Evaluation of Bayesian Networks for Predicting Spawning Habitat Quality of Chinook Salmon in Data-Poor Watersheds. (2nd January 2017)

• Record Type: Journal Article
• Authors: Brumbaugh, Steven M.; Coleman, Ronald
• Is Part Of: North American journal of fisheries management. Volume 37:Number 1(2017)
• Page Start: 176
• Page End: 188
• Publication Date: 2017-01-02
• Subjects: Fishery management -- United States -- Periodicals; 333.956097305
• DOI: 10.1080/02755947.2016.1254125
• Languages: English
• ISSNs: 0275-5947

Abstract

Abstract: California's native salmonid populations are in decline, as evidenced by the 2008 fishing closures on one historically abundant species, Chinook Salmon Oncorhynchus tshawytscha . One major impact on spring-run Chinook Salmon within the Central Valley has been the modification of natal rivers. Bayesian networks are one modeling method that could help researchers to understand these systems and direct restoration efforts. We constructed a Bayesian network for Deer Creek, in Tehama County, to assess its utility as a tool for guiding restoration of spawning habitats for spring-run Chinook Salmon. We applied this network on a riffle–pool subreach scale to determine the suitability of each reach for spawning, indicated by the probability of redd presence. To evaluate the network we conducted sensitivity analyses and thereby determined the influence of each variable and the degree to which each variable informed the probability of redd presence. Sensitivity analyses were run for networks trained with two different stream alignments, one derived from the National Hydrography Dataset from the U.S. Geological Survey and one derived from tracing aerial imagery. We also conducted a Mann–Whitney test comparing redd densities from subreaches predicted to be good with those predicted to be poor for four fish passage condition scenarios. Of the four scenarios we modeled with the network, three exhibited significantly higher redd densities in subreaches designated as good by the network. Our results indicate that Bayesian networks can be used to predict habitat use and prioritize restoration in a data-poor northern California watershed. Received November 30, 2015; accepted October 20, 2016