Changeset 337 for trunk/LMDZ.MARS/README

Timestamp:

Nov 2, 2011, 5:32:28 PM (13 years ago)

Author:

acolaitis

Message:

10/10/2011 == AC

*
This commit aims at increasing the thermals speed, especially for large tracer number configurations. The idea behind this commit is to advect non-active conserved variables outside of the sub-timestep of the thermals. Because these variables are not used in thermals computation, we can decouple them:

momentum: can be decoupled because we assume a constant ratio between horizontal velocity in alimentation layer and maximum vertical velocity in the thermals.

tracer: can be decoupled because we do not take condensation of any tracer into account and hence do not liberate latent heat nor form clouds in the thermals.

temperature: cannot be decoupled (of course)
*

D 336 libf/phymars/thermcell_dqupdown.F90
^{---------------- Deleted and replaced by a simpler version. Notes about downdraft advection are still available from revision 336 of SVN in thermcell_dqupdown.}

A 0 libf/phymars/thermcell_dqup.F90
^{---------------- New upward advection for tracers and momentum in thermals. Several changes are done compared to the old approach:}

• Updraft quantities are not longer computed by making hypothesis on the amount of advected air.
  • In general, the formalism for updraft computation is much simpler and clearer.
• Tracer tendancies are no longer computed using the conservation equation. Instead, we use the divergence of an approximated turbulent flux of the concerned quantity, where downdraft are also neglected.

M 336 libf/phymars/thermcell_main_mars.F90
<sup>---------------- The Main does not call anymore thermcell_dqupdown, which it was doing 2+tracer_number times per subtimestep (140 times per physical step for a 2 tracer config)

M 336 libf/phymars/calltherm_mars.F90</sup>
---------------- Entrainment, detrainment and mass-fluxes are recomputed in the sub-timestep loop. Their final value after iterations is used by the new advection routine to compute tracer and momentum fluxes.

* Results

• Conservation of tracers has been assessed over 1 yr in 1D and found to be comparable to that obtained with the simple convective adjustment. (it actually seems to be better by a factor of 10%!)
• GCM speed-up is of about 20% compared to the old thermal configuration, for a 2 tracer case.
• Advection of sharp tracer profiles has been successfully observed, similar to the old method.

File:

• trunk/LMDZ.MARS/README (modified) (1 diff)

trunk/LMDZ.MARS/README

@@ -1120 +1120 @@
 --> Added settings which works for 35 levels in inidissip.F according to FL. Commented for the moment.
 --> Checked compilation and run, looks fine. Note that 'jmars.20101220' is needed.
+
+== 10/10/2011 == AC
+
+***********
+This commit aims at increasing the thermals speed, especially for large tracer number configurations. The idea behind this commit is to advect non-active conserved variables outside of the sub-timestep of the thermals. Because these variables are not used in thermals computation, we can decouple them:
+
+momentum: can be decoupled because we assume a constant ratio between horizontal velocity in alimentation layer and maximum vertical velocity in the thermal
+s.
+
+tracer: can be decoupled because we do not take condensation of any tracer into account and hence do not liberate latent heat nor form clouds in the thermal
+s.
+
+temperature: cannot be decoupled (of course)
+***********
+
+D             336   libf/phymars/thermcell_dqupdown.F90
+^----------------   Deleted and replaced by a simpler version. Notes about downdraft advection are still available from revision 336 of SVN in thermcell_dqu
+pdown.
+
+A               0   libf/phymars/thermcell_dqup.F90
+^----------------   New upward advection for tracers and momentum in thermals. Several changes are done compared to the old approach:
+                          - Updraft quantities are not longer computed by making hypothesis on the amount of advected air.
+                          - In general, the formalism for updraft computation is much simpler and clearer.
+                          - Tracer tendancies are no longer computed using the conservation equation. Instead, we use the divergence
+                          of an approximated turbulent flux of the concerned quantity, where downdraft are also neglected.
+
+M             336   libf/phymars/thermcell_main_mars.F90
+^----------------   The Main does not call anymore thermcell_dqupdown, which it was doing 2+tracer_number times per subtimestep (140 times per physical step for a 2 tracer config)
+
+M             336   libf/phymars/calltherm_mars.F90
+^----------------   Entrainment, detrainment and mass-fluxes are recomputed in  the sub-timestep loop. Their final value after iterations is used by the new
+ advection routine to compute tracer and momentum fluxes.
+
+*********** Results
+
+- Conservation of tracers has been assessed over 1 yr in 1D and found to be comparable to that obtained with the simple convective adjustment. (it actually
+seems to be better by a factor of 10%!)
+- GCM speed-up is of about 20% compared to the old thermal configuration, for a 2 tracer case.
+- Advection of sharp tracer profiles has been successfully observed, similar to the old method.
+