Outflow of reservoir/lake

[hboisgon]

Maybe a little related to #90 but I have a question on how the reservoir/lake outflow are being given back to the wflow model.
What I see from the code is that the outflow is added into qin of the downstream cell rather than directly changing the q/q_av value of the cell itself (see few line below):

Wflow.jl/src/flow.jl

Lines 284 to 298 in e3de2bc

	sf.q[v] = kinematic_wave(
	sf.qin[v],
	sf.q[v],
	sf.qlat[v] + inflow,
	sf.α[v],
	sf.β,
	adt,
	sf.dl[v],
	)
	if !isnothing(sf.reservoir) && sf.reservoir_index[v] != 0
	# run reservoir model and copy reservoir outflow to inflow (qin) of
	# downstream river cell
	i = sf.reservoir_index[v]
	update(sf.reservoir, i, sf.q[v] + inflow_wb[v], adt)

The thing is that then for the reservoir/lake outlet cell, if you save q_av you get a completely hypothetic value and the outflow of the lake is masked as an external inflow to the downstream cell instead. So if you would make a mass balance over the surface flow cells as a postprocess (eg coupling to Delwaq) then you have to manually overwrite the volume and q_av of the lake outlet to get the mass balances right. I think it would be easier and maybe more intuitive if for these cells both q/q_av/volume would be overwritten by the lake/reservoir data ?
This would also fix the problem of modelling a lake without a downstream node.

[verseve]

For the reservoirs and lakes we did take a very similar approach as in the wflow Python version. q and q_av are actually not completely hypothetical at a reservoir or lake cell, these values represent the inflow generated by river flow. The reason to also include the river (kinematic wave) of the reservoir/lake cell for the inflow calculation, is for the case when the reservoir/lake area is small compared to the cell resolution.

I think it would be easier and maybe more intuitive if for these cells both q/q_av/volume would be overwritten by the lake/reservoir data ?

This is possible, then we have to take the river flow from the upstream cell. The advantage is that this will indeed also fix the problem of modelling a reservoir/lake without a downstream node. For the case with reservoirs/lakes that are relatively small compared to the cell resolution I would say the current implementation is better. I am not so sure that overwriting volume from the reservoir/lake model is a good idea (mixing of information). I think just using the volume from the lake/reservoir is more explicit for a mass balance calculation (e.g. for local inertial flow h is used as a boundary condition, and would be good to have volume in line with h).

@JoostBuitink: would be good to hear your opinion on this!

[hboisgon]

Thanks for your quick reply @verseve !
I agree that for volume then this indeed can be kept to what it is.

I'm not sure it happens so often with the classic resolution of wflow that the lake/reservoir area has a much smaller resolution than the cell itself? I mean qlat and q from upstream would already be considered so the lake would have to be really small to get significant errors (1/3 of the cell or less?). The global lake model is anyhow only good for large lakes (>10km2)

In a way you could also still run the kinematic wave first but then instead of adding the lake outflow to the dowsntream overwrite the q and q_av from the intermediate kinematic wave instead?

[verseve]

I'm not sure it happens so often with the classic resolution of wflow that the lake/reservoir area has a much smaller resolution than the cell itself? I mean qlat and q from upstream would already be considered so the lake would have to be really small to get significant errors (1/3 of the cell or less?). The global lake model is anyhow only good for large lakes (>10km2)

Yes, I agree, for the SBM concept, where we often run with a ~1 km resolution (default) this is not an issue I would say. I am not sure about the model resolution of other concepts like FLEXTopo (@laurenebouaziz ?) or HBV, it is probably close to SBM. Anyway, if we change this, we can also explain this in more detail in the documentation.

In a way you could also still run the kinematic wave first but then instead of adding the lake outflow to the dowsntream overwrite the q and q_av from the intermediate kinematic wave instead?

I am not sure if I understand this question completely. I think you propose to overwrite q and q_av of the kinematic wave in the same cell as the reservoir/lake outlet location with the reservoir/lake outflow? This is indeed possible, but for the inflow (from the river) to the reservoir/lake we then have to take the upstream cell(s). Otherwise, the reservoir/lake outflow is also used as inflow. Or, we have to account for both in one cell, not sure if it is worth to go in that direction.

[laurenebouaziz]

yes I think we can in general assume the flextopo resolution to be similar to the sbm resolution (and same for hbv I would say)

[hboisgon]

Otherwise, the reservoir/lake outflow is also used as inflow.

Oh yes indeed, I forgot that part, so ignore what I said, then just indeed overwrite q and q_av