| [1] | 1 | SUBROUTINE timestep |
|---|
| 2 | |
|---|
| 3 | !------------------------------------------------------------------------------! |
|---|
| [258] | 4 | ! Current revisions: |
|---|
| [1] | 5 | ! ----------------- |
|---|
| [316] | 6 | ! |
|---|
| [667] | 7 | ! |
|---|
| [1] | 8 | ! Former revisions: |
|---|
| 9 | ! ----------------- |
|---|
| [3] | 10 | ! $Id: timestep.f90 668 2010-12-23 13:22:58Z suehring $ |
|---|
| [110] | 11 | ! |
|---|
| [668] | 12 | ! 667 2010-12-23 12:06:00Z suehring/gryschka |
|---|
| 13 | ! Exchange of terminate_coupled between ocean and atmosphere via PE0 |
|---|
| 14 | ! Minimum grid spacing dxyz2_min(k) is now calculated using dzw instead of dzu |
|---|
| 15 | ! |
|---|
| [623] | 16 | ! 622 2010-12-10 08:08:13Z raasch |
|---|
| 17 | ! optional barriers included in order to speed up collective operations |
|---|
| 18 | ! |
|---|
| [392] | 19 | ! 343 2009-06-24 12:59:09Z maronga |
|---|
| 20 | ! Additional timestep criterion in case of simulations with plant canopy |
|---|
| 21 | ! Output of messages replaced by message handling routine. |
|---|
| 22 | ! |
|---|
| [226] | 23 | ! 222 2009-01-12 16:04:16Z letzel |
|---|
| 24 | ! Implementation of a MPI-1 Coupling: replaced myid with target_id |
|---|
| 25 | ! Bugfix for nonparallel execution |
|---|
| 26 | ! |
|---|
| [110] | 27 | ! 108 2007-08-24 15:10:38Z letzel |
|---|
| 28 | ! modifications to terminate coupled runs |
|---|
| 29 | ! |
|---|
| [3] | 30 | ! RCS Log replace by Id keyword, revision history cleaned up |
|---|
| 31 | ! |
|---|
| [1] | 32 | ! Revision 1.21 2006/02/23 12:59:44 raasch |
|---|
| 33 | ! nt_anz renamed current_timestep_number |
|---|
| 34 | ! |
|---|
| 35 | ! Revision 1.1 1997/08/11 06:26:19 raasch |
|---|
| 36 | ! Initial revision |
|---|
| 37 | ! |
|---|
| 38 | ! |
|---|
| 39 | ! Description: |
|---|
| 40 | ! ------------ |
|---|
| 41 | ! Compute the time step under consideration of the FCL and diffusion criterion. |
|---|
| 42 | !------------------------------------------------------------------------------! |
|---|
| 43 | |
|---|
| 44 | USE arrays_3d |
|---|
| 45 | USE control_parameters |
|---|
| 46 | USE cpulog |
|---|
| 47 | USE grid_variables |
|---|
| 48 | USE indices |
|---|
| 49 | USE interfaces |
|---|
| 50 | USE pegrid |
|---|
| 51 | USE statistics |
|---|
| 52 | |
|---|
| 53 | IMPLICIT NONE |
|---|
| 54 | |
|---|
| 55 | INTEGER :: i, j, k |
|---|
| 56 | |
|---|
| [318] | 57 | REAL :: div, dt_diff, dt_diff_l, dt_plant_canopy, & |
|---|
| 58 | dt_plant_canopy_l, & |
|---|
| 59 | dt_plant_canopy_u, dt_plant_canopy_v, dt_plant_canopy_w, & |
|---|
| 60 | dt_u, dt_v, dt_w, lad_max, percent_change, & |
|---|
| [316] | 61 | u_gtrans_l, vabs_max, value, v_gtrans_l |
|---|
| [1] | 62 | |
|---|
| 63 | REAL, DIMENSION(2) :: uv_gtrans, uv_gtrans_l |
|---|
| 64 | REAL, DIMENSION(nzb+1:nzt) :: dxyz2_min |
|---|
| 65 | |
|---|
| [667] | 66 | |
|---|
| 67 | |
|---|
| [1] | 68 | CALL cpu_log( log_point(12), 'calculate_timestep', 'start' ) |
|---|
| 69 | |
|---|
| 70 | ! |
|---|
| 71 | !-- Determine the maxima of the velocity components. |
|---|
| [667] | 72 | CALL global_min_max( nzb, nzt+1, nysg, nyng, nxlg, nxrg, u, 'abs', & |
|---|
| [1] | 73 | u_max, u_max_ijk ) |
|---|
| [667] | 74 | CALL global_min_max( nzb, nzt+1, nysg, nyng, nxlg, nxrg, v, 'abs', & |
|---|
| [1] | 75 | v_max, v_max_ijk ) |
|---|
| [667] | 76 | CALL global_min_max( nzb, nzt+1, nysg, nyng, nxlg, nxrg, w, 'abs', & |
|---|
| [1] | 77 | w_max, w_max_ijk ) |
|---|
| 78 | |
|---|
| 79 | ! |
|---|
| 80 | !-- In case maxima of the horizontal velocity components have been found at the |
|---|
| 81 | !-- bottom boundary (k=nzb), the corresponding maximum at level k=1 is chosen |
|---|
| 82 | !-- if the Dirichlet-boundary condition ('mirror') has been set. This is |
|---|
| 83 | !-- necessary, because otherwise in case of Galilei-transform a far too large |
|---|
| 84 | !-- velocity (having the respective opposite sign) would be used for the time |
|---|
| 85 | !-- step determination (almost double the mean flow velocity). |
|---|
| 86 | IF ( ibc_uv_b == 0 ) THEN |
|---|
| 87 | IF ( u_max_ijk(1) == nzb ) THEN |
|---|
| 88 | u_max = -u_max |
|---|
| 89 | u_max_ijk(1) = nzb + 1 |
|---|
| 90 | ENDIF |
|---|
| 91 | IF ( v_max_ijk(1) == nzb ) THEN |
|---|
| 92 | v_max = -v_max |
|---|
| 93 | v_max_ijk(1) = nzb + 1 |
|---|
| 94 | ENDIF |
|---|
| 95 | ENDIF |
|---|
| 96 | |
|---|
| 97 | ! |
|---|
| 98 | !-- In case of Galilei-transform not using the geostrophic wind as translation |
|---|
| 99 | !-- velocity, compute the volume-averaged horizontal velocity components, which |
|---|
| 100 | !-- will then be subtracted from the horizontal wind for the time step and |
|---|
| 101 | !-- horizontal advection routines. |
|---|
| 102 | IF ( galilei_transformation .AND. .NOT. use_ug_for_galilei_tr ) THEN |
|---|
| 103 | IF ( flow_statistics_called ) THEN |
|---|
| 104 | ! |
|---|
| 105 | !-- Horizontal averages already existent, just need to average them |
|---|
| 106 | !-- vertically. |
|---|
| 107 | u_gtrans = 0.0 |
|---|
| 108 | v_gtrans = 0.0 |
|---|
| 109 | DO k = nzb+1, nzt |
|---|
| 110 | u_gtrans = u_gtrans + hom(k,1,1,0) |
|---|
| 111 | v_gtrans = v_gtrans + hom(k,1,2,0) |
|---|
| 112 | ENDDO |
|---|
| 113 | u_gtrans = u_gtrans / REAL( nzt - nzb ) |
|---|
| 114 | v_gtrans = v_gtrans / REAL( nzt - nzb ) |
|---|
| 115 | ELSE |
|---|
| 116 | ! |
|---|
| 117 | !-- Averaging over the entire model domain. |
|---|
| 118 | uv_gtrans_l = 0.0 |
|---|
| 119 | DO i = nxl, nxr |
|---|
| 120 | DO j = nys, nyn |
|---|
| 121 | DO k = nzb+1, nzt |
|---|
| 122 | uv_gtrans_l(1) = uv_gtrans_l(1) + u(k,j,i) |
|---|
| 123 | uv_gtrans_l(2) = uv_gtrans_l(2) + v(k,j,i) |
|---|
| 124 | ENDDO |
|---|
| 125 | ENDDO |
|---|
| 126 | ENDDO |
|---|
| 127 | uv_gtrans_l = uv_gtrans_l / REAL( (nxr-nxl+1)*(nyn-nys+1)*(nzt-nzb) ) |
|---|
| 128 | #if defined( __parallel ) |
|---|
| [622] | 129 | IF ( collective_wait ) CALL MPI_BARRIER( comm2d, ierr ) |
|---|
| [1] | 130 | CALL MPI_ALLREDUCE( uv_gtrans_l, uv_gtrans, 2, MPI_REAL, MPI_SUM, & |
|---|
| 131 | comm2d, ierr ) |
|---|
| 132 | u_gtrans = uv_gtrans(1) / REAL( numprocs ) |
|---|
| 133 | v_gtrans = uv_gtrans(2) / REAL( numprocs ) |
|---|
| 134 | #else |
|---|
| 135 | u_gtrans = uv_gtrans_l(1) |
|---|
| 136 | v_gtrans = uv_gtrans_l(2) |
|---|
| 137 | #endif |
|---|
| 138 | ENDIF |
|---|
| 139 | ENDIF |
|---|
| 140 | |
|---|
| 141 | IF ( .NOT. dt_fixed ) THEN |
|---|
| 142 | ! |
|---|
| 143 | !-- Variable time step: |
|---|
| 144 | ! |
|---|
| 145 | !-- For each component, compute the maximum time step according to the |
|---|
| 146 | !-- FCL-criterion. |
|---|
| 147 | dt_u = dx / ( ABS( u_max - u_gtrans ) + 1.0E-10 ) |
|---|
| 148 | dt_v = dy / ( ABS( v_max - v_gtrans ) + 1.0E-10 ) |
|---|
| 149 | dt_w = dzu(MAX( 1, w_max_ijk(1) )) / ( ABS( w_max ) + 1.0E-10 ) |
|---|
| 150 | |
|---|
| 151 | ! |
|---|
| 152 | !-- Compute time step according to the diffusion criterion. |
|---|
| 153 | !-- First calculate minimum grid spacing which only depends on index k |
|---|
| 154 | !-- Note: also at k=nzb+1 a full grid length is being assumed, although |
|---|
| 155 | !-- in the Prandtl-layer friction term only dz/2 is used. |
|---|
| 156 | !-- Experience from the old model seems to justify this. |
|---|
| 157 | dt_diff_l = 999999.0 |
|---|
| 158 | |
|---|
| 159 | DO k = nzb+1, nzt |
|---|
| [667] | 160 | dxyz2_min(k) = MIN( dx2, dy2, dzw(k)*dzw(k) ) * 0.125 |
|---|
| [1] | 161 | ENDDO |
|---|
| 162 | |
|---|
| 163 | !$OMP PARALLEL private(i,j,k,value) reduction(MIN: dt_diff_l) |
|---|
| 164 | !$OMP DO |
|---|
| 165 | DO i = nxl, nxr |
|---|
| 166 | DO j = nys, nyn |
|---|
| 167 | DO k = nzb+1, nzt |
|---|
| 168 | value = dxyz2_min(k) / ( MAX( kh(k,j,i), km(k,j,i) ) + 1E-20 ) |
|---|
| 169 | |
|---|
| 170 | dt_diff_l = MIN( value, dt_diff_l ) |
|---|
| 171 | ENDDO |
|---|
| 172 | ENDDO |
|---|
| 173 | ENDDO |
|---|
| 174 | !$OMP END PARALLEL |
|---|
| 175 | #if defined( __parallel ) |
|---|
| [622] | 176 | IF ( collective_wait ) CALL MPI_BARRIER( comm2d, ierr ) |
|---|
| [1] | 177 | CALL MPI_ALLREDUCE( dt_diff_l, dt_diff, 1, MPI_REAL, MPI_MIN, comm2d, & |
|---|
| 178 | ierr ) |
|---|
| 179 | #else |
|---|
| 180 | dt_diff = dt_diff_l |
|---|
| 181 | #endif |
|---|
| 182 | |
|---|
| 183 | ! |
|---|
| 184 | !-- In case of non-cyclic lateral boundaries, the diffusion time step |
|---|
| 185 | !-- may be further restricted by the lateral damping layer (damping only |
|---|
| 186 | !-- along x and y) |
|---|
| 187 | IF ( bc_lr /= 'cyclic' ) THEN |
|---|
| 188 | dt_diff = MIN( dt_diff, 0.125 * dx2 / ( km_damp_max + 1E-20 ) ) |
|---|
| 189 | ELSEIF ( bc_ns /= 'cyclic' ) THEN |
|---|
| 190 | dt_diff = MIN( dt_diff, 0.125 * dy2 / ( km_damp_max + 1E-20 ) ) |
|---|
| 191 | ENDIF |
|---|
| 192 | |
|---|
| 193 | ! |
|---|
| [316] | 194 | !-- Additional timestep criterion with plant canopies: |
|---|
| 195 | !-- it is not allowed to extract more than the available momentum |
|---|
| 196 | IF ( plant_canopy ) THEN |
|---|
| [318] | 197 | |
|---|
| 198 | dt_plant_canopy_l = 0.0 |
|---|
| 199 | DO i = nxl, nxr |
|---|
| 200 | DO j = nys, nyn |
|---|
| 201 | DO k = nzb+1, nzt |
|---|
| 202 | dt_plant_canopy_u = cdc(k,j,i) * lad_u(k,j,i) * & |
|---|
| 203 | SQRT( u(k,j,i)**2 + & |
|---|
| 204 | ( ( v(k,j,i-1) + & |
|---|
| 205 | v(k,j,i) + & |
|---|
| 206 | v(k,j+1,i) + & |
|---|
| 207 | v(k,j+1,i-1) ) & |
|---|
| 208 | / 4.0 )**2 + & |
|---|
| 209 | ( ( w(k-1,j,i-1) + & |
|---|
| 210 | w(k-1,j,i) + & |
|---|
| 211 | w(k,j,i-1) + & |
|---|
| 212 | w(k,j,i) ) & |
|---|
| 213 | / 4.0 )**2 ) |
|---|
| 214 | IF ( dt_plant_canopy_u > dt_plant_canopy_l ) THEN |
|---|
| 215 | dt_plant_canopy_l = dt_plant_canopy_u |
|---|
| 216 | ENDIF |
|---|
| 217 | dt_plant_canopy_v = cdc(k,j,i) * lad_v(k,j,i) * & |
|---|
| 218 | SQRT( ( ( u(k,j-1,i) + & |
|---|
| 219 | u(k,j-1,i+1) + & |
|---|
| 220 | u(k,j,i) + & |
|---|
| 221 | u(k,j,i+1) ) & |
|---|
| 222 | / 4.0 )**2 + & |
|---|
| 223 | v(k,j,i)**2 + & |
|---|
| 224 | ( ( w(k-1,j-1,i) + & |
|---|
| 225 | w(k-1,j,i) + & |
|---|
| 226 | w(k,j-1,i) + & |
|---|
| 227 | w(k,j,i) ) & |
|---|
| 228 | / 4.0 )**2 ) |
|---|
| 229 | IF ( dt_plant_canopy_v > dt_plant_canopy_l ) THEN |
|---|
| 230 | dt_plant_canopy_l = dt_plant_canopy_v |
|---|
| 231 | ENDIF |
|---|
| 232 | dt_plant_canopy_w = cdc(k,j,i) * lad_w(k,j,i) * & |
|---|
| 233 | SQRT( ( ( u(k,j,i) + & |
|---|
| 234 | u(k,j,i+1) + & |
|---|
| 235 | u(k+1,j,i) + & |
|---|
| 236 | u(k+1,j,i+1) ) & |
|---|
| 237 | / 4.0 )**2 + & |
|---|
| 238 | ( ( v(k,j,i) + & |
|---|
| 239 | v(k,j+1,i) + & |
|---|
| 240 | v(k+1,j,i) + & |
|---|
| 241 | v(k+1,j+1,i) ) & |
|---|
| 242 | / 4.0 )**2 + & |
|---|
| 243 | w(k,j,i)**2 ) |
|---|
| 244 | IF ( dt_plant_canopy_w > dt_plant_canopy_l ) THEN |
|---|
| 245 | dt_plant_canopy_l = dt_plant_canopy_w |
|---|
| 246 | ENDIF |
|---|
| 247 | ENDDO |
|---|
| 248 | ENDDO |
|---|
| 249 | ENDDO |
|---|
| 250 | |
|---|
| 251 | IF ( dt_plant_canopy_l > 0.0 ) THEN |
|---|
| [320] | 252 | ! |
|---|
| 253 | !-- Invert dt_plant_canopy_l and apply a security timestep factor 0.1 |
|---|
| [318] | 254 | dt_plant_canopy_l = 0.1 / dt_plant_canopy_l |
|---|
| [320] | 255 | ELSE |
|---|
| 256 | ! |
|---|
| 257 | !-- In case of inhomogeneous plant canopy, some processors may have no |
|---|
| 258 | !-- canopy at all. Then use dt_max as dummy instead. |
|---|
| 259 | dt_plant_canopy_l = dt_max |
|---|
| [318] | 260 | ENDIF |
|---|
| [320] | 261 | |
|---|
| [316] | 262 | ! |
|---|
| [318] | 263 | !-- Determine the global minumum |
|---|
| 264 | #if defined( __parallel ) |
|---|
| [622] | 265 | IF ( collective_wait ) CALL MPI_BARRIER( comm2d, ierr ) |
|---|
| [318] | 266 | CALL MPI_ALLREDUCE( dt_plant_canopy_l, dt_plant_canopy, 1, MPI_REAL, & |
|---|
| 267 | MPI_MIN, comm2d, ierr ) |
|---|
| 268 | #else |
|---|
| 269 | dt_plant_canopy = dt_plant_canopy_l |
|---|
| 270 | #endif |
|---|
| [316] | 271 | |
|---|
| 272 | ELSE |
|---|
| 273 | ! |
|---|
| 274 | !-- Use dt_diff as dummy value to avoid extra IF branches further below |
|---|
| 275 | dt_plant_canopy = dt_diff |
|---|
| 276 | |
|---|
| 277 | ENDIF |
|---|
| 278 | |
|---|
| 279 | ! |
|---|
| 280 | !-- The time step is the minimum of the 3-4 components and the diffusion time |
|---|
| [1] | 281 | !-- step minus a reduction to be on the safe side. Factor 0.5 is necessary |
|---|
| 282 | !-- since the leap-frog scheme always progresses by 2 * delta t. |
|---|
| 283 | !-- The user has to set the cfl_factor small enough to ensure that the |
|---|
| 284 | !-- divergences do not become too large. |
|---|
| 285 | !-- The time step must not exceed the maximum allowed value. |
|---|
| 286 | IF ( timestep_scheme(1:5) == 'runge' ) THEN |
|---|
| [316] | 287 | dt_3d = cfl_factor * MIN( dt_diff, dt_plant_canopy, dt_u, dt_v, dt_w ) |
|---|
| [1] | 288 | ELSE |
|---|
| [316] | 289 | dt_3d = cfl_factor * 0.5 * & |
|---|
| 290 | MIN( dt_diff, dt_plant_canopy, dt_u, dt_v, dt_w ) |
|---|
| [1] | 291 | ENDIF |
|---|
| 292 | dt_3d = MIN( dt_3d, dt_max ) |
|---|
| 293 | |
|---|
| 294 | ! |
|---|
| 295 | !-- Remember the restricting time step criterion for later output. |
|---|
| [316] | 296 | IF ( MIN( dt_u, dt_v, dt_w ) < MIN( dt_diff, dt_plant_canopy ) ) THEN |
|---|
| [1] | 297 | timestep_reason = 'A' |
|---|
| [316] | 298 | ELSEIF ( dt_plant_canopy < dt_diff ) THEN |
|---|
| 299 | timestep_reason = 'C' |
|---|
| [1] | 300 | ELSE |
|---|
| 301 | timestep_reason = 'D' |
|---|
| 302 | ENDIF |
|---|
| 303 | |
|---|
| 304 | ! |
|---|
| 305 | !-- Set flag if the time step becomes too small. |
|---|
| 306 | IF ( dt_3d < ( 0.00001 * dt_max ) ) THEN |
|---|
| 307 | stop_dt = .TRUE. |
|---|
| [108] | 308 | |
|---|
| [320] | 309 | WRITE( message_string, * ) 'Time step has reached minimum limit.', & |
|---|
| 310 | '&dt = ', dt_3d, ' s Simulation is terminated.', & |
|---|
| 311 | '&old_dt = ', old_dt, ' s', & |
|---|
| 312 | '&dt_u = ', dt_u, ' s', & |
|---|
| 313 | '&dt_v = ', dt_v, ' s', & |
|---|
| 314 | '&dt_w = ', dt_w, ' s', & |
|---|
| 315 | '&dt_diff = ', dt_diff, ' s', & |
|---|
| 316 | '&dt_plant_canopy = ', dt_plant_canopy, ' s', & |
|---|
| 317 | '&u_max = ', u_max, ' m/s k=', u_max_ijk(1), & |
|---|
| 318 | ' j=', u_max_ijk(2), ' i=', u_max_ijk(3), & |
|---|
| 319 | '&v_max = ', v_max, ' m/s k=', v_max_ijk(1), & |
|---|
| 320 | ' j=', v_max_ijk(2), ' i=', v_max_ijk(3), & |
|---|
| 321 | '&w_max = ', w_max, ' m/s k=', w_max_ijk(1), & |
|---|
| 322 | ' j=', w_max_ijk(2), ' i=', w_max_ijk(3) |
|---|
| [258] | 323 | CALL message( 'timestep', 'PA0312', 0, 1, 0, 6, 0 ) |
|---|
| [108] | 324 | ! |
|---|
| 325 | !-- In case of coupled runs inform the remote model of the termination |
|---|
| 326 | !-- and its reason, provided the remote model has not already been |
|---|
| 327 | !-- informed of another termination reason (terminate_coupled > 0) before. |
|---|
| [222] | 328 | #if defined( __parallel ) |
|---|
| [108] | 329 | IF ( coupling_mode /= 'uncoupled' .AND. terminate_coupled == 0 ) THEN |
|---|
| 330 | terminate_coupled = 2 |
|---|
| [667] | 331 | IF ( myid == 0 ) THEN |
|---|
| 332 | CALL MPI_SENDRECV( & |
|---|
| 333 | terminate_coupled, 1, MPI_INTEGER, target_id, 0, & |
|---|
| 334 | terminate_coupled_remote, 1, MPI_INTEGER, target_id, 0, & |
|---|
| 335 | comm_inter, status, ierr ) |
|---|
| 336 | ENDIF |
|---|
| 337 | CALL MPI_BCAST( terminate_coupled_remote, 1, MPI_INTEGER, 0, comm2d, ierr) |
|---|
| [108] | 338 | ENDIF |
|---|
| [222] | 339 | #endif |
|---|
| [1] | 340 | ENDIF |
|---|
| 341 | |
|---|
| 342 | ! |
|---|
| 343 | !-- Ensure a smooth value (two significant digits) of the timestep. For |
|---|
| 344 | !-- other schemes than Runge-Kutta, the following restrictions appear: |
|---|
| 345 | !-- The current timestep is only then changed, if the change relative to |
|---|
| 346 | !-- its previous value exceeds +5 % or -2 %. In case of a timestep |
|---|
| 347 | !-- reduction, at least 30 iterations have to be performed before a timestep |
|---|
| 348 | !-- enlargement is permitted again. |
|---|
| 349 | percent_change = dt_3d / old_dt - 1.0 |
|---|
| 350 | IF ( percent_change > 0.05 .OR. percent_change < -0.02 .OR. & |
|---|
| 351 | timestep_scheme(1:5) == 'runge' ) THEN |
|---|
| 352 | |
|---|
| 353 | ! |
|---|
| 354 | !-- Time step enlargement by no more than 2 %. |
|---|
| 355 | IF ( percent_change > 0.0 .AND. simulated_time /= 0.0 .AND. & |
|---|
| 356 | timestep_scheme(1:5) /= 'runge' ) THEN |
|---|
| 357 | dt_3d = 1.02 * old_dt |
|---|
| 358 | ENDIF |
|---|
| 359 | |
|---|
| 360 | ! |
|---|
| 361 | !-- A relatively smooth value of the time step is ensured by taking |
|---|
| 362 | !-- only the first two significant digits. |
|---|
| 363 | div = 1000.0 |
|---|
| 364 | DO WHILE ( dt_3d < div ) |
|---|
| 365 | div = div / 10.0 |
|---|
| 366 | ENDDO |
|---|
| 367 | dt_3d = NINT( dt_3d * 100.0 / div ) * div / 100.0 |
|---|
| 368 | |
|---|
| 369 | ! |
|---|
| 370 | !-- Now the time step can be adjusted. |
|---|
| 371 | IF ( percent_change < 0.0 .OR. timestep_scheme(1:5) == 'runge' ) & |
|---|
| 372 | THEN |
|---|
| 373 | ! |
|---|
| 374 | !-- Time step reduction. |
|---|
| 375 | old_dt = dt_3d |
|---|
| 376 | dt_changed = .TRUE. |
|---|
| 377 | ELSE |
|---|
| 378 | ! |
|---|
| 379 | !-- For other timestep schemes , the time step is only enlarged |
|---|
| 380 | !-- after at least 30 iterations since the previous time step |
|---|
| 381 | !-- change or, of course, after model initialization. |
|---|
| 382 | IF ( current_timestep_number >= last_dt_change + 30 .OR. & |
|---|
| 383 | simulated_time == 0.0 ) THEN |
|---|
| 384 | old_dt = dt_3d |
|---|
| 385 | dt_changed = .TRUE. |
|---|
| 386 | ELSE |
|---|
| 387 | dt_3d = old_dt |
|---|
| 388 | dt_changed = .FALSE. |
|---|
| 389 | ENDIF |
|---|
| 390 | |
|---|
| 391 | ENDIF |
|---|
| 392 | ELSE |
|---|
| 393 | ! |
|---|
| 394 | !-- No time step change since the difference is too small. |
|---|
| 395 | dt_3d = old_dt |
|---|
| 396 | dt_changed = .FALSE. |
|---|
| 397 | ENDIF |
|---|
| 398 | |
|---|
| 399 | IF ( dt_changed ) last_dt_change = current_timestep_number |
|---|
| 400 | |
|---|
| 401 | ENDIF |
|---|
| 402 | CALL cpu_log( log_point(12), 'calculate_timestep', 'stop' ) |
|---|
| 403 | |
|---|
| 404 | END SUBROUTINE timestep |
|---|
