limit Visbeck eddy length to dcEdge (AMOC PR 4/6) #4868
Conversation
| @@ -611,8 +611,8 @@ subroutine ocn_GM_compute_Bolus_velocity(statePool, & | |||
| c_Visbeck = sqrt(sumN2*ltsum) | |||
| sumRi = sumRi / (ltsum + 1.0E-11_RKIND) | |||
|
|
|||
| eddyLength(iEdge) = min(c_Visbeck/(1.0E-15_RKIND + abs(fEdge(iEdge))), & | |||
There was a problem hiding this comment.
@vanroekel this array was created in the original Visbeck addition. Both eddyLength and eddyTime are used only in these three lines within the iEdge loop, so these arrays could both easily be replaced by scalars. The only reason to keep them is for writing them out. Do you have a preference if these are arrays or scalars?
There was a problem hiding this comment.
@mark-petersen no real preference on scalar v array. I had them as arrays for diagnostics but I don't think we need them for the duration.
mark-petersen
changed the title
| @@ -611,8 +611,8 @@ subroutine ocn_GM_compute_Bolus_velocity(statePool, & | |||
| c_Visbeck = sqrt(sumN2*ltsum) | |||
| sumRi = sumRi / (ltsum + 1.0E-11_RKIND) | |||
|
|
|||
| eddyLength(iEdge) = min(c_Visbeck/(1.0E-15_RKIND + abs(fEdge(iEdge))), & | |||
There was a problem hiding this comment.
@mark-petersen no real preference on scalar v array. I had them as arrays for diagnostics but I don't think we need them for the duration.
|
Ran a 5 year |
|
Ran a 5-year test with this branch for the SORRM: |
|
Visbeck baseline test (ran with hash tag b404a46) with SORRM was completed. Complete analysis here: https://web.lcrc.anl.gov/public/e3sm/diagnostic_output/ac.milena/20220411.CRYO1850.ne30pg2_SOwISC12to60E2r4.chrysalis.CryoBranchBaselineGMVisbeck/Years1-5/ |
|
Clearly, the Visbeck eddy length limiter makes a small difference for SORRM. |
|
I'll post the analogous plots for the low-res comparison, adding global MOC: Visbeck baseline: Visbeck limiter (this PR): |
…ext (PR #4868) Limit Visbeck eddy length to dcEdge When using config_GM_closure='Visbeck' (Visbeck et al 1997), limit the eddy length scale to be no less than dcEdge, the distance between cells. This ensures that gmBolusKappa does not have very small values due to the eddyLength factor. When using config_GM_closure='Visbeck' it is not BFB. However, all E3SM runs use config_GM_closure='constant' or 'EdenGreatbatch', but never Visbeck, so this PR is BFB for all current tests. [BFB]
|
Passed:
merged to next |
|
merged to master |
When using
config_GM_closure='Visbeck'(Visbeck et al 1997), limit the eddy length scale to be no less than dcEdge, the distance between cells. This ensures thatgmBolusKappadoes not have very small values due to theeddyLengthfactor.All E3SM runs use
config_GM_closure='constant'orconfig_GM_closure='EdenGreatbatch', but never Visbeck, so this PR is BFB for all tests. When usingconfig_GM_closure='Visbeck'it is not BFB.[BFB]