Considerations for the use of laboratory-based and field-based estimates of environmental tolerance in water management decisions for an endangered salmonid

Abstract

Water infrastructure development and operation can provide for human economic activity, health, and safety. However, this infrastructure can also impact native fish populations resulting in regulatory protections that can, in turn, alter operations. Conflict over water allocation for ecological function and human use has come to the forefront at Shasta Reservoir, the largest water storage facility in California, USA. Shasta Reservoir supports irrigation for a multibillion-dollar agricultural industry, provides water for urban and domestic use, provides flood protection for downstream communities, and power generation as part of the larger Central Valley Project in California. Additionally, an endangered run of Chinook Salmon relies on cold water management at the dam for successful spawning and egg incubation. Tradeoffs between these uses can be explored through application of models that assess biological outcomes associated with flow and temperature management scenarios. However, the utility of models for predicting outcomes of management decisions is contingent on the data used to construct them, and data collected to evaluate their predictions. We evaluated laboratory and field data currently available to parameterize temperature-egg survival models for winter run Chinook Salmon that are used to inform Shasta Dam operations. Models based on both laboratory and field data types had poor predictive performance which appears to limit their value for management decisions. The sources of uncertainty that led to poor performance were different for each data type (field or laboratory) but were rooted in the fact that neither data set was collected with the intention to be used in a predictive model. Our findings suggest that if a predictive model is desired to evaluate operational tradeoffs, data should be collected for the specific variables and life stages desired, over an appropriate range of values, and at sufficient frequency to achieve the needed level of precision to address the modeling objective.