f.cam6_3_160.FMTHIST_ne30.unstable_Ri_mixing.001

cecilehannay commented 4 months ago

Purpose:

@JulioTBacmeister noted that HB diffusion performs mixing for stable conditions (Ri>0) as well as background mixing (see outcommented code from subroutine austausch_atm below that computes mixing coefficients):

  kvf(:ncol,:)           = 0.0_r8
  ! Compute the free atmosphere vertical diffusion coefficients: kvh = kvq = kvm.
  do k = ntop, nbot - 1
     do i = 1, ncol
        if( ri(i,k) < 0.0_r8 ) then
           fofri = sqrt( max( 1._r8 - 18._r8 * ri(i,k), 0._r8 ) )
        else
!phl           fofri = 1.0_r8 / ( 1.0_r8 + 10.0_r8 * ri(i,k) * ( 1.0_r8 + 8.0_r8 * ri(i,k) ) )
           fofri = 0.0_r8
        end if
        kvn = ml2(k) * sqrt(s2(i,k))
        kvf(i,k+1) = kvn * fofri !phl
!phl        kvf(i,k+1) = max( zkmin, kvn * fofri )
     end do
  end do

We would like to turn that off!

Description:

Same as https://github.com/NCAR/amwg_dev/issues/530 with source mod:

/glade/u/home/pel/src/for-unstable-Ri-hb/pbl_utils.F90

Case directory:

Locally (if still available): /glade/campaign/cesm/cesmdata/cseg/runs/cesm2_0/f.cam6_3_160.FMTHIST_ne30.unstable_Ri_mixing.001
On github: https://github.com/NCAR/amwg_dev/tree/f.cam6_3_160.FMTHIST_ne30.unstable_Ri_mixing.001

Sandbox:
Locally (if still available): /glade/work/hannay/cesm_tags/cam6_3_160
On github: https://github.com/ESCOMP/CAM/tree/cam6_3_160

Diagnostics:
ADF diags (if available) https://webext.cgd.ucar.edu/FMTHIST/f.cam6_3_160.FMTHIST_ne30.unstable_Ri_mixing.001/atm/

Output:
Raw output on derecho (if still available) /glade/derecho/scratch/hannay/archive/f.cam6_3_160.FMTHIST_ne30.unstable_Ri_mixing.001/
CAM Climos: /glade/campaign/cgd/amp/amwg/climo/f.cam6_3_160.FMTHIST_ne30.unstable_Ri_mixing.001

Contacts: @cecilehannay @JulioTBacmeister @PeterHjortLauritzen

Extra details:

See run request (https://github.com/NCAR/amwg_dev/issues/537)

cecilehannay commented 4 months ago

@PeterHjortLauritzen

@PeterHjortLauritzen: I am getting the error:

ERROR: negative moist layer thickness.  timestep or remap time too large

RIght now we have

 se_hypervis_subcycle       =  3 
 se_nsplit      =   2  
 se_rsplit      =  3

Do we want to change these?

PeterHjortLauritzen commented 3 months ago

OK (it ran 25297 time-steps stably before crashing!). Lets introduce a slight del4 sponge:

se_sponge_del4_lev = 4 se_sponge_del4_nu_fac = 3.375 se_sponge_del4_nu_div_fac = 3.375

(why 3.375? ... maximum value without having to increase se_hypervis_subcycle)

NCAR / amwg_dev

f.cam6_3_160.FMTHIST_ne30.unstable_Ri_mixing.001 #541