Identifying Relationships Between Broad-Scale Habitat Characteristics and Listed Pacific Salmon Abundance:
Use and Misuse of Models in Recovery Planning

Dr. Ashley Steel
NW Fisheries Science Center Watershed Program,
National Marine Fisheries Service

Many clusters of Pacific salmon populations are currently listed as threatened or endangered under the Endangered Species Act. The NW Fisheries Science Center, NMFS, Seattle, is charged with providing scientific guidance and quantitative tools for developing sound recovery plans for these unique fish. I will present one of the models being developed to assist watershed planners in deciding what kinds of habitat to protect or restore first. Habitat quality and quantity are key variables influencing salmon population trends; yet, basic relationships between fish abundance and broad-scale habitat characteristics such as geology and land-use have not been established. Temporal variability in population indices makes habitat-productivity relationships difficult to detect in any one stream or in any one year. We have found that, despite annual fluctuations in escapement, certain areas within watersheds consistently produce the majority of spawners.

To quantify this relationship, we spatially linked redd counts to multiple layers of habitat data available at broad spatial scales (geology, mean annual air temperature, road density, land use and forest cover). We assessed statistical significance of observed patterns using mixed linear models. Mixed models allow us to estimate the correlation between redd counts taken over time at the same site, thus improving our estimation of model parameters. We have applied this technique to examine patterns in chinook salmon distribution in the Salmon River basin, ID, coho salmon distribution in the Snohomish River basin, WA, and winter steelhead distribution in the Willamette River basin, OR. In each basin, we identified a suite of models that associate habitat characteristics with the reaches that consistently support a large number of spawners. These models can be used to identify locations within the basin that might support unusually high numbers of fish, to prioritize field data collection, or to predict relative fish abundance in areas that are currently inaccessible due to man-made barriers. I will present the final models for each basin and discuss past and potential use and misuse of these models in recovery planning.