Habitat assessments were conducted in an intermountain watershed at three spatial extents to explore ways to predict the presence of tubificid oligochaetes likely to support the parasite Myxobolus cerebralis, cause of salmonid whirling disease. Stream reaches with six different reach slope characteristics were selected using GIS. The aquatic habitat in 60 reaches selected at random was measured and classified into distinct habitat units. Within the habitat units, areas of microhabitat with depositional fine sediments were chosen, measured, and core samples were removed to characterize the sediments and benthic oligochaetes. Two tubificids, Tubifex spp. and Limnodrilus hoffmeisteri, were abundant and co-occurred in silt-clay and fine sand sediments in these habitats. Models were posed and tested to predict the presence and relative abundance of tubificids using habitat characteristics from the three spatial extents: reach, habitat unit, and microhabitat. At the reach extent, tubificids were associated with low-reach slope and with slow water habitats. Within habitat units, tubificids were associated with higher percentages of fine sediments and higher stream width:depth ratios. In microhabitat cores, the presence of silt-clay sediments was positively associated with higher average stream width:depth ratios. Since ecological relationships are often scale dependent and stream systems have a natural hierarchy, predictive habitat models such as these that use measures from several scales may help researchers and managers more efficiently identify and quantify aquatic communities at highest risk of infection by the M. cerebralis parasite.

ABSTRACT1. The distribution and composition of in-stream habitats are reflections of landscape scale geomorphic and climatic controls. Correspondingly, Pacific salmon (Oncorhynchus spp.) are largely adapted to and constrained by the quality and complexity of those in-stream habitat conditions. The degree to which lands have been fragmented and managed can disrupt these patterns and affect overall habitat availability and quality.2. Eleven in-stream habitat features were modelled as a function of landscape composition. In total, 121 stream reaches within coastal catchments of Oregon were modelled. For each habitat feature, three linear regression models were applied in sequence; final models were composed of the immutable and management-influenced landscape predictors that best described the variability in stream habitat.3. Immutable landscape predictors considered proxies for stream power described the majority of the variability seen in stream habitat features. Management-influenced landscape predictors, describing the additional human impacts beyond that which was inherently entwined with the immutable predictors, explained a sizeable proportion of variability. The largest response was seen in wood volume and pool frequency.4. By using a sequential linear regression analysis, management-influenced factors could be segregated from natural gradients to identify those stream habitat features that may be more sensitive to land-use pressures. These results contribute to the progressing notion that the conservation of freshwater resources is best accomplished by investigating and managing stream systems from a landscape perspective.

Distribution of fishes, both occupancy and abundance, is often correlated with landscape-scale characteristics (e.g., geology, climate, and human disturbance). Understanding these relationships is essential for effective conservation of depressed populations. We used landscape characteristics to explain the distribution of coho salmon ( Oncorhynchus kisutch ) in the Oregon Plan data set, one of the first long-term, probabilistic salmon monitoring data sets covering the full range of potential habitats. First we compared data structure and model performance between the Oregon Plan data set and two published data sets on coho salmon distribution. Most of the variation in spawner abundance occurred between reaches but much also occurred between years, limiting potential model performance. Similar suites of landscape predictors are correlated with coho salmon distribution across regions and data sets. We then modeled coho salmon spawner distribution using the Oregon Plan data set and determined that landscape characteristics could not explain presence vs. absence of spawners but that the percentage of agriculture, winter temperature range, and the intrinsic potential of the stream could explain some variation in abundance (weighted average R2 = 0.30) where spawners were present. We conclude that the previous use of nonrandom monitoring data sets may have obscured understanding of species distribution, and we suggest minor modifications to large-scale monitoring programs.