J. Daniel Weller, R.D. (Dan) Moore, Josephine C. Iacarella

Water temperature is a key feature of freshwater ecosystems but comprehensive datasets are severely lacking, a limiting factor in research and management of freshwater species and habitats. An existing statistical stream temperature model developed for British Columbia, Canada, was refit to predict August mean stream temperatures, a common index of stream thermal regime also used in thermalscapes developed for the western US. Thermalscapes of predicted August mean stream temperature were produced for 680,000 km of stream network at approximately 400 m intervals. Temperature predictions were averaged for 20-year periods from 1981-2100 to produce 86 scenarios: one for each historical period (i.e., 1981-2000, 2001-2020), and 21 for each future period (i.e., six global climate models and an ensemble average under three representative concentration pathways). The final model performance was consistent with other published regional-scale statistical models (R^2 = 0.79, RMSE = 1.53°C, MAE = 1.18°C), particularly given the relative paucity of data, large geographic extent, and range of climatic and physiographic conditions. Model results suggested an average increase of August mean stream temperature of 2.9 ± 1.0°C (RCP 4.5 ensemble mean ± SD) by end of century, with significant heterogeneity in predicted temperatures and warming rates across the province. Compared to stream temperature predictions from the western US, the predictions for BC showed good agreement at cross-border streams (Pearson’s r = 0.91), suggesting the possible integration of both products for a full western North America thermalscape. These stream thermalscapes for British Columbia address a major data deficiency in freshwater ecosystems and have potential applications to stream ecology, species distribution modelling, and evaluation of climate change impacts.