Reconstruction of Pacific salmon abundance from riparian tree-ring growth

We use relationships between modern Pacific salmon (Oncorhynchus spp.) escapement (migrating adults counted at weirs or dams) and riparian tree-ring growth to reconstruct the abundance of stream-spawning salmon over 150-350 years. After examining nine sites, we produced reconstructions for five mid-order rivers and four salmon species over a large geographic range in the Pacific Northwest: chinook (O. tschwatcha) in the Umpqua River, Oregon, USA; sockeye (O. nerka) in Drinkwater Creek, British Columbia, Canada; pink (O. gorbuscha) in Sashin Creek, southeastern Alaska, USA; chum (O. keta) in Disappearance Creek, southeastern Alaska, USA; and pink and chum in the Kadashan River, southeastern Alaska, USA. We first derived stand-level, non-climatic growth chronologies from riparian trees using standard dendroecology methods and differencing. When the chronologies were compared to 18-55 years of adult salmon escapement we detected positive, significant correlations at five of the nine sites. Regression models relating escapement to tree-ring growth at the five sites were applied to the differenced chronologies to reconstruct salmon abundance. Each reconstruction contains unique patterns characteristic of the site and salmon species. Reconstructions were validated by comparison to local histories (e.g., construction of dams and salmon canneries) and regional fisheries data such as salmon landings and aerial surveys and the Pacific Decadal Oscillation climate index. The reconstructions capture lower-frequency cycles better than extremes and are most useful for determination and comparison of relative abundance, cycles, and the effects of interventions. Reconstructions show lower population cycle maxima in both Umpqua River chinook and Sashin Creek pink salmon in recent decades. The Drinkwater Creek reconstruction suggests that sockeye abundance since the mid-1990s has been 15-25% higher than at any time since 1850, while no long-term deviations from natural cycles are detected for salmon in the Kadashan River or in Disappearance Creek. Decadal-scale cycles in salmon abundance with periods of 25-68 years were detected in all of the reconstructions. This novel approach provides river-specific, long-term perspectives on salmon abundance and cycles. Additionally, it provides a new frame of reference for maintaining and rebuilding individual stocks and for striking a balance between societal demands and the limited, always-changing salmon resource.