• SFWS HOME
• Graduate Degrees Offered
• Faculty Research Expertise
• Research Assistantships
• Graduate Student Projects
• Contact the GPO Office
• FAQ

Modeling differences in detectability for trail cameras

Precise and accurate estimates of population parameters such as age structure, production, and density are necessary in determining habitat and harvest management strategies for any wildlife species.  Surveys employing remote cameras are becoming an increasingly popular tool for estimating these parameters; however, most camera studies fail to incorporate detectability.  Thus, they underestimate the parameters of interest. The purpose of this study was to determine the sources of heterogeneity in detectability for trail cameras which incorporate a passive infrared triggering system sensitive to heat and motion.  Images were collected at four baited sites in the Conecuh National Forest, Alabama using three identically oriented cameras (Penn’s Woods DS-04, (this is not an endorsement)) at each site operating continuously over the same seven-day period.  Detectability was estimated for four groups of animals based on phylogenetic class and body size.  Our hypotheses of detection considered variation among bait sites, differences among cameras, as well as interactive and additive models.  The best model (AICc weight = 0.9998) estimated different rates of detection among each camera and site in addition to different detection rates for four animal groups.  Factors that explain this variability might include poor manufacturing tolerances, behavior, and species specific infrared radiation.  PIR sensitivity plays a role in detection of different sizes of animals.  Surveys using trail cameras with a PIR system must incorporate detectability for individual cameras.  Incorporating time-lapse triggering systems into survey designs may avoid this pitfall.

Using time-lapse cameras to estimate abundance and structure of Eastern wild turkeys (Meleagris gallopavo sylvesteris) in Alabama

Eastern wild turkey populations have exploded in recent decades due to intense efforts of rocket netting and relocation by wildlife biologists, land managers, and the public.  Large turkey populations have resulted in increased harvest, and questions have arisen as to the sustainability of this harvest.  Current population estimation techniques are fraught with bias and do not provide acgh results for managers to prudently make judgments about how many birds to harvest. We performed a statistically rigorous population survey using time-lapse cameras and bait to estimate age and sex ratios, abundance, and annual poult production to assess harvest sustainability in the state of Alabama.  Hypotheses of density in relation to habitat characteristics at the landscape level were developed a priori from the literature to determine the sources of variability that cause unequal distribution of wild turkeys at an ecoregion scale. The objectives of the study are to 1) estimate the abundance (through estimates of density and incorporation of models of detectability) of turkeys using repeated time lapse camera surveys in a nine county area in southwest Alabama; 2) estimate annual production (poults per hen) and age and sex structure of the population; and 3) determine sources of heterogeneity in habitat that cause bias in estimates of turkey density and detectability.  Upon completion, this survey will provide land managers with critical information required to maintain current population levels of wild turkeys through sustainable harvest.

Movements of Eastern wild turkeys (Meleagris gallopavo sylvesteris) in response to hunting pressure and bait stations

Two critical assumptions were identified with the camera surveys in southwest Alabama.  First we assume that the sampling framework allows some probability that every turkey can be counted.  This assumption is addressed by using a sampling grid smaller than the home range of most turkeys during the late summer survey period.  The second critical assumption is that the distribution of turkeys across the surveyed area is unaffected by the bait used to hold them at the survey site long enough to capture several photographs.

Public lands are important to turkey hunters and nearly every Wildlife Management Area in Alabama receives significant turkey hunting pressure.  Research suggests that satisfaction among turkey hunters is correlated with not only hunting success, but hearing turkeys gobble during the hunting season.  Anecdotal information from WFFD personnel suggests that gobbling by turkeys declines dramatically after the first few days of the hunting season.  This may result from changes in gobbling behavior, movements by adult gobblers in response to hunting pressure, or high hunting mortality among gobblers.  High gobbler harvest is acceptable as long as it does not affect the mating success of females (i.e., productivity is unaffected) or reduce the number of adult males below levels that provide a satisfactory hunting experience.  Presently, we assume that harvest on public lands is taking place at acceptable levels, but without better information on population size and structure, harvest rates, and gobbler behavior that assumption cannot be validated.

Two objectives resulted from these two issues:  1) Determine whether the bait stations used in the automated camera surveys influence the distribution of successful and unsuccessful female turkeys enough to bias survey results on Barbour Wildlife Management Area (BWMA). 2) Determine whether changes in behavior or harvest rates are responsible for the decrease in gobbling behavior on BWMA.

Turkeys (40 gobblers and 40 hens) will be captured using baited-cannon nets.  We will collect 20-40 locations for each gobbler prior to and then following the start of turkey season (March 16-April 30).  Estimates of each gobbler’s home range prior to and following the beginning of hunting season will be compared to the BCWMA boundaries to determine whether gobbler movements (size and distribution of home range) change in relation to the BCWMA boundary.   Mortality and harvest rates will be estimated via maximum likelihood methods.  We will collect 20-40 locations for each hen prior to and then following establishing bait stations.  We will estimate each hen’s home range prior to and following establishing bait stations to determine whether hen movements (size and distribution of home range) change in relation to locations of bait.   Mortality and harvest rates will be estimated via maximum likelihood methods.