| description abstract | In the Model Parameter Estimation Experiment (MOPEX) project, after calibration of model parameters, complex rainfall?runoff hydrological models (HMs) simulated streamflow better than land surface models (LSMs), including the Soil?Water?Atmosphere?Plant (SWAP) model. A possible explanation for this is that the LSMs may not have been well calibrated. To test this statement, different strategies to calibrate SWAP using daily streamflow from 12 MOPEX basins were investigated. Optimization of parameter values was performed using a combination of an automated optimization algorithm and manual efforts. For automated optimization, two different global optimization algorithms were used: 1) a random search technique and 2) a shuffled complex evolution method developed by the University of Arizona (SCE-UA). Two objective functions, based on the Nash?Sutcliffe coefficient of efficiency and the mean systematic error, were applied. The number of calibrated parameters ranged from 10 to 15. All adjusted parameters were kept within a reasonable range so as not to violate physical constraints while providing a close match between simulated and measured daily streamflow. The results of streamflow simulations with different sets of optimal parameters were compared with each other, with observations, and with simulation results obtained by the HMs that participated in the MOPEX project. The new SWAP calibration strategies resulted in significant improvement of SWAP streamflow simulations, which came close to the best HM results. It was confirmed that model performance depends greatly on the calibration strategy and that the land surface model SWAP, with appropriate calibration, can simulate runoff with the accuracy that is comparable to the accuracy of hydrological models. | |
