The identification of optimum model parameters may be influenced by temporal or event-specific changes of optimum parameter ranges and the length and information content of calibration data. These effects were studied for the hydrological model WaSiM-ETH in a 170 km² catchment. Based on a Monte-Carlo simulation including seven model parameters, we investigated temporal and state dependent changes of parameter identifiability using the DYNIA algorithm. The effect of data length was studied using a modified DYNIA approach based on a growing window algorithm. The DYNIA analysis revealed temporal changes of identifiability for the snow melt runoff parameter cmelt, which is only identifiable during winter runoff, and for the drainage density parameter drd. The drd parameter was closely related to observed discharge (or catchment moisture), when re-ordering the time series by discharge. Such dependencies probably result from processes not included in model equations. The growing window analysis shows that more than one year of data did not result in improved identification of model parameters cmelt and drd. Using the re-ordered data series, good identifiability of cmelt was bound to high discharges, while identifiability of drd changed with the addition of further values in descending or ascending order. The methodology revealed structural problems with regard to the parameter drd, which are not yet completely understood and require further investigation.