Integrating Downscaled Climate Predictions for Native Salmonid Conservation: Lessons from Probabilistic Models

The need to assimilate and integrate climate change projections into species vulnerability assessments and conservation planning has become a pressing concern for biologists and managers.  The rapidly increasing volume of climate data can make this a daunting task, and highlights the need for an organizing framework that formally considers potential climate impacts along with existing stressors.  Graphical probabilistic models, such as Bayesian belief networks (BBNs) are a useful evaluation and decision support framework in this regard, because they are relatively easy to build and implement, can integrate different types of information (e.g., opinion, empirical, output from other models), can incorporate relevant ecological and physical processes, and can explicitly represent uncertainty.   Using bull trout (Salvelinus confluentus) and inland cutthroat trout (Oncorhynchus clarkii) as species-specific examples, we show how BBNs can be developed and implemented to evaluate vulnerability at different spatial scales using downscaled climate projections for temperature and streamflow.   We find that this process can lead to counterintuitive predictions, even in cases where the target species is strongly influenced by a single factor such as water temperature.  Thus, what may initially be perceived as a straightforward response to anticipated climate change (i.e., stream warming) must be placed in a proper abiotic and biological context to provide robust and transparent assessments, and to identify where our limited understanding of pattern or process needs to be augmented by additional research.