Data from: From a line in the sand to a landscape of decisions: a Hierarchical Diversity Decision Framework (HiDDeF) for estimating and communicating biodiversity loss along anthropogenic gradients
Voss, Kristofor A.; King, Ryan S.; Bernhardt, Emily S. (2016), Data from: From a line in the sand to a landscape of decisions: a Hierarchical Diversity Decision Framework (HiDDeF) for estimating and communicating biodiversity loss along anthropogenic gradients, Dryad, Dataset, https://doi.org/10.5061/dryad.m1787
1. In setting water quality criteria, managers must choose thresholds for stressors that are protective of aquatic biodiversity. Setting such thresholds requires making implicit judgments about the degree of biodiversity loss that managers are willing to accept. 2. We present a new modeling approach, the Hierarchical Diversity Decision Framework model (HiDDeF) that explicitly communicates the sensitivity of water quality benchmarks to these implicit judgements. We apply HiDDeF to a dataset of stream macroinvertebrate abundances across 218 sites in southwestern West Virginia, USA where alkaline mine drainage increases streamwater conductivity and leads to the loss of sensitive taxa throughout regional river networks. 3. By integrating responses of individual taxa within a flexible hierarchical framework, HiDDeF reliably predicts macroinvertebrate assemblages across the full range of conductivities observed in the training dataset but requires only a fraction of the sites required in previous studies. HiDDeF results suggest that the current conductivity benchmark (300 μS/cm) for regional streams translates to 50% loss in abundance for at least one-quarter of regional macroinvertebrate taxa. 4. HiDDeF produces a “decision landscape” that allows decision makers to assess sensitivity of proposed benchmarks to their choice of protective level. HiDDeF allows users to investigate both individual and community level responses to environmental gradients and generates output that includes a comprehensive summary of uncertainty in model parameters.