Hybrid Parallel and Sequential Models

Complementary models that work in parallel with the operational flow forecasting model and aim at reducing its error are becoming more and more popular during the last decade. One of such approaches uses a parallel architecture (model ensembles, i.e. several models running in parallel); the disadvantage is that it becomes difficult to speak about the “model state space”. Another way of combining models is a sequential combination of models like it is done in a data-assimilation process that updates model states (or only its outputs). Contrasting these approaches as “parallel” and “sequential”, and comparing their performance may be useful for selecting the model architecture for operational forecasting.