Automated calibration applied to a GIS-based flood simulation model using PEST

This paper describes an automated calibration procedure applied to the GIS-based distributed WetSpa (Water and Energy Transfer between Soil, Plant and Atmosphere) model by incorporating a modelindependent parameter estimator PEST (Parameter ESTimation). This calibration approach is applied to estimate the most sensitive parameters of the model with observed flow hydrographs as the calibration target. The best set of parameters is selected from within reasonable ranges by adjusting the values until the discrepancies between observed and simulated hydrographs is reduced to a minimum in the weighted least squares sense. A case study is performed in the Margecany catchment, 1133 km2, situated in the upstream part of the Hornad River basin, Slovakia. The parameter set obtained from the automated calibration procedure provides a good fit compared to observed flow hydrographs, indicating that the automated calibration scheme is a credible alternative to the manual approach. parameter values must be defined so that a given parameter mainly reflects the significant and systematic variation for a certain process, thus reducing significantly the number of free parameters that need to be adjusted subsequently. As pointed out by Refsgaard (1997), the important points considered in a parameterization procedure should include: (1) to select parameter classes (topography, soil type, land use, climatology, etc.) for easily associating parameter values, (2) to evaluate parameters explicitly which can be assessed from field data or need some kind of calibration, and (3) to keep a low number of real calibration parameters both from practical and methodological points of view. This paper describes the calibration process for the GIS-based WetSpa hydrological model using the PEST automated calibration routine. The procedures of integrating PEST with WetSpa, the methods of dealing with the calibration of spatial model parameters, and the steps in the estimation process, including model parameterization, choice of calibration parameters, and specification of calibration and evaluation criteria, are presented. The approach is illustrated for model calibration of the Margecany catchment, Slovakia, with 10 years observed daily flow hydrographs at the basin outlet as the calibration target. Calibration results show that this scheme has a good performance in estimation of model parameters, and can be a credible alternative to the manual approach. However, a further manual refinement is necessary to avoid the ill-posed problems associated with the direct inverse procedures.