Optimizing landbird surveys for detecting population and spatial dynamics

Abstract

Landbird populations are undergoing concurrent changes in population size, spatial distribution, and phenology. The sensitivity of landbird monitoring programs to detect and distinguish these varied processes is of critical importance. Consequently, these efforts require inference methods that are efficient and fully leverage information about spatial, population, and phenological dynamics. The development of efficient inference methods can be addressed in part through a thorough understanding of how the data are actually generated, the application of sampling methods that attempt to maximize encounter probability, and the tailoring of sampling methods to maximize sensitivity to specific inference objectives. Chapter one of this dissertation is concerned with accommodating temporary emigration in spatial distance sampling models. Model-based distance sampling is commonly used to understand spatial variation in the density of wildlife species. The standard approach is to assume that individuals are distributed uniformly in space and model spatial variation in abundance using plot-level effects. Thinned point process models for surveys of unmarked populations (spatial distance sampling) frame the sampling process in terms of the individual encounter in space and, consequently, are expected to offer greater sensitivity for understanding spatial processes. However, existing spatial distance sampling approaches are conditioned on the assumption that all individuals are present and available for sampling. Temporary emigration of individuals can therefore result in biased estimates of abundance. Herein, I extend spatial distance sampling models to accommodate temporary emigration. A simulation study indicated more precise and less biased estimation under the spatial distance sampling model compared to models that assume a uniform distribution of individuals and assess spatial variation in abundance using plot-level effects. An applied example involving two arctic-breeding passerines indicated considerably stronger inference under the spatial distance sampling model than standard distance sampling models. Chapter two is concerned with the capacity of subarctic passerines to adjust their arrival timing to relatively extreme variation in spring conditions. I assessed interannual variation in passerine arrival timing in Denali National Park, Alaska from 1995-2015, a period that included both the warmest and coldest recorded mean spring temperatures for the park. Neotropical-Nearctic migrants varied in terms of the flexibility of their arrival timing, but generally showed plastic phenologies, suggesting resilience under extreme spring conditions. In comparison, Nearctic-Nearctic migrants showed similar or greater plasticity in arrival timing. A majority of species showed synchronous-asynchronous fluctuation in arrival (i.e., synchronous arrival in some years, asynchronous in others) in combination with various levels of the mean response (i.e., early, average, and late arrival), suggesting the presence of interactions between environmental conditions at multiple scales and inter-individual variation. Overall, these findings suggest that monitoring of the mean-variance relationship may lead to a deeper understanding of the factors shaping phenological responses. Chapter three is concerned with developing efficient inference methods for inventorying and monitoring cliff-nesting raptor populations. In nest occupancy studies of cliff-nesting raptors, the standard approach is to allocate a level of survey effort that is assumed to ensure that the occupancy state is known with certainty. However, allocating effort in this manner is inefficient, particularly at landscape scales, constraining our capacity for effective management of these species. To increase survey efficiency and expand the spatial inference of these studies, I developed two versions of a multi-state, time-removal model, one for long-term monitoring studies and another for population inventories or single-season surveys in which there is no prior knowledge of nest locations. For long-term monitoring of species with alternative nests, I formulated a version of the model that accounts for state uncertainty at the territory-level caused by a failure to observe all nests within a territory. Simulation studies indicated generally low to moderate relative bias under the monitoring and inventory models. In addition, I applied the monitoring model to a long-term study of golden eagles (Aquila chrysaetos) in Alaska and demonstrate that the maximum effort spent on any nesting territory could be reduced by up to almost 90% of that recommended by standard protocols.