The aim of this study is to analyze the impact of a new version of the Canadian Regional Ensemble Prediction System (REPS), including surface condition uncertainties, on precipitation forecasts and analysis. In the REPS, the Canadian version of the ISBA land-surface scheme is coupled to Environment Canada's Numerical Weather Prediction model (GEM). In the first part of the study, stochastic perturbations were produced to perturb the surface parameters (e.g., vegetation fraction, leaf area index, sea-ice fraction) and prognostic variables (e.g., soil moisture, soil temperature at different layers) during twenty summer days in 2009. The impacts of this new REPS were quantified with regard to precipitation forecasting up to 48-h lead-time relative to the old version (without surface perturbations) and the deterministic model. In the second part of the study, two versions of the REPS were used to drive the Canadian Precipitation Analysis (CaPA) system to produce 6 h rainfall accumulations at a resolution of 15 km over North America. The spatial interpolation technique is based on statistical interpolation using short‐range precipitation forecasts from the Canadian Meteorological Centre's (CMC) regional model as the background and rain‐gauge measurements from the surface network. By using the REPS to drive the CaPA, the influence of the surface and atmospheric uncertainties on precipitation analysis products could be analyzed in relation to the deterministic model analysis.