A novel method for frequency analysis of high water temperatures using temperature duration curves in a partially regulated watershed.

Water temperature is the key variable for aquatic environmental assessments, and in the absence of long-term observational records, simulations provide essential data for further environmental analysis. In this study, we investigate the upper percentiles of a novel Temperature Duration Curve (TDC) as a proxy to characterize high water temperatures in a partially regulated watershed. The study area comprises the 47,200 km Nechako Watershed of British Columbia, Canada, where nine hydrometric stations measure streamflow and water temperature. Four sites represent regulated flows, while five sites are unregulated. Using daily streamflow and water temperature observations and ERA5-Land air temperature data, the Air2Stream model was calibrated for historical periods from 1950 to 2023. Simulations were also conducted using naturalized flows for the regulated Nechako River at Vanderhoof and Isle Pierre to reconstruct naturalized water temperatures. A frequency analysis was performed on observed, simulated, and naturalized water temperatures for different return periods and TDC percentiles. The analysis included exceedances of 0, 1, 5, and 10 % for high temperatures and the 50th percentile (median) for mean temperatures. Results showed robust model performance with RMSE <1.5 °C (NSE > 0.9 and KGE > 0.8) for all stations during calibration. The frequency curves showed reduced high temperatures for the Nechako River at Vanderhoof due to ecological flow releases, but increased mean temperatures for return periods >2 years. This novel approach suggests that thermal mandates should be related to Q and Q, which had the best agreement between observations and simulations. The method is applicable globally across various basin sizes and hydrological regimes, making it a valuable tool for assessing water temperature dynamics in diverse environments.