Human societies are dependent on the functioning of ecosystems and the associated services they provide. However, the combinations and strengths of stressors acting upon both nat- ural and managed ecosystems is growing, with abrupt non-linear shifts in ecosystem state and function increasingly of concern. Early warning signal (EWS) indicators have emerged as a conceptually generic tool for predicting these shifts but are inconsistent when applied to real-world data. In this thesis, I explore the functionality of cutting edge EWS tech- niques in highly complex and multivariate empirical systems to guide their practical usage in conservation and ecosystem management. I first describe the performance of classical univariate EWS approaches in a disease system involving multiple sequential outbreaks, to highlight the impact of multiple ‘regime shifts’ on EWS ability. I then present a new R package which consolidates current EWS research techniques, pulling together univariate, multivariate and machine learning approaches to regime shift and resilience loss detection. Using this toolbox, I identify which techniques are most robust across transitioning and non- transitioning lake plankton communities and reveal how - despite their widespread use - the mathematical requirements for EWSs are often not met. I then challenge the assumption that generic trait information is an indicator of ecosystem stability change, showing that plankton functional diversity changes synchronously with ecosystem state. Finally, I inves- tigate alternative methodologies for quantifying ecosystem resilience away from equilibrium and show that these alternative metrics are predictable in their behaviour across simulated systems. This thesis therefore provides ecosystem managers guidance regarding when and where EWSs are informative, while also suggesting viable best-practice for their usage.
Complexity and regime shifts: testing the predictability of ecosystem transitions
O'Brien, D. A. (Author). 7 May 2024
Student thesis: Doctoral Thesis › Doctor of Philosophy (PhD)