Predicting streamflow variables at ungauged sites using spatially-referenced regression methods

The objective of our work is to classify watersheds based on hydrologic flow regime properties chosen for their relevance to biological assessment. Here, the flow regime is defined based on variables derived from streamflow at the gauged sites that quantify the following characteristics: a) magnitude b) frequency c) duration d) rate of change of flow and e) timing of the streamflow. The classification process involves prediction of the hydrologic variables at ungauged sites. Traditionally, the estimation of hydrologic variables at ungauged sites has used simple multivariate regression. Here we explore the use of spatially referenced regression for the estimation of flow regime variables at ungauged sites. Spatially Referenced Regression on Watershed Attributes (SPARROW) is a widely used model developed by US Geological Survey (USGS) that uses spatiallylinked regressions between its independent variables (watershed attributes) and water quality attributes (e.g. transport rate of phosphorous, nitrogen etc). The important aspect of this model is that the mass balance of each constituent is accounted for and the model can be adopted for any quantity which is conservative. We are motivated by the SPARROW methodology because of its use of the stream network information to maintain mass balance within its regression framework. Here we extend SPARROW to some of the streamflow regime variables we are using to classify watersheds for biological assessment. We will use Hydro Climatic Data Network (HCDN) streamflow gauges in the Pacific North West to develop a SPARROW like model to predict the flow regime variables from spatially referenced watershed attributes. The preliminary results from this work will be presented