Estimating stream temperature in the Willamette River Basin, northwestern Oregon—A regression-based approach

Abstract

The alteration of thermal regimes, including increased temperatures and shifts in seasonality, is a key challenge to the health and survival of federally protected cold-water salmonids in streams of the Willamette River basin in northwestern Oregon. To better support threatened fish species, the U.S. Army Corps of Engineers (USACE) and other water managers seek to improve the thermal regime in the Willamette River and key tributaries downstream of USACE dams by utilizing strategically timed flow releases from USACE dams. To inform flow management decisions, regression relations were developed for 12 Willamette River basin locations below USACE dams relating stream temperature with streamflow and air temperature utilizing publicly available datasets spanning 2000–18. The resulting relations provide simple tools to investigate stream temperature responses to changes in streamflow and climatic conditions in the Willamette River system. Regression relations on the Willamette River and key tributaries show that, at locations sufficiently distant from the direct temperature influence of upstream dam releases, air temperature and streamflow are reasonable proxies to predict the 7-day average of the daily mean (7dADMean) and 7-day average of the daily maximum (7dADMax) water temperature with errors generally ≤1 degrees Celsius (°C). To account for seasonal variations in the relation between air temperature, streamflow, and stream temperature, a transition-smoothed, seasonal regression approach was used. Stream temperature is inversely correlated with streamflow in all seasons except “winter” (January–March), when it is relatively independent. Stream temperature is positively correlated with air temperature in all seasons, but the slope decreases at very low or very high air temperatures. Generally, fit is best for seasonal models “winter” (January–March), “spring” (April–May), “summer” (June–August), and “early autumn” (September–October). Error in “autumn” (November–December) is larger, probably due to variation in the onset timing of winter storms. Simulated results from a climatological analysis of predicted stream temperature suggest that, excluding extremes and accounting for some seasonal variability, the 7dADMean and 7dADMax stream temperature sensitivity to air temperature and streamflow varies by location on the river. To investigate the potential range of stream temperature variability based on historical air temperature and streamflow conditions, stream temperature predictions were calculated using synthetic time series comprised of daily temperature values representing the 0.10, 0.33, 0.50, 0.67, and 0.90 quantile of air temperature and streamflow from 1954 (the year meaningful streamflow augmentation began) to 2018. Results show that from a “very hot” (0.90 quantile) and “very dry” (0.10 quantile) year to a “very cool” (0.10 quantile) and “very wet” (0.90; all quantiles from 1954 to 2018) year, the stream temperature sensitivity to air temperature and streamflow is about 3 °C at Harrisburg (river mile 161.0) and increases to about 5 °C at Keizer (river mile 82.2). While the number of days exceeding regulatory criteria are fewer in cooler, wetter years than in warmer, dryer years, the models suggest that the Willamette River will likely continue to exceed the State of Oregon maximum water-temperature criterion of 18 °C for sustained periods from late spring to early autumn and that the flow management practices evaluated in this study, while effective at influencing stream temperature, likely cannot prevent many or all such exceedances. As modeled for 2018, a representative very hot year with normal to below-normal streamflow, stream temperature sensitivity to changes in streamflow of ±100 to ±1000 cubic feet per second produced mean monthly temperature changes from 0.0 to 1.4 °C at Keizer, Albany, and Harrisburg during summer. For a specified change in flow, temperature sensitivity is greater at upstream locations where streamflow is less than that at downstream locations because the change in streamflow is a greater percentage of total streamflow at upstream locations. Similarly, temperature response to a set change in flow is greater in the summer and early autumn low-flow season than in spring when flows are higher. The regression models developed in this study thus indicate that flow management is likely to have a greater effect on stream temperature at upstream locations (such as Harrisburg or Albany) and during the low-flow season than at downstream locations (such as Keizer) or during periods of higher streamflow.