Helfand Monin-Obukhov derivatives for GEOSldas offline runs #918
Conversation
biljanaorescanin
added
enhancement
| PSL = PSMB * (1. - (DZE*MAPL_GRAV)/(MAPL_RGAS*(TA+TC(:,N)) ) ) / & | ||
| (1. + (DZE*MAPL_GRAV)/(MAPL_RGAS*(TA+TC(:,N)) ) ) | ||
|
|
||
| CALL helfsurface( UWINDLMTILE,VWINDLMTILE,TA,TC(:,N),QA,QC(:,N),PSL,PSMB,Z0T(:,N),lai, & |
There was a problem hiding this comment.
is this intentional to call this function three times? or it should be
if (incl_HelfandMo_extra_derives ==0) then
call once
else if(incl_HelfandMO_extra_derivs ==1) then
call your addition ( twice)
endif
| t2m,q2m,u2m,v2m,t10m,q10m,u10m,v10m,u50m,v50m,CHOOSEZ0) | ||
|
|
||
| ! for helfsurface | ||
| if(CATCH_INTERNAL_STATE%CHOOSEMOSFC==1 .and. incl_HelfandMO_extra_derivs ==1) then |
There was a problem hiding this comment.
CATCH_INTERNAL_STATE%CHOOSEMOSFC==1 is redundant here
|
The crash is due to small_TC which is way to small. TC is around 270. and the smal_tc = 0.0001 is just roundoff error. This eventually leads to NaN of TC and crash. It should work if small_TC=0.1 . But I think the better way is to use relative perturb factor, so each tile has the same perturbation. When I choose factor = 0.01, the program has run for 1 year. /gpfsm/dnb34/wjiang/LDAS_exp/model/exp_helfand_sles15_small_factor |
- changed default to NOT use extra derivs - renamed resource variable for extra derivs - reordered and moved variable declarations etc for consistency
|
Replaced with new PR #999 |
3 participants
This PR adds Helfand M-O derivatives needed for GEOSldas offline runs.
Almost all GEOSldas offline runs are run using Louis scheme choice due to the lack of these derivatives for Helfand M-O option. This is not in sink to what AGCM uses which is Helfand always.
Surface layer turbulence scheme0 : Louis (MERRA, Fortuna-DAS, SMAP NRv4/4.1/5/7.2/8.1/9.1/10.0)1 : Helfand Monin-Obukhov (Fortuna-AR5, Ganymed, Heracles, Icarus-3_2, MERRA-2)GEOSagcm=>CHOOSEMOSFC: 1
Derivatives added are:
This PR is NOT zero diff for GEOSldas runs with CHOOSEMOSFC: 1 but is zero diff for AGCM runs.
NOTE: Long test runs are in progress I will update PR once testing is done.