Boosted regression tree analysis of juvenile salmonids and their freshwater habitat in the Pacific Northwest

Background/Question/Methods

Identifying crucial freshwater juvenile habitat is a key component in the restoration plans for several endangered salmonid species in the Pacific Northwest. In 2011, the Columbia Habitat Monitoring Program (CHaMP) was implemented in several basins to measure a wide variety of habitat characteristics. At the same time, samples of juvenile Chinook (Oncorhynchus tshawytscha) and steelhead (Oncorhynchus mykiss) were taken at many of the same sites to estimate fish densities. This goal of this study was to determine how to best predict fish densities based on about forty habitat metrics. We used boosted regression trees (BRT) in our analysis because of their ability to capture the potential non-linear relationships between habitat metrics and fish densities, as well as account for unknown interactions between habitat metrics. The results of the BRT analysis were then used to predict fish densities at all locations where habitat measurements were taken, but fish sampling was not conducted.

Results/Conclusions

Habitat requirements differed between species, with Chinook densities being best predicted by metrics related to flow and pools, while steelhead densities depended more on the particular basin than anything else. One of the most interesting results from the BRT analysis is the identification of habitat thresholds, specific points along a habitat gradient where fish densities suddenly shift. Resource managers can use such thresholds to define restoration goals, and identify factors that may be limiting salmonid abundances.