Open MattReimer opened 7 years ago
MattReimer
commented
7 years ago Peter A Mchugh Yo boyz to men,
That's a bit of a conundrum. But in light of Jensen's Inequality ;) methinks that resolution should remain constant across flows, at least within a particular visit running at different Qs...maybe. But this means that things get bigger as they go up.
Yours truly,
Spice drop
MattReimer
commented
7 years ago Eric Hey guys,
Just spent some time looking into the code (MN's portion of code for the NREI model -- the hydraulic portion, not mine), and it looks like I was wrong about the grid expanding/contracting. Yesterday I had something like this in mind: https://www.youtube.com/watch?v=7YYTgENyJxs when trying do describe what I thought would happen.
What actually happens is that the grid will add/subtract points in the XY plane and then add/subtract vertical layers in the z direction according to the same spacing rules as before as the wetted channel expands/contracts with the higher/lower discharges. By "the same spacing rules as before" I mean by the same spacing rules as the "standard" rules for the "standard" flow (the one measured during the CHaMP survey). So...absolute number of 3D grid points will NOT be constant across flows. Point spacing in XY and the spacing of the vertical layers in Z will both be maintained across flows. As per PM's comment, this sounds favorable to the scenario I wrongly thought we were in yesterday. Thanks for that nod to Jensen's inequality, Pete. Can't get away from that thing! This does mean that if flow differences are large (e.g. 3x or 1/3x difference in comarison to the "standard" flow), we should expect much longer/shorter runtimes because 3D grid size (the best determinant of runtime) will change as flow changes.
Sorry for any undue stress! Realized it could be a problem and thought I'd draw attention to it, but hadn't had a chance to dig in further yet.
Hope you guys are well.
From Eric: Some food for thought here that came up in recent discussion with Shubha.
3D grids in the nrei model (the points at which nrei calculations are performed) are built (size/resolution) using parameters based on results of the CHaMP survey at the site in question.
Now that we are simulating multiple flows (or will be soon?), we might want to consider whether or not the Champ-survey-sourced 3D parameters are still appropriate?
If we recycle the 3D grid-building parameters from the Champ survey, I believe the 3D grid will basically just expand/contract to the new flow (i.e., the absolute number of 3D points will not change between the flows, but the resolution in xyz will differ). Perhaps this is a good way to handle things?
If, however, the ranges of flows to be simulated are quite large, this may not be optimal. Some experimentation may be needed in the future to get a better handle on this. Do we have guidelines for what flow ranges will be simulated? I am guessing they will top out at bankfull?
Not necessarily proposing that we do anything about this now, but do want the ideas to be out there as something that should be thought about.