Biometrics and Population Modeling III

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

10:30AM Estimating Bird Species’ Responses to Sandhill Habitat Restoration in the Southeastern United States: Single-Visit Surveys, an Unbalanced Design, and Varying Methodologies
Bryan L. Nuse; Matt Elliott; Clinton T. Moore
Survey data are rarely collected under ideal conditions, and broad-scale, coordinated survey efforts involving multiple entities may be subject to variation in methodological approaches. From an analysis perspective such aberrations in methods vary on a continuum from negligible, to differences so large that a meta-analysis approach might be considered. We consider a set of surveys conducted by a group of four states in the southeastern US (AL, FL, GA, MS), during a project to restore sandhill habitat and monitor bird species’ responses in 17 distinct state-managed areas. A blocked before-after-control-impact design was generally followed, although there was severe imbalance in many of the blocks; not all sites were revisited the same number of times; and interspersion was minimal or absent in some areas. Only one visit was made to each site in a given year, but a removal design was used for point counts, and distance information was collected: the data thus adhered to a zero-inflated version of the abundance model of Amundson et al. (2014). However, point counts varied in duration and total area, and in the number and widths of distance bands. The dataset thus displayed a moderate amount of methodological inconsistency, which we attempted to accommodate within a single Bayesian hierarchical model. The Bayesian framework allowed fairly straightforward and transparent synthesis of the data into an ANOVA model with which we estimated treatment effects. We also accounted for model uncertainty with a Bayesian variable selection technique, allowing inference about treatment effects upon either occupancy, abundance, or both. Several species of conservation concern displayed positive responses to sandhill restoration efforts, including Bachman’s Sparrow (Peucaea aestivalis), Eastern Bluebird (Sialia sialis), Northern Bobwhite (Colinus virginianus), Prairie Warbler (Setophaga discolor) and Yellow-breasted Chat (Icteria virens).
10:50AM Timing of Predation and Carnivore Movement in an Urban-Wildland Interface
Frances Buderman; Mevin Hooten; Mat Alldredge
Carnivores inhabiting the urban-wildland interface encounter a different set of challenges than those in wildland areas. While threats stemming from interactions with anthropogenic landscape features increase in urban and residential areas, these areas can provide an easily accessible and abundant prey base. However, as more time passes since a predation event, carnivores may be more likely to engage in risky behavior, entering residential areas that they would not enter if they had fed recently. Time-varying responses to landscape features can be modeled using a varying-coefficient model in a continuous-time discrete-space (CTDS) framework. Unlike resource- and step-selection functions, the CTDS model makes inference on the unobserved movement path. In addition, the response variable in the CTDS model is the transition rate (speed and directionality) of an individual, not use of an area. We used a hierarchical CTDS model to assess whether cougars in the Colorado Front Range, a rapidly urbanizing area of Colorado, responded differently to landscape features as the time since a potential predation event occurred. We modeled cougar response to geomorphological landscape features, as well as a set of covariates that approximated risk- and reward-based landscape features. In addition, we analyzed the time-varying interaction between prey utilization and potential kill site locations and the amount of residential development. We found that cougar movement rates varied as a function of time since potential predation event. For example, individuals moved more slowly as distance to roofed structure decreased as the time since predation event increased. In addition, movement rate as a function of time since potential predation event varied between residential and non-residential areas. In conclusion, time-varying responses can highlight movement behavior that would be unobserved in a traditional static response framework and information about prey availability may be useful in predicting where and when mountain lions may enter residential areas.
11:10AM A Bayesian Hierarchical Model of Loggerhead Sea Turtle Nest Success Response to Extreme High Tide Events
Alicia Wilson
Threats to coastal Georgia barrier islands from global climate change include increasing vulnerability to storm surge, flooding and high tide events. The Georgia coast is prime reproductive habitat for loggerhead sea turtles (Caretta caretta) and 2,000 to 3,000 nests are laid annually on approximately 14 of the state’s barrier islands. Loggerhead sea turtles are listed as threatened under the U.S. Endangered Species Act and threats include incidental capture in fishing nets, coastal development and human disturbance. The objectives of this study were to quantify the changes in frequency of high tide events and assess effects of tide dynamics on nest success of sea turtles using four of Georgia’s islands. We collected data on nest elevation, number of inundation events, microhabitat elevations and daily high tide levels for 711 nests from Ossabaw Island, Sapelo Island, Little St. Simon’s Island and Jekyll Island. We used a hierarchical model within a Bayesian framework to estimate the relationships between these physical attributes and nest success. We modeled nest success as a function of nest elevation and a zero-inflated process of nest destruction driven by repeated inundation events. Number of inundation events were in turn modeled as a function of nest elevation and number of high tide events during incubation. Finally, nest elevation was modeled as a function of tide height on the night of the turtle’s emergence. By understanding the processes that drive nest site selection, we can better evaluate nest success and better inform decision making for conservation measures, including decisions about nest relocation procedures. Predicting tidal trends will also enable us to better understand consequences of an increasing frequency of extreme high tides on sea turtle demography and population recovery.
11:30AM Estimating the Survival of Dependent Young from Repeated Counts: A Hierarchical Bayesian Approach
Timothy P. Lyons; Kirk W. Stodola; Thomas J. Benson
Accurate estimates of juvenile survival are vital to wildlife management, but obtaining such estimates can be difficult. Common methods to monitor survival, such as leg bands or radio-transmitters, may be impractical for some species due to limited resources or concerns about animal welfare. Methods that estimate survival from repeated counts are available but may perform poorly with large group numbers or small sample sizes. Furthermore, the number of visits needed to obtain desired levels of precision may be impractical or would require undue disturbance for cryptic species. To address these problems, we developed a hierarchical Bayesian model that accurately estimates period survival and detection probability from a minimum of two counts of unmarked young accompanied by a marked adult. We simulated data under a variety of scenarios to examine model performance, including varying detection probability, sample size, number of visits, as well as the violation of closure between repeated counts. We then applied this model to estimate chick survival of ring-necked pheasants (Phasianus colchicus) in east-central Illinois 2015-2016. Our simulation study showed bias was usually < 5%, but the model failed to converge when detection was < 0.4 and yielded biased estimates under gross violations of the closure assumption. In most cases, increasing the number of visits or minimizing the time between repeated counts sufficiently addressed these problems, but we identified scenarios where the use of this model may not be appropriate. During our study, chick survival to 20d post-hatch was generally high, and averaged 68% in 2015 and 84% in 2016. Precipitation during the first 5d post-hatch had a strong negative effect on chick survival. Overall, our results suggest that the combination of structured counts and an appropriate statistical model can accurately estimate the survival of dependent young in pheasants, and may be effective for other species as well.
11:50AM Data Weighting for Integrated Population Models
Charles R. Henderson; Paul M. Lukacs; Mark A. Hurley
Integrated population models (IPM) enable managers and researchers to analyze different types of data together, leading to an increase in the understanding of the demographic processes that drive population trajectories and ultimately providing better information for management decisions. Currently, all data types included in an IPM have equal influence on the abundance estimates generated by the model; however, the relative quality of different types of data varies. We sought to create a weighting system that reflects the relative quality of each data type. We used the amount of variance and the accuracy of underlying assumptions of the models associated with each data type as our measures of quality. We used data previously gathered by IDFG (Idaho Department of Fish and Game) about mule deer (Odocoileus hemionus) populations to construct and test this method. Data types included in this analysis are survival, sightability, composition, and harvest. We devised a weighting system based on improved variance structures that incorporates model assumptions and estimate precision within the current IPM framework. The expected result is a better representation of the variance around the abundance estimates generated by the IPM. This method of data weighting should ensure that each data type influences the IPM estimate of abundance in proportion to its relative quality and allows different quality data types to be exploited to their full potential. In addition, the weights will allow managers and researchers to compare the relative quality of different data types and allocate monitoring resources accordingly.


Contributed Paper
Location: Albuquerque Convention Center Date: September 27, 2017 Time: 10:30 am - 12:10 pm