Abstract. The spatially distributed MESH hydrologic model (Pietroniro et al., 2007) was successfully calibrated and then validated for the prediction of snow water equivalent (SWE) and streamflow in the Reynolds Creek Experimental Watershed in Idaho, USA. The tradeoff between fitting to SWE versus streamflow data was assessed and showed that both could be simultaneously predicted with good quality by the MESH model. Not surprisingly, calibrating to only one objective (e.g. SWE) yielded poor simulation results for the other objective (e.g. streamflow). The multiobjective calibration problem in this study was efficiently solved via a simple weighted objective function approach and analyses showed that the approach yielded a balanced solution between the objectives. Our approach therefore eliminated the need to rely on a potentially more computationally intensive evolutionary multiobjective algorithm to approximate the entire tradeoff surface between objectives. Additional calibration experiments showed that for our calibration computational budget (2000 model evaluations), the autocalibration procedure would fail without being initialized to a model parameter set carefully determined for this specific case study. This study serves as a benchmark for MESH model simulation accuracy which can be compared with future versions of MESH.
A statistical basis for multiobjective calibration
