Search

The logistic regression models that we previously used to predict where fall Chinook salmon Oncorhynchus tshawytscha would spawn in the Hanford Reach of the Columbia River were based on so-called static variables (i.e., the riverbed surface substrate, riverbed slope, and time-averaged velocity and depth [representing the velocity and depth in each cell associated with the 50% exceedance flow]). Not all habitat predicted to be used for spawning contained redds, and "one explanation for the overprediction is that the models did not incorporate streamflow fluctuation. Streamflow fluctuation occurs daily in the Hanford Reach owing to load-following operations (power generation to meet short-term electrical demand) at Priest Rapids Dam, a hydroelectric dam located at the upper end of the reach. Daily flow fluctuations could change the hydraulic characteristics to which fall Chinook salmon respond in selecting redd sites. The purpose of this study was to determine whether incorporating metrics of flow variability would improve modeling of spawning habitat availability. Flow variability was represented by so-called dynamic variables (i.e., the standard deviations of velocity and depth for each habitat cell over a 60-d time segment during the fall spawning period in 1994, 1995, and 2001). Both the static and dynamic models were correct at least 85% of the time in predicting habitat as spawning or nonspawning. However, incorporation of the dynamic variables into the logistic regression models reduced the amount of overpredicted habitat by 42% in 1994, 32% in 1995, and 25% in 2001. For example, the area predicted to be suitable but not used in 1994 decreased from 20.8 ha for the static model to 12.1 ha for the dynamic model for all sites combined; similar results were found for 1995 and 2001. This represents an improvement in our ability to accurately predict suitable fall Chinook salmon spawning habitat in the Hanford Reach of the Columbia River. [ABSTRACT FROM AUTHOR]