Version 61 (modified by fabien.roquet, 9 years ago) (diff)

Last edited Timestamp?

Author : Gurvan Madec & Fabien Roquet

Ticket : #927

Branch : 2012/dev_r3309_LOCEAN12_Ediag

LOCEAN .12 - Energy diagnostics

Motivation: output 3D trends of tracers, momentum, kinetic energy and potential energy.
Status : the extraction of trends terms exists, but not the 3D output of the trends
Main tasks :

(1) implement the 3D output of tracers and momentum trends using iom_put
(2) compute and output the 3D trends of PE and KE
(3) validatation + documentation

Science Reviewer: NOCS guy?
System Reviewer: NOCS guy?
Deadline: spring 2012
Priority: high
Depends on: gurvan disponibilities
Principal Investigator : Gurvan Madec and Fabien Roquet (gurvan.madec@…)

Detail of the implementation

Step I : add the 3D output

see revision 3316

trdmod_oce module and namtrd namelist

logical flags added in namlist namtrd which now controls what is done with the trends.
All the types of treatment of a given trend are available at the same time. The memory requirement will only increase due to the time averaged arrays defined in IOM.

LOGICAL , PUBLIC  ::   ln_dyn_trd   = .FALSE.        !: (T) 3D momentum             trends or (F) not    
LOGICAL , PUBLIC  ::   ln_tra_trd   = .FALSE.        !: (T) 3D tracer               trends or (F) not    
LOGICAL , PUBLIC  ::   ln_PE_trd    = .FALSE.        !: (T) 3D Potential Energy     trends or (F) not    
LOGICAL , PUBLIC  ::   ln_KE_trd    = .FALSE.        !: (T) 3D Kinetic   Energy     trends or (F) not    
LOGICAL , PUBLIC  ::   ln_vor_trd   = .FALSE.        !: (T) 3D barotropic vorticity trends or (F) not    
LOGICAL , PUBLIC  ::   ln_glo_trd   = .FALSE.        !: (T) global domain averaged diag for T, T!^2, KE, and PE    
LOGICAL , PUBLIC  ::   ln_dyn_mld   = .FALSE.        !: (T) 2D tracer   trends averaged over the mixed layer     
LOGICAL , PUBLIC  ::   ln_tra_mld   = .FALSE.        !: (T) 2D momentum trends averaged over the mixed layer

Add these new logical in the namelist. =⇒>> CAUTION only in the ORCA2_LIM directory
NB: here is the new name set in revision number 3318

trdtra module

Add a systematic mask of the trend.
Change the comments to better describe the purpose of this module. Its purpose is:

'TRA' case: to regroup T & S trends and send them to trd_mod, with, in case of advection, transform the incoming advective fluxes into advctive trend (U.grad[T])
'TRC' case: send trend to ted_mod_trc, with, in case of advection, transform the incoming advective fluxes into advective trend
all cases : mask the trend ( ==⇒>> PROBABLY add in the module a lbc_lnk so that the trend is defined everywhere )

dynadv_cen2 and _ubs modules

change jpdyn_trd_had into jpdyn_trd_keg. Now in flux form _keg corresponds to the horizontal advection trends and _rvo to the metric terms

dynnxt module

add the output using sum of the total dyn trend (except asselin time filter) ("utrd_tot", "vtrd_tot") and of the asselin time filter trend ("utrd_atf", "vtrd_atf") but with a shift by one time step

dynvor, trdvor and trdmod_oce modules

suppress the call to trd_mod in the jpdyn_trd_dat case (computation of beta.V) add add the calculation of beta.V term in 'trdvor' in jpvor_pvo case. And obviously suppress jpdyn_trd_dat from trdmod_oce
Also suppress the jpdyn_trd_had case horizontal advection for the dynamics is 'keg' + 'vor' ; in case of flux form, 'had' is put in 'keg' and the metric terms is put in 'vor'
there is now only 10 trends on the dynamics instead of 12

trdmod module

1- introduce the new logical namelist parameters
2- introduce new subroutines : trd_budget : computation of the domain averaged T,T2, PE, KE trends formerly computes in trd_mod routine)
trd_3Diom: output of the 3D trends using IOM

trdicp module

add in trd_twr routine the computation of the vertical diffusive trend on T & S in case of iso-neutral diffusion (ln_traldf_iso=T). These trends ("ttrd_zdfp", "strd_zdfp") name zdfp for "PURE" vertical diffusion trends are output so that by difference with "zdf" trends we can access to the vertical contribution of the iso-neutral operator

iodef.xml file

add all the trends nick name : =⇒>> CAUTION only in the ORCA2_LIM directory

Step II : simplification of the structure

trdicp and trdicp_oce modules

see revision 3317

1- suppress trdicp_oce module (put required variables in trdmod_oce parameters)
2- use the jptra_trd_… instead of jpicp_… 3- move trd_budget subroutine from trdmod to trdicp.F90. Suppress trd_icp routine (i.e. trd_2d, trd_3d) as the work is now simply done in trd_budget.
4- rescan all dyn/tra trend indices. add jptra_trd_zdfp for "PURE Kz dissusive trend when ln_traldf_iso=T (see also changes in tranxt where a call to trd_tra is done just before the swap, so that PURE Kz trends can be diagnosed in tra_trd. See also xml file). 5- jptra_trd_nsr and jptra_trd_cdt are 3D trends as their incorporate both surface forcing AND runoff, the later being possibly spread in depth (=⇒>> probably to be changed)
6- suppress the key_trddyn key trdtra from trdicp

trdtra - trddyn - trdmod - trdvor modules : impact on TRA , DYN , TRD and almost all TOP_SRC modules

see revision 3318

Change the logic : split DYN and TRA trend diagnostics in separate modules

move the call to trd_… from the end of step to the select CASE(jp…_atf) i.e. when the last trend is send to trdtra or trddyn

Change the name of the trend indices from jptra_trd_xxx to jptra_xxx and from jpdyn_trd_yyy to jpdyn_yyy. This impact TRA,DYN, TRD but also all TOP_SRC

0- move trdmod_oce into trd_oce
1- trdvor: move the trdvor 1st part from trdmod to trevor module. Suppress the call of trd_vor from step (now called in trddyn for each trend, with the output done for the last trend (_atf).
2- create a trddyn that manage the distribution to iom, global mean, KE, vor, mld
3- trdtra in TRC case pre-process advective trends and then call trdmod_trc ; in TRA case: it regroups T & S trend in one and calla local routine to manage the distribution to iom, global mean, PE, mld
4- trdmod now contains only the trdmod_init routine. Rename all as trdini.F90 and ted_init subroutine 5- suppress all key_trd… keys (thanks to dynamical allocation) except for the top key: key_trdmld_trc. =⇒>> To be done

Step III : addition of 3D KE diagnostics

trdken module

see revision 3325

0- bug correction: in trddyn.F90 (ji,jj) =⇒> (:,:) in surface stress computation

1- create a trdken.F90 module which compute the KE trends at T-point and out them using IOM

TO BE DONE diagnose the KE local dissipation and associated diffusive fluxes ; diagnose bottom friction in all cases (implicit/explicit) ; in flux form, remove the KE*divh(U) from the H/V advection trends

NB: Missing the KE trends in the xml files ; Bug in trdtra: 2 call to ted_tra_mng in case of jptra_zdfp. Both corrected in Step IV release (see below)

Step IV : addition of 3D PE diagnostics

trdpen module

see revision 3326 and revision 3327 0- correct a bug in trdtra and trdken

1- add xml name for both PE and KE 2- create a trdpen.F90 module which compute the PE trends at T-point and out them using IOM

3- introduce in eosbn2 a subroutine, eos_drau_dtds, that computes the partial derivative off in situ density with respect to T and S (NB: only coded the 2 linear eos, TO BE DONE for the UNESCO eos (nn_eos=0)

4- correct a few bugs and adopt almost systematic CASE structure in trdtra (see revision 3327)

NB: to be verified: the addition of a ssh term in PE for NOT lk_vvl case

NB: runoff put as a source term in the divergence computation since v3.3 (see divcur), implication for source term in PE, but also tracer trends ? Think about that''''''

Pending issues :

atmospheric pressure gradient trend not taken into account (see dynspg.F90 To be done !!!'''

kpp non-local trend put in zdf trends !!! this will not work ! a additional trend term should be add To be done !!!'''

problems to be solved: vvl case for tracer sad trends ; flux form case for had (keg) and zad momentum trends

add separate modules for each option …

create the momentum diag over the ML

reshape trdtra so that T and S are treated separately in all ted routine (including mld diag…) create umask_i and vmask_i (2D) fields that mirror task_i field but for the velocity points

Changes done by Fabien R.

In trdtra.F90:

  • removed semi-colon on lines 118 and 267.
  • replaced ln_glo_trd by ln_PE_trd in line 239.
  • Moved wrk_alloc and wrk_dealloc in the IF (.NOT.lk_vvl ) structure.

In trddyn.F90:

  • Added wrk_alloc, wrk_dealloc and lbc_lnk in trd_dyn_iom.

Remove unused variables ztswu, ztswv.

  • Following norm in the code, utrd_bfr and vtrd_bfr should be filled only when ln_bfrimp=.FALSE. (non implicit bottom friction).
  • Implicit bottom friction is calculated at the end of each timestep, in dynzdf_imp.

The diagnostic of bottom friction in the implicit case is saved in separate variable names to avoid confusion: utrd_bfri and vtrd_bfri.

Similarly, diagnostics of wind stress inputs are provided in 2D fields utrd_tau and vtrd_tau.

In step.F90:

  • Added "IF( .NOT.ln_bfrimp)" before the "CALL dyn_bfr( kstp )" to clarify that dyn_bfr is called only if bottom friction is explicit, as proposed by gm.

In trdglo.F90:

  • In trd_glo_init, corrected definition for tvolt:

tvolt = tvolt + SUM( e1e2t(:,:) * fse3t(:,:,jk) * tmask(:,:,jk) * tmask_i(:,:) )

  • Treatment of bfr has been clarified: if bfr is implicit (ln_bfrimp=.TRUE.), the bottom friction term remain 0, as bottom friction is implicitly included in the vertical diffusion term.
  • Correction of outputs of dynamic trends and hke.
  • What is the test: "pressure gradient u2 = - 1/rau0" ? Identity is not verified in my configuration!
  • Wind stress is always zero (although not in the 3d diag provided in trddyn.F90)
  • TODO Once trddyn and trdtra are ready, use them for trdglo computations…
  • In glo_dyn_wri, missing zcof to compute density flux at w-point?!
  • Conversion is calculated as -g*div(rho*U)/rho0 !! Not sure what the link with -rho*g*w…

In trd_oce.F90

INTEGER, PUBLIC, PARAMETER ::   jptot_dyn  = 13     !: Total trend nb: change it when adding/removing one indice below    !                               ===============     !      
INTEGER, PUBLIC, PARAMETER ::   jpdyn_hpg  =  1     !: hydrostatic pressure gradient     
INTEGER, PUBLIC, PARAMETER ::   jpdyn_spg  =  2     !: surface     pressure gradient    
INTEGER, PUBLIC, PARAMETER ::   jpdyn_keg  =  3     !: kinetic energy gradient  or horizontal advection    
INTEGER, PUBLIC, PARAMETER ::   jpdyn_rvo  =  4     !: relative  vorticity      or metric term    
INTEGER, PUBLIC, PARAMETER ::   jpdyn_pvo  =  5     !: planetary vorticity    
INTEGER, PUBLIC, PARAMETER ::   jpdyn_zad  =  6     !: vertical advection    
INTEGER, PUBLIC, PARAMETER ::   jpdyn_ldf  =  7     !: horizontal diffusion       
INTEGER, PUBLIC, PARAMETER ::   jpdyn_zdf  =  8     !: vertical   diffusion    
INTEGER, PUBLIC, PARAMETER ::   jpdyn_bfr  =  9     !: bottom  stress     
INTEGER, PUBLIC, PARAMETER ::   jpdyn_atf  = 10     !: Asselin time filter    
INTEGER, PUBLIC, PARAMETER ::   jpdyn_tau  = 11     !: surface stress    
INTEGER, PUBLIC, PARAMETER ::   jpdyn_bfri = 12     !: implicit bottom friction (ln_bfrimp=.TRUE.)    
INTEGER, PUBLIC, PARAMETER ::   jpdyn_ken  = 13     !: use for calculation of KE
  • where jpdyn_tau and jpdyn_bfri are diagnostics of surface and bottom stress, respectively.
  • jpdyn_ken has been added for the calculation of KE in dynnxt.F90 (done just before the swap)

In trd_ken.F90:

  • l110: DO ji = 2, jpj should be DO ji = 2, jpi
  • Same in l133, l152 and l178.
  • In l183, replaced "ketrd_bfr" by "ketrd_bfri" to avoid confusion with "ketrd_bfr".
  • Important: contribution from atf is calculated with un and vn, after they have been updated.
    A small error is made: the contribution of ATF term at time t is recorded with trends taken at t-1.
  • In trd_ken_init, the 3 variables fse3x_n were replaced by their constant value fse3x for the non-vvl case.
  • Definition of KE (kinetic_energy in xml) at time t+½: KE(t+½)=rau0*u(t)*u(t+1)/2, calculated during dynnxt.F90, using ktrd=jpdyn_ken.
  • l. 109: multiplication by rau0 of zke to obtain KE trends in W/m3, and KE in J/m3.
  • remove r1_2_rau0 variable, which is no more used.
  • Add subroutine ken_conv_P2K to compute conversion rate.

In dynspg_flt:

  • Add a diagnostic of the explicit and implicit (due to filter) contributions to SPG.
  • sshimp: diagnostic of the ssh modification due to filter.

In dynnxt.F90

  • computation of z1_2dt must be put before the IF( ln_dyn_trd ) block (l. 194 and 195)
  • Add call to ken_conv_P2K (+USE trdken) and save conv rate in ketrd_convP2K (l. 200)

In eosbn2.F90

  • Add Vallis equation of state (Vallis 2006, p34-35): most simple form of EOS which accounts for thermobaricity, cabelling and compressibility.
  • STRANGE: nn_eos=0 case is a modified version of the Jackett and Mc Dougall? (1995) EOS!! Numerically close though…
  • Add original Jackett and McDougall? (1995) EOS (nn_eos=-1)
  • Buoyancy frequency calculation: exact formulation.
  • Same for alpha and beta
  • eos_alpha_beta added, which provides alpha and beta
  • compute_pen added to provide PE anomaly, and its partial derivatives with respect to T and S, needed in PE diagnostic (trdpen).
    PE anomaly is the anomaly of PE: (PE - rau0*gz)/(rau0*gz). For a linear case, PE anomaly is equal to density anomaly (nice isn't it?)
    Done for linear case, and Vallis case.
    Not ready yet for realistic cases… (discussion with Trevor is needed!)
    When z=0, PE anomaly is always defined asymptotically, converging toward the density anomaly value.
  • TODO add global arrays for alpha and beta, computed at the same time than bn2? Indeed, they are used in several places, and are needed to compute bn2 in any case…
  • BUG I don't understand why, but the linear case 1 is very slow, contrary to the others. Could it be coming from eos_alpbet?

In trdtra.F90

  • Add 'trdt(:,:,:) = ptrd(:,:,:) * tmask(:,:,:)' in CASE jptra_bbc in trd_tra (l. 103)

In trdpen.F90

  • replace petrd_ldf by petrd_zdf in l. 102
  • renamed petrd_for as petrd_nsr for consistency with trd_oce (l. 107)
  • remove lines on petrd_sad in CASE jptra_atf in trd_pen: petrd_sad cannot be called 2 times. Furthermore, contribution of asselin filter on ssh must be diagnosed in ssh_nxt, not in tra_nxt!!
  • Remove '( nn_eos == 0 .OR. nn_eos == 3 ) .AND. ' in the condition to call pen_ddt_dds in trd_pen (l. 79). These two coefficients are updates at each timestep whatever the eos (very quick anyway for the linear case!)
  • Add 'CALL iom_put( "potential_energy", zpe )' in trd_pen, just after the call to pen_ddt_dds.
  • Replace ptrdx by ptrdy and remove multiplication by fsde3w to compute pe trends in trd_pen (l. 88-89)

Bug during compilation:

  • make: Circular dynhpg.o ← trddyn.o dependency dropped

patched by compiling a first time with the line 'USE trddyn' commented in dynhpg.F90, then recompiling with 'USE trddyn' decommented.


  • In Annex A,
    • change label name Apdx_A_grad_p in Apdx_A_grad_p2 l. 418 and 430.
    • sign error for vertical advection in Eq. A.18 (tracer equation)
  • In Annex C, sign errors corrected in discretization of HPG (p. 293)
  • In Chap_TRA,
    • update part 5.8 on equation of state to include reference to Vallis2006+JM95 true formulation
  • In Chap_DYN, correction of title 6.1.2, and references


Testing could consider (where appropriate) other configurations in addition to NVTK].

NVTK Tested '''YES/NO'''
Other model configurations '''YES/NO'''
Processor configurations tested [ Enter processor configs tested here ]
If adding new functionality please confirm that the

New code doesn't change results when it is switched off

and ''works'' when switched on

(Answering UNSURE is likely to generate further questions from reviewers.)

'Please add further summary details here'

  • Processor configurations tested
  • etc——

Bit Comparability

Does this change preserve answers in your tested standard configurations (to the last bit) ? '''YES/NO '''
Does this change bit compare across various processor configurations. (1xM, Nx1 and MxN are recommended) '''YES/NO'''
Is this change expected to preserve answers in all possible model configurations? '''YES/NO'''
Is this change expected to preserve all diagnostics?

,,''Preserving answers in model runs does not necessarily imply preserved diagnostics. ''

If you answered '''NO''' to any of the above, please provide further details:

  • Which routine(s) are causing the difference?
  • Why the changes are not protected by a logical switch or new section-version
  • What is needed to achieve regression with the previous model release (e.g. a regression branch, hand-edits etc). If this is not possible, explain why not.
  • What do you expect to see occur in the test harness jobs?
  • Which diagnostics have you altered and why have they changed?Please add details here……..

System Changes

Does your change alter namelists? '''YES/NO '''
Does your change require a change in compiler options? '''YES/NO '''

If any of these apply, please document the changes required here…….


''Please ''summarize'' any changes in runtime or memory use caused by this change……''

IPR issues

Has the code been wholly (100%) produced by NEMO developers staff working exclusively on NEMO? '''YES/ NO '''

If No:

  • Identify the collaboration agreement details
  • Ensure the code routine header is in accordance with the agreement, (Copyright/Redistribution? etc).Add further details here if required……….