Colloquium

Thesis colloquium Caspar Cluitmans

In this research, techniques for the calibration of WALRUS were identified. The developed calibration approach was validated on four different lowland catchments.

Organisator Hydrology and Quantitative Water Management
Datum

do 18 september 2014 10:30 tot 11:30

Locatie Atlas, building number 104
Droevendaalsesteeg 4
104
6708 PB Wageningen
Zaal/kamer Atlas 1

 

For this study, techniques for parameter estimation of the Wageningen Lowland Runoff Simulator (WALRUS) were identified and validated. The Levenberg-Marquardt calibration algorithm was selected as a suitable technique for parameter estimation. The used Levenberg-Marquardt package required hard boundary values to be set for each parameter that hat to be fitted. To obtain suitable values for the lower- and upper-boundaries, so-called ’rough-runs’ were performed. The ’rough-runs’ prove to be able to give some insight into what suitable boundary values might be. Its successful implementation however depends on the quality of a priori information on the possible optimal parameter values. Validation of the calibration approach showed that it had to be expanded with regularisation, to obtain more realistic simulations. The dependency on a priori information on the possible optimal parameter values however increased, due to the regularisation method’s use of centre-parameters. A multi-start method was added to the approach so that local minima could be detected. The fully developed calibration approach was validated on four lowland catchments. One of these catchments had good a priori information on the possible optimal parameter values. For this catchment good results were obtained. For the other three catchments the above mentioned a priori information was of lower quality. For these catchments it was not possible to obtain good results. A sensitivity analysis and the results of the validation showed that cG, cV, cQ and cS showed behaviour that could indicate that they are sensitive to changes. The sensitivity analysis was also successfully validated as tool for uncertainty analysis. WALRUS simulations were coupled to see if it was possible to model a large catchment without introducing a hydraulic component. This was possible when only a few sub-catchments were coupled.