Recovery of anadromous salmonid populations is complicated by the fact that these fish have complex life-histories. Habitat valuation and capacity methods need to account for spatiotemporal variability in temperature, geomorphic features, and a species’ thermal sensitivity mediated by biological interactions. We examined this interplay in a case study of steelhead trout (Oncorhynchus mykiss) and Chinook Salmon (O. tshawytscha) in California’s Eel River watershed. We estimated habitat suitability and fish capacity for each salmonid run and freshwater life stage during average, cool, and warm years in each of the watershed’s subbasins, including a high-elevation subbasin upstream of an impassable dam. Our estimates varied depending on whether we accounted for exposure to the Sacramento pikeminnow (Ptychocheilus grandis), an introduced predator/competitor. Our results indicate that the dammed subbasin could provide an important cool-water refuge during warm years and from pikeminnow, potentially improving the productivity and resilience of multiple anadromous salmonid populations. Our approach can be applied in any setting where spatially explicit habitat metrics can be estimated and population specific and life-stage specific habitat criteria can be specified.

Summary

The attached shapefile contains information used to assess salmonid habitat suitability and fish capacity for over 10,000 stream segments in the Eel River watershed, California. Specifically, the shapefile contains spatial and/or temporal data for stream temperature, geomorphic channel type, salmonid accessibility and potential occupancy, and stream segment characteristics needed to estimate capacity (e.g. stream segment length and wetted width. Each stream segment is ~1 km in length. Below, we summarize each dataset briefly. For full details, refer to our full paper.

Stream temperature

Mean monthly stream temperature for each reach was obtained from FitzGerald et al. (2021). Briefly, FitzGerald et al. (2021) predicted stream temperature using a spatial stream network (SSN) model, a specialized statistical regression model that accounts for spatial autocorrelation in temperatures due to stream-network structure and geographic proximity Additional modeling details can be found in FitzGerald et al. (2021). The model was then used to predict monthly mean stream temperatures for every river km in the Eel River Basin for every year in the study time period (attribute column "Month_Year"). Reaches classified as manmade lakes and reservoirs were removed because they involve different thermal dynamics that are not well-represented by the SSN model. The temperature predictions and habitat suitability analyses therefore do not include reaches that are currently inundated (e.g. by Lake Pillsbury, created by Scott Dam). A single river km in the Eel River Basin showed abnormally high predictions (sometimes > 10°C higher than the next highest stream temperature in the Basin), and this outlier was removed from all subsequent analyses.

Geomorphic channel types

We classified each reach by geomorphic channel type. To do this, we generated a fine-grained hydrography with channel gradients and catchment areas from a 10 m DEM. We then spatially joined the finer-grained hydrography to the stream network that was used in the temperature modeling, summarizing the mean gradient and catchment area for each 1 km reach. Then, channel morphology types were assigned using channel gradient and catchment area from a classification tree ("ChanType"). 

Potentially accessible streams for salmonids

We defined the potential spatial distribution for each run/life stage from historical population boundaries, accessibility of reaches, and channel type. First, historical population boundaries were defined from a study on salmonid biogeographic breaks that showed that steelhead trout and Chinook Salmon in the Eel River Basin are divided into historical populations that generally reflect watershed subbasins ("Pop_SW", "Pop_SS", and "Pop_CF"). We did not analyze any subbasins that were historically uninhabited for a given run. Next, we removed reaches beyond the limits of anadromy for each species ("STL_access" and "CHK_access"). Anadromous limits were defined as upstream of physical impassable barriers (e.g. large waterfalls) or upstream of species-specific barriers inferred from stream gradient; we excluded the currently impassable Scott Dam. The remaining reaches, including those in the currently blocked Upper Mainstem, are hereafter referred to as ‘potentially accessible’ (e.g., "Pop_SW" == 'Upper Mainstem' & "STL_access" == 1).

Estimating capacity

To estimate juvenile rearing capacity, we expanded the Unit Characteristic Method (UCM), applied by Cooper et al. (2020) to the Upper Mainstem, to all subbasins in the Eel River Basin. The UCM is a capacity model that multiplies baseline fish density by unit area of stream habitat, then adjusts the density by habitat scalar values based on parameters describing local conditions (e.g. cover, depth, pH) for each habitat unit type, such as fastwater, flatwater, and pools. In the Eel River Basin, empirical measures of local conditions (excluding stream segment length, stream temperature, and wetted width; see below) were only available for reaches throughout the Upper Mainstem (Cooper et al. 2020). Cooper et al. (2020) categorized each reach surveyed by channel gradient and upstream watershed area and measured habitat characteristics for each reach to estimate the appropriate scalar for local conditions in the Upper Mainstem. Following their approach, we first applied the same reach categorization scheme throughout the Eel River Basin for each stream segment (attribute "ReachCateg"), and, assuming that local conditions in the Upper Mainstem are representative of the entire Eel River Basin, we then assigned the averaged habitat values by reach category (Cooper et al. 2020) to the appropriate stream segment. Stream segment length ("length_m") and monthly temperature ("Month_Year") were extracted from our stream temperature modeling.

The absolute capacity of a reach is given by the product of its capacity density and the reach area, the product of average wetted width and channel length of stream segment. We modeled wetted width each month in order to better predict how reach area changes throughout the year. We then fit linear models for each month from observed wetted widths and bankfull widths. Monthly models generally performed well (r² range: 0.61-0.84) and better than an annual model (r² = 0.60), so we used the fitted monthly models to predict wetted width each month throughout the Basin ("width_Month"). 

Please refer to the README file