using restart dumps between each year to avoid moving back to initial conditions, while allowing large parallel runs
