Effects of temporal variability on HBV model calibration

Abstract

This study aimed to investigate the effects of temporal variability on the optimization of the Hydrologiska Byråns Vattenbalansavedlning (HBV) model, as well as the calibration performance using manual optimization and average parameter values. By applying the HBV model to the Jiangwan Catchment, whose geological features include lots of cracks and gaps, simulations under various schemes were developed: short, medium-length, and long temporal calibrations. The results show that, with long temporal calibration, the objective function values of the Nash-Sutcliffe efficiency coefficient (NSE), relative error (RE), root mean square error (RMSE), and high flow ratio generally deliver a preferable simulation. Although NSE and RMSE are relatively stable with different temporal scales, significant improvements to RE and the high flow ratio are seen with longer temporal calibration. It is also noted that use of average parameter values does not lead to better simulation results compared with manual optimization. With medium-length temporal calibration, manual optimization delivers the best simulation results, with NSE, RE, RMSE, and the high flow ratio being 0.5636, 0.1223, 0.9788, and 0.8547, respectively; and calibration using average parameter values delivers NSE, RE, RMSE, and the high flow ratio of 0.4811, 0.4676, 1.0210, and 2.7840, respectively. Similar behavior is found with long temporal calibration, when NSE, RE, RMSE, and the high flow ratio using manual optimization are 0.5253, −0.0692, 1.0580, and 0.9800, respectively, as compared with 0.4903, 0.2248, 1.0962, and 0.5479, respectively, using average parameter values. This study shows that selection of longer periods of temporal calibration in hydrological analysis delivers better simulation in general for water balance analysis.