Unresolved issue:
Advection term, gradient should operate on velocity, not mass
transport!
2012-07-12:
Checkerboard due to advection,
imminent catastrophe.
The method was NONLIN=7, i.e. the formula uses (M·Grad) M/H
(NONLIN=2 or 3: (M/H·Grad) M
)
Have a look at the instruction
file with which this was produced.
We had already reduced the time interval to 0.6 subcritical. Eddy
viscosity is not low. Depth is changing dramatically there, however.
Here's the advection term's x-component:
B.t.w. at the boundaries the advection term was formulated according
to nonlin 2 (3)
2012-07-11:
otemt1 did not step far
otem92 did not copy the UPD option from tide to tide, it rather
reset it to a default 'NNY' or 'NYY'. This implied that user update
was only included in the Active-boundary data set for the M2. otem92
did not fail on this condition, nor did otem16, only otemt1's
watchdog began to bark and bite.
The offending values were 999. × √2.
It was helpful to set the bite_wd
level in otemt1 very high and sample the elevation grid with the
mouse.
It was even more helpful to send OC ^A in order to trace the active boundary
calculation.
Remedy:
otem92 copies the previous UPD option now (unless another 'UPD:'
option is found on the tide line.)
otem92 will stop on missing data at an active boundary. A stop
message is sent to STDERR
2012-07-11:
Big elevation difference at an
active boundary.
Reason: Eddy viscosity parameter too high.
Remedy:
With Fricp(4) = 5.e3 the difference was much less bold.
2012-07-10, FRIH:
otemt1 bombed at the southern coast
of the Seattle strait, near that cape. In running up to the
catastrophy a checkerboard pattern was developing.
Remedy:
Straight passive-boundary formulation of the advection term seemed
to have had errors. Determination of the upwind grid offset must be
done differently for these boundaries. Seemed to have helped.
2012-07-10:
Wrong conclusion. The checkerboard
pattern was first thought to originated due to a sign error
in the eddy viscosity term. What happens when you attack the bull at
both horn and tail is that you think the horns has stabbed you while
it was only the tail that slapped into your face.
No, the eddy term was o.k., the straightening of the advection term
made all the change.
Appendix
2012-07-11 User
update KATT/itoz-upd.dat
provides lines for copying tide information into parts of the active
boundary where the global model has land nodes.
otem92b.ins namelist:
itoz_upd_u=21
qrwd_itoz_upd_u=.true.
itoz_opt='YNN'
Open-file block:
21 B KATT/itoz-upd.dat