Models of hydrocarbon reservoirs are often used to support management decisions about field development and redevelopment. Typical modelling workflows result in a reservoir and simulation model with a property distribution generally comparable to the well data, and this is often considered sufficient. The current study tests this assumption on the fluvial reservoir in the Upper Lunde Member of the Snorre Field where sedimentary heterogeneities at multiple scales influence reservoir properties such as porosity and permeability. This work shows that, by describing and modelling the sedimentary heterogeneities at several length scales in the reservoir, and by using a flow-based local upscaling method, the resulting porosity and permeability distribution at the scale of the reservoir and simulation model are significantly different from porosity and permeability distribution at the well data scale; the variance tends to reduce and, for permeability, the distribution type is changed from log-normal to normal. Reservoir property distributions based on multiscale modelling, sensitive to the representative elementary volumes for permeability, and upscaled in a realistic sedimentological framework, give a better representation of the effective permeability architecture.