Abstract

The threat to biodiversity from human land use and climate change has spurred the growth of
ecological subfields centered on conservation. Conserving biodiversity requires the ability to
accurately link biodiversity patterns to ecological processes and generalize across systems and
species. Traditionally, this has been attempted with univariate modeling of a diversity measure or
with ordination analyses, neither of which allows for the explicit modeling of processes that act
at the species level. Conceptual, statistical and computational improvements have led to the
development of more explicit, hypothesis-driven models such as species and joint species
distribution models. However, biodiversity patterns are known to be scale-dependent, and the
issue of spatial scale is commonly left unaddressed in these models. This is especially true for
community ecology, where multiple species distributions are modelled simultaneously.
Biodiversity patterns are ubiquitous in ecological systems and can be found across a wide range
of spatial scales, and the choice of spatial scale in a study affects the ability to accurately detect
and link pattern to process. From the broadest perspective, my thesis investigates scale
dependence in biodiversity patterns at three levels: landscape, species, and genetic diversity.
Chapter 2 uses a previously published dataset of bee species occurrence and abundance to
investigate and model the scale dependency of species-landcover relationships in a
heterogeneous landscape. In this chapter, I found that scale dependency in bee species-landcover
associations is more pervasive and complex than previously assumed. Building off these results,
Chapter 3 finds that scale dependency of species-landcover associations is not a species-specific
trait but depends on the combination of species and landcover, so that the scale of effect cannot
be predicted by body size, sociality, or inferred from closely related species. Similarly, Chapter 4
finds that the degree of spatial structuring in neutral genetic variation within species cannot be
compared across populations, species, or landscapes. Overall, my thesis concludes that we
cannot reliably assess patterns in biodiversity unless we address the complexity of scale
dependence. In a bee community specifically, species-landcover relationships in an agricultural
ecosystem should be managed at both local and landscape scales. It is paramount that scale
dependence be addressed in ecological models to avoid erroneous inference about biodiversity
processes.