V5.1.0 lake fixes #161
V5.1.0 lake fixes #161
Conversation
…area, and zero out lateral inflow (temporarily). Prints needs to be removed beofre primetime.
… inflows are still hard-coded to 0.
| @@ -94,7 +94,7 @@ subroutine LEVELPOOL(ln,qi0,qi1,qo1,ql,dt,H,ar,we,maxh,wc,wl,oe,oc,oa) | |||
| real :: dh, dh1, dh2, dh3 ! Depth in weir, and height function for 3 order RK | |||
| real :: It, Itdt_3, Itdt_2_3 ! inflow hydrographs | |||
| real :: maxWeirDepth !maximum capacity of weir | |||
| real :: qldum !dummy ql to 0 out | |||
| !real :: hdiff_vol, qdiff_vol ! water balance check variables | |||
There was a problem hiding this comment.
If adding new variables to the code, can we make an effort to move away from F77 style/conventions? Pick a descriptive name and add a descriptive inline comment with units. Is qdiff_vol a flow? A Rate? A Volume? Similiarly, hdiff_vol is not informative, I can't determine if it is a height, head, or volume. I recommend adding them each on their own line with elaborated names and inline comments.
There was a problem hiding this comment.
Sure, although these were really just our internal testing checks that I left in just in case we need to activate again. This is not really how we should be doing mass balance checks on this routine, which is why it is commented out. Might be cleaner just to remove completely - this is really just a quick patch to get these updates into ops since I know you all are working on a full refactor for next round...
|
@laurareads can you include some more details on your offline tests/evaluation mentioned above by @aubreyd ? |
|
We have tested these changes across CONUS, evaluating at all USGS gauge locations for a 4-year period. Statistics do not show alarming differences, and answer changes are within expected ranges. The testing results shown here (answer diffs) are expected since the lake code changes will result in different behavior while the model is spinning up/lakes reaching equilibrium. |
Committing a few lake fixes:
(1) Properly enclosing weir equation exponent in parentheses.
(2) Setting uniform lake surface area consistent with vertical wall assumption (may change in the future).
(3) Zeroing out qlateral inputs for reach-based methods for now as this was likely improperly specified (qlateral inputs are more important for gridded).
(4) Yates' updates to Runge-Kutta solution formulation.
We left in a quick & dirty diag water budget tracking that is commented out for now. Won't get full closure at any given sub-timestep due to approximate RK solution, but should even out to 0s. This should eventually be moved out of the channel routing timestep.
Engineering tests passed. Expected answer changes in channel and lake fluxes only. These answer changes over a multi-year CONUS run were evaluated offline by the Wonderful Read.