Characterizing forest harvest mosaics and quantifying their effects on stream temperatures

Forest harvest can warm streams, but quantifying the relationship between forest harvest and stream temperature at the watershed level is difficult because forest harvest is spatiotemporally variable and controls can override or confound these relationships. I used a spatial stream network modeling framework to describe the dominant controls of summer stream temperature in the North Thompson River watershed, and test whether spatiotemporally explicit forest harvest metrics, and their interactions, better characterize relationships between forest harvest and stream temperature compared to conventional metrics. I created a base model for describing summer stream temperature by comparing 18 candidate predictors known to control stream temperatures across three summer months and four response metrics. The final 12 base models included 10 predictors that consistently explained more variation in temperature and improved fit. The base model described the majority of variation in summer stream temperatures for all 12 response metrics. I compared model performance and explained variance for models that included one of six forest harvest metrics, spanning three spatial representations and two temporal summaries, with or without an interaction. Each forest harvest model included a main or interactive effect of forest harvest and the 10 base model covariates. Conventional forest harvest metrics largely failed to explain more variation in stream temperatures. Forest harvest metrics that accounted for flow-distance, harvest recovery, and elevation, largely explained more variation in summer stream temperatures and improved model fit. Models with interactions between forest harvest and elevation² performed the best, improving fit for 7 to 12 of 12 responses depending on harvest metric. For the best performing response, July maximum temperature, warming peaked at ~1,285 m with an estimated increase of 1.0 °C per standard deviation of cumulative upstream catchment harvest (9.6%). Taken together, these results demonstrate that the spatiotemporal layout of harvest and its location across an elevation gradient can greatly influence the relationship between harvest and stream temperature. Importantly, these findings show large positive associations between forest harvest and stream temperature at mid-to-high elevations, where current regulations may have the least riparian protections and where climate change is projected to impact streams differentially.