round goes the wheel
sometimes the simplest things can make the greatest difference. you know we had been struggling for a while, getting a version of the upper mississippi model up and running in ecomsed, that would reproduce the work done by those "other consultants" back in the day. and there were instabilities during low flow years, particularly 1988, that we just couldn't get rid of. pool 4 was okay, fortunately, as water quality in lake pepin is what everybody cares about the most anyway, so we can run the sensitivity analyses as required. but we had to get pool2 and pool 3 right, eventually.
i tried a bunch of things, including combing the old ecomsed code for hardwire fixes, which turned up a number of clues (like arbitrarily reducing wind stress by 50%). it seemed to behave better with no weather forcings at all, and although heat flux plays a role, really the problem had something to do with wind effects, especially at low flows. we became suspicious of the clamped boundary condition at the downstream (lock and dam) ends, and the possibility of wave reflection problems.
so yesterday i became convinced that it would be worth trying one of the radiation-type boundary condition options in ecomsed. to make a long story short, what worked was a reid & bodine type of open boundary condition, which uses a lagrange multiplier to allow the boundary level to change a little from the specified value, so that longwave radiation energy can get the hell out of there instead of being reflected back into the model domain and causing all sorts of havoc. that was most likely what was happening before. interestingly, when i look at animations of water surface elevation there are now some furious oscillations near the dam, but the scale is exaggerated; the fluctuations are of order 5 cm, which is probably close enough to true wave height, so one could argue that the model is actually simulating what is actually happening. i like that argument.
i can't tell you the deep feeling of relief when i looked at this output, and saw the way the green line (our new result) fell right on top of the pink line (their old result), instead of wagging up and down and all around like our previous results. it's enough to make you feel like you did something right, or that you know what you're doing. or can learn from your mistakes, or some other clichéd observation about life.
on the bike front, i'm delirious about my soon-to-be-built track wheels: sew-ups at last! right now i have only the rims, but a pair of record pista 36-hole hubs are on their way from germany, of all places. i splurged a little on the hubs, after getting a good deal on the rims, because i figured that 1) a good pair of track hubs would not be cheap under any circumstances, b) i've been good, and iii) i've got a birthday coming up. so happy birthday, crunchy! the rims are pictured here; after the hubs arrive, and steve at great lakes builds them up for me (whaddaya think, 4-cross on the back, 3-cross on the front?), i'll post them here. and then: to the vélodrome!!
i tried a bunch of things, including combing the old ecomsed code for hardwire fixes, which turned up a number of clues (like arbitrarily reducing wind stress by 50%). it seemed to behave better with no weather forcings at all, and although heat flux plays a role, really the problem had something to do with wind effects, especially at low flows. we became suspicious of the clamped boundary condition at the downstream (lock and dam) ends, and the possibility of wave reflection problems.
so yesterday i became convinced that it would be worth trying one of the radiation-type boundary condition options in ecomsed. to make a long story short, what worked was a reid & bodine type of open boundary condition, which uses a lagrange multiplier to allow the boundary level to change a little from the specified value, so that longwave radiation energy can get the hell out of there instead of being reflected back into the model domain and causing all sorts of havoc. that was most likely what was happening before. interestingly, when i look at animations of water surface elevation there are now some furious oscillations near the dam, but the scale is exaggerated; the fluctuations are of order 5 cm, which is probably close enough to true wave height, so one could argue that the model is actually simulating what is actually happening. i like that argument.
i can't tell you the deep feeling of relief when i looked at this output, and saw the way the green line (our new result) fell right on top of the pink line (their old result), instead of wagging up and down and all around like our previous results. it's enough to make you feel like you did something right, or that you know what you're doing. or can learn from your mistakes, or some other clichéd observation about life.
on the bike front, i'm delirious about my soon-to-be-built track wheels: sew-ups at last! right now i have only the rims, but a pair of record pista 36-hole hubs are on their way from germany, of all places. i splurged a little on the hubs, after getting a good deal on the rims, because i figured that 1) a good pair of track hubs would not be cheap under any circumstances, b) i've been good, and iii) i've got a birthday coming up. so happy birthday, crunchy! the rims are pictured here; after the hubs arrive, and steve at great lakes builds them up for me (whaddaya think, 4-cross on the back, 3-cross on the front?), i'll post them here. and then: to the vélodrome!!
posted by cicadashell at
10:13 AM
0 Comments:
Post a Comment
<< Home