Biometrics and Population Modeling I

Contributed Paper
ROOM: Rooms 16 – Acoma, 32 – Teseque and 15 – Zuni Combined
SESSION NUMBER: 24
 

1:10PM Combining Individual-Based and Dynamic-Landscape Metapopulation Models to Identify Climate Change Threat to a Regional Songbird Population
Thomas W. Bonnot; W. Andrew Cox; Frank R. Thompson III; Joshua J. Millspaugh
Predicting the impacts of global change on wildlife populations for conservation planning requires approaches that comprehensively model important biological processes at scales at which they interact with climate. We combined individual-based and metapopulation models to estimate direct effects of climate change on the population viability of a common forest songbird (Acadian flycatcher [Empidonax virescens]) through the year 2100 across the Central Hardwoods ecoregion, a 39.5 million hectare area of temperate and broadleaf forests in the United States. We estimated territory-level productivity based on empirically-derived demographic parameters that varied as a function of landscape and climatic factors (e.g., forest cover, daily temperature) and integrated those results into dynamic landscape metapopulation models that projected growth of the regional populations over time. Models projected reduced rates at which nests fledged young and substantial declines in annual productivity with increased warming under three scenarios of future climate change. Average annual productivity across the region declined by an average of 5% under mild (MRI-CGCM3-2.6), 7% under moderate (CanESM2-4.5), and 30% under severe (GFDL-CM3-8.5) climate change scenarios. These patterns produced a range of population declines across scenarios. For example, under severe climate change we predicted a 45% risk the population would experience a 95% decline. Our results predict that climate change poses a substantial threat to flycatcher viability in the Central Hardwoods. These results highlight the importance of considering direct effects of climate change on demographics when assessing vulnerability of species and in conservation planning.
1:30PM Extending the Use of Conditional Occupancy Models to Understand the Role of Competitive Exclusion in an Endangered Plethodontid Salamander
Staci M. Amburgey; David A.W. Miller; Evan H.C. Grant
Models of species distributions have historically focused on the role of climate as a primary ecological filter determining species distributions, but recent research has highlighted the importance of species interactions in determining range boundaries. Species interactions can be modeled in an occupancy framework by conditioning occupancy and detection probabilities of one species upon the probabilities of another. However, these relationships are frequently spatially structured with occupancy probabilities varying by proximity to occupied neighboring sites. We compared the use of 2-species conditional occupancy models to occupancy models with univariate and multivariate conditional autoregressive (CAR) random effects that allow for spatial relationships in these species interactions. We applied this approach to two species of terrestrial plethodontid salamanders where surface activity and thus detectability make observational studies difficult to conduct. The Shenandoah salamander (Plethodon shenandoah) is an endangered terrestrial salamander endemic to only three mountain peaks in Shenandoah National Park, Virginia. The range of this species is encompassed by the range of the red-backed salamander (P. cinereus), and previous work has focused on the potential role of competitive exclusion and interspecific aggression in setting this range edge. We analyzed two years of transect surveys to gain better spatial resolution of the range boundary where these two species meet. Range overlap was historically assumed to be minimal between the species, and raw data showed that both species were not frequently simultaneously observed. Non-spatial conditional occupancy models showed some overlap between the two species, but spatial occupancy models including a CAR random effect gave a higher resolution view of this range edge and the zone of co-occurrence. Individual characteristics of detected animals in these zones can be used to draw inference about competitive exclusion and co-occurrence. This understanding of local level pressures in setting range limits is necessary for management decisions of this endangered species.
1:50PM Estimating Denning Date of Wolves with Daily Movement and GPS Location Fix Failure
Pat Walsh; Suresh Sethi; Bryce Lake; Buck Mangipane; Ryan Neilson; Stacy Lowe
We used Global Positioning System (GPS) radiotelemetry data from 7 breeding female wolves (Canis lupus; n = 14 dennings) in 3 regions across Alaska, USA, during 2008–2011 to develop and compare methods for estimating the onset of denning, and thus infer timing of parturition. We developed and tested 2 estimators based on a combination of GPS radiocollar location-fix failure and distance traveled between locations. We developed a quantitative method employing Generalized Additive Models to smooth time series of wolf data to estimate denning onset. In contrast, 3 study authors with first-hand experience with the study wolves implemented a subjective method of estimating denning onset by visual inspection of detection and distance traveled data. We then tested the visual method for repeatability by subjecting it to 10 wolf experts not associated with this study. Side-by-side comparison of estimators indicates that denning onset can be precisely measured using GPS detection success and distance traveled. Furthermore, the visual-inspection method was simple and rapid to implement and yielded more accurate (relative to assumed dates of denning onset) and precise results compared to the quantitative estimator. Although the Generalized Additive Model based approach had the advantage of estimating denning onset objectively following a set of prescribed rules in a statistical inferential framework, we found the method required significant technical capacity to implement and did not represent an improvement over simple visual-inspection–based estimates of denning onset.
2:10PM Modeling Plague Dynamics to Optimize Vaccination Strategies
Robin E. Russell; Martin D. Grunnill; Katherine L.D. Richgels; Tonie E. Rocke
Recently an oral vaccine for sylvatic plague was field-tested on 29 paired sites located on prairie dog (Cynomys spp) colonies throughout the Intermountain West. Preliminary results suggest that oral vaccination can provide protection from plague; however, the seasonal timing of vaccine distribution influences the levels of bait uptake, and therefore, herd immunity. We developed a spatially-explicit SIR (susceptible-infected-resistant) model of plague transmission dynamics parameterized with field and laboratory data to explore the effects of several scenarios of bait distribution on plague transmission dynamics. Results from spatial capture recapture models are integrated into plague transmission models to provide information on animal movement kernels and contact rates. Distributing baits later in the summer or fall, when juveniles are better able to compete with adults for baits and forage is less available, is more effective at reducing the impacts of plague on prairie dog colonies. However, the dynamics of prairie dog populations, including the number of contacts and size of activity center, strongly influence the outcomes of model scenarios, and the optimal spatial distribution and timing of baiting.
2:30PM An Evolving Sightability Model for a Low-Density Moose Population Using Distance Sampling
Rachel E. Wheat; Jacqueline Frair
Over a three-year period, we adjusted aerial surveys for a low-density population of moose (Alces alces) within the six million-acre Adirondack Park, NY, USA, aiming to improve abundance estimates. The Adirondack Park is the largest publicly protected area in the contiguous United States, and is along the southern range margin for moose, making this a critical region for documenting population fluctuations in the species. In 2015, the first-ever survey for moose in Adirondack Park applied distance sampling with random but adaptive transects, which provided broad geographic coverage at low survey intensity (176 10-km long transects). Adaptive segments increased search intensity by a factor of 1.5 on average and accounted for 20% of the total moose detections, but detections were too sparse for formal density estimation (n = 15 groups total). In 2016, we utilized a stratified random design (80 10×3-km survey blocks) and tripled local survey intensity with 3 parallel 10-km transects per block. Although we achieved an 80% increase in moose detections (n = 27 groups), we still lacked sufficient power to estimate detection probability and density with precision. In 2017, we focused exclusively on gaining a sufficient number of moose detections to estimate a probability of detection with precision. We repeatedly visited areas known to contain moose to obtain sightings under different survey conditions, resulting in 89 moose group detections. Maximizing detections allowed us to build a standalone detection function that incorporates covariates affecting sightability, such as land cover type and group size, that improve the precision of density estimates. In addition, the detection function can be applied to future surveys, enabling density estimation even in years when data might be sparse. We suggest that for rare species, distance sampling might best be used in combination with a standalone sightability model to increase the precision of density estimates.
2:50PM Refreshment Break
3:20PM Evaluating the Roles of Propagule Size, Life History, and Environmental Conditions on the Establishment Success of an Invasive Species Using Transient Dynamics Models
Michael A. Tabak; Ryan S. Miller
Species are commonly released into new environments for two main reasons: by managers to promote global populations of threatened species (i.e., translocations) and by those who wish to increase the global distribution of species for recreation (e.g., rainbow trout and wild ungulates). Once species are introduced, their establishment success depends on propagule pressure, environmental conditions, transient population dynamics, and the life history of the species. We sought to develop a method for predicting the probability of establishment for newly introduced species based on these four factors. We used recently developed transient population dynamics models and vital rate data (i.e., survival and fecundity) for a commonly moved and globally invasive species (wild pigs; Sus scrofa) under different forage conditions to simulate population dynamics following introduction based on environmental conditions and propagule size. Population growth rate and probability of establishment were low in an environment with conditions similar to what wild pigs experience in their natural range, but in anthropogenic environments (environments similar to their invasive range and those where diets can be subsidized by agricultural crops), growth rate and establishment probability were high. Wild pigs appear to switch from slow- to fast life history strategies as the environment improves; juveniles become more important for population growth in environments with better forage. Our results have important implications for management plans, as we identify the age class that is most beneficial to target for eradication programs. As population dynamics in newly introduced species are likely to be dominated by transient effects, we argue that our approach is relevant to modeling the dynamics of such species. Our approach is also valuable for attempts to model the dynamics of populations whose vital rates depend on environmental conditions that fluctuate (e.g., pulsed resource consumers).
3:40PM A Range-Wide Species Status Assessment to Evaluate Past Trends and Future Viability of the Red-Cockaded Woodpecker
Stephanie DeMay; Michael Marshall; Will McDearman; Jeffrey Walters
The United States Fish and Wildlife Service (USFWS) has recently adopted a Species Status Assessment (SSA) Framework to assess the biological needs, current condition, and future viability of species receiving federal protection or proposed to receive protection under the Endangered Species Act (ESA). An SSA is a scientific assessment of resilience, redundancy, and representation to support USFWS listing, recovery, consultation, and other decisions pursuant to the ESA with greater efficiency, defensibility, and transparency. We conducted an SSA for the Red-Cockaded Woodpecker (Picoides borealis), a federally endangered species native to pine forests in the southeastern United States. To assess population resilience, we compiled data from populations widely distributed across the species’ range and performed linear regression on the annual population growth rate r to model the effects of management actions and habitat characteristics on past population growth. Best fit models were used to stochastically simulate future population dynamics under various management scenarios. Over the past several decades, population sizes have been generally increasing, and model results reinforce the conservation reliance of this species. Active management, specifically installation and maintenance of artificial tree cavities and midstory management, is indicated as crucial to population persistence and growth. The importance of active management for woodpeckers is strongest for smaller populations (<30 active groups), which tend to receive less consistent active management than larger populations on government-managed properties. Results from this range-wide analysis will provide key information to both USFWS decision-makers operating within the ESA and partnering property managers working to conserve Red-Cockaded Woodpecker populations.
4:00PM Simulations Inform Design of Regional Occupancy-Based Monitoring for a Sparsely Distributed, Mobile Species
Quresh S. Latif; Martha M. Ellis; Victoria A. Saab; Kim Mellen-McLean
Sparsely distributed, mobile species attract management concern. Insufficient information on population trends, however, challenges conservation and funding prioritization. Occupancy-based methods are cost effective and therefore attractive for broad-scale trend monitoring, but appropriate sampling design and inference depend on particulars of the study system. We employed spatially explicit simulations to inform regional occupancy-based monitoring of white-headed woodpeckers (Picoides albolvartus), a sparsely distributed, mobile species threatened by habitat decline and degradation. We used population simulations based on knowledge of species ecology to compare statistical power and trend estimation error under alternative scenarios. We found maximum power when using occupancy as an index of abundance, which was best served by estimates from single-survey data with auxiliary data to account for detectability. Assuming the baseline sampling allocation scheme (surveying all transects with 10 points each every year), sampling effort needed to achieve adequate power to observe a long-term population trend (≥ 80% chance to observe a 2% yearly population decline over 20 years) consisted of monitoring ≥ 120 transects using the single-survey approach or ≥ 90 transects using a repeat-survey approach. Alternate allocation schemes improved statistical power and trend estimates over the baseline, including surveying a subset (33%) of transects each year (i.e., a panel design) and surveying fewer points per transect in exchange for a larger spatial sample. Considering this case study, single-survey methods (with separate evaluation of detectability), panel designs, and aligning sampling resolution with home range size could likely benefit broad-scale occupancy-based monitoring of other sparsely distributed and mobile species.
4:20PM Identifying Limits to Red-Cockaded Woodpecker Population Growth
Emily Brown; Cynthia Ragland; Paige F. B. Ferguson
The Red-cockaded Woodpecker (Picoides borealis; RCW) is listed as Endangered under the United States Endangered Species Act. The largest population of RCW in Alabama is found in the Oakmulgee Ranger District of the Talladega National Forest. Efforts are underway in the Oakmulgee to restore longleaf ecosystems and conserve RCW through methods such as controlled burns, restoring mixed stands to longleaf, and installing artificial cavities. Despite these efforts, the number of active RCW clusters has not exceeded 120, although the District’s Recovery Plan objective is 394 active clusters. Our objectives are to identify factors limiting RCW population growth and identify management methods that could reduce these limitations. We held four structured decision making workshops with representatives from the United States Forest Service, the Animal and Plant Health Inspection Service, the Birmingham Audubon Society, the Longleaf Alliance, and local residents. We built a decision network that predicted the relative likelihood of a range of management options to meet stakeholder objectives, including increase the number of RCW clusters. In addition, we collected field data related to factors the decision network identified as influencing the number of RCW clusters. Cavity insert installation had the greatest probability of increasing the number of RCW clusters, and prescribed burning was most likely to meet the combination of stakeholder objectives. There was some support for midstory removal meeting stakeholder objectives. The decision network suggested that the number of RCW clusters is affected by helper and breeder survival, recruitment rates, food availability, and herbaceous understory. RCW population growth in the Oakmulgee may be limited by a combination of the number of available cavities, rates of survival and recruitment, and the structure and composition of vegetation. The decision network based on stakeholder objectives will be the framework for addressing future questions.
4:40PM Intervals of Resource Selection Across Telemetered Animals Enhances Interpretations of Habitat Requirements
Brian S. Cade; Douglas S. Ouren
We developed a method to summarize relative selection ratios from resource selection functions (RSFs) estimated on individual telemetered animals that provides useful interpretations and predictions that account for variation among individual animals. Our approach permits a variety of informative summaries about the resources that are less to more highly selected by any to most individual animals (any proportion of individuals) and where those resources are located. We used the procedure for an analysis of resource selection by the Crawford subpopulation of Gunnison’s sage-grouse (Centrocercus minimus) in southwestern Colorado, USA. We estimated exponential forms of a RSF for the breeding season for 8 female grouse located with GPS collars up to 14 times a day for 1-3 years during 2011-2014. Our RSFs for 14 bird-year combinations were based on logistic regression models using b-splines on continuous predictors (NDVI, heat load index, distance to water, distance to road) to capture nonlinear responses, four vegetation categories, and three vegetation treatment categories. The RSF models for each of the 14 bird-years were used to estimate relative selection ratios for each of the available sample locations (n = 47139) which were then rescaled into their quantiles [0, 1]. The 14 quantiles of relative selection ratios for each available sample observation then were used to determine resources and their locations that had higher relative selection ratios (quantiles >0.6) for at least one bird-year and for 90% of the bird-years (13 of 14). Resources that were consistently highly selected in 90% of the bird-years were more restricted in values (e.g., NDVI = 0.2 – 0.4) and occurred on only 5% of the available area, whereas resources that were highly selected in at least one bird-year included a greater range of values (e.g., NDVI = 0.1 – 0.6) on 73% of the available area.

 

Contributed Paper
Location: Albuquerque Convention Center Date: September 24, 2017 Time: 1:10 pm - 5:00 pm