Decision Trees for Incorporating Hypothesis Tests of Hydrologic Alteration into Hydropower-Ecosystem Tradeoffs

Short streamflow records make it difficult to determine the extent to which discharge changes in excess of ecological thresholds are due to dam operations or natural variability. Unnecessary changes to reservoir operating rules can reduce off-stream benefits, whereas no changes to rules when thresholds are exceeded can degrade downstream riverine ecosystems. We introduce a Bayesian decision tree approach to a hypothetical hydropower-ecosystem decision problem that compares expected in-stream and off-stream losses resulting from incorrect decisions. Expected losses are computed using loss probabilities derived using Bayes' theorem, type I and II errors, and prior probabilities of alteration. Decision-tree recommendations compared with those from deterministic and null hypothesis significance testing under a variety of conditions illuminate the benefits of including valuations of hydropower and ecological losses as well as type II error probabilities in reservoir operation decisions. This is the first study to both introduce and demonstrate the value of Bayesian decision trees for addressing tradeoffs between hydropower and ecosystem benefits and losses.