Open costachris opened 2 years ago
Similar failure for the LES_driven_SCM
case on cfsites 12-15 (which represent moderately-deep to deep convection).
For site 15, the input values to TD.PhaseEquil_pθq
right before failure are:
(p0_c[k], en.H.values[k], en.QT.values[k]) = (74138.3896242084, -3008.236921241473, 0.0)
(a_bulk_c, up.H.bulkvalues[k], gm.H.values[k]) = (0.9000000000000001, 684.4972004591178, 315.22378828905926)
Values of up.H.bulkvalues[k]
should typically be ~300 and values of a_bulk_c
~0.01
Timeseries of thetal_mean
and updraft_area
for site 15:
Description: Certain combinations of namelist parameters lead to
DomainError
ormaxiter reached in saturation_adjustment_given_pθq
, stemming from values of H<0 being fed into thePhaseEquil_pθq
function when performing saturation adjustment. Occasionally θ_liq_ice=NaN, q_tot=NaN, T=NaN as well.https://github.com/CliMA/TurbulenceConvection.jl/blob/971a265708beabbd0a1d359ba1ab3a0ae07994d2/src/update_aux.jl#L82
This is encountered in the first few iterations of the simulation.
TODO: Investigate root cause.
Stacktrace:
. . .
. . .