Changeset 561 for trunk

Timestamp:

Mar 5, 2012, 6:03:06 PM (13 years ago)

Author:

acolaitis

Message:

Thermals. Added a second instance of subtimestep variation depending on global timestep. Without active clouds without active dust, going from 48 to 96 timestep per days can be compensated in the thermals by cutting the subtimestep from 35 to 10. Resulting thermal structure is better, the model is also observed to be more stable. In this configuration, the cost of doubling the timestep is of 10% only.

File:

• trunk/LMDZ.MARS/libf/phymars/calltherm_interface.F90 (modified) (2 diffs)

trunk/LMDZ.MARS/libf/phymars/calltherm_interface.F90

@@ -77 +77 @@
       REAL d_u_the(ngridmx,nlayermx),d_v_the(ngridmx,nlayermx)
       REAL dq2_the(ngridmx,nlayermx)
-      INTEGER isplit,nsplit_thermals
-      REAL r_aspect_thermals
+      INTEGER isplit
+      INTEGER,SAVE :: nsplit_thermals
+      REAL, SAVE :: r_aspect_thermals
       REAL fact
       REAL zfm_therm(ngridmx,nlayermx+1),zdt
@@ -191 +192 @@
 ! **********************************************************************
 
-!         r_aspect_thermals     ! ultimately conrols the amount of mass going through the thermals
-                                ! decreasing it increases the thermals effect. Tests at gcm resolution
-                                ! shows that too low values destabilize the model
-                                ! when changing this value, one should check that the surface layer model
-                                ! outputs the correct Cd*u and Ch*u through changing the gustiness coefficient beta
-
+!         r_aspect_thermals     ! Mainly control the shape of the temperature profile
+                                ! in the surface layer. Decreasing it goes toward
+                                ! a convective-adjustment like profile.
+!         nsplit_thermals       ! Sub-timestep for the thermals. Very dependant on the
+                                ! chosen timestep for the radiative transfer.
+                                ! It is recommended to run with 96 timestep per day and 
+                                ! iradia = 1., configuration in which thermals can run
+                                ! very well with a sub-timestep of 10.
+         IF (firstcall) THEN
 #ifdef MESOSCALE
-         !! valid for timesteps < 200s
-         nsplit_thermals=2
-         r_aspect_thermals=0.7
+            !! valid for timesteps < 200s
+            nsplit_thermals=4
+            r_aspect_thermals=0.7
 #else
-         nsplit_thermals=35
-         r_aspect_thermals=1.5
+            IF ((ptimestep .lt. 3699.*24./96.) .and. (iradia .eq. 1)) THEN
+               nsplit_thermals=10
+            ELSE
+               nsplit_thermals=35
+            ENDIF
+            r_aspect_thermals=1.5
 #endif
-         call getin("nsplit_thermals",nsplit_thermals)
-         call getin("r_aspect_thermals",r_aspect_thermals)
+            call getin("nsplit_thermals",nsplit_thermals)
+            call getin("r_aspect_thermals",r_aspect_thermals)
+         ENDIF
 
 ! **********************************************************************