Updating Spatial Stream Network Models of August Stream Temperature for the Washington Coast Salmon Recovery Region

Background - In 2017, Isaak et al. (2017) published the NorWeST model of mean August stream temperatures for the Pacific Northwest that provided aquatic resource managers, researchers, conservation organizations, and other stakeholders a valuable and easily accessible geospatial tool for conservation planning. Despite the many contributions of this model, an analysis of local conditions in the Chehalis River Basin identified that the NorWeST model predictions were on average 2.7°C lower than observed current stream temperatures in this basin. A similar analysis of regional conditions from the Washington Coast Salmon Recovery Region followed a similar trend where 83% of NorWeST predictions were cooler than observations and underpredictions were on average 2.1°C and up to 5.6°C cooler than observations. These discrepancies coupled with increased availability of stream temperature data in Western Washington since NorWeST was published warrants the development of an updated regional model, informed with additional data, which should improve understanding of summer stream temperature conditions within the Washington Coast Salmon Recovery Region.

Purpose of the study –The overall goal of this project is to develop an updated spatial stream network model for mean August stream temperature for rivers of the Washington Coast Salmon Recovery Region, specifically Water Resource Inventory Areas (WRIAs) 20-24. The resulting Washington Coast Thermalscape Model includes data sets from 564 stream temperature monitoring locations and downscaled climate projections from 5th Climate Model Intercomparison Program (CMIP5), thus improving the spatial coverage and updating the climate change projections available when the NorWeST model was published.

Methodological approach – Stream temperature data were collated from multiple organizations collecting data in our study area. Data were examined for outliers and organized into a database for modeling. The modeling steps (e.g., geospatial preprocessing, covariate selection and evaluation, and model fitting and evaluation) essentially followed the approach of Isaak et al. (2017) We evaluated models for subregions in addition to the region in its entirety (e.g., North Coast [WRIAs 20 and 21], Chehalis Basin [WRIAs 22 and 23], and Willapa Bay [WRIA 24], and full Washington Coast [WRIAs 20- 24]). A model selection approach was followed to select the best fitting models at each spatial scale and then examined spatial and temporal factors associated with summer stream temperature. Model performance was then compared among all models and the best performing model was selected to predict August stream temperatures for the region under current and future climate scenarios. Model structure was similar to that previously developed for the NorWeST model of summer stream temperatures in the Pacific Northwest.

Results – Fit of the Thermalscape Model was evaluated for sub-regions as well as for the region in its entirety. The final regional spatial stream network model showed high predictive performance in terms of precision (root mean square predicted error = 0.6°C, r2 = 0.95) and accuracy (mean absolute prediction error = 0.01°C) and was selected to develop Thermalscapes of current and future stream temperature conditions for the Washington Coast Salmon Recovery Region. Significant warming effects on mean August stream temperatures (p<0.05) were associated with August air temperature, cumulative drainage area, and lake % and significant cooling effects were associated with stream flow, elevation, riparian canopy cover, and mean annual precipitation. Current maximum, minimum and mean August stream temperature across the Washington Coast Salmon Recovery Region estimated at 25.0, 1.8, and 14.9°C, respectively. Late century predictions (2080s) of maximum, minimum, and mean August stream temperature across the Washington Coast were 25.8, 2.4, and 15.6°C, respectively.

Conclusions – Model results are a geospatial resource available for many applications of riverine research and fish and habitat management on the Washington Coast. The results also highlight areas where uncertainty in stream temperature projections is high, and this information may inform priorities for future temperature monitoring activity.were