A Two-Stage Information-Theoretic Approach to Modeling Landscape-Level Attributes and Maximum Recruitment of Chinook Salmon in the Columbia River Basin

Many anadromous salmonid stocks in the Pacific Northwest are at their lowest recorded levels, which has raised questions regarding their long-term persistence under current conditions. There are a number of factors, such as freshwater spawning and rearing habitat, that could potentially influence their numbers. Therefore, we used the latest advances in information-theoretic methods in a twostage modeling process to investigate relationships between landscape-level habitat attributes and maximum recruitment of 25 index stocks of chinook salmon (Onocorhynchus tshawytscha) in the Columbia River basin. Our first-stage model selection results indicated that the Ricker-type, stock recruitment model with a constant Ricker a, i.e., recruitsper-spawner at low numbers of fish) across stocks was the only plausible one given these data, which contrasted with previous unpublished findings. Our second-stage results revealed that maximum recruitment of chinook salmon had a strongly negative relationship with percentage of surrounding subwatersheds categorized as predominantly containing U.S. Copyright c ©2002 Rocky Mountain Mathematics Consortium