Understanding Temporal and Spatial Trends of In-Stream Salmonid Habitat in the Interior Columbia River Basin

The 2004 and 2008 Biological Opinions (BiOps) on the Federal Columbia River Power System identified offsite mitigation actions, largely in the form of habitat restoration, as a means to offset mortality imposed by the FCRPS on anadromous salmonids. For the past seven years the Integrated Status and Effectiveness Monitoring Program (ISEMP) has collected in-stream habitat data in the Wenatchee subbasin of the Upper Columbia River Basin in Washington state as a means to meet BiOp needs for listed salmon.  As part of this effort, the ISEMP has focused on identifying a diverse suite of habitat metrics that are sensitive to change, robust to observer error, and associated to steelhead and Chinook salmon response metrics. In order to meet these objectives, ISEMP used a spatially balanced sampling design consisting of one annual panel of 25 sites and five rotating panels of 25 sites. Additionally, the sampling design was adapted in 2008 to address crew variability (3 separate visits to 25 sites in single summer). Habitat sampling at each site included geomorphic (e.g. channel unit metrics, slope and channel type), fish cover, undercut banks, riparian vegetation metrics, bank stability, and substrate measurements. Using these data, we explored statistical support for time trends and site classification schemes as factors explaining variability in both univariate and multivariate habitat data. We also assessed crew variability within these same model frameworks.  Our results suggest that site classification schemes such as Strahler order and geomorphic unit types were good descriptors of habitat variability while spatially related factors, such as watershed, were less effective. We also found strong support for time trends in both univariate and multivariate habitat responses despite the relatively short time-series of habitat data. Not surprisingly, crew variability was highly dependent on the habitat metric considered. Similar statistical and response designs are being used in the Lemhi (ID) and Entiat (WA) basins and results for variance partitioning for these basins are forthcoming.