Multisite calibration of a semi-distributed hydrologic and thermal model in a large Canadian watershed

Abstract

ABSTRACT This study compares single-site and multisite calibration methods with upscaled parameters for thermal modelling. Multisite calibration and parameter upscaling are novel for calibrating a hydrological–thermal model. To this end, model calibration and simulations are carried out using a semi-distributed hydrological–thermal model (CEQUEAU). First, a hydrological calibration is carried out. The same calibrated hydrological model outputs are used for all thermal calibrations. The covariance matrix adaptation evolution strategy (CMA-ES) single-objective optimization algorithm is applied to calibrate the thermal module using observed water temperatures at seven monitoring sites within the Nechako Watershed in British Columbia, Canada. Single-site calibration provided the best performance for individual sites, but these sets of parameters performed poorly when applied to the other sites. The multisite calibration and the upscaled methods perform adequately for all sites with overall Root Mean Square Error (RMSE)<2°C as an acceptable threshold. However, the multisite method resulted in the best performance metrics and shortest computing time.