Modeling Stream Temperature and Flow from Gridded Climate Datasets in Alaska's Yukon and Kuskokwim Basins

Authors: Shaftel, Rebecca S, Alaska Center for Conservation Science, University of Alaska Anchorage; Feddern, Megan L., College of Fisheries and Ocean Sciences, University of Alaska Fairbanks; Schoen, Erik, International Arctic Research Center, University of Alaska Fairbanks; Cunningham, Curry, College of Fisheries and Ocean Sciences, University of Alaska Fairbanks; von Biela, Vanessa R., U.S. Geological Survey, Alaska Science Center; McAfee, Stephanie, Department of Geography, University of Nevada Reno; Falke, Jeffrey A., U.S. Geological Survey, Alaska Cooperative Fish and Wildlife Research Unit

Video Presentation

Abstract

Stream temperature and streamflow are critical controls on freshwater habitat dynamics and are important for understanding climate impacts on freshwater resources. In Alaska, using empirical stream temperature and streamflow datasets for research poses several challenges: data are often unavailable for an area of interest, datasets are typically of short durations, and sites are managed independently across agencies and organizations. Fortunately, advances in gridded climate products and downscaled climate projections provide alternatives for quantifying freshwater habitat conditions in remote regions like Alaska. For this project, we reviewed products and validated models against in situ data to develop more complete historical time series of stream temperature and streamflow. Specifically, our objectives included: 1) developing a list of gridded or modeled products available for Alaska, 2) comparing products with a focus on streamflow and temperature, and 3) comparing three different stream temperature models. We validated alternative products representing streamflow and temperature using empirical datasets associated with a case study of Chinook Salmon habitat in the Yukon and Kuskokwim basins. Our results indicated that a global modeled streamflow product had a strong positive correlation to observed streamflow (mean r = 0.79 for 11 sites). Boosted regression tree models that included daily gridded air temperatures along with other covariates had the highest positive correlation to observed stream temperatures (r = 0.97 for 31 sites) compared to other commonly used models and good prediction accuracy (mean RMSE = 0.7 degC). Overall, we found several products that could be used to develop accurate time series of freshwater habitat conditions in Alaska and utilized for fisheries research. Potential applications include predicting aquatic species distributions, spread of invasive species, food availability, fish growth potential, and generally informing fish responses to climate change for research and management.

Citation

Please use the following citation when citing this presentation: