A Regional Stream Temperature Model for Assessing Climate Change Impacts on Thermal Habitat in Massachusetts

This report describes the development of a regional model for predicting stream temperatures at gauged and ungauged streams and rivers in Massachusetts based on air temperature and spatial landscape characteristics. The model was used to project changes in stream temperature and thermal habitat resulting from rising air temperatures due to climate change. The model was comprised of two components: 1) a non-linear regression model representing the functional relationship between air and water temperatures at a single location, and 2) a machine learning model (XGBoost) for estimating the parameters of the air-water temperature model spatially based on landscape characteristics. Together, these models demonstrated strong performance in predicting weekly water temperatures with an RMSE of 1.3 degC and Nash Sutcliffe Efficiency (NSE) of 0.97 based on a 25% hold-out (i.e., validation) subset of the observed data. The model was used to predict mean July stream temperatures for each stream reach across the state (excluding large rivers with drainage areas > 10,000 km2 ) based on the 30-yr (1971-2000) normal mean air temperature to represent baseline conditions. Mean July water temperatures were also predicted for a series of climate change projections based on the RCP 8.5 emissions scenario over four 30-yr averaging periods centered on 2030, 2050, 2070, and 2090. For both the baseline and climate change scenarios, the predicted stream temperatures were used to classify each reach based on thermal habitat thresholds for cold- and cool-water fish species. Lastly, the final dataset of predicted temperatures and thermal classes across the state were summarized with maps and figures showing changes in the distribution of thermal classes using the total river miles for each major drainage basin and stream order. Statewide, the model predicts a reduction in cold water habitat (mean July temperature 50%) being warm-water habitat (> 22.30 degC). The projected stream temperatures and thermal classifications generated by this project will be a valuable dataset for researchers and resource managers to assess potential climate change impacts on thermal habitats across the state. With this spatially continuous dataset, researchers and managers can identify specific reaches or basins projected to be the most resilient to climate change, and prioritize them for protection or restoration. As more datasets become available, in particular high-resolution hydrography with a consistent routing network and additional streamflow data at locations co-located with existing temperature sensors, this model can be readily extended and adapted to increase its spatial extent and resolution, and to incorporate flow data for assessing the impacts of not only rising air temperatures but also changing precipitation patterns. Results from this study are provided in the `mastath-output-20230713.zip` zip file, which contains a geopackage and CSV file containing a geospatial layer of flowlines (i.e., stream reaches) with attributes providing both model inputs and outputs. See README.txt within the zip file for more details