Now showing 1 - 1 of 1
ItemRobust prediction and decision strategies for managing extinction risks under climate changeBaumgartner, John Bruno ( 2016)Effective management of biodiversity requires decision strategies that are robust to the uncertainty embodied in predictions of habitat suitability and environmental change. This is particularly relevant in the context of climate change, which may interact with existing threats in unexpected ways. Predictive modelling has become important for addressing questions about climate change impacts. In particular, correlative species distribution models (SDMs) are popular for predicting species' fates, and have been noted as effective tools for guiding conservation decisions. However, SDM predictions are uncertain due to our imperfect understanding of the processes underlying species-environment associations, and, crucially, imprecision in predictions of regional climate change. While this is widely recognised, SDM prediction uncertainty is frequently overlooked, and practical approaches to handling this uncertainty are rare. When SDMs are used to investigate questions of species' persistence during times of environmental change, failure to consider uncertainty about the arrangement and quality of habitat may lead to flawed inferences and ineffective management. It is therefore essential that we improve our understanding of key uncertainties, and develop methods that explicitly handle uncertainty in a way that promotes sensible management decisions. In this thesis, I explore these issues through case studies of the mountain pygmy-possum, Burramys parvus, in the alpine region of south-eastern Australia. I draw on a range of quantitative tools and classical decision theory to: (1) determine the magnitude of uncertainty about habitat suitability due to SDM predictor choice, and how this varies under climate change; (2) develop a framework for identifying the optimal spatial allocation of resources for species' conservation under climate change, given uncertain predictions of habitat suitability; (3) explore the utility of abundance time series for improving our understanding of environmental dynamics influencing populations; (4) combine SDMs and models of population dynamics with decision theory to assess the extent to which predictions are refined by explicitly including population processes; and (5) develop a suite of open source software tools that facilitate common ecological modelling tasks, making rigorous investigation of climate change questions more computationally efficient and feasible. I found that standard approaches to model evaluation obscure key differences amongst competing SDMs, suggesting that consideration of ecological relevance during model construction is essential. I showed that despite extensive uncertainty about future habitat, conservation actions can be prioritised in a way that reflects managers' appetites for risk and reward. I demonstrated that for spatially-structured populations, hierarchical models can reveal the spatial scales at which environmental processes control population growth. Regional synchrony in population dynamics is evident for B. parvus, but local, density-independent environmental forces are more important in determining abundance trajectories. Finally, I demonstrated that habitat change is an unreliable surrogate for a species' response to climate change. Predictions about the distribution and quality of future habitat for B. parvus are uncertain. However, this is an inevitable challenge when forecasting species' fates. Importantly, it does not preclude effective management. The way forward is to recognise and account for uncertainty in ecological models, thereby enabling sensible conservation decisions for species impacted by climate change.