Biometrics & Population Modeling I

Contributed Paper
ROOM: HCCC, Room 19

8:10AM Regularization: Squeezing the Most Out of Your Data
Daniel Walsh; Dennis Heisey; Robin Russell; Andrew Norton; Daniel Storm
Wildlife biologists are often faced with the challenge of making management decisions in the face of substantial uncertainty. In these situations, it is of paramount importance that the utmost information be extracted from the available data. One tool in ecological modeling that can be leveraged to maximize the knowledge gained from sparse data is regularization. Regularization has a rich history in statistics and machine learning where it has been successfully applied to a wide array of challenging problems, but it has been largely underappreciated in the wildlife field. To help make regularization techniques more accessible, we provide a heuristic motivation for regularization using a one-way anova as a simple motivating example. We then detail how regularization can be implemented using Bayesian hierarchical models, and how these models can accommodate model selection uncertainty and model-averaging automatically. To demonstrate the application of regularization techniques in a hierarchical Bayesian framework and the benefit they provide, we apply them to age-at-harvest modeling. Age-at-harvest information is commonly collected by wildlife management agencies for a wide variety of hunted species. This information often forms the basis for monitoring demographics and determining license quotas. Additionally, for many species (e.g., large carnivores, fur-bearers, etc.) age-at-harvest data may represent the sole source of information available. This represents a challenging analytical problem for which traditionally auxiliary data has been suggested as a requisite for making inference. We use simulation techniques to show how regularization methods permit estimation of key demographic parameters in a rigorous manner even in the absence of auxiliary information. This provides wildlife managers with a valuable method to reduce uncertainty and inform decisions in data-poor systems.
8:30AM Synthesizing Multiple Data Streams to Estimate Abundance and Detection Probability in a Highly Wary, Long-Lived Apex Predator: an Integrated Modeling Approach
Abby J. Lawson; Clint T. Moore; Beth E. Ross; Patrick GR Jodice
American alligators (Alligator mississippiensis) are a species of ecological, economic, and cultural importance in the southeastern United States. Though several states within their distribution have successfully implemented sustainable harvest programs over multiple decades, management decisions are often executed despite sweeping demographic uncertainties. Alligator demographic parameter estimates are difficult to obtain due to their large body size, long lifespan (80+ years), and highly wary behavior towards common monitoring methodology. We synthesized mark-recapture, telemetry, and replicated count data from South Carolina to develop the first-known integrated population model (IPM) for alligators. We used data from a long-term mark-recapture study (1981-2016, n=802 individuals) to create a multistate modeling framework that included six live stages and dead recoveries. We then integrated the most parsimonious multistate model with replicated spotlight count surveys (n=115, 2013-2016). Within the count component of our IPM, we used a modified dynamic N-mixture model and expressed expected abundances partly as functions of state-transition probabilities from the multistate model. Previously published alligator detection probability estimates are typically very low (<0.10) due to detection error and temporary emigration (e.g., submergence, movements)— even within short time periods. To model the detection process, we included environmental covariates known to influence submergence patterns and time-varying covariates that quantified movement rates from a spatiotemporally-overlapping GPS-based telemetry study. The inclusion of both mark-recapture data and telemetry-derived covariates that explicitly accounted for temporary emigration greatly enhanced our ability to obtain reliable survival, recruitment, and abundance estimates in a Bayesian IPM framework. Our study is among the first to provide demographic parameter estimates that account for imperfect detection in crocodilians and is a recent application of telemetry data to model detection in abundance estimation. Here we present the results of our integrated modeling efforts and discuss management applications (e.g., harvest), potentially to other populations or cryptic species.
8:50AM Design Considerations for Estimating Survival Rates with Standing Age Structures
Rebecca L. Taylor; Mark S. Udevitz
Survival rate estimates are critical to understanding the dynamics and status of a population, and they are often inferred from samples of the population’s age structure. A recently developed method uses time series of standing age-structure data with information about population growth rate or fecundity to provide explicit maximum likelihood estimators of age-specific survival rates, without assuming population stability or stationarity. We explored properties of these estimators and determined sample size requirements for the estimators to achieve desired levels of precision, limit bias, and limit the probability a rate will be inestimable or its estimate inadmissible (>1). We show that estimating combined rates for adjacent age classes is an effective method of overcoming sensitivity to sampling noise in situations where collecting a larger sample of data is not feasible. Finally, we developed a 5-step method to help wildlife biologists determine whether to estimate survival rates with this method or whether their data would best be analyzed by other methods.
9:10AM Examining the Utility of Time to Event Abundance Estimation for Low Density Species
Kenneth Loonam
Abundance estimation is a common task in wildlife biology, but techniques to estimate the abundance of low density, difficult to detect species are limited, often requiring intensive field effort and incurring high costs. Remote cameras offer an effective means of detecting these species, but most abundance estimation methods using remote camera data rely on a portion of the population being marked or uniquely identifiable. Methods to estimate the abundance of populations without identifiable or marked individuals using remote cameras have assumptions that are difficult to meet in field studies. The recent application of time to event modelling to abundance estimation relaxes these assumptions, requiring only random movement with respect to the cameras, an estimate of movement rates, and a demographically closed population (Moeller 2017), assumptions that can still be difficult to meet in low density populations. To examine the utility of the time to event model for estimating the abundance of low density populations, I use simulated walk models to test the robustness of the time to event model to violations of these assumptions including (1) territoriality, (2) camera placement biased toward high use habitats, (3) changing population size during the survey, and (4) incorrect estimations of movement speed. I estimate the magnitude of biases introduced by violating the time to event assumptions and test methods that correct for the biases that are introduced. Biased camera placement, open populations, and incorrect movement estimates bias abundance estimats but can be accounted for with supplemental data sources.
9:30AM Opportunity Costs of the Flagship Approach to Farmland Conservation
John M. Yeiser; John J. Morgan; Danna L. Baxley; Richard B. Chandler; James A. Martin
Conserving biodiversity on farmlands despite the increasing demands of human populations is a global imperative. The Conservation Reserve Program has been a major avenue for managing wildlife on farmlands, especially game birds like Northern Bobwhite (Colinus virginianus). Northern Bobwhites are considered a “flagship” species for grassland conservation because they are of great recreational and cultural value, have relatively abundant revenue sources, and have similar habitat needs to other grassland birds of conservation concern. However, agricultural landscapes are complex networks of production, natural, and semi-natural land cover. The functional value of a patch type to two similar species may vary because of differing life history requirements. Furthermore, the spatial scale at which land management influences populations (the scale of effect) differs among species. Given these potential conflicting factors, we sought to investigate the opportunity costs of managing landscapes for Northern Bobwhite on a suite of similar species. We used an open-population distance-sampling model with an embedded predictor of scale of effect to estimate the relationship between population growth and landscape structure. We used a novel decision support tool to predict spatially explicit densities under competing management scenarios. We found that several species had similar responses to landscape structure as Northern Bobwhite, but the scale of effect of landscape structure on population growth varied. For example, the area surrounding sites that mattered to local Northern Bobwhite population growth was a 12 km radius circle, while for Dickcissels it was a 20 km radius circle. We predicted that managing landscapes for Northern Bobwhite would incur opportunity costs of -82% for Dickcissels, 0.8% for Eastern Meadowlarks, and -5.8% for Field Sparrows. Although the flagship approach is assumed to benefit a suite of species, we provide evidence that there are opportunity costs that need to be considered when planning and implementing farmland conservation.


Contributed Paper
Location: Huntington Convention Center of Cleveland Date: October 8, 2018 Time: 8:10 am - 9:50 am