Wednesday, August 5, 2009 - 3:20 PM

COS 72-6: Using heirarchical regression to explain juvenile density in terms of number of spawners and habitat metrics in Oregon's coho salmon (Oncorhynchus kisutch)

Yasmin Lucero, Northwest Fisheries Science Center

Background/Question/Methods

Pacific salmon (genus Oncorhynchus) face an uncertain future throughout their range, with several sub-populations listed under the Endangered Species Act. Because salmon rear in freshwater, their fate is closely tied to the status of the watershed. Like many Pacific salmon, Oregon's coho (O. kisutch) populations are threatened by altered flow regimes (due to hydropower and water diversions for agriculture), harmful sediment loads (due to land use practices such as forestry, grazing and agriculture), and rising mean stream temperatures (due to a warming climate and loss of riparian vegetation). Many of these habitat changes can be mitigated by restoration and management of water and land use. However, such actions are difficult and costly, and to be cost-effective managers require strong quantitative tools for predicting the population consequences of altering habitat. Towards the end of developing such tools, the Oregon Department of Fish and Wildlife has collected---and made publicly available---datasets for sites throughout Oregon that include juvenile coho density, number of spawners and habitat characteristics over a ten year period.
Results/Conclusions

I present a two part analysis of the Oregon dataset for coho salmon. I begin with an ordination analysis to reduce the dimensionality of the habitat description. The dataset includes upwards of 30 habitat variates, several of which are closely related, e.g. the number of deep pools versus the total number of pools at a site. I present the principle components and loadings that emerge from the ordination analysis. I illustrate the limits to and advantages of using ordination components to remove redundancy in the data in lieu of the more commonplace strategy of removing variates that appear co-linear. I then show a heirarchical regression analysis to explain juvenile density in terms of habitat metrics and spawner density, grouping by subpopulations and year. This analysis yields a measurement of the adequacy of currently collected habitat metrics to explain variation in juvenile density. I also show how this juvenile/habitat relationship varies across subgroups and years.