Ungulate space-use and movement in response to changes in reproductive status and hunting pressure
Open Access
- Author:
- Gundermann, Kathleen
- Graduate Program:
- Wildlife and Fisheries Science (MS)
- Degree:
- Master of Science
- Document Type:
- Master Thesis
- Date of Defense:
- July 22, 2022
- Committee Members:
- Bradley Cardinale, Program Head/Chair
Frances E Buderman, Thesis Advisor/Co-Advisor
Ephraim Mont Hanks, Committee Member
Duane R Diefenbach, Committee Member - Keywords:
- space-use
movement
ungulates - Abstract:
- Ecological systems contain many complex components, including individual variation, non-linear interactions, scale multiplicity, and spatial heterogeneity. In addition, it is often difficult to observe ecological processes directly because obtaining individually identifiable and repeatable observations of wild animals over long durations can be logistically difficult and expensive. For example, in traditional mark-recapture or resight studies, the individual is only observed when it is re-captured or re-detected by an observer. These methods result in spatially and temporally disconnected observations because they are restricted to the spatial location of a trap or observer during periods of active observation. However, as wildlife telemetry technology has advanced, so has our ability to study individual wild animals, providing critical insights into animal behavior and ecology that are not spatially or temporally constrained. Contemporary telemetry devices can collect high-frequency observations of animal locations over more extended periods and with reduced measurement error. Even so, the animal's location does not necessarily provide direct information about an individual's behavior at a particular location or when there is a shift in the unknown or latent behavioral state. As quantitative methods continue to develop, researchers can identify latent behavioral states and them to important resource use areas, such as calving grounds, foraging areas, or migration corridors. These latent behaviors can manifest in location data in multiple ways, either through a change in physical location or changes in movement-related metrics. For example, movement behavior is typically quantified using step lengths (the straight-line distance between two points) and turning angles (the change in direction between three successive steps). However, a change in location does not necessarily result in a change in movement metrics and vice versa. For example, relocating a core area can be achieved without a detectable change in movement metrics. Meanwhile, the geographic location of the core-use area may not shift, but if the area contracts, then quantities such as step lengths and turning angles will change. Additionally, behavioral changes may be related to biological factors such as birthing events, migration, external factors such as disturbance or intraspecific interactions. It is important to understand when these behavioral changes occur and what variables are related to changes in movement behavior as they can inform conservation and management. There are multiple statistical methods to identify latent behavioral states. For example, hidden Markov models (HMM) are now commonly used to identify the temporally varying behavioral state of an individual given movement-related quantities, namely step length (speed) and turning angle. The development of easy-to-use R packages has facilitated this increase in the use of HMMs. However, HMMs are a general modeling framework that is not restricted to analyzing step lengths and turning angles. In addition, HMMs can be viewed as an extension of a much simpler latent state model, referred to as a change-point model, which contains multiple latent states but only a single (or fixed a priori) point at which the individual transitions from one state to another. It is important to consider the type of latent state models and select an ecologically appropriate quantitative framework for the behavior of interest. For example, a change-point model may be appropriate if the individual transitions between behaviors once during a period of interest, such as during a birthing or migratory event. In contrast, an HMM may be more appropriate if an individual transitions between behaviors throughout the study duration. Researchers must consider the ecological questions they are asking when choosing the most appropriate analytic tool. To advance our ability to quantify changes in movement behavior, we present a holistic Bayesian framework for estimating latent behavioral states that vary in the quantities being modeled. Our framework contains two general models, one based on individual locations, and one based on movement metrics; both may be useful for determining when transitions in latent behavioral states occur. Although other work has focused on identifying state-transitions with these metrics, the two have not been compared directly and neither response has been identified as the best practice in identifying behavioral shifts. Our framework allows the comparison of these two models across species and individuals, depending on species-specific ecology or significant within-species individual variation in movement behavior. We first applied this framework to identify the timing of a single behavioral shift, related to parturition, in white-tailed deer (Odocoileus virginianus, n = 19) and elk (Cervus canadensis, n = 37) in central Pennsylvania. We equipped each population with global positioning system (GPS) collars and vaginal implant transmitters (VITs) and compared the ability of the change-point models based on either movement quantities or spatial locations to detect the timing of parturition events. To summarize the ability of each model to estimate the timing of parturition, we defined two levels of success; a Level 1 success occurred when the 95% credible interval of the estimated change-point fell within +/- six hours of the known parturition event. A Level 2 success occurred when the 95% credible interval fell within +/- twelve hours of the known parturition event. The behavioral change-point model detected the parturition event more consistently (Level 1; 0% deer, 24% elk; Level 2 6% deer, 32% elk) than the geographic change-point model (Level 1; 0% deer and elk; Level 2 0% deer, 8% elk). Additionally, we conducted a simulation study to determine the effect that frequency of observations, duration of observations following a behavioral change, and the relative size of the behavioral change had on the ability of the model to detect the true change-point. We recommend that researchers interested in detecting a single change-point obtain data at least 12 hours after the change has occurred and with a 15-minute fix interval. Researchers and managers can use our change-point models across species to identify the timing of parturition events or other changes in behavior that occur at a single point in time if individuals exhibit a consistent and distinct change in the quantity being modeled. We then applied hidden-Markov models to identify multiple behavioral shifts and understand how behavior varies with predator activity in a population of deer (n=63) in Pennsylvania. We used GPS collars to monitor deer at hourly intervals during the 2016-2019 rifle hunting season in Pennsylvania. We developed two Bayesian hidden-Markov models, based on the framework described above, to evaluate deer behavior in a temporally and spatially varying landscape of fear. These models allow individuals to switch between two generating distributions, depending on conditions at a given time step, that each describes a different latent behavior. In both models, the shift in behavioral states was a function temporally and spatially varying predictors. We compared the ability of these models to two simpler models that did not account for the effects of hunting season. Our results indicate that hunting season may not be a contributing factor to behavioral transitions in individuals. We found while most individuals are changing their behavior through time, hunting season did not explain this transition well across most individuals. Further investigation into the spatial and temporal features that influence white-tailed deer space use and movement is needed to understand the impact of hunter presence during periods of high mortality pressure.