Biometrics and Population Modeling IV

Contributed Paper
ROOM: Rooms 16 – Acoma, 32 – Teseque and 15 – Zuni Combined
SESSION NUMBER: 104
 

1:10PM Novel Application of Adaptive Sampling Principles to Spatial Capture-Recapture
Alec Wong; Jeffrey A. Royle; Angela Fuller
Rare species present inherent challenges to data collection, particularly when the species is spatially clustered over large areas, such that the encounter frequency of the organism is low. These challenges can be ameliorated by the use of adaptive sampling, whereby a probabilistic sampling method (simple-random, systematic, etc.) is employed first, but additional sites are added in the vicinity of sites where some index variable exceeds an a priori threshold. We applied these principles to the spatially explicit capture-recapture (SCR) analytical framework and assessed the performance of the adaptive SCR model (AS-SCR) under a two-stage adaptive sampling procedure by simulation. In the AS-SCR scenario, SCR sampling was informed by a count index assumed to be proportional to local population size. The decision to re-visit the site and apply SCR sampling was dependent upon whether the index variable exceeded the a priori threshold. For comparison, we also evaluated the informative sampling scenario (AS-SCR without modeling index information), a full standard SCR (F-SCR) scenario as a reference baseline, and SCR applied at a simple random sample of sites (SRS-SCR) equal to AS-SCR. Standard capture-recapture data were simulated and analyzed using the aforementioned models and sampling scenarios. We observed minimal bias and comparable variance with respect to parameter estimates provided by the standard SCR model and sampling implementation, but with substantially reduced effort: with AS-SCR, the data were drawn from approximately 50% of the sites sampled in F-SCR, in expectation, indicating significant cost saving potential by applying adaptive sampling.
1:30PM Open Population Spatial Capture-Recapture Models – Promises and Challenges Capture-Recapture Models – Promises and Challenges
Rahel Sollmann; Beth Gardner; Jerrold Belant
With continued biodiversity loss and habitat fragmentation, the need for assessment of long-term population dynamics and population monitoring of threatened species is growing. Capture-recapture methods provide a powerful way to estimate population size and dynamics while accounting for imperfect individual detection. Spatial capture-recapture (SCR) models are a fairly recent expansion of capture-recapture methods. They are of particular interest for rare and wide-ranging animals as they make efficient use of capture-recapture data, while being robust to study design changes. Relatively few studies have implemented open SCR models and to date, very few have explored potential issues in defining these models. We develop a series of simulation studies to examine the effects of the state space definition and between primary period movement model on demographic parameter estimation. The results of our simulation study show that movement and survival are confounded in open SCR models when little information is known about between primary period movements of animals. Particularly, when activity centers are modeled as random across primary periods, estimates of survival increase with increasing state space. This effect is ameliorated when modeling shifts in activity centers between primary periods as a Markovian process. We demonstrate the implications of these difficulties on a multi-year hair snare data set of black bears (Ursus americanus) and bobcats (Lynx rufus) in Michigan. In general, we suggest that open SCR models can provide an efficient and flexible framework for long-term monitoring of populations, when detection data contain sufficient information to model between-year shifts in home ranges. Augmenting these models with telemetry information to inform between-year movement may be a promising avenue for further model development.
2:10PM Monitoring Matters
Douglas H. Johnson; Glen Sargeant
Monitoring is critically important for assessing the status of systems, such as animal populations, habitat conditions, or ecosystems in general. Monitoring that is unguided by specific hypotheses has come under criticism from some and given the pejorative label, “surveillance monitoring.” Certainly, if some hypothesis is in need of testing, or some management action is planned, appropriate data should be collected to evaluate the hypothesis or assess the action. But gathering such data is a short-term and focused endeavor, whereas monitoring as generally defined is a long-term activity sensitive to a wider variety of changes in the system. I cite several examples of monitoring programs that provided valuable insight into critical issues and were essential in addressing consequences of unanticipated events such as global climate change and the Exxon-Valdez and Deepwater Horizon oil spills. I discuss other topics as well, such as why “trend” is a poorly defined concept, why seeking unbiased estimators can be a poor strategy, what to do if a monitored population expands beyond the area where it had been surveyed, and how new-and-improved survey methods can be reconciled with long-term historical data based on older methods.
2:30PM Generalized Spatial Partial Identity – a Generalization and Extension of Current Spatial Capture-Recapture Models with Latent Individual Identities
Ben C. Augustine; J. Andrew Royle; Marcella J. Kelly
Current spatial capture-recapture (SCR) models can accommodate a range of information about individual identity from all identities of captured individuals being known in typical spatial capture-recapture to no identities known in Chandler-Royle unmarked SCR. Intermediate cases are spatial mark-resight (SMR), the Chandler-Clark SCR-occupancy hybrid, an in development random identity thinning model, and the 2-flank spatial partial identity model (SPIM). In all but typical SCR, some or all of the individual identities are latent and the precision of the density estimate is a function of how much uncertainty there is in the latent individual identities. Here, we present the generalized SPIM (genSPIM), of which all models previously listed are special cases. GenSPIM allows for individual observations to be connected to other observations (observation A is certainly from the same individual as observation B) and for groups of connected observations to be excluded from matching with other groups of connected observations (observations A and B certainly do not match observations C and D). We show that genSPIM can fit the special cases listed above and allow for more information about individual identities to be used to further reduce the uncertainty in individual identities and thus increase the precision of density estimates. For example, in any of these models, we may have information that excludes matches between unmarked individuals such as sex, age class, or ear tag color that is currently not being used. Further, in SMR with natural marks, it often cannot be determined if two “marked” individuals are not actually the same individual and so only one can be used. GenSPIM allows both sets of linked observations to be used. Finally, we show that genSPIM can be used to model partial genotypes and that density can be estimated using a fraction of the loci at a fraction of the cost.
2:50PM Refreshment Break
3:20PM Identifying Unique Individuals Using Natural Pelt Patterns to Estimate Population Abundance with Mark-Resight Models
Ben S. Teton; Jesse S. Lewis; Christina Tombach Wright; Mike White; Hillary Young
Invasive wild pigs (Sus scrofa) cause extensive damage to ecological resources, agricultural land, and private property throughout much of the United States. Since at least 1990, the 270,000-acre, privately owned Tejon Ranch in the Tehachapi Mountains of California has been occupied by a population of wild pigs. To facilitate wildlife management planning, the Tejon Ranch Conservancy (TRC) has developed a method of identifying and cataloguing individual wild pigs using natural markings alone, for the purpose of mark-resight population estimation. This method was tested over a period of 16 months using an array of 48 trap-cameras across a 48km² survey grid. Wild pig images captured during this period were standardized based on image quality and flank distinctiveness. Thresholds of identifiability were established to account for the inherent variability of natural markings between individuals. The majority of wild pigs in this region are homogeneous in appearance, with indistinct black pelts and few defining features, yet this method was able to reliably identify approximately 20% of all individuals that encountered the camera array. Capture histories of these identifiable individuals were recorded and catalogued in order to facilitate estimation of population abundance through time with standard mark-resight analytical methods. As this method requires no trapping or tagging of any kind, it may be utilized as a cost and resource efficient alternative to traditional mark-resight techniques that rely on ear-tags or neck-bands for individual identification. To better enable comparison of these methods, the results of analyses from natural mark photo-ID are compared with those derived from a traditional trapping and tagging effort that took place contemporaneously across the same survey grid. This study uses wild pigs to highlight the potential of a simple, natural mark photo-ID catalogue as a management tool that could be widely applied to other species found in remote wilderness settings.
3:40PM Novel Methods to Estimate Abundance of Unmarked Animals Using Camera Trap Data
Anna K. Moeller; Paul M. Lukacs; Jon Horne
Abundance estimates are central to the field of ecology and are an important tool for wildlife managers. To estimate deer and elk abundances, most western wildlife agencies rely on aerial survey methods, which can be expensive, dangerous, and difficult to implement. One step toward noninvasive abundance estimation is the use of passive “traps” such as remote cameras or acoustic recording devices. While these have been used successfully to estimate abundance from individually identifiable animals, current methods are lacking when species have no natural markings. To address these issues, we developed three methods for estimating abundance of unmarked animals using camera trap data. First, we used random sampling theory to use pictures at multiple cameras as snapshot samples of density. Then we expanded this theory within a time-to-event framework to leverage the relationship between abundance and encounter rate. Intuitively, as the number of animals in an area increases, the trapping rate increases. We developed two new models to estimate abundance from trapping rate in time and space. We tested all three methods on simulated data and used them to estimate elk abundance in Idaho. In simulations with known population size, all three methods produced unbiased estimates of abundance, regardless of animal movement rate. In one study area, they produced abundance estimates that compared favorably to the most recent aerial survey estimate. In the other study area, these models allowed us to estimate elk abundance where this has never before been possible. Our three methods provide a new framework for estimating abundance from unmarked animals and for utilizing camera data most efficiently. They have wide applications across species; biologists can select the method that best meets their specific circumstances. All three methods greatly reduce the amount of data required for analysis, making them practical management tools.
4:00PM Single Survey Bobcat Density Estimation Using Fecal DNA and Spatial Capture Recapture
Dana J. Morin; Lisette P. Waits; David C. McNitt; Marcella J. Kelly
Bobcat (Lynx rufus) population density estimates are necessary to inform management and conservation yet are difficult to obtain for such cryptic species. We implemented a single-survey, closed session scat sampling protocol to estimate bobcat density using fecal DNA and spatial capture-recapture at two sites in Virginia, USA. We estimated density for five sessions at one study site (three summer sessions and two winter sessions 2011 through 2013). However, we only estimated density for three summer sessions at the second site due to low detection rates and few spatial recaptures in the winter. Summer densities ranged from 6.65 to 13.96 bobcats/100 km2 across both sites (posterior modes), and some estimates suffered from poor precision. We suggest the summer session estimates are representative of the resident population, that differences in density between summer and winter are representative of potential net recruitment, and that differences between consecutive summer sessions are representative of the net recruitment realized for the population (dependent on survival and emigration). We also estimated realized density surfaces to demonstrate how spatial heterogeneity in local densities can be used for spatially-explicit harvest, for evaluating habitat associations, or for identifying areas with potential for increased predation or human-wildlife conflict. Finally, we assessed factors affecting precision in estimates and provide recommendations to improve detection and reduce credible intervals that may be applicable across the bobcat range and to other species. The single catch-effort transect methodology represents an improvement in monitoring bobcat populations specifically, or as part of concurrent monitoring for multiple carnivore species.
4:20PM Optimal Dynamic Sampling of a Spreading Population
Perry J. Williams; Mevin B. Hooten; Jamie N. Womble; George G. Esslinger; Michael R. Bower
Population spread is a dynamic process that changes in space and time. Wildlife survey designs that ignore these dynamics may not capture essential spatio-temporal variability of a process. Alternatively, dynamic survey designs explicitly incorporate knowledge of ecological processes, the associated uncertainty in those processes, and can be optimized with respect to management objectives. Additionally, dynamic designs permit the flexibility to optimally add or remove sampling locations, which makes dynamic designs ideal for monitoring expanding or contracting populations. Sea otters (Enhydra lutris) began colonizing Glacier Bay National Park in 1993 and have increased in abundance and distribution since then; abundance estimates have increased from 5 otters to more than 5,000 otters and they are now distributed across most of the Park. Sea otters were recently identified as a vital sign for long-term ecological monitoring by the National Park Service due to their role as a keystone apex predator, and their influence in structuring nearshore marine communities. Our objective was to develop a framework for optimal dynamic sampling of a spreading population, and apply our framework to select a survey design for estimating the distribution and abundance of sea otters in Glacier Bay National Park. We first developed a dynamic spatio-temporal model that allowed estimation of prediction error associated with distribution and abundance of sea otters in the future, given past data. We then identified a dynamic optimal design for future surveys. We optimized the survey design based on minimizing uncertainty in model-based predictions of future abundance and distribution of sea otters. Our framework is flexible, and can be adjusted annually to align with changing monitoring budgets.

 

Contributed Paper
Location: Albuquerque Convention Center Date: September 27, 2017 Time: 1:10 pm - 5:00 pm