A key point to bear in mind is that even the Operational run includes estimates, since there are more grid points than there is actual data. If you start with a degree of uncertainty that can only grow over time: the ensemble suite attempts to rest the stability of the evolution by tweaking some of the opening estimates.
The higher resolution should be better at picking out smaller changes that can evolve to influence the bigger picture: eg the sudden appearance of that small feature which delayed the plunge south of the Arctic air mass.
Originally Posted by: Gandalf The White