Shifting Signals: Correlations Among Freshwater, Marine And Climatic Indices Often Investigated In Pacific Salmon Studies

The common practice of incorporating environmental indices into population models has greatly advanced our understanding of ecological systems. Unfortunately, we are increasingly seeing published correlations between population indicators and environmental indices breaking down when tested with new data. Examining how the correlations among indices change over time could help explain underlying causal mechanisms, which ultimately strengthen the basis for prediction of population indicators. For migratory animals such as anadromous salmon (Oncorhynchus spp.), the habitat conditions they experience can affect their lifetime fitness and population viability. We analyzed 43 freshwater, marine, and climate indices associated with 72 river sites and five coastal ecoregions inhabited by Chinook and coho salmon (O. tshawytscha and O. kisutch) in the western USA. Utilizing long time series (ranging from 32 to 124 years), we examined spatial and temporal patterns in correlations through hierarchical clustering across sites and non-stationarity across time. Individual river sites clustered into two Northwest and one Southwest groups. Northwest sites generally showed stronger correlations between freshwater and climate indices, while Southwest sites showed stronger correlations within freshwater or within marine/climate indices. For a closer examination at shorter periods, we parsed the time series into 10-year windows and showed how pairwise correlations changed over time with spring-summer Pacific Decadal Oscillation index in the Northwest and with spring flow in the Southwest. Stronger correlations across multiple indices tended to occur when large-scale climatic events (e.g., Oceanic Nino and Pacific Decadal Oscillation indices) were in-phase, and phase transitions (e.g., from positive to negative) occurred in the same 10-year window. In a third analysis, we assessed how well indices provided unique vs. confounding/complex information and had consistent vs. varying relationships based on the mean and variance of 10-year correlations. Across index types, the variance in correlations tended to be lowest in marine vs. climate indices, higher among freshwater indices, and highest for freshwater vs. marine/climate indices. Yet, the mean strength of correlations for freshwater vs. marine/climate indices was still comparable to those among freshwater ones. Overall, identifying time periods when correlations tend to change will help interpret historical and projected population indicators. Spatial trends in the strength of correlations also indicate that the level of confounding effects among indices can differ regionally. We emphasize the importance of knowing the strength and variability of correlations among indices, and their representativeness of ecological processes in the context of combined phases of multiple climatic indices.