Understanding and predicting ecological responses to climate change is crucial if we are to manage for the detrimental consequences that might ensue in its wake. This thesis looks to develop some new ideas and tools for ecological prediction under climate change, focusing on species traits rather than species themselves as units of prediction. Chapter One begins by reviewing the basis for a trait-based approach to prediction, presenting evidence from natural and experimental systems that responses to climate change cluster by traits. Chapter Two undertakes a proof-of-concept modelling study, using a well-known dataset of contemporary phenological changes under warming to test for trait-based links in the strength and direction of species responses. Chapter Three addresses the critical issue of transferability. One of the strongest justifications for the use of a trait-based approach is that inferences may extend more generally beyond the focal species and system. To test this, it develops trait-based models for long-term datasets of first-arrival dates for migratory birds in two neighbouring US states and assesses cross-applicability between them.
Having investigated the functionality and transferability of the trait-based approach, I then explore the bounds of its utility. Chapter Four uses a detailed record of community-wide changes in species abundance under ten years of experimental climate change to assess whether changes in abundance cluster by traits. It also re-evaluates the prevailing hypothesis guiding researchers’ interpretations of ecological changes in this system. Chapter Five designs and implements an experiment building on the results of Chapter Four. This experiment tests the role of climatic effects on nitrogen-fixation in driving system dynamics via controlled removal of this trait from experimental communities. Overall, this thesis sheds new light on the role of traits in ecological responses to climate change, highlighting opportunities and limitations for using traits to organise our thinking over prediction and adaptation.