Despite advancements in hydrological modelling, quantifying inherent uncertainties in simulation and forecasting remains essential. These uncertainties arise from sources such as initial conditions, input data, parameter estimation, and model structure. While the hydrological community has increasingly focused on uncertainty assessment, most studies concentrate on input data and parameter uncertainty within specific models, leaving model structure uncertainty unexplored. This study introduces an ensemble-based approach to assess hydrological distributed model uncertainty, emphasizing model structure and input data uncertainties concurrently. The study leverages the Raven hydrological modelling framework to create an ensemble of hydrological model structures. This structural ensemble fed noise-perturbed forcing inputs to represent input data uncertainty. The forward greedy method aids in selecting a pool of models from the ensemble, enhancing reliability and reducing the model count. This method is employed to refine the model pool by ensuring that each criterion meets the predefined performance standards. The approach is demonstrated over the southwest portion of the Saint-Laurent watershed in Canada, evaluating model ensembles against observed streamflow. This study advances the understanding of hydrological model uncertainty assessment and emphasizes the significance of a comprehensive, multimodel approach that accounts for structural, input data, and calibration uncertainties for robust streamflow simulations and forecasts. The findings highlight how employing multimodel ensembles is crucial for minimizing different sources of uncertainty, as opposed to relying on a solitary model.