Geology topics

Kylee Dunham

Publications and source records attributed to Kylee Dunham.

3 recordsLinked to original sources

USGS2016

Evaluating models of population process in a threatened population of Steller’s eiders: A retrospective approach

The Alaskan breeding population of Steller’s eiders ( Polysticta stelleri ) was listed as threatened under the Endangered Species Act in 1997 in response to perceived declines in abundance throughout their breeding and nesting range. Aerial surveys suggest the breeding population is small and highly variable in number, with zero birds counted in 5 of the last 25 years. Research was conducted to evaluate competing population process models of Alaskan-breeding Steller’s eiders through comparison of model projections to aerial survey data. To evaluate model efficacy and estimate demographic parameters, a Bayesian state-space modeling framework was used and each model was fit to counts from the annual aerial surveys, using sequential importance sampling and resampling. The results strongly support that the Alaskan breeding population experiences population level nonbreeding events and is open to exchange with the larger Russian-Pacific breeding population. Current recovery criteria for the Alaskan breeding population rely heavily on the ability to estimate population viability. The results of this investigation provide an informative model of the population process that can be used to examine future population states and assess the population in terms of the current recovery and reclassification criteria.

Open-File Report↗

USGS2016

Viability of the Alaskan breeding population of Steller’s eiders

The U.S. Fish and Wildlife Service is tasked with setting objective and measurable criteria for delisting species or populations listed under the Endangered Species Act. Determining the acceptable threshold for extinction risk for any species or population is a challenging task, particularly when facing marked uncertainty. The Alaskan breeding population of Steller’s eiders ( Polysticta stelleri ) was listed as threatened under the Endangered Species Act in 1997 because of a perceived decline in abundance throughout their nesting range and geographic isolation from the Russian breeding population. Previous genetic studies and modeling efforts, however, suggest that there may be dispersal from the Russian breeding population. Additionally, evidence exists of population level nonbreeding events. Research was conducted to estimate population viability of the Alaskan breeding population of Steller’s eiders, using both an open and closed model of population process for this threatened population. Projections under a closed population model suggest this population has a 100 percent probability of extinction within 42 years. Projections under an open population model suggest that with immigration there is no probability of permanent extinction. Because of random immigration process and nonbreeding behavior, however, it is likely that this population will continue to be present in low and highly variable numbers on the breeding grounds in Alaska. Monitoring the winter population, which includes both Russian and Alaskan breeding birds, may offer a more comprehensive indication of population viability.

Open-File Report↗

USGS2016

Effects of model complexity and priors on estimation using sequential importance sampling/resampling for species conservation

We examined the effects of complexity and priors on the accuracy of models used to estimate ecological and observational processes, and to make predictions regarding population size and structure. State-space models are useful for estimating complex, unobservable population processes and making predictions about future populations based on limited data. To better understand the utility of state space models in evaluating population dynamics, we used them in a Bayesian framework and compared the accuracy of models with differing complexity, with and without informative priors using sequential importance sampling/resampling (SISR). Count data were simulated for 25 years using known parameters and observation process for each model. We used kernel smoothing to reduce the effect of particle depletion, which is common when estimating both states and parameters with SISR. Models using informative priors estimated parameter values and population size with greater accuracy than their non-informative counterparts. While the estimates of population size and trend did not suffer greatly in models using non-informative priors, the algorithm was unable to accurately estimate demographic parameters. This model framework provides reasonable estimates of population size when little to no information is available; however, when information on some vital rates is available, SISR can be used to obtain more precise estimates of population size and process. Incorporating model complexity such as that required by structured populations with stage-specific vital rates affects precision and accuracy when estimating latent population variables and predicting population dynamics. These results are important to consider when designing monitoring programs and conservation efforts requiring management of specific population segments.

Ecological Modelling↗