• Salmonid distribution models to support restoration planning across the fragmented Chehalis River basin, WA

      Walther, Eric J.; Westley, Peter; Zimmerman, Mara; Falke, Jeffrey; Seitz, Andrew (2021-05)
      Understanding the factors that influence the distribution of species through time and across space is a fundamental goal of ecology and crucial information needed to effectively manage and recover populations. Anthropogenic fragmentation of habitat disrupts ecological processes and is an on-going threat to species persistence across taxa. River ecosystems are particularly vulnerable to disruptions in connectivity and are the focus of extensive restoration efforts and financial investment. For example, over $300 million/year is invested towards restoration in the Columbia River basin. However, restoration is often impeded by knowledge gaps in distribution that can result in omitting locations that would benefit from restoration. For mobile species within dendritic freshwater networks, the boundary that demarcates the total quantity of available habitat can be defined by the upper limit of occurrence (ULO) and is a useful metric for assessing the extent of habitat to consider for restoration. The first goal of this work was to identify the ULO boundary for three socially and ecologically important anadromous fishes (Oncorhynchus spp.) in a subset of representative streams across a complex river network in southwestern Washington State, USA, and quantify the relationship of the ULO with landscape attributes for these species. Extensive field surveys covering 669 river km across two years documented the ULO in 115 terminal streams (i.e., uppermost independent stream segment within a stream network) for coho salmon (O. kisutch), 97 terminal streams for steelhead trout (O. mykiss), and 57 terminal streams for chum salmon (O. keta). The landscape attributes associated with these ULO locations varied among species. For example, precipitation was an important predictor only for coho salmon, whereas slope was an important predictor only for steelhead trout. In contrast, drainage area, elevation, and geology were important predictors for all species; while the direction was the same for drainage area and elevation, the magnitude of the effect of each landscape attribute varied among species. I demonstrated that large-scale landscape attributes can accurately and consistently detect species-specific distribution boundaries across broad and diverse habitat (percent correct classification:78%-89%; area under the receiver operating characteristic curve: 0.87-0.96). The ability to quantify landscape attributes related to distribution boundaries illuminates how the biology and life history of a species is captured across the landscape. The second goal of this work was to predict the range of occurrence as a function of landscape attributes for coho salmon, steelhead trout, and chum salmon across a range of probability decision thresholds, that reflect different risk-tolerance scenarios and determine whether stream reaches are within or outside the range of occurrence. Generalized linear mixed models were used to compare the quantity of currently described distribution used in restoration planning in the basin and quantify the amount of habitat inaccessible due to anthropogenic barriers. The change in amount of habitat within the predicted range of occurrence across probability decision thresholds ranged from 60%-74% among species. Differences between the model predictions and the currently described distribution for each species ranged from -14% to 171%, which on a whole indicates that the amount of habitat being considered for restoration is currently underestimated. As predicted, species with a greater range of occurrence (e.g., coho salmon) had a greater percentage of predicted suitable habitat inaccessible due to anthropogenic barriers (coho salmon:17.4%-28.8%, 0.75-0.25 PDT; steelhead trout:10.2%-17.5%; chum salmon: 3.9%-12.3%), and the locations of these barriers varied among species. Modelling species distributions at multiple levels of risk-tolerance allows practitioners to weigh the ecological benefits and financial investment when considering locations for restoration. Ultimately, the effective consideration of restoration actions requires tools such that managers can weigh the trade-offs of their decisions given that not all actions equitably benefit all species.