The easy part: the parallel run is a new computer programme - intended to be an improbable the current Op.
The Op starts with what is thought to be the best approximation of the state of the atmosphere at T0. But there are always some estimates in that starting point. The various ensemble runs just tweak those estimates to test how it affects the evolution.
IIRC the Control is run at a lower resolution.
The professionals don’t just focus on the Op from one source; they use the ensembles and other sources.
Originally Posted by: Gandalf The White