ThibHlln
commented
3 years ago Also, the current open water component only routes streamflow internally to a mesh element, it does not transfer it to its neighbour according to the flow direction. I think that an openwater component additional to GR4 should be used, that would route the surface_runoff.
In addition, the inter-basin transfer of GR4 is not actually transferred anywhere. It should be transferred to its neighbour(s), probably as "subsurface_runoff"?
GR4J features a production bucket ($S$) conceptually corresponding to the runoff production, which would map onto the subsurface component. Then, the GR4J model features a Nash cascade, i.e. series of buckets ($S{h, 1}$... $S{h, nres}$), conceptually corresponding to the runoff routing, and it features a routing bucket ($R$) conceptually corresponding to runoff routing too (maybe the channel routing?). These two runoff routing components are not explicitly labelled as land vs. river routing, so deciding whether they should belong to the sub-surface component or the open water component is not clear-cut.
In fact, the GR5J model (siblings model of GR4J with one additional parameter, also a rainfall-runoff model) is sometimes used in a semi-distributed manner (i.e. applied individually to each river sub-basin, rather than the whole catchment), and the modellers use a linear lag propagation model to route the model outflow to the downstream sub-basin (see https://doi.org/10.5194/piahs-373-87-2016). This would motivate the fact that the whole of the runoff routing of GR4J (Nash cascade + routing store) should belong in the sub-surface component, and nothing should belong in the open water component.