1 | SUBROUTINE timestep |
---|
2 | |
---|
3 | !------------------------------------------------------------------------------! |
---|
4 | ! Actual revisions: |
---|
5 | ! ----------------- |
---|
6 | ! |
---|
7 | ! |
---|
8 | ! Former revisions: |
---|
9 | ! ----------------- |
---|
10 | ! $Log: timestep.f90,v $ |
---|
11 | ! Revision 1.21 2006/02/23 12:59:44 raasch |
---|
12 | ! nt_anz renamed current_timestep_number |
---|
13 | ! |
---|
14 | ! Revision 1.20 2005/04/23 09:45:38 raasch |
---|
15 | ! fcl_factor renamed cfl_factor |
---|
16 | ! |
---|
17 | ! Revision 1.19 2005/03/26 21:11:53 raasch |
---|
18 | ! In case of non-cyclic lateral boundary conditions, the increased diffusivity |
---|
19 | ! in the lateral damping layer is regarded in the minimum timestep calculation. |
---|
20 | ! |
---|
21 | ! Revision 1.18 2004/01/30 10:40:02 raasch |
---|
22 | ! Timestep increment limitation in case of Runge-Kutta removed |
---|
23 | ! |
---|
24 | ! Revision 1.17 2004/01/28 15:34:03 raasch |
---|
25 | ! Setting of steering factors for the prognostic equations moved to new |
---|
26 | ! routine timestep_scheme_steering now called in routine leap_frog. |
---|
27 | ! |
---|
28 | ! Revision 1.16 2003/03/16 09:49:47 raasch |
---|
29 | ! Two underscores (_) are placed in front of all define-strings |
---|
30 | ! |
---|
31 | ! Revision 1.15 2002/12/19 16:19:54 raasch |
---|
32 | ! Two mpi_allreduce calls for determining mean horizontal velocity reduced to |
---|
33 | ! one call. Calculation of minimum timestep for diffusion optimized by |
---|
34 | ! determining minimum grid spacing in a single k loop. |
---|
35 | ! |
---|
36 | ! Revision 1.14 2001/03/30 07:55:12 raasch |
---|
37 | ! Translation of remaining German identifiers (variables, subroutines, etc.) |
---|
38 | ! |
---|
39 | ! Revision 1.13 2001/01/30 22:10:11 letzel |
---|
40 | ! All comments translated into English. |
---|
41 | ! |
---|
42 | ! Revision 1.12 2001/01/22 08:17:33 raasch |
---|
43 | ! Module test_variables removed |
---|
44 | ! |
---|
45 | ! Revision 1.11 2000/01/10 10:10:30 raasch |
---|
46 | ! Translationsgeschwindigkeit fuer Galilei-Transformation wird nur noch |
---|
47 | ! bedingt berechnet |
---|
48 | ! |
---|
49 | ! Revision 1.10 1999/02/05 09:19:06 raasch |
---|
50 | ! Zeitschrittsteuerung beruecksichtigt jetzt auch Wahl des reinen Euler- |
---|
51 | ! Verfahrens |
---|
52 | ! |
---|
53 | ! Revision 1.9 1998/09/22 17:30:41 raasch |
---|
54 | ! Ausfuehrlichere Informationen bei Abbruch durch Unterschreitung des |
---|
55 | ! minimalen Zeitschritts |
---|
56 | ! |
---|
57 | ! Revision 1.8 1998/07/06 12:33:52 raasch |
---|
58 | ! Korrektur der ermittelten horizontalen Geschwindigkeitsmaxima, falls diese |
---|
59 | ! bei k=0 gefunden wurden (wegen Spiegelungsrandbedingung treten dort negative |
---|
60 | ! Werte auf, die bei Galilei-Transformation den Zeitschritt erheblich |
---|
61 | ! einschraenken). + USE test_variables |
---|
62 | ! |
---|
63 | ! Revision 1.7 1998/04/21 15:57:46 raasch |
---|
64 | ! Implementierung der Galilei-Transformation |
---|
65 | ! |
---|
66 | ! Revision 1.6 1998/03/25 13:57:20 raasch |
---|
67 | ! dt in dt_3d umbenannt |
---|
68 | ! |
---|
69 | ! Revision 1.5 1998/03/09 16:25:10 raasch |
---|
70 | ! Schoenheitskorrekturen |
---|
71 | ! |
---|
72 | ! Revision 1.4 1998/03/03 08:02:15 raasch |
---|
73 | ! Leap-Frog auch nach Zeitschrittaenderung erlaubt |
---|
74 | ! |
---|
75 | ! Revision 1.3 1997/09/12 06:30:27 raasch |
---|
76 | ! Einbau des Diffusions-Kriteriums |
---|
77 | ! |
---|
78 | ! Revision 1.2 1997/08/26 06:34:55 raasch |
---|
79 | ! Fehler bei w_max_ijk - Index |
---|
80 | ! |
---|
81 | ! Revision 1.1 1997/08/11 06:26:19 raasch |
---|
82 | ! Initial revision |
---|
83 | ! |
---|
84 | ! |
---|
85 | ! Description: |
---|
86 | ! ------------ |
---|
87 | ! Compute the time step under consideration of the FCL and diffusion criterion. |
---|
88 | !------------------------------------------------------------------------------! |
---|
89 | |
---|
90 | USE arrays_3d |
---|
91 | USE control_parameters |
---|
92 | USE cpulog |
---|
93 | USE grid_variables |
---|
94 | USE indices |
---|
95 | USE interfaces |
---|
96 | USE pegrid |
---|
97 | USE statistics |
---|
98 | |
---|
99 | IMPLICIT NONE |
---|
100 | |
---|
101 | INTEGER :: i, j, k |
---|
102 | |
---|
103 | REAL :: div, dt_diff, dt_diff_l, dt_u, dt_v, dt_w, percent_change, & |
---|
104 | u_gtrans_l, value, v_gtrans_l |
---|
105 | |
---|
106 | REAL, DIMENSION(2) :: uv_gtrans, uv_gtrans_l |
---|
107 | REAL, DIMENSION(nzb+1:nzt) :: dxyz2_min |
---|
108 | |
---|
109 | |
---|
110 | CALL cpu_log( log_point(12), 'calculate_timestep', 'start' ) |
---|
111 | |
---|
112 | ! |
---|
113 | !-- Determine the maxima of the velocity components. |
---|
114 | CALL global_min_max( nzb, nzt+1, nys-1, nyn+1, nxl-1, nxr+1, u, 'abs', & |
---|
115 | u_max, u_max_ijk ) |
---|
116 | CALL global_min_max( nzb, nzt+1, nys-1, nyn+1, nxl-1, nxr+1, v, 'abs', & |
---|
117 | v_max, v_max_ijk ) |
---|
118 | CALL global_min_max( nzb, nzt+1, nys-1, nyn+1, nxl-1, nxr+1, w, 'abs', & |
---|
119 | w_max, w_max_ijk ) |
---|
120 | |
---|
121 | ! |
---|
122 | !-- In case maxima of the horizontal velocity components have been found at the |
---|
123 | !-- bottom boundary (k=nzb), the corresponding maximum at level k=1 is chosen |
---|
124 | !-- if the Dirichlet-boundary condition ('mirror') has been set. This is |
---|
125 | !-- necessary, because otherwise in case of Galilei-transform a far too large |
---|
126 | !-- velocity (having the respective opposite sign) would be used for the time |
---|
127 | !-- step determination (almost double the mean flow velocity). |
---|
128 | IF ( ibc_uv_b == 0 ) THEN |
---|
129 | IF ( u_max_ijk(1) == nzb ) THEN |
---|
130 | u_max = -u_max |
---|
131 | u_max_ijk(1) = nzb + 1 |
---|
132 | ENDIF |
---|
133 | IF ( v_max_ijk(1) == nzb ) THEN |
---|
134 | v_max = -v_max |
---|
135 | v_max_ijk(1) = nzb + 1 |
---|
136 | ENDIF |
---|
137 | ENDIF |
---|
138 | |
---|
139 | ! |
---|
140 | !-- In case of Galilei-transform not using the geostrophic wind as translation |
---|
141 | !-- velocity, compute the volume-averaged horizontal velocity components, which |
---|
142 | !-- will then be subtracted from the horizontal wind for the time step and |
---|
143 | !-- horizontal advection routines. |
---|
144 | IF ( galilei_transformation .AND. .NOT. use_ug_for_galilei_tr ) THEN |
---|
145 | IF ( flow_statistics_called ) THEN |
---|
146 | ! |
---|
147 | !-- Horizontal averages already existent, just need to average them |
---|
148 | !-- vertically. |
---|
149 | u_gtrans = 0.0 |
---|
150 | v_gtrans = 0.0 |
---|
151 | DO k = nzb+1, nzt |
---|
152 | u_gtrans = u_gtrans + hom(k,1,1,0) |
---|
153 | v_gtrans = v_gtrans + hom(k,1,2,0) |
---|
154 | ENDDO |
---|
155 | u_gtrans = u_gtrans / REAL( nzt - nzb ) |
---|
156 | v_gtrans = v_gtrans / REAL( nzt - nzb ) |
---|
157 | ELSE |
---|
158 | ! |
---|
159 | !-- Averaging over the entire model domain. |
---|
160 | uv_gtrans_l = 0.0 |
---|
161 | DO i = nxl, nxr |
---|
162 | DO j = nys, nyn |
---|
163 | DO k = nzb+1, nzt |
---|
164 | uv_gtrans_l(1) = uv_gtrans_l(1) + u(k,j,i) |
---|
165 | uv_gtrans_l(2) = uv_gtrans_l(2) + v(k,j,i) |
---|
166 | ENDDO |
---|
167 | ENDDO |
---|
168 | ENDDO |
---|
169 | uv_gtrans_l = uv_gtrans_l / REAL( (nxr-nxl+1)*(nyn-nys+1)*(nzt-nzb) ) |
---|
170 | #if defined( __parallel ) |
---|
171 | CALL MPI_ALLREDUCE( uv_gtrans_l, uv_gtrans, 2, MPI_REAL, MPI_SUM, & |
---|
172 | comm2d, ierr ) |
---|
173 | u_gtrans = uv_gtrans(1) / REAL( numprocs ) |
---|
174 | v_gtrans = uv_gtrans(2) / REAL( numprocs ) |
---|
175 | #else |
---|
176 | u_gtrans = uv_gtrans_l(1) |
---|
177 | v_gtrans = uv_gtrans_l(2) |
---|
178 | #endif |
---|
179 | ENDIF |
---|
180 | ENDIF |
---|
181 | |
---|
182 | IF ( .NOT. dt_fixed ) THEN |
---|
183 | ! |
---|
184 | !-- Variable time step: |
---|
185 | ! |
---|
186 | !-- For each component, compute the maximum time step according to the |
---|
187 | !-- FCL-criterion. |
---|
188 | dt_u = dx / ( ABS( u_max - u_gtrans ) + 1.0E-10 ) |
---|
189 | dt_v = dy / ( ABS( v_max - v_gtrans ) + 1.0E-10 ) |
---|
190 | dt_w = dzu(MAX( 1, w_max_ijk(1) )) / ( ABS( w_max ) + 1.0E-10 ) |
---|
191 | |
---|
192 | ! |
---|
193 | !-- Compute time step according to the diffusion criterion. |
---|
194 | !-- First calculate minimum grid spacing which only depends on index k |
---|
195 | !-- Note: also at k=nzb+1 a full grid length is being assumed, although |
---|
196 | !-- in the Prandtl-layer friction term only dz/2 is used. |
---|
197 | !-- Experience from the old model seems to justify this. |
---|
198 | dt_diff_l = 999999.0 |
---|
199 | |
---|
200 | DO k = nzb+1, nzt |
---|
201 | dxyz2_min(k) = MIN( dx2, dy2, dzu(k)*dzu(k) ) * 0.125 |
---|
202 | ENDDO |
---|
203 | |
---|
204 | !$OMP PARALLEL private(i,j,k,value) reduction(MIN: dt_diff_l) |
---|
205 | !$OMP DO |
---|
206 | DO i = nxl, nxr |
---|
207 | DO j = nys, nyn |
---|
208 | DO k = nzb+1, nzt |
---|
209 | value = dxyz2_min(k) / ( MAX( kh(k,j,i), km(k,j,i) ) + 1E-20 ) |
---|
210 | |
---|
211 | dt_diff_l = MIN( value, dt_diff_l ) |
---|
212 | ENDDO |
---|
213 | ENDDO |
---|
214 | ENDDO |
---|
215 | !$OMP END PARALLEL |
---|
216 | #if defined( __parallel ) |
---|
217 | CALL MPI_ALLREDUCE( dt_diff_l, dt_diff, 1, MPI_REAL, MPI_MIN, comm2d, & |
---|
218 | ierr ) |
---|
219 | #else |
---|
220 | dt_diff = dt_diff_l |
---|
221 | #endif |
---|
222 | |
---|
223 | ! |
---|
224 | !-- In case of non-cyclic lateral boundaries, the diffusion time step |
---|
225 | !-- may be further restricted by the lateral damping layer (damping only |
---|
226 | !-- along x and y) |
---|
227 | IF ( bc_lr /= 'cyclic' ) THEN |
---|
228 | dt_diff = MIN( dt_diff, 0.125 * dx2 / ( km_damp_max + 1E-20 ) ) |
---|
229 | ELSEIF ( bc_ns /= 'cyclic' ) THEN |
---|
230 | dt_diff = MIN( dt_diff, 0.125 * dy2 / ( km_damp_max + 1E-20 ) ) |
---|
231 | ENDIF |
---|
232 | |
---|
233 | ! |
---|
234 | !-- The time step is the minimum of the 3 components and the diffusion time |
---|
235 | !-- step minus a reduction to be on the safe side. Factor 0.5 is necessary |
---|
236 | !-- since the leap-frog scheme always progresses by 2 * delta t. |
---|
237 | !-- The user has to set the cfl_factor small enough to ensure that the |
---|
238 | !-- divergences do not become too large. |
---|
239 | !-- The time step must not exceed the maximum allowed value. |
---|
240 | IF ( timestep_scheme(1:5) == 'runge' ) THEN |
---|
241 | dt_3d = cfl_factor * MIN( dt_diff, dt_u, dt_v, dt_w ) |
---|
242 | ELSE |
---|
243 | dt_3d = cfl_factor * 0.5 * MIN( dt_diff, dt_u, dt_v, dt_w ) |
---|
244 | ENDIF |
---|
245 | dt_3d = MIN( dt_3d, dt_max ) |
---|
246 | |
---|
247 | ! |
---|
248 | !-- Remember the restricting time step criterion for later output. |
---|
249 | IF ( dt_diff > MIN( dt_u, dt_v, dt_w ) ) THEN |
---|
250 | timestep_reason = 'A' |
---|
251 | ELSE |
---|
252 | timestep_reason = 'D' |
---|
253 | ENDIF |
---|
254 | |
---|
255 | ! |
---|
256 | !-- Set flag if the time step becomes too small. |
---|
257 | IF ( dt_3d < ( 0.00001 * dt_max ) ) THEN |
---|
258 | stop_dt = .TRUE. |
---|
259 | IF ( myid == 0 ) THEN |
---|
260 | PRINT*,'+++ time_step: Time step has reached minimum limit.' |
---|
261 | PRINT*,' dt = ', dt_3d, ' s Simulation is terminated.' |
---|
262 | PRINT*,' old_dt = ', old_dt, ' s' |
---|
263 | PRINT*,' dt_u = ', dt_u, ' s' |
---|
264 | PRINT*,' dt_v = ', dt_v, ' s' |
---|
265 | PRINT*,' dt_w = ', dt_w, ' s' |
---|
266 | PRINT*,' dt_diff = ', dt_diff, ' s' |
---|
267 | PRINT*,' u_max = ', u_max, ' m/s k=', u_max_ijk(1), & |
---|
268 | ' j=', u_max_ijk(2), ' i=', u_max_ijk(3) |
---|
269 | PRINT*,' v_max = ', v_max, ' m/s k=', v_max_ijk(1), & |
---|
270 | ' j=', v_max_ijk(2), ' i=', v_max_ijk(3) |
---|
271 | PRINT*,' w_max = ', w_max, ' m/s k=', w_max_ijk(1), & |
---|
272 | ' j=', w_max_ijk(2), ' i=', w_max_ijk(3) |
---|
273 | ENDIF |
---|
274 | ENDIF |
---|
275 | |
---|
276 | ! |
---|
277 | !-- Ensure a smooth value (two significant digits) of the timestep. For |
---|
278 | !-- other schemes than Runge-Kutta, the following restrictions appear: |
---|
279 | !-- The current timestep is only then changed, if the change relative to |
---|
280 | !-- its previous value exceeds +5 % or -2 %. In case of a timestep |
---|
281 | !-- reduction, at least 30 iterations have to be performed before a timestep |
---|
282 | !-- enlargement is permitted again. |
---|
283 | percent_change = dt_3d / old_dt - 1.0 |
---|
284 | IF ( percent_change > 0.05 .OR. percent_change < -0.02 .OR. & |
---|
285 | timestep_scheme(1:5) == 'runge' ) THEN |
---|
286 | |
---|
287 | ! |
---|
288 | !-- Time step enlargement by no more than 2 %. |
---|
289 | IF ( percent_change > 0.0 .AND. simulated_time /= 0.0 .AND. & |
---|
290 | timestep_scheme(1:5) /= 'runge' ) THEN |
---|
291 | dt_3d = 1.02 * old_dt |
---|
292 | ENDIF |
---|
293 | |
---|
294 | ! |
---|
295 | !-- A relatively smooth value of the time step is ensured by taking |
---|
296 | !-- only the first two significant digits. |
---|
297 | div = 1000.0 |
---|
298 | DO WHILE ( dt_3d < div ) |
---|
299 | div = div / 10.0 |
---|
300 | ENDDO |
---|
301 | dt_3d = NINT( dt_3d * 100.0 / div ) * div / 100.0 |
---|
302 | |
---|
303 | ! |
---|
304 | !-- Now the time step can be adjusted. |
---|
305 | IF ( percent_change < 0.0 .OR. timestep_scheme(1:5) == 'runge' ) & |
---|
306 | THEN |
---|
307 | ! |
---|
308 | !-- Time step reduction. |
---|
309 | old_dt = dt_3d |
---|
310 | dt_changed = .TRUE. |
---|
311 | ELSE |
---|
312 | ! |
---|
313 | !-- For other timestep schemes , the time step is only enlarged |
---|
314 | !-- after at least 30 iterations since the previous time step |
---|
315 | !-- change or, of course, after model initialization. |
---|
316 | IF ( current_timestep_number >= last_dt_change + 30 .OR. & |
---|
317 | simulated_time == 0.0 ) THEN |
---|
318 | old_dt = dt_3d |
---|
319 | dt_changed = .TRUE. |
---|
320 | ELSE |
---|
321 | dt_3d = old_dt |
---|
322 | dt_changed = .FALSE. |
---|
323 | ENDIF |
---|
324 | |
---|
325 | ENDIF |
---|
326 | ELSE |
---|
327 | ! |
---|
328 | !-- No time step change since the difference is too small. |
---|
329 | dt_3d = old_dt |
---|
330 | dt_changed = .FALSE. |
---|
331 | ENDIF |
---|
332 | |
---|
333 | IF ( dt_changed ) last_dt_change = current_timestep_number |
---|
334 | |
---|
335 | ENDIF |
---|
336 | |
---|
337 | CALL cpu_log( log_point(12), 'calculate_timestep', 'stop' ) |
---|
338 | |
---|
339 | |
---|
340 | END SUBROUTINE timestep |
---|