A Framework for Developing Flow-Ecology Relationships in Washington State

Environmental classification is a first step towards quantifying the ecological tradeoffs of flow regulation by creating the framework necessary for analyzing effects of flow variability on riverine biota.  Our study presents a spatially explicit hydrogeomorphic classification of streams and rivers in Washington State, U.S.A., and investigates how projected climate change is likely to affect flow regimes in the future.  We calculated 99 hydrologic metrics from 15 years of continuous daily discharge data for 64 gages with negligible upstream impact, which were entered into a Bayesian mixture model to classify flow regimes into seven major classes described by their dominant flow source: groundwater (GW), rainfall (RF), rain-with-snow (RS), snow-and-rain (S&R), snow-with-rain (SR), snowmelt (SM), and ultra-snowmelt (US).  The largest class sizes were represented by the transitional RS and S&R classes (14 and 12 gages, respectively), which are ubiquitous in temperate, mountainous landscapes found in Washington.  We used a recursive partitioning algorithm and Random Forests to predict flow class based on a suite of environmental and climate variables.  Overall classification success was 75%, and the model was used to predict normative flow classes at the reach-scale for the entire state.  Application of future climate change scenarios to the model inputs indicated shifts of varying magnitude from snow- to rain-dominated flow classes.  Lastly, a geomorphic classification was developed using a Digital Elevation Model and climatic data to assign stream segments as either dominantly able or unable to migrate, which was cross tabulated with the flow types to produce a 14-tier hydrogeomorphic classification.  The hydrogeomorphic classification provides a framework upon which empirical flow alteration-ecological response relationships and associated environmental flow prescriptions can be developed using ecological information collected throughout the region.