Calibration of the Variable Infiltration Capacity Model from Hyper-Resolution to the Regional Scale

Abstract:

The Variable Infiltration Capacity (VIC, Liang et al., 1994) model has been used for a broad range of applications, in hydrology as well as in the fields of climate and global change. Calibration of the often distributed application of the model is difficult, certainly in the light of the ongoing discussion on applying global models at hyper-resolution (Wood et al., 2011). 
To improve the calibration procedure for VIC applied at grid resolutions varying from meso-scale catchments to the 1 km ‘hyper’resolution now used in several global modeling studies, the parameters of the model are studied in more detail with specific focus on scale effects. A lumped VIC-model was constructed for three nested basins: the Rietholzbach (3.4 km2), Jonschwil (492 km2) and the Thur basin (1700 km2) in Switzerland. With the DELSA sensitivity analysis method (Rakovec et al., 2013) it was shown that parameter sensitivity does not change over scale. Extensive calibration of the lumped models using the DREAM algorithm (Vrugt et al., 2008) revealed that most of the calibrated parameter values of the three basins were within each others uncertainty bound based on the converged part of the posterior. This information was used for calibration of the distributed VIC models, which where constructed for the Thur basin at a grid resolution of 1x1 km, 5x5 km and 10x10 km.