Analyses of potential factors affecting survival of juvenile salmonids volitionally passing through turbines at McNary and John Day Dams, Columbia River

Abstract

This report describes analyses of data from radio- or acoustic-tagged juvenile salmonids passing through hydro-dam turbines to determine factors affecting fish survival. The data were collected during a series of studies designed to estimate passage and survival probabilities at McNary (2002-09) and John Day (2002-03) Dams on the Columbia River during controlled experiments of structures or operations at spillways. Relatively few tagged fish passed turbines in any single study, but sample sizes generally were adequate for our analyses when data were combined from studies using common methods over a series of years. We used information-theoretic methods to evaluate biological, operational, and group covariates by creating models fitting linear (all covariates) or curvilinear (operational covariates only) functions to the data. Biological covariates included tag burden, weight, and water temperature; operational covariates included spill percentage, total discharge, hydraulic head, and turbine unit discharge; and group covariates included year, treatment, and photoperiod. Several interactions between the variables also were considered. Support of covariates by the data was assessed by comparing the Akaike Information Criterion of competing models. The analyses were conducted because there was a lack of information about factors affecting survival of fish passing turbines volitionally and the data were available from past studies. The depth of acclimation, tag size relative to fish size (tag burden), turbine unit discharge, and area of entry into the turbine intake have been shown to affect turbine passage survival of juvenile salmonids in other studies. This study indicates that turbine passage survival of the study fish was primarily affected by biological covariates rather than operational covariates. A negative effect of tag burden was strongly supported in data from yearling Chinook salmon at John Day and McNary dams, but not for subyearling Chinook salmon or juvenile steelhead. The negative effect of tag burden in data we examined from yearling Chinook salmon supports the recent findings from laboratory studies of barotrauma effects. A curvilinear (quadratic) effect of turbine unit discharge was weakly supported in data from subyearling Chinook salmon at John Day Dam. The maximum survival from those data was estimated to occur at a discharge of 15.9 thousand cubic feet per second, but the estimate was imprecise (95 percent confidence interval of -1.7-33.7 thousand cubic feet per second). This estimate is within the range of 1 percent of peak turbine operating efficiency (12.0-21.6 thousand cubic feet per second), but is lower than the 17.2 thousand cubic feet per second discharge at peak operating efficiency (at a head of 102 feet near the median in the data we examined). Effects of water temperature were supported in four of the five examined data sets and were strongly supported in all but one. Spill percentage, head, and total discharge received weak or moderate support in some cases. The results are consistent with those of several controlled field experiments of turbine discharge. Studies based on the Hi-Z Turb'N tag (balloon tag) often show small, generally statistically insignificant, differences in survival at different turbine discharge levels. Some studies also show that a quadratic equation can be well fit to the relation of survival and turbine unit discharge. The lack of support for the operational covariates in most of the data sets we examined may be due to the small effect turbine discharge has even in controlled studies, the observational nature of the data we used, and the evaluation method. We assessed support of the data for models of linear and quadratic effects, whereas controlled experiments often statistically compare the point estimates of survival from each operational treatment studied. The results of our analyses suggest tag burden should be minimized or controlled for in analyses of future stu