Ecological monitoring in the age of spatial awareness: Novel approaches to estimating status and trend in the presence of spatial complexity

Spatially-explicit approaches to designing, analyzing, and interpreting ecological studies have been widely described. However, there remains a lag in implementation, particularly in the arena of long-term monitoring. This is somewhat surprising given that monitoring studies are particularly prone to inheriting spatial structure from past land use and management unit boundaries, in addition to the omnipresent environmental gradients and endogenous processes such as dispersal that also induce spatial structure. One reason for this lag is that spatially-explicit statistical techniques are complex, computationally intensive, and often difficult to implement. Another reason seems to be the inherently multi-disciplinary nature of monitoring -- few of us are trained specifically to do monitoring. Monitoring teams are assembled from local subject-matter experts, statisticians, and decision-makers for specific projects, but rarely and only recently have large-scale monitoring programs assembled permanent teams focused solely on the design and implementation of ecological monitoring. Monitoring programs therefore often follow parochial methods reflecting the usual and accustomed research methods for a given organism, slowing the dissemination of novel approaches to long-term monitoring.;In an effort to advance the state of the science of ecological monitoring, I pursued 3 case studies, each presenting novel taxonomic and spatio-temporal challenges. While each case study addressed its own narrow set of objectives and hypotheses, taken together the collection provides a "tour" of cutting-edge sampling design and analytical motivating examples for dealing with spatial complexity in the environment. Chapter 1 presents an application of spatially-balanced sampling to the specific challenge of monitoring bats with remotely-deployed acoustic detectors along a stream stratified by restoration treatment and control designations. The approach taken is demonstrated to be practically and statistically efficient, and provides a substantial improvement over common alternatives. Chapter 2 develops a trend model for a population of the facultative wetland plant, camas lily (Camassia quamash) that is also undergoing restoration. Using Bayesian hierarchical modeling techniques, the study provides a method for dealing with complex anisotropy (spatial non-stationarity across the study area) resulting from patterns of past land use. The resulting temporal trend model is rich with spatial information. By way of sequential model development, I demonstrate how the strength of evidence for trend is fundamentally influenced by the amount of spatial complexity included in the model. Finally, in Chapter 3 I develop a dynamic distribution model for the little brown bat (Myotis lucifugus) across Oregon and Washington. Using a Bayesian occupancy modeling approach, I account for temporal and spatial dependence, as well as imperfect detection, all issues of critical importance for bat surveys. While the primary motivation is to establish baseline trend in distributional patterns against which future anticipated declines can be measured, I also address the species-energy hypothesis by including forest cover and net primary productivity as model predictors. Results of the study reveal the importance of accounting for imperfect detection, as well as a positive trend in little brown bat occurrence probabilities across the productivity gradient in the region. This latter finding has important implications for the potential impacts of climate change on the species.;These examples demonstrate both the technical and conceptual challenges as well as the value of taking a spatially-explicit approach to monitoring. Importantly, each of these studies does more than just account for spatial autocorrelation as a statistical nuisance; they effectively harness the spatial information inherent in the system at hand and provide deeper insights into the ecological processes at work. Despite the added complexity, these spatial approaches to monitoring should be of tremendous interest to land managers and decision-makers. Given the complexity of contemporary natural resource management and conservation decision-making, it is no longer sufficient for monitoring programs to simply report on whether a given resource is "going up or down". Rather, monitoring programs increasingly must provide insights into "why" and "how" resources are changing over time. The kinds of spatially-explicit approaches to estimating status and trend explored in this dissertation have resulted in robust methods and powerful results that I expect will contribute to the evolution of ecological monitoring and further its advancement "into the age of spatial awareness".