth.
A variety of non-invasive techniques, such as identifying the presence of animal
sign, to monitor
rare or elusive species have been developed. However, any monitoring program for
rare or
elusive species must (1) sample usually because of the large spatial size of the
range of the
species and (2) account for the probability of detecting the presence of a
species at a sampling
site because of imperfect detection probabilities. One method of confirming
presence of a
species is by conducting surveys for fecal sign. However, this method is limited
only to species
that have distinct fecal morphological characteristics from other species. For
species whose
feces have similar morphological characteristics, DNA analysis has been
identified as a means of
differentiating among species, and has been developed for lagomorphs.
The goals of this project were to delineate the geographic distribution and
identify coarse-scale
habitat associations of snowshoe hares across northern Pennsylvania. Randomly
selected sites
from across northern Pennsylvania were visited and lagomorph fecal pellets
collected as well
direct evidence of the presence of hares, such as tracks in snow and visual
observations. Fecal
pellets were identified whether from hares by extracting and analyzing DNA and
habitat
characteristics of each sampling site was documented.
We sampled 213 of 240 selected sites (56 conifer, 56 deciduous, 50 mixed, and 51
transitional)
and 34 additional woody transitional sites during January–April 2004.
Transitional habitat was
areas in which tree vegetation was regenerating. Eighteen sites were discarded
because they fell
outside defined habitats of the study area. Nine sites could not be sampled
because of
treacherous terrain, permission was not granted by landowners, or deep snow.
Lagomorph sign
was undetected at 144 sites. Lagomorph pellets, tracks, or both were detected at
62 sites.
Snowshoe hare sign was positively identified at 18 of these 62 sites. Sign
(primarily pellets) at
44 sites either was cottontail sign, or could not be distinguished in the field
between hares or
cottontails. Sign at 7 sites was too indistinct to classify as lagomorph or non-lagomorph.
Lagomorph sign was detected at 12 (including positively identified hare sign at
3) of the 34
additional transitional sites. To estimate detection probabilities, 24 sites
with lagomorph sign
were resampled by different technicians 1–3 times, with and without snow
cover. Some sign,
either pellets, tracks, or direct observation, of snowshoe hares were observed
in 47 sites, or
18.73% of the sites visited.
Snowshoe hare pellets were most easily detected in transitional habitat; the
detectability of
snowshoe hare pellets in transitional habitat is estimated to be more than three
times that in the
forested habitats. Snowshoe hare tracks were most detectable in full snow, and
were not easily
detected when there was no or partial snow on the ground. Not surprisingly, the
detectability of
this elusive animal by direct sighting was very small, at about 2%.
The estimated occurrence of snowshoe hares in conifer habitat was 15.9% (SE =
6.57), in
deciduous habitat was 5.7% (SE = 3.28), in mixed deciduous-conifer habitat was
38.1% (SE =
10.44), and in transitional habitat 26.8% (SE = 6.31).
Based on harvest data from the Pennsylvania Game Commission, the range of
snowshoe hares in Pennsylvania includes counties of the northern tier of the
state and extends
south to Maryland in counties that encompass the Laurel Highlands. The largest
concentrations
of sites where hares were detected were distributed similarly to the harvest
data, which
were Warren, McKean, Forest, and Elk counties in the west and the Poconos in the
east.
Duane R. Diefenbach, Principal Investigator
Click here for a pdf of the final report (1.1 MB)
Executive Summary
In winter 2005 the Pennsylvania Department of Conservation and Natural Resources, Bureau of Forestry conducted 10 aerial surveys of portions of 8 state forests to count white-tailed deer using forward-looking infrared (FLIR) technology. Aerial counts of deer using FLIR technology have been conducted in Pennsylvania, but relatively little research has been conducted on the ability of these counts to provide accurate population estimates. The Pennsylvania Game Commission conducted a literature review of FLIR technology and concluded, as applied to white-tailed deer, the technique provided inconsistent results because of differences in methods and equipment. The most recent peer-reviewed research evaluating FLIR technology in a deciduous forest environment reported that nine aerial surveys missed 11-69% of the deer.
I used four repeated surveys of the Dents Run area in Elk State Forest to estimate the precision (repeatability) of FLIR aerial surveys for deer. I then used this estimate of precision to investigate the ability (statistical power) of FLIR aerial surveys to detect various declines in deer abundance.
The data used in the analysis were from flights conducted on 13 and 14 April 2005 when each of these nights two complete surveys of the study area were conducted. On 13 April, 63 and 73 deer were observed (10.0 and 11.6 deer/sq. mi). On 14 April, 70 and 77 deer were observed (11.1 and 12.2 deer/sq. mi). This resulted in a mean count of 70.75 deer (11.2 deer/sq. mi), a standard deviation of 5.9090, and a coefficient of variation of 0.08352.
The power analysis indicated that three replicate surveys per year could detect a >20% decline in deer abundance with >80% probability. This is equivalent to detecting a decline from 10 deer/sq. mile to 8 deer/sq. mile, or on a 50 square mile study area a decline from 500 deer to 400 deer. Two replicate surveys per year could detect a >50% decline in abundance, and six replicate surveys could detect a >10% decline. The number of replicate surveys required for making deer harvest management decisions would depend on the research or management objectives of a specific project. It is likely that most wildlife managers and administrators would prefer to be able to detect population declines of 20% with good confidence (>80% power).
This analysis did not address whether FLIR aerial surveys provide unbiased population estimates. Estimating bias requires knowledge of the true number of deer on the study area, which would be an expensive undertaking. By accepting the assumption that bias is constant among observers and years, FLIR aerial surveys could be used to monitor trends in deer abundance. However, research to date suggests bias may be a concern for FLIR surveys of deer in deciduous forests (e.g., Haroldson et al. 2003) and more research to directly address this issue is warranted. If bias is not constant, a greater number of replicate surveys would be required to detect the same population change.
These results demonstrate that variability in counts of deer from FLIR aerial surveys needs to be incorporated in statistical analyses to ensure correct inferences are made about changes in deer abundance. That is, a change in the number of deer observed from one year to the next based on single FLIR surveys simply could be caused by inherent variability in the number of deer counted using this technique, unless the decline was exceedingly large.
Duane R. Diefenbach, Principal Investigator
Justin K. Vreeland, Research Associate
Click here for a pdf of the final report (1.3 MB)
Abundant white-tailed deer (Odocoileus virginianus) at Fort Indiantown Gap
National Guard Training Center (FIG-NGTC) present a safety hazard to aircraft at
Muir Army Field, reduce understory vegetation in forested habitats to the detriment
of military training activities, and can adversely affect forest regeneration after
silvicultural treatments. To best manage deer populations, baseline data
regarding the deer population and habitat condition are required to assess the
efficacy of future management actions designed to reduce the deer population.
We designed this study to test the ability of catch-effort models to estimate
the deer population size on FIG-NGTC. Catch-effort models are appealing because
they simply require hunter effort and harvest data to estimate population size,
which are data that can be collected readily, especially on a military
installation that closely regulates access. However, catch-effort models may not
provide good (accurate and precise) estimates of abundance if hunting pressure
does not vary and hunters harvest few deer. Consequently, we designed this study
to compare population estimates obtained from a more reliable, but expensive,
method to assess whether catch-effort models are feasible on FIG-NGTC.
We captured and radio-collared deer using Clover traps, rocket nets, dart guns,
and drop nets during December - April of 2003 and 2004. Attached to
radio-collars were unique alpha-numeric two-digit characters that could be used
to identify individual animals during spotlight surveys. We used sightings of
radio-collared and uncollared deer during spotlight surveys to estimate
abundance using Bowden's estimator (Bowden and Kufeld 1995). This estimator
requires less restrictive assumptions during sighting surveys than other
mark-sight estimators. During the regular rifle season we used hunter check-in
data collected at Range Control to determine hunter effort and harvest. These
data were used in a catch-effort model (Gould and Pollock 1997) to estimate
abundance.
In 2003 we had 35 radio-collared deer available prior to the spring spotlight
surveys and 30 deer available during spotlight surveys conducted prior to the
fall hunting season. We estimated there were 1,266 deer (95% CI = 810 - 1,979)
or 48.1 deer per square mile (95% CI = 31 - 75) in spring 2003. In fall 2003 we
estimated 1,021 deer (95% CI = 812 -1,284; 38.8 deer/sq. mile). In 2004, we had
55 deer available during the spring spotlight surveys and estimated 766 deer
(95% CI = 611 - 961) or 29.1 deer per square mile (95% CI = 23 - 37). Prior to
the 2004 regular rifle season we estimated 1,485 deer (95% CI = 1,105 - 1,995)
or 56.5 deer per square mile (95% CI = 42 - 76). Too few radio-collared,
antlered deer were observed during spotlight surveys to separately estimate
abundance of antlered and antlerless deer.
We used hunter-trips as a measure of hunter effort and found that this varied
greatly among days and training areas. On some days, usually Saturday, the
density of hunter-trips on some training areas exceeded 100 hunter-trips per
square mile, although hunter-trip densities on most training areas were <12
hunter-trips per square mile. Daily hunter effort, during the regular rifle
season, ranged from 52 - 429 hunter trips during 2003 and 30 - 438 hunter trips
in 2004. During the regular rifle season 26 antlered and 97 antlerless deer were
harvested in 2003 and 28 antlered and 99 antlerless deer were harvested in 2004.
Based on our estimates of population size and harvest, harvest rates were 9 -
12% during the regular rifle season and <15% for all deer seasons combined.
Catch-effort population estimates did not provide reasonable estimates of
abundance ( < 200) for either year. One reason the catch-effort model did not
work was because harvest rates were too low. Also, analysis of the density of
hunter-trips on training areas suggest that hunter density may be so great on
some training areas that deer may move onto adjacent areas closed to hunting.
This is possible because training areas are small (median = 0.38 sq. miles = 240
acres) and many training areas receive little or no hunting. Consequently, the
abundance estimates from the catch-effort model may be accurate for the areas
being hunted but the hunting data are not representative of the complete
installation.
Given the existing methods by which hunting is implemented on FIG-NGTC, we do
not believe catch-effort models will be useful for estimating population size
unless hunting effort is more evenly distributed among training areas and better
data on hunter effort and harvest are collected. However, in the course of our
study of deer abundance, hunter effort, and deer harvest, we provide the
following recommendations, which could increase hunter success rates, overall
harvest, and hunter satisfaction:
1. Greater oversight of hunters during sign-out procedures will be required if better data are to be obtained regarding the training areas where hunters hunt. In particular, correctly recording where they actually hunted, and whether they harvested a deer would be especially useful information.
2. Requiring hunters to present their deer at a check station would greatly improve the accuracy of deer harvest estimates. Furthermore, this would provide valuable data on the sex, age, and physical characteristics of harvested deer.
3. Hunter success rates likely were adversely affected by the high hunter
densities on some training areas on certain days, especially Saturdays. The
following recommendations could increase hunter satisfaction and success rates:
a. Limit hunter density at any given time in a given training area to <12
hunters per square mile (1 hunter per 50 acres).
b. Maximize the number of training areas open to hunting, especially during the
regular rifle season when hunter participation and harvest efficiency is
greatest.
c. Consider consolidating smaller training areas to provide hunters with greater
flexibility in where they can hunt.
This is a project of the Human Dimensions Research Unit at The Pennsylvania State University
Duane R. Diefenbach, Co-Investigator
James C. Finley, Co-Investigator
A. E. Luloff, Co-Investigator
Gary J. San Julian, Co-Investigator
Richard C. Stedman, Co-Investigator
Harry C. Zinn, Co-Investigator
Craig W. Swope, Graduate Research Assistant
Studies of Pennsylvania's hunters suggest they believe public lands have lower deer densities and greater hunting pressures which together contribute to lower hunter success than experienced on private lands (Diefenbach et al., in press, Human Dimensions of Wildlife). Beyond such perceptions, however, little is known about hunter's actual behavior on public lands. For example, how far do they hunt from roads and how far do they walk? Knowledge of these practices is essential to public game and land agencies charged with developing effective tools for managing deer. This project was designed to provide such information.
There has been extensive research conducted on free-ranging deer and other North American big game species, but hunter field behavior or the factors that influence this behavior is less well understood. Our research in the Sproul State Forest (hereafter, Sproul) was based on three integrated protocols designed to estimate hunter density, distribution, movements, habitat use, characteristics, and attitudes, all of which can be used on large areas with unrestricted access.
Members of The Pennsylvania State University Human Dimensions Unit (HDU) spent two deer hunting seasons (2001 and 2002) collecting data on hunter movement on the Sproul. Following each season, HDU contacted hunters using either a mail or telephone survey to learn about hunting attitudes and experiences on the Sproul.
This multi-method study increased interest in deer management issues in the Commonwealth. Essential information about hunting opportunities on public lands has been generated and this study provides insights into hunter behavior on public land. Its innovative research design can be implemented elsewhere for studies that simultaneously monitor deer and hunter movement and will contribute to the development of creative solutions to deer management. Results of this research clearly have contributed to an enlightened debate over deer numbers and hunting opportunities on public versus private lands.
This project used aerial surveys, in conjunction with distance sampling techniques and a geographic information system (GIS) database of landscape characteristics, to generate estimates of hunter density and a map of hunter distribution and habitat use. The distribution of global position system (GPS) units to hunters so that their locations could be systematically recorded facilitated this effort. In addition, hunters were asked to complete a simple field questionnaire.
Although aerial surveys are limited to discrete points in time and relate only to aggregations of hunters, GPS unit carriers provide information on hunter habitat use and distribution at different times of day across the landscape. When coupled to information gathered via traditional mail and telephone surveys, we are better able to assess how hunter characteristics (e.g., age, physical condition, and attitudes) are related to field behavior.
The Sproul was chosen as the study site because it is representative of large tracts of contiguous forested public land commonly referred to as Big Woods. Forested habitats across the Commonwealth often exhibit evidence of the effects of deer overabundance. Questions exist about the relationship between those areas most severely damaged and hunter success - that is, is damage more common in areas where hunter access is difficult or hunting effectiveness is low?
It is generally thought that many such sites are big woods areas where hunter use or access, especially for harvesting antlerless deer, is low or difficult. The latter can result in a forested landscape characterized by moderate to high hunting pressure in some areas, and deer refuges with little or no hunting pressure in others. Such refugia may retain deer densities large enough to continue to degrade habitat. Until the implementation of our study, natural resource managers in Pennsylvania knew relatively little about hunter behavior in big woods areas.
Our research focused on two objectives during the two years of fieldwork:
(1) Monitor hunter movement and distribution on a large tract of public land where hunter numbers are not restricted; and
(2) Explore hunter concerns, motivations, strategies, and habitat use when hunting in this landscape.
Click here for a copy of the Final Report and Appendix F
The implications of inter-annual movements of a migratory songbird on annual
survival estimationMatthew R. Marshall, Co-Investigator
Duane R. Diefenbach, Co-Investigator
Many Neotropical migratory songbirds exhibit a high degree of site fidelity between breeding seasons facilitating estimation of annual survival rates through traditional capture-mark-recapture methodology. However, it is recognized that a proportion of the surviving individuals do not return to precisely the same site each breeding season and some birds move off of the defined study area. Recognition of this phenomenon has led to terminology such as “apparent survival” since the estimates include both survival and permanent emigration. Our objective was to utilize a five year banding and movement study of 423 Prothonotary Warblers (Protonotaria citrea) from the White River National Wildlife Refuge, Arkansas to evaluate the extent to which these inter-annual movements lead to underestimates of “true” survival rates. Using a FORTRAN program that simulates a five year capture-mark-recapture study that explicitly incorporates inter-annual movements and permanent emigration, we were able to demonstrate that these movements can result in negatively biased apparent survival estimates of 14% (percent relative bias) compared to known survival rates. Furthermore, these biases differed among the sexes (9% and 22%, male and female, respectively) due to differences in site fidelity and movement patterns. This means that our observed apparent survival rates of 0.52 and .039 for male and female Prothonotary Warblers, respectively, may actually be “true” survival rates of 0.57 and 0.50. Differences of this magnitude can have profound effects on the predictions of population persistence through time, source/sink dynamics, and aspects of life-history theory that are of interest and concern to wildlife conservation and management. We are currently investigating alternative study designs that incorporate information about inter-annual movements and attempt to obtain estimates of true survival.
Survival
Rates, Cause-specific Mortality, and Landscape Influence on Survival of
White-tailed Deer Fawns in Northcentral Pennsylvania
Justin K. Vreeland, M.S. in Wildlife and Fisheries Science
Dr. Duane R. Diefenbach, Advisor
Bret D. Wallingford, Cooperator, Pennsylvania Game Commission
Estimates of survival and cause-specific mortality of white-tailed deer (Odocoileus virginianus) fawns are important to population management, but are unknown for Pennsylvania. Sources of fawn mortality likely include predation, other natural causes excluding predation, legal harvest, poaching, collisions with vehicles and farm machinery, and accidents. However, in what proportions fawns die from these causes is unknown in Pennsylvania. Habitat type, extent, and arrangement can influence predator and prey communities and their interactions, and therefore also might influence fawn survival. However, influence of habitat characteristics on fawn survival has not been investigated. Therefore, we quantified cause-specific mortality, survival rates, and habitat characteristics related to survival of white-tailed deer fawns in a forested landscape (QWA) in northern central Pennsylvania with presumed poor habitat condition and greater predator density, and a separate, agricultural landscape (PV) in central Pennsylvania with presumed better habitat condition and lesser predator density.
Using foot searches in PV and vehicles searches in QWA, we captured
neonatal fawns in May and June 2000 and 2001.
Fawns were fitted with expandable, releasable radiocollars designed to
transmit for ≥9 months. We
monitored fawns at least weekly from capture until death, transmitter or collar
failure, or the end of the study. We
developed 13 models of fawn survival and used Akaike’s Information Criteria (AIC)
and the known fates procedure in computer program MARK to model survival through
9 weeks. We created circular buffer
areas corresponding to the median areas of study-site- and year-specific 95%
fixed-kernel home ranges for fawns at 9 weeks after capture and centered these
buffer areas on the median location for each fawn. Using a geographic information system, we calculated edge
density, road density, proportion of buffer area in annual and perennial
herbaceous land cover, and habitat patch diversity within fawn buffer areas.
We used logistic regression models and AIC to evaluate the relation
between these 4 habitat characteristics and fawn survival.
We captured 110 fawns in PV and 108 fawns in QWA. In the best (DAIC = 0) logistic regression model, only study site and fawn mass at capture were related to fawn survival, with fawns in PV and heavier fawns more likely to survive. None of the 4 metrics of habitat composition and configuration was related to fawn survival. Of known-fate models, the best (DAICc = 0, AICc weight = 95.0%) model suggested fawn survival differed between QWA and PV through time. Survival at one week post-capture was 83% in PV (82.7%, 95% CI = 74.5–88.7%) and in QWA (83.3%, 95% CI = 75.1–89.2%). Survival at 9 weeks after capture was 72.4% (95% CI = 63.3–80.0%) in PV and 57.2% (95% CI = 47.5–66.3%) in QWA. Survival at 26 weeks after capture was 58.6% (95% CI = 48.8–67.7%) in PV and 45.6% (95% CI = 36.0–55.6%) in QWA. Thirty-four-week survival was 52.9% (95% CI = 42.7–62.8%) in PV and 37.9% (95% CI = 27.7%–49.3%) in QWA.
Within 34 weeks of capture, 106 of 218 monitored fawns died and 21 were censored. Of 98 fawns radio-tagged in 2000, 51 died within 34 weeks of capture and 7 were censored. Of 120 fawns radio-tagged in 2001, 55 died within 34 weeks of capture and 14 were censored. For both study sites combined, predation was the greatest source of mortality, accounting for deaths of 22.5% (95% CI = 17.6–28.8) of captured fawns and 46.2% (95% CI = 37.6–56.7) of mortalities through 34 weeks. Natural causes, excluding predation, were the second leading cause of death, accounting for deaths of 13.3% (95% CI = 9.5–18.6) of captured fawns and 27.4% (95% CI = 20.1–37.3) of mortalities. Vehicle accidents accounted for deaths of 9 fawns. Hunting accounted for deaths of 7 monitored fawns. Predation rates were greater in QWA, where 83.7% of predation events occurred. Mortality rates from other sources of mortality did not differ between QWA and PV, but 62.1% of deaths by natural causes, excluding predation, occurred in PV. We attributed 32.7% and 36.7% of predation events to black bears (Ursus americanus) and coyotes (Canis latrans), respectively. Bobcats (Lynx rufus) and unidentified predators accounted for 6.1% and 24.5% of predation events, respectively
White-tailed deer fawn survival in a forested and an agricultural landscape in central Pennsylvania is comparable to fawn survival in other forested and agricultural regions in northern portions of the white-tailed deer’s range. Fawn survival may be greater in agricultural landscapes where habitat quality is presumed greater and predator densities may be less than in forested landscapes where habitat condition may be poorer and predators may be more abundant. However, the influence of landscape condition on fawn survival requires further study with replicate landscapes over larger geographic scales. Mortality from predation and other natural causes, excluding predation, are the dominant sources of mortality to fawns in Pennsylvania. In heavily forested regions in Pennsylvania where black bear densities are great, black bears may be at least as efficient predators of fawns as are coyotes. Collisions with vehicles and farm machinery, hunting and other legal means of take, poaching, and accidents play a comparatively minor role in fawn survival in Pennsylvania.
We detected no relation between fawn survival and habitat characteristics at home-range scales. However, landscape ecology likely plays an important role in fawn survival both directly through habitat type and arrangement, and indirectly by influencing predator distribution and activity. Future studies of fawn survival should consider the landscape context through replicated studies of the effect of landscape composition and configuration on fawn survival.
More Detailed Sources of Information on this Project
Diefenbach, D. R., C. O. Kochanny, J. K. Vreeland, and B. D. Wallingford. 2003. Evaluation of an expandable, breakaway radiocollar for white-tailed deer fawns. Wildlife Society Bulletin 31:756-761.
Vreeland, J. K., D. R. Diefenbach, B. D. Wallingford. 2004. Survival rates, mortality causes, and habitats of Pennsylvania white-tailed deer fawns. Wildlife Society Bulletin 32:542–553.
Population
Demographics of a Suppressed Wild Turkey Population in Pennsylvania
Mark A. Lowles, M.S. in Wildlife and Fisheries Science
Dr. Duane R. Diefenbach, Advisor
Mary Jo Casalena, Cooperator, Pennsylvania Game Commission
Eastern wild turkey (Meleagris gallopavo sylvestris) populations in Turkey Management Area (TMA) 7b, located in south-central Pennsylvania, are presumed to have been in decline since the late 1980s. The Pennsylvania Game Commission (PGC) created this management area from the southeastern portion of TMA 7 in 1995 with a shorter, one-week fall hunting season in an attempt to increase turkey population size. The turkey population did not respond to this management action. To investigate the factors responsible for the continuing suppression of this population, a two and a half year study was conducted in which wild turkey hens were captured and fitted with radio transmitters.
Nesting incubation rates were less than similar studies (69%) in 2000, but nest success was good with 72% of hens producing at least one live poult. In 2001, nest initiation rates were greater (88%) than in 2000 but nest success was poor, with only 42% of hens producing at least one live poult.
During the first year of the study, (August 1999-August 2000) we observed an annual survival rate for hens (25.2 %, 95% CI = 19.1- 33.3%) substantially less than survival rates reported in similar studies. Hen survival was greater in the second year, (August 2000-August 2001) with an annual survival rate of 55.8% (95 % CI = 44.7 - 69.7%), but still less than most similar studies. We detected no mortality because of starvation or disease. Of 163 hens monitored during the study predation was the major cause of mortality: 38 hens (48.1%) died from predation, 13 (16.5%) were legally harvested, 4 (5.1%) were poached, 4 (5.1%) died from rodenticide poisoning, and one (1%) died from a vitamin A deficiency. Nineteen hens (24.1%) were killed by unknown causes and the fates of 18 hens were unknown because of radio transmitter failure. Poult survival in 2000 (11.8%, SE = 0.049) and 2001 (23.3%, SE = 0.037) was poor.
We used estimates of population parameters from this study as input for a population model. We conducted a sensitivity analysis to assess which population parameter had the greatest influence on population decline. The population model was most sensitive to changes in sub-adult mortality. Eighty percent sub-adult mortality was identified as the level required for a stable population, which is a lower mortality rate than was observed over the course of our research.
We analyzed habitat data from the Pennsylvania Bureau of Forestry (BOF) to characterize edge habitat at different spatial scales surrounding nests. We used logistic regression to investigate associations among three edge metrics. Only two of 12 edge metrics at different spatial scales had P-values <0.05, which is similar to what we would expect for a random data set, leading us to conclude that the habitat characteristics we measured were not related to nest success. We measured habitat variables at nest sites in an attempt to identify which variables may be related to nesting success. The principle component analysis we conducted on habitat characteristics within 10 m of the nest showed no association among the variables we measured and nest success.
It is possible that conditions for turkey population expansion were favorable when turkeys were re-introduced in the late 1970s and early 1980s. We believe predator populations were greatly reduced because of a rabies epidemic and the record high prices being paid for raw fur. These conditions happened to coincide with the reintroduction of turkeys to the Michaux State Forest. Those favorable conditions no longer exist and the seemingly low turkey densities may be what the habitat is capable of supporting.
The present management strategy for turkeys has been to shorten the fall either-sex hunting season from two weeks to one. Habitat improvements, in the form of permanent herbaceous openings (PHO) have been constructed throughout the forest. Future management and research should evaluate PHOs to investigate whether they are being used by turkeys as claimed, and if there is a difference in the recruitment rate of hens with poults that use PHOs compared to those brooded elsewhere, or even if hens select areas with PHOs to rear broods. Different regimens of maintenance of the PHOs also should be investigated. Investigation into predator densities and population sizes could provide empirical evidence about the possible effect of predators on turkey population trends.