Modeling Realities: Towards an Improved Understanding of Population Responses to Changes in Complex Environmental Drivers

Abstract

Population models are powerful tools that have long been used to explore the interplay between population dynamics and environmental drivers, including human-induced changes in environmental conditions. Yet these models often require simplifying assumptions that have the potential to break down when the relationships between demographic vital rates (survival, growth, reproduction, and recruitment) and drivers become complex or when multiple co-occurring drivers interact. As humans continue to alter global ecosystems, there is a growing need to ensure that models are able to accurately capture population responses to increasingly complicated driver scenarios, including the potential for nonlinear relationships and nonadditive interactions among co-occurring drivers. In this dissertation, I explore this question of model accuracy in three ways. First, I (along with collaborators) extend a common post-hoc analysis tool – the life table response experiment – to incorporate potential second-order population responses to drivers and compare the resulting accuracy of this extended approach to the standard method that assumes only linear relationships. Second, I combine global climate model projections, a climate vs. wildfire model, and empirical demographic data for a threatened North Carolina endemic plant (mountain golden heather; Hudsonia montana) in order to assess the extent to which climate change influences population dynamics directly, as well as indirectly by altering wildfire probability. Finally, I examine the outputs of a handful of population models built using either different underlying data or different assumptions in order to characterize how management recommendations can be shaped by various aspects of data limitation. The results of the first two chapters demonstrate that extending models to account for more complex drivers and population responses can both increase the accuracy of population projections and improve the quality of our inferences about the underlying sources of observed population changes in response to shifting environmental conditions. On the other hand, the results of the third chapter reveal the extent to which modeling output is contingent on the types of assumptions necessitated when data are sparse and highlights the need to balance preconceptions of optimal management with the realities of scientific uncertainty.