Development of a HEC-RAS temperature model for the North Santiam River, northwestern Oregon

Abstract

A one-dimensional, unsteady streamflow and temperature model (HEC-RAS) of the North Santiam and Santiam Rivers was developed by the U.S. Geological Survey to be used in conjunction with previously developed two-dimensional hydrodynamic water-quality models (CE-QUAL-W2) of Detroit and Big Cliff Lakes upstream of the study area. In conjunction with the output from the previously developed models, the HEC-RAS model can simulate streamflows and temperatures within acceptable limits (mean error [bias] near zero; typical streamflow errors less than 5 percent; typical water temperature errors less than 1.0 °C) for the length of the North Santiam River downstream of Big Cliff Dam under a series of potential future conditions in which dam structures and/or dam operations are modified to improve temperature conditions for threatened and endangered fish. Although a two-dimensional (longitudinal, vertical) CE-QUAL-W2 model for the North Santiam and Santiam Rivers downstream of Big Cliff Dam exists, that model proved unstable under highly variable flow conditions. The one-dimensional HEC-RAS model documented in this report can better simulate cross-sectional-averaged stream temperatures under a wide range of flow conditions. The model was calibrated using 2011 streamflow and temperature data. Measured data were used as boundary conditions when possible, although several lateral inflows and their associated water temperatures, including the South Santiam River, were estimated using statistical models. Streamflow results showed high accuracy during low-flow periods, but predictions were biased low during large storm events when unmodeled ephemeral tributaries contributed to the actual streamflow. Temperature results showed low annual bias against measured data at two locations on the North Santiam River and one location on the Santiam River. Mean absolute errors using 2011 hourly data ranged from 0.4 to 0.7 °C. Model results were checked against 2012 data and showed a positive bias at the Santiam River station (+0.6 ˚C). Annual mean absolute errors using 2012 hourly data ranged from 0.4 to 0.8 °C. Much of the error in temperature predictions resulted from the model’s inability to accurately simulate the full range of diurnal fluctuations during the warmest months. Future iterations of the model could be improved by the collection and inclusion of additional streamflow and temperature data, especially near the mouth of the South Santiam River. Presently, the model is able to predict hourly and daily water temperatures under a wide variety of conditions with a typical error of 0.8 and 0.7 °C, respectively.