This thesis is closely related to the framework of the operational flood forecasting system of the Tyrolean River Inn (HoPI), in which the hydrological model HQsim is applied for tributary catchments. The present studies pursue the issue, whether the integration of additional snow data into the optimization process decreases parameter and prognosis uncertainty of HQsim. The analyses start with a sensitivity analysis of the modelled runoff output attributable to the simulated snow cover performance. A variance reduction in the model output due to a well-calibrated snow module could be shown. The effect of additional snow cover data, considered within the model optimization, is analyzed subsequently. Two methods to quantify model uncertainty, established in hydrological sciences, are applied to confront the results of a classical uni-objective optimization with the results of a multi-objective optimization regarding runoff and snow cover data. To start with the results of the GLUE method, no homogeneous prognosis improvement with only minor differences between the results of both optimization approaches is recognizable. Yet, the residual analysis of the runoff simulations and the goodness-of-fit values of the snow cover simulation feature obvious spatial patterns. Since uncertainties cannot be separated further into model predictive uncertainty and total predictive uncertainty within the GLUE method, a proper Bayesian Inference considering heteroscedastic residuals is computed. The results unfortunately show irregular characteristics of the predictive intervals, due to a recently identified programming error within the source code of HQsim. Inferential improvements are still noticeable for the use of the multi-objective likelihood. Finally, the regional model behaviour characteristics, analyzed by the GLUE method, are also confirmed by the proper Bayesian inference. The effects of the programming error on the performed analyses are still analyzed and lead to the assumption, that further simulation improvements, including narrower uncertainty intervals using the multi-objective likelihood, can be assumed by an optimized adjusted model. Thus, similar modeling results oriented to several mountain ranges of Tyrol should also be more distinctive.