SUBROUTINE data_output_2d( mode, av ) !--------------------------------------------------------------------------------! ! This file is part of PALM. ! ! PALM is free software: you can redistribute it and/or modify it under the terms ! of the GNU General Public License as published by the Free Software Foundation, ! either version 3 of the License, or (at your option) any later version. ! ! PALM is distributed in the hope that it will be useful, but WITHOUT ANY ! WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR ! A PARTICULAR PURPOSE. See the GNU General Public License for more details. ! ! You should have received a copy of the GNU General Public License along with ! PALM. If not, see . ! ! Copyright 1997-2014 Leibniz Universitaet Hannover !--------------------------------------------------------------------------------! ! ! Current revisions: ! ----------------- ! ! ! Former revisions: ! ----------------- ! $Id: data_output_2d.f90 1329 2014-03-21 11:09:15Z maronga $ ! ! 1327 2014-03-21 11:00:16Z raasch ! parts concerning iso2d output removed, ! -netcdf output queries ! ! 1320 2014-03-20 08:40:49Z raasch ! ONLY-attribute added to USE-statements, ! kind-parameters added to all INTEGER and REAL declaration statements, ! kinds are defined in new module kinds, ! revision history before 2012 removed, ! comment fields (!:) to be used for variable explanations added to ! all variable declaration statements ! ! 1318 2014-03-17 13:35:16Z raasch ! barrier argument removed from cpu_log. ! module interfaces removed ! ! 1311 2014-03-14 12:13:39Z heinze ! bugfix: close #if defined( __netcdf ) ! ! 1308 2014-03-13 14:58:42Z fricke ! +local_2d_sections, local_2d_sections_l, ns ! Check, if the limit of the time dimension is exceeded for parallel output ! To increase the performance for parallel output, the following is done: ! - Update of time axis is only done by PE0 ! - Cross sections are first stored on a local array and are written ! collectively to the output file by all PEs. ! ! 1115 2013-03-26 18:16:16Z hoffmann ! ql is calculated by calc_liquid_water_content ! ! 1076 2012-12-05 08:30:18Z hoffmann ! Bugfix in output of ql ! ! 1065 2012-11-22 17:42:36Z hoffmann ! Bugfix: Output of cross sections of ql ! ! 1053 2012-11-13 17:11:03Z hoffmann ! +qr, nr, qc and cross sections ! ! 1036 2012-10-22 13:43:42Z raasch ! code put under GPL (PALM 3.9) ! ! 1031 2012-10-19 14:35:30Z raasch ! netCDF4 without parallel file support implemented ! ! 1007 2012-09-19 14:30:36Z franke ! Bugfix: missing calculation of ql_vp added ! ! 978 2012-08-09 08:28:32Z fricke ! +z0h ! ! Revision 1.1 1997/08/11 06:24:09 raasch ! Initial revision ! ! ! Description: ! ------------ ! Data output of horizontal cross-sections in netCDF format or binary format ! compatible to old graphic software iso2d. ! Attention: The position of the sectional planes is still not always computed ! --------- correctly. (zu is used always)! !------------------------------------------------------------------------------! USE arrays_3d, & ONLY: dzw, e, nr, p, pt, q, qc, ql, ql_c, ql_v, ql_vp, qr, qsws, & rho, sa, shf, tend, ts, u, us, v, vpt, w, z0, z0h, zu, zw USE averaging USE cloud_parameters, & ONLY: hyrho, l_d_cp, precipitation_amount, precipitation_rate, prr, & pt_d_t USE control_parameters, & ONLY: cloud_physics, data_output_2d_on_each_pe, data_output_xy, & data_output_xz, data_output_yz, do2d, & do2d_xy_last_time, do2d_xy_n, do2d_xy_time_count, & do2d_xz_last_time, do2d_xz_n, do2d_xz_time_count, & do2d_yz_last_time, do2d_yz_n, do2d_yz_time_count, & ibc_uv_b, icloud_scheme, io_blocks, io_group, & message_string, netcdf_data_format, & ntdim_2d_xy, ntdim_2d_xz, ntdim_2d_yz, psolver, section, & simulated_time, simulated_time_chr, time_since_reference_point USE cpulog, & ONLY: cpu_log, log_point USE grid_variables, & ONLY: dx, dy USE indices, & ONLY: nbgp, nx, nxl, nxlg, nxr, nxrg, ny, nyn, nyng, nys, nysg, & nz, nzb, nzt USE kinds USE netcdf_control USE particle_attributes, & ONLY: particle_advection_start, particles, prt_count, & prt_start_index USE pegrid IMPLICIT NONE CHARACTER (LEN=2) :: do2d_mode !: CHARACTER (LEN=2) :: mode !: CHARACTER (LEN=4) :: grid !: CHARACTER (LEN=25) :: section_chr !: CHARACTER (LEN=50) :: rtext !: INTEGER(iwp) :: av !: INTEGER(iwp) :: ngp !: INTEGER(iwp) :: file_id !: INTEGER(iwp) :: i !: INTEGER(iwp) :: if !: INTEGER(iwp) :: is !: INTEGER(iwp) :: iis !: INTEGER(iwp) :: j !: INTEGER(iwp) :: k !: INTEGER(iwp) :: l !: INTEGER(iwp) :: layer_xy !: INTEGER(iwp) :: n !: INTEGER(iwp) :: ns !: INTEGER(iwp) :: psi !: INTEGER(iwp) :: s !: INTEGER(iwp) :: sender !: INTEGER(iwp) :: ind(4) !: LOGICAL :: found !: LOGICAL :: resorted !: LOGICAL :: two_d !: REAL(wp) :: mean_r !: REAL(wp) :: s_r3 !: REAL(wp) :: s_r4 !: REAL(wp), DIMENSION(:), ALLOCATABLE :: level_z !: REAL(wp), DIMENSION(:,:), ALLOCATABLE :: local_2d !: REAL(wp), DIMENSION(:,:), ALLOCATABLE :: local_2d_l !: REAL(wp), DIMENSION(:,:,:), ALLOCATABLE :: local_pf !: REAL(wp), DIMENSION(:,:,:), ALLOCATABLE :: local_2d_sections !: REAL(wp), DIMENSION(:,:,:), ALLOCATABLE :: local_2d_sections_l !: #if defined( __parallel ) REAL(wp), DIMENSION(:,:), ALLOCATABLE :: total_2d !: #endif REAL(wp), DIMENSION(:,:,:), POINTER :: to_be_resorted !: NAMELIST /LOCAL/ rtext ! !-- Immediate return, if no output is requested (no respective sections !-- found in parameter data_output) IF ( mode == 'xy' .AND. .NOT. data_output_xy(av) ) RETURN IF ( mode == 'xz' .AND. .NOT. data_output_xz(av) ) RETURN IF ( mode == 'yz' .AND. .NOT. data_output_yz(av) ) RETURN ! !-- For parallel netcdf output the time axis must be limited. Return, if this !-- limit is exceeded. This could be the case, if the simulated time exceeds !-- the given end time by the length of the given output interval. IF ( netcdf_data_format > 4 ) THEN IF ( mode == 'xy' .AND. do2d_xy_time_count(av) + 1 > & ntdim_2d_xy(av) ) THEN WRITE ( message_string, * ) 'Output of xy cross-sections is not ', & 'given at t=', simulated_time, '&because the', & ' maximum number of output time levels is exceeded.' CALL message( 'data_output_2d', 'PA0384', 0, 1, 0, 6, 0 ) RETURN ENDIF IF ( mode == 'xz' .AND. do2d_xz_time_count(av) + 1 > & ntdim_2d_xz(av) ) THEN WRITE ( message_string, * ) 'Output of xz cross-sections is not ', & 'given at t=', simulated_time, '&because the', & ' maximum number of output time levels is exceeded.' CALL message( 'data_output_2d', 'PA0385', 0, 1, 0, 6, 0 ) RETURN ENDIF IF ( mode == 'yz' .AND. do2d_yz_time_count(av) + 1 > & ntdim_2d_yz(av) ) THEN WRITE ( message_string, * ) 'Output of yz cross-sections is not ', & 'given at t=', simulated_time, '&because the', & ' maximum number of output time levels is exceeded.' CALL message( 'data_output_2d', 'PA0386', 0, 1, 0, 6, 0 ) RETURN ENDIF ENDIF CALL cpu_log (log_point(3),'data_output_2d','start') two_d = .FALSE. ! local variable to distinguish between output of pure 2D ! arrays and cross-sections of 3D arrays. ! !-- Depending on the orientation of the cross-section, the respective output !-- files have to be opened. SELECT CASE ( mode ) CASE ( 'xy' ) s = 1 ALLOCATE( level_z(nzb:nzt+1), local_2d(nxlg:nxrg,nysg:nyng) ) IF ( netcdf_data_format > 4 ) THEN ns = 1 DO WHILE ( section(ns,s) /= -9999 .AND. ns <= 100 ) ns = ns + 1 ENDDO ns = ns - 1 ALLOCATE( local_2d_sections(nxlg:nxrg,nysg:nyng,1:ns) ) local_2d_sections = 0.0 ENDIF ! !-- Parallel netCDF4/HDF5 output is done on all PEs, all other on PE0 only IF ( myid == 0 .OR. netcdf_data_format > 4 ) THEN CALL check_open( 101+av*10 ) ENDIF IF ( data_output_2d_on_each_pe ) THEN CALL check_open( 21 ) ELSE IF ( myid == 0 ) THEN #if defined( __parallel ) ALLOCATE( total_2d(-nbgp:nx+nbgp,-nbgp:ny+nbgp) ) #endif ENDIF ENDIF CASE ( 'xz' ) s = 2 ALLOCATE( local_2d(nxlg:nxrg,nzb:nzt+1) ) IF ( netcdf_data_format > 4 ) THEN ns = 1 DO WHILE ( section(ns,s) /= -9999 .AND. ns <= 100 ) ns = ns + 1 ENDDO ns = ns - 1 ALLOCATE( local_2d_sections(nxlg:nxrg,1:ns,nzb:nzt+1) ) ALLOCATE( local_2d_sections_l(nxlg:nxrg,1:ns,nzb:nzt+1) ) local_2d_sections = 0.0; local_2d_sections_l = 0.0 ENDIF ! !-- Parallel netCDF4/HDF5 output is done on all PEs, all other on PE0 only IF ( myid == 0 .OR. netcdf_data_format > 4 ) THEN CALL check_open( 102+av*10 ) ENDIF IF ( data_output_2d_on_each_pe ) THEN CALL check_open( 22 ) ELSE IF ( myid == 0 ) THEN #if defined( __parallel ) ALLOCATE( total_2d(-nbgp:nx+nbgp,nzb:nzt+1) ) #endif ENDIF ENDIF CASE ( 'yz' ) s = 3 ALLOCATE( local_2d(nysg:nyng,nzb:nzt+1) ) IF ( netcdf_data_format > 4 ) THEN ns = 1 DO WHILE ( section(ns,s) /= -9999 .AND. ns <= 100 ) ns = ns + 1 ENDDO ns = ns - 1 ALLOCATE( local_2d_sections(1:ns,nysg:nyng,nzb:nzt+1) ) ALLOCATE( local_2d_sections_l(1:ns,nysg:nyng,nzb:nzt+1) ) local_2d_sections = 0.0; local_2d_sections_l = 0.0 ENDIF ! !-- Parallel netCDF4/HDF5 output is done on all PEs, all other on PE0 only IF ( myid == 0 .OR. netcdf_data_format > 4 ) THEN CALL check_open( 103+av*10 ) ENDIF IF ( data_output_2d_on_each_pe ) THEN CALL check_open( 23 ) ELSE IF ( myid == 0 ) THEN #if defined( __parallel ) ALLOCATE( total_2d(-nbgp:ny+nbgp,nzb:nzt+1) ) #endif ENDIF ENDIF CASE DEFAULT message_string = 'unknown cross-section: ' // TRIM( mode ) CALL message( 'data_output_2d', 'PA0180', 1, 2, 0, 6, 0 ) END SELECT ! !-- Allocate a temporary array for resorting (kji -> ijk). ALLOCATE( local_pf(nxlg:nxrg,nysg:nyng,nzb:nzt+1) ) ! !-- Loop of all variables to be written. !-- Output dimensions chosen if = 1 l = MAX( 2, LEN_TRIM( do2d(av,if) ) ) do2d_mode = do2d(av,if)(l-1:l) DO WHILE ( do2d(av,if)(1:1) /= ' ' ) IF ( do2d_mode == mode ) THEN ! !-- Store the array chosen on the temporary array. resorted = .FALSE. SELECT CASE ( TRIM( do2d(av,if) ) ) CASE ( 'e_xy', 'e_xz', 'e_yz' ) IF ( av == 0 ) THEN to_be_resorted => e ELSE to_be_resorted => e_av ENDIF IF ( mode == 'xy' ) level_z = zu CASE ( 'lpt_xy', 'lpt_xz', 'lpt_yz' ) IF ( av == 0 ) THEN to_be_resorted => pt ELSE to_be_resorted => lpt_av ENDIF IF ( mode == 'xy' ) level_z = zu CASE ( 'lwp*_xy' ) ! 2d-array IF ( av == 0 ) THEN DO i = nxlg, nxrg DO j = nysg, nyng local_pf(i,j,nzb+1) = SUM( ql(nzb:nzt,j,i) * & dzw(1:nzt+1) ) ENDDO ENDDO ELSE DO i = nxlg, nxrg DO j = nysg, nyng local_pf(i,j,nzb+1) = lwp_av(j,i) ENDDO ENDDO ENDIF resorted = .TRUE. two_d = .TRUE. level_z(nzb+1) = zu(nzb+1) CASE ( 'nr_xy', 'nr_xz', 'nr_yz' ) IF ( av == 0 ) THEN to_be_resorted => nr ELSE to_be_resorted => nr_av ENDIF IF ( mode == 'xy' ) level_z = zu CASE ( 'p_xy', 'p_xz', 'p_yz' ) IF ( av == 0 ) THEN IF ( psolver /= 'sor' ) CALL exchange_horiz( p, nbgp ) to_be_resorted => p ELSE IF ( psolver /= 'sor' ) CALL exchange_horiz( p_av, nbgp ) to_be_resorted => p_av ENDIF IF ( mode == 'xy' ) level_z = zu CASE ( 'pc_xy', 'pc_xz', 'pc_yz' ) ! particle concentration IF ( av == 0 ) THEN IF ( simulated_time >= particle_advection_start ) THEN tend = prt_count CALL exchange_horiz( tend, nbgp ) ELSE tend = 0.0 ENDIF DO i = nxlg, nxrg DO j = nysg, nyng DO k = nzb, nzt+1 local_pf(i,j,k) = tend(k,j,i) ENDDO ENDDO ENDDO resorted = .TRUE. ELSE CALL exchange_horiz( pc_av, nbgp ) to_be_resorted => pc_av ENDIF CASE ( 'pr_xy', 'pr_xz', 'pr_yz' ) ! mean particle radius IF ( av == 0 ) THEN IF ( simulated_time >= particle_advection_start ) THEN DO i = nxl, nxr DO j = nys, nyn DO k = nzb, nzt+1 psi = prt_start_index(k,j,i) s_r3 = 0.0 s_r4 = 0.0 DO n = psi, psi+prt_count(k,j,i)-1 s_r3 = s_r3 + particles(n)%radius**3 * & particles(n)%weight_factor s_r4 = s_r4 + particles(n)%radius**4 * & particles(n)%weight_factor ENDDO IF ( s_r3 /= 0.0 ) THEN mean_r = s_r4 / s_r3 ELSE mean_r = 0.0 ENDIF tend(k,j,i) = mean_r ENDDO ENDDO ENDDO CALL exchange_horiz( tend, nbgp ) ELSE tend = 0.0 END IF DO i = nxlg, nxrg DO j = nysg, nyng DO k = nzb, nzt+1 local_pf(i,j,k) = tend(k,j,i) ENDDO ENDDO ENDDO resorted = .TRUE. ELSE CALL exchange_horiz( pr_av, nbgp ) to_be_resorted => pr_av ENDIF CASE ( 'pra*_xy' ) ! 2d-array / integral quantity => no av CALL exchange_horiz_2d( precipitation_amount ) DO i = nxlg, nxrg DO j = nysg, nyng local_pf(i,j,nzb+1) = precipitation_amount(j,i) ENDDO ENDDO precipitation_amount = 0.0 ! reset for next integ. interval resorted = .TRUE. two_d = .TRUE. level_z(nzb+1) = zu(nzb+1) CASE ( 'prr*_xy' ) ! 2d-array IF ( icloud_scheme == 1 ) THEN IF ( av == 0 ) THEN CALL exchange_horiz_2d( precipitation_rate ) DO i = nxlg, nxrg DO j = nysg, nyng local_pf(i,j,nzb+1) = precipitation_rate(j,i) ENDDO ENDDO ELSE CALL exchange_horiz_2d( precipitation_rate_av ) DO i = nxlg, nxrg DO j = nysg, nyng local_pf(i,j,nzb+1) = precipitation_rate_av(j,i) ENDDO ENDDO ENDIF ELSE IF ( av == 0 ) THEN CALL exchange_horiz_2d( prr(nzb+1,:,:) ) DO i = nxlg, nxrg DO j = nysg, nyng local_pf(i,j,nzb+1) = prr(nzb+1,j,i) * hyrho(nzb+1) ENDDO ENDDO ELSE CALL exchange_horiz_2d( prr_av(nzb+1,:,:) ) DO i = nxlg, nxrg DO j = nysg, nyng local_pf(i,j,nzb+1) = prr_av(nzb+1,j,i) * & hyrho(nzb+1) ENDDO ENDDO ENDIF ENDIF resorted = .TRUE. two_d = .TRUE. level_z(nzb+1) = zu(nzb+1) CASE ( 'prr_xy', 'prr_xz', 'prr_yz' ) IF ( av == 0 ) THEN CALL exchange_horiz( prr, nbgp ) DO i = nxlg, nxrg DO j = nysg, nyng DO k = nzb, nzt+1 local_pf(i,j,k) = prr(k,j,i) ENDDO ENDDO ENDDO ELSE CALL exchange_horiz( prr_av, nbgp ) DO i = nxlg, nxrg DO j = nysg, nyng DO k = nzb, nzt+1 local_pf(i,j,k) = prr_av(k,j,i) ENDDO ENDDO ENDDO ENDIF resorted = .TRUE. IF ( mode == 'xy' ) level_z = zu CASE ( 'pt_xy', 'pt_xz', 'pt_yz' ) IF ( av == 0 ) THEN IF ( .NOT. cloud_physics ) THEN to_be_resorted => pt ELSE DO i = nxlg, nxrg DO j = nysg, nyng DO k = nzb, nzt+1 local_pf(i,j,k) = pt(k,j,i) + l_d_cp * & pt_d_t(k) * & ql(k,j,i) ENDDO ENDDO ENDDO resorted = .TRUE. ENDIF ELSE to_be_resorted => pt_av ENDIF IF ( mode == 'xy' ) level_z = zu CASE ( 'q_xy', 'q_xz', 'q_yz' ) IF ( av == 0 ) THEN to_be_resorted => q ELSE to_be_resorted => q_av ENDIF IF ( mode == 'xy' ) level_z = zu CASE ( 'qc_xy', 'qc_xz', 'qc_yz' ) IF ( av == 0 ) THEN to_be_resorted => qc ELSE to_be_resorted => qc_av ENDIF IF ( mode == 'xy' ) level_z = zu CASE ( 'ql_xy', 'ql_xz', 'ql_yz' ) IF ( av == 0 ) THEN to_be_resorted => ql ELSE to_be_resorted => ql_av ENDIF IF ( mode == 'xy' ) level_z = zu CASE ( 'ql_c_xy', 'ql_c_xz', 'ql_c_yz' ) IF ( av == 0 ) THEN to_be_resorted => ql_c ELSE to_be_resorted => ql_c_av ENDIF IF ( mode == 'xy' ) level_z = zu CASE ( 'ql_v_xy', 'ql_v_xz', 'ql_v_yz' ) IF ( av == 0 ) THEN to_be_resorted => ql_v ELSE to_be_resorted => ql_v_av ENDIF IF ( mode == 'xy' ) level_z = zu CASE ( 'ql_vp_xy', 'ql_vp_xz', 'ql_vp_yz' ) IF ( av == 0 ) THEN IF ( simulated_time >= particle_advection_start ) THEN DO i = nxl, nxr DO j = nys, nyn DO k = nzb, nzt+1 psi = prt_start_index(k,j,i) DO n = psi, psi+prt_count(k,j,i)-1 tend(k,j,i) = tend(k,j,i) + & particles(n)%weight_factor / & prt_count(k,j,i) ENDDO ENDDO ENDDO ENDDO CALL exchange_horiz( tend, nbgp ) ELSE tend = 0.0 END IF DO i = nxlg, nxrg DO j = nysg, nyng DO k = nzb, nzt+1 local_pf(i,j,k) = tend(k,j,i) ENDDO ENDDO ENDDO resorted = .TRUE. ELSE CALL exchange_horiz( ql_vp_av, nbgp ) to_be_resorted => ql_vp ENDIF IF ( mode == 'xy' ) level_z = zu CASE ( 'qr_xy', 'qr_xz', 'qr_yz' ) IF ( av == 0 ) THEN to_be_resorted => qr ELSE to_be_resorted => qr_av ENDIF IF ( mode == 'xy' ) level_z = zu CASE ( 'qsws*_xy' ) ! 2d-array IF ( av == 0 ) THEN DO i = nxlg, nxrg DO j = nysg, nyng local_pf(i,j,nzb+1) = qsws(j,i) ENDDO ENDDO ELSE DO i = nxlg, nxrg DO j = nysg, nyng local_pf(i,j,nzb+1) = qsws_av(j,i) ENDDO ENDDO ENDIF resorted = .TRUE. two_d = .TRUE. level_z(nzb+1) = zu(nzb+1) CASE ( 'qv_xy', 'qv_xz', 'qv_yz' ) IF ( av == 0 ) THEN DO i = nxlg, nxrg DO j = nysg, nyng DO k = nzb, nzt+1 local_pf(i,j,k) = q(k,j,i) - ql(k,j,i) ENDDO ENDDO ENDDO resorted = .TRUE. ELSE to_be_resorted => qv_av ENDIF IF ( mode == 'xy' ) level_z = zu CASE ( 'rho_xy', 'rho_xz', 'rho_yz' ) IF ( av == 0 ) THEN to_be_resorted => rho ELSE to_be_resorted => rho_av ENDIF CASE ( 's_xy', 's_xz', 's_yz' ) IF ( av == 0 ) THEN to_be_resorted => q ELSE to_be_resorted => s_av ENDIF CASE ( 'sa_xy', 'sa_xz', 'sa_yz' ) IF ( av == 0 ) THEN to_be_resorted => sa ELSE to_be_resorted => sa_av ENDIF CASE ( 'shf*_xy' ) ! 2d-array IF ( av == 0 ) THEN DO i = nxlg, nxrg DO j = nysg, nyng local_pf(i,j,nzb+1) = shf(j,i) ENDDO ENDDO ELSE DO i = nxlg, nxrg DO j = nysg, nyng local_pf(i,j,nzb+1) = shf_av(j,i) ENDDO ENDDO ENDIF resorted = .TRUE. two_d = .TRUE. level_z(nzb+1) = zu(nzb+1) CASE ( 't*_xy' ) ! 2d-array IF ( av == 0 ) THEN DO i = nxlg, nxrg DO j = nysg, nyng local_pf(i,j,nzb+1) = ts(j,i) ENDDO ENDDO ELSE DO i = nxlg, nxrg DO j = nysg, nyng local_pf(i,j,nzb+1) = ts_av(j,i) ENDDO ENDDO ENDIF resorted = .TRUE. two_d = .TRUE. level_z(nzb+1) = zu(nzb+1) CASE ( 'u_xy', 'u_xz', 'u_yz' ) IF ( av == 0 ) THEN to_be_resorted => u ELSE to_be_resorted => u_av ENDIF IF ( mode == 'xy' ) level_z = zu ! !-- Substitute the values generated by "mirror" boundary condition !-- at the bottom boundary by the real surface values. IF ( do2d(av,if) == 'u_xz' .OR. do2d(av,if) == 'u_yz' ) THEN IF ( ibc_uv_b == 0 ) local_pf(:,:,nzb) = 0.0 ENDIF CASE ( 'u*_xy' ) ! 2d-array IF ( av == 0 ) THEN DO i = nxlg, nxrg DO j = nysg, nyng local_pf(i,j,nzb+1) = us(j,i) ENDDO ENDDO ELSE DO i = nxlg, nxrg DO j = nysg, nyng local_pf(i,j,nzb+1) = us_av(j,i) ENDDO ENDDO ENDIF resorted = .TRUE. two_d = .TRUE. level_z(nzb+1) = zu(nzb+1) CASE ( 'v_xy', 'v_xz', 'v_yz' ) IF ( av == 0 ) THEN to_be_resorted => v ELSE to_be_resorted => v_av ENDIF IF ( mode == 'xy' ) level_z = zu ! !-- Substitute the values generated by "mirror" boundary condition !-- at the bottom boundary by the real surface values. IF ( do2d(av,if) == 'v_xz' .OR. do2d(av,if) == 'v_yz' ) THEN IF ( ibc_uv_b == 0 ) local_pf(:,:,nzb) = 0.0 ENDIF CASE ( 'vpt_xy', 'vpt_xz', 'vpt_yz' ) IF ( av == 0 ) THEN to_be_resorted => vpt ELSE to_be_resorted => vpt_av ENDIF IF ( mode == 'xy' ) level_z = zu CASE ( 'w_xy', 'w_xz', 'w_yz' ) IF ( av == 0 ) THEN to_be_resorted => w ELSE to_be_resorted => w_av ENDIF IF ( mode == 'xy' ) level_z = zw CASE ( 'z0*_xy' ) ! 2d-array IF ( av == 0 ) THEN DO i = nxlg, nxrg DO j = nysg, nyng local_pf(i,j,nzb+1) = z0(j,i) ENDDO ENDDO ELSE DO i = nxlg, nxrg DO j = nysg, nyng local_pf(i,j,nzb+1) = z0_av(j,i) ENDDO ENDDO ENDIF resorted = .TRUE. two_d = .TRUE. level_z(nzb+1) = zu(nzb+1) CASE ( 'z0h*_xy' ) ! 2d-array IF ( av == 0 ) THEN DO i = nxlg, nxrg DO j = nysg, nyng local_pf(i,j,nzb+1) = z0h(j,i) ENDDO ENDDO ELSE DO i = nxlg, nxrg DO j = nysg, nyng local_pf(i,j,nzb+1) = z0h_av(j,i) ENDDO ENDDO ENDIF resorted = .TRUE. two_d = .TRUE. level_z(nzb+1) = zu(nzb+1) CASE DEFAULT ! !-- User defined quantity CALL user_data_output_2d( av, do2d(av,if), found, grid, & local_pf, two_d ) resorted = .TRUE. IF ( grid == 'zu' ) THEN IF ( mode == 'xy' ) level_z = zu ELSEIF ( grid == 'zw' ) THEN IF ( mode == 'xy' ) level_z = zw ELSEIF ( grid == 'zu1' ) THEN IF ( mode == 'xy' ) level_z(nzb+1) = zu(nzb+1) ENDIF IF ( .NOT. found ) THEN message_string = 'no output provided for: ' // & TRIM( do2d(av,if) ) CALL message( 'data_output_2d', 'PA0181', 0, 0, 0, 6, 0 ) ENDIF END SELECT ! !-- Resort the array to be output, if not done above IF ( .NOT. resorted ) THEN DO i = nxlg, nxrg DO j = nysg, nyng DO k = nzb, nzt+1 local_pf(i,j,k) = to_be_resorted(k,j,i) ENDDO ENDDO ENDDO ENDIF ! !-- Output of the individual cross-sections, depending on the cross- !-- section mode chosen. is = 1 loop1: DO WHILE ( section(is,s) /= -9999 .OR. two_d ) SELECT CASE ( mode ) CASE ( 'xy' ) ! !-- Determine the cross section index IF ( two_d ) THEN layer_xy = nzb+1 ELSE layer_xy = section(is,s) ENDIF ! !-- Update the netCDF xy cross section time axis. !-- In case of parallel output, this is only done by PE0 !-- to increase the performance. IF ( simulated_time /= do2d_xy_last_time(av) ) THEN do2d_xy_time_count(av) = do2d_xy_time_count(av) + 1 do2d_xy_last_time(av) = simulated_time IF ( myid == 0 ) THEN IF ( .NOT. data_output_2d_on_each_pe & .OR. netcdf_data_format > 4 ) & THEN #if defined( __netcdf ) nc_stat = NF90_PUT_VAR( id_set_xy(av), & id_var_time_xy(av), & (/ time_since_reference_point /), & start = (/ do2d_xy_time_count(av) /), & count = (/ 1 /) ) CALL handle_netcdf_error( 'data_output_2d', 53 ) #endif ENDIF ENDIF ENDIF ! !-- If required, carry out averaging along z IF ( section(is,s) == -1 .AND. .NOT. two_d ) THEN local_2d = 0.0 ! !-- Carry out the averaging (all data are on the PE) DO k = nzb, nzt+1 DO j = nysg, nyng DO i = nxlg, nxrg local_2d(i,j) = local_2d(i,j) + local_pf(i,j,k) ENDDO ENDDO ENDDO local_2d = local_2d / ( nzt -nzb + 2.0) ELSE ! !-- Just store the respective section on the local array local_2d = local_pf(:,:,layer_xy) ENDIF #if defined( __parallel ) IF ( netcdf_data_format > 4 ) THEN ! !-- Parallel output in netCDF4/HDF5 format. IF ( two_d ) THEN iis = 1 ELSE iis = is ENDIF #if defined( __netcdf ) ! !-- For parallel output, all cross sections are first stored !-- here on a local array and will be written to the output !-- file afterwards to increase the performance. DO i = nxlg, nxrg DO j = nysg, nyng local_2d_sections(i,j,iis) = local_2d(i,j) ENDDO ENDDO #endif ELSE IF ( data_output_2d_on_each_pe ) THEN ! !-- Output of partial arrays on each PE #if defined( __netcdf ) IF ( myid == 0 ) THEN WRITE ( 21 ) time_since_reference_point, & do2d_xy_time_count(av), av ENDIF #endif DO i = 0, io_blocks-1 IF ( i == io_group ) THEN WRITE ( 21 ) nxlg, nxrg, nysg, nyng WRITE ( 21 ) local_2d ENDIF #if defined( __parallel ) CALL MPI_BARRIER( comm2d, ierr ) #endif ENDDO ELSE ! !-- PE0 receives partial arrays from all processors and !-- then outputs them. Here a barrier has to be set, !-- because otherwise "-MPI- FATAL: Remote protocol queue !-- full" may occur. CALL MPI_BARRIER( comm2d, ierr ) ngp = ( nxrg-nxlg+1 ) * ( nyng-nysg+1 ) IF ( myid == 0 ) THEN ! !-- Local array can be relocated directly. total_2d(nxlg:nxrg,nysg:nyng) = local_2d ! !-- Receive data from all other PEs. DO n = 1, numprocs-1 ! !-- Receive index limits first, then array. !-- Index limits are received in arbitrary order from !-- the PEs. CALL MPI_RECV( ind(1), 4, MPI_INTEGER, & MPI_ANY_SOURCE, 0, comm2d, & status, ierr ) sender = status(MPI_SOURCE) DEALLOCATE( local_2d ) ALLOCATE( local_2d(ind(1):ind(2),ind(3):ind(4)) ) CALL MPI_RECV( local_2d(ind(1),ind(3)), ngp, & MPI_REAL, sender, 1, comm2d, & status, ierr ) total_2d(ind(1):ind(2),ind(3):ind(4)) = local_2d ENDDO ! !-- Relocate the local array for the next loop increment DEALLOCATE( local_2d ) ALLOCATE( local_2d(nxlg:nxrg,nysg:nyng) ) #if defined( __netcdf ) IF ( two_d ) THEN nc_stat = NF90_PUT_VAR( id_set_xy(av), & id_var_do2d(av,if), & total_2d(0:nx+1,0:ny+1), & start = (/ 1, 1, 1, do2d_xy_time_count(av) /), & count = (/ nx+2, ny+2, 1, 1 /) ) ELSE nc_stat = NF90_PUT_VAR( id_set_xy(av), & id_var_do2d(av,if), & total_2d(0:nx+1,0:ny+1), & start = (/ 1, 1, is, do2d_xy_time_count(av) /), & count = (/ nx+2, ny+2, 1, 1 /) ) ENDIF CALL handle_netcdf_error( 'data_output_2d', 54 ) #endif ELSE ! !-- First send the local index limits to PE0 ind(1) = nxlg; ind(2) = nxrg ind(3) = nysg; ind(4) = nyng CALL MPI_SEND( ind(1), 4, MPI_INTEGER, 0, 0, & comm2d, ierr ) ! !-- Send data to PE0 CALL MPI_SEND( local_2d(nxlg,nysg), ngp, & MPI_REAL, 0, 1, comm2d, ierr ) ENDIF ! !-- A barrier has to be set, because otherwise some PEs may !-- proceed too fast so that PE0 may receive wrong data on !-- tag 0 CALL MPI_BARRIER( comm2d, ierr ) ENDIF ENDIF #else #if defined( __netcdf ) IF ( two_d ) THEN nc_stat = NF90_PUT_VAR( id_set_xy(av), & id_var_do2d(av,if), & local_2d(nxl:nxr+1,nys:nyn+1), & start = (/ 1, 1, 1, do2d_xy_time_count(av) /), & count = (/ nx+2, ny+2, 1, 1 /) ) ELSE nc_stat = NF90_PUT_VAR( id_set_xy(av), & id_var_do2d(av,if), & local_2d(nxl:nxr+1,nys:nyn+1), & start = (/ 1, 1, is, do2d_xy_time_count(av) /), & count = (/ nx+2, ny+2, 1, 1 /) ) ENDIF CALL handle_netcdf_error( 'data_output_2d', 447 ) #endif #endif do2d_xy_n = do2d_xy_n + 1 ! !-- For 2D-arrays (e.g. u*) only one cross-section is available. !-- Hence exit loop of output levels. IF ( two_d ) THEN two_d = .FALSE. EXIT loop1 ENDIF CASE ( 'xz' ) ! !-- Update the netCDF xz cross section time axis. !-- In case of parallel output, this is only done by PE0 !-- to increase the performance. IF ( simulated_time /= do2d_xz_last_time(av) ) THEN do2d_xz_time_count(av) = do2d_xz_time_count(av) + 1 do2d_xz_last_time(av) = simulated_time IF ( myid == 0 ) THEN IF ( .NOT. data_output_2d_on_each_pe & .OR. netcdf_data_format > 4 ) & THEN #if defined( __netcdf ) nc_stat = NF90_PUT_VAR( id_set_xz(av), & id_var_time_xz(av), & (/ time_since_reference_point /), & start = (/ do2d_xz_time_count(av) /), & count = (/ 1 /) ) CALL handle_netcdf_error( 'data_output_2d', 56 ) #endif ENDIF ENDIF ENDIF ! !-- If required, carry out averaging along y IF ( section(is,s) == -1 ) THEN ALLOCATE( local_2d_l(nxlg:nxrg,nzb:nzt+1) ) local_2d_l = 0.0 ngp = ( nxrg-nxlg+1 ) * ( nzt-nzb+2 ) ! !-- First local averaging on the PE DO k = nzb, nzt+1 DO j = nys, nyn DO i = nxlg, nxrg local_2d_l(i,k) = local_2d_l(i,k) + & local_pf(i,j,k) ENDDO ENDDO ENDDO #if defined( __parallel ) ! !-- Now do the averaging over all PEs along y IF ( collective_wait ) CALL MPI_BARRIER( comm2d, ierr ) CALL MPI_ALLREDUCE( local_2d_l(nxlg,nzb), & local_2d(nxlg,nzb), ngp, MPI_REAL, & MPI_SUM, comm1dy, ierr ) #else local_2d = local_2d_l #endif local_2d = local_2d / ( ny + 1.0 ) DEALLOCATE( local_2d_l ) ELSE ! !-- Just store the respective section on the local array !-- (but only if it is available on this PE!) IF ( section(is,s) >= nys .AND. section(is,s) <= nyn ) & THEN local_2d = local_pf(:,section(is,s),nzb:nzt+1) ENDIF ENDIF #if defined( __parallel ) IF ( netcdf_data_format > 4 ) THEN ! !-- Output in netCDF4/HDF5 format. !-- Output only on those PEs where the respective cross !-- sections reside. Cross sections averaged along y are !-- output on the respective first PE along y (myidy=0). IF ( ( section(is,s) >= nys .AND. & section(is,s) <= nyn ) .OR. & ( section(is,s) == -1 .AND. myidy == 0 ) ) THEN #if defined( __netcdf ) ! !-- For parallel output, all cross sections are first !-- stored here on a local array and will be written to the !-- output file afterwards to increase the performance. DO i = nxlg, nxrg DO k = nzb, nzt+1 local_2d_sections_l(i,is,k) = local_2d(i,k) ENDDO ENDDO #endif ENDIF ELSE IF ( data_output_2d_on_each_pe ) THEN ! !-- Output of partial arrays on each PE. If the cross !-- section does not reside on the PE, output special !-- index values. #if defined( __netcdf ) IF ( myid == 0 ) THEN WRITE ( 22 ) time_since_reference_point, & do2d_xz_time_count(av), av ENDIF #endif DO i = 0, io_blocks-1 IF ( i == io_group ) THEN IF ( ( section(is,s) >= nys .AND. & section(is,s) <= nyn ) .OR. & ( section(is,s) == -1 .AND. & nys-1 == -1 ) ) & THEN WRITE (22) nxlg, nxrg, nzb, nzt+1 WRITE (22) local_2d ELSE WRITE (22) -1, -1, -1, -1 ENDIF ENDIF #if defined( __parallel ) CALL MPI_BARRIER( comm2d, ierr ) #endif ENDDO ELSE ! !-- PE0 receives partial arrays from all processors of the !-- respective cross section and outputs them. Here a !-- barrier has to be set, because otherwise !-- "-MPI- FATAL: Remote protocol queue full" may occur. CALL MPI_BARRIER( comm2d, ierr ) ngp = ( nxrg-nxlg+1 ) * ( nzt-nzb+2 ) IF ( myid == 0 ) THEN ! !-- Local array can be relocated directly. IF ( ( section(is,s) >= nys .AND. & section(is,s) <= nyn ) .OR. & ( section(is,s) == -1 .AND. nys-1 == -1 ) ) & THEN total_2d(nxlg:nxrg,nzb:nzt+1) = local_2d ENDIF ! !-- Receive data from all other PEs. DO n = 1, numprocs-1 ! !-- Receive index limits first, then array. !-- Index limits are received in arbitrary order from !-- the PEs. CALL MPI_RECV( ind(1), 4, MPI_INTEGER, & MPI_ANY_SOURCE, 0, comm2d, & status, ierr ) ! !-- Not all PEs have data for XZ-cross-section. IF ( ind(1) /= -9999 ) THEN sender = status(MPI_SOURCE) DEALLOCATE( local_2d ) ALLOCATE( local_2d(ind(1):ind(2), & ind(3):ind(4)) ) CALL MPI_RECV( local_2d(ind(1),ind(3)), ngp, & MPI_REAL, sender, 1, comm2d, & status, ierr ) total_2d(ind(1):ind(2),ind(3):ind(4)) = & local_2d ENDIF ENDDO ! !-- Relocate the local array for the next loop increment DEALLOCATE( local_2d ) ALLOCATE( local_2d(nxlg:nxrg,nzb:nzt+1) ) #if defined( __netcdf ) nc_stat = NF90_PUT_VAR( id_set_xz(av), & id_var_do2d(av,if), & total_2d(0:nx+1,nzb:nzt+1),& start = (/ 1, is, 1, do2d_xz_time_count(av) /), & count = (/ nx+2, 1, nz+2, 1 /) ) CALL handle_netcdf_error( 'data_output_2d', 58 ) #endif ELSE ! !-- If the cross section resides on the PE, send the !-- local index limits, otherwise send -9999 to PE0. IF ( ( section(is,s) >= nys .AND. & section(is,s) <= nyn ) .OR. & ( section(is,s) == -1 .AND. nys-1 == -1 ) ) & THEN ind(1) = nxlg; ind(2) = nxrg ind(3) = nzb; ind(4) = nzt+1 ELSE ind(1) = -9999; ind(2) = -9999 ind(3) = -9999; ind(4) = -9999 ENDIF CALL MPI_SEND( ind(1), 4, MPI_INTEGER, 0, 0, & comm2d, ierr ) ! !-- If applicable, send data to PE0. IF ( ind(1) /= -9999 ) THEN CALL MPI_SEND( local_2d(nxlg,nzb), ngp, & MPI_REAL, 0, 1, comm2d, ierr ) ENDIF ENDIF ! !-- A barrier has to be set, because otherwise some PEs may !-- proceed too fast so that PE0 may receive wrong data on !-- tag 0 CALL MPI_BARRIER( comm2d, ierr ) ENDIF ENDIF #else #if defined( __netcdf ) nc_stat = NF90_PUT_VAR( id_set_xz(av), & id_var_do2d(av,if), & local_2d(nxl:nxr+1,nzb:nzt+1), & start = (/ 1, is, 1, do2d_xz_time_count(av) /), & count = (/ nx+2, 1, nz+2, 1 /) ) CALL handle_netcdf_error( 'data_output_2d', 451 ) #endif #endif do2d_xz_n = do2d_xz_n + 1 CASE ( 'yz' ) ! !-- Update the netCDF yz cross section time axis. !-- In case of parallel output, this is only done by PE0 !-- to increase the performance. IF ( simulated_time /= do2d_yz_last_time(av) ) THEN do2d_yz_time_count(av) = do2d_yz_time_count(av) + 1 do2d_yz_last_time(av) = simulated_time IF ( myid == 0 ) THEN IF ( .NOT. data_output_2d_on_each_pe & .OR. netcdf_data_format > 4 ) & THEN #if defined( __netcdf ) nc_stat = NF90_PUT_VAR( id_set_yz(av), & id_var_time_yz(av), & (/ time_since_reference_point /), & start = (/ do2d_yz_time_count(av) /), & count = (/ 1 /) ) CALL handle_netcdf_error( 'data_output_2d', 59 ) #endif ENDIF ENDIF ENDIF ! !-- If required, carry out averaging along x IF ( section(is,s) == -1 ) THEN ALLOCATE( local_2d_l(nysg:nyng,nzb:nzt+1) ) local_2d_l = 0.0 ngp = ( nyng-nysg+1 ) * ( nzt-nzb+2 ) ! !-- First local averaging on the PE DO k = nzb, nzt+1 DO j = nysg, nyng DO i = nxl, nxr local_2d_l(j,k) = local_2d_l(j,k) + & local_pf(i,j,k) ENDDO ENDDO ENDDO #if defined( __parallel ) ! !-- Now do the averaging over all PEs along x IF ( collective_wait ) CALL MPI_BARRIER( comm2d, ierr ) CALL MPI_ALLREDUCE( local_2d_l(nysg,nzb), & local_2d(nysg,nzb), ngp, MPI_REAL, & MPI_SUM, comm1dx, ierr ) #else local_2d = local_2d_l #endif local_2d = local_2d / ( nx + 1.0 ) DEALLOCATE( local_2d_l ) ELSE ! !-- Just store the respective section on the local array !-- (but only if it is available on this PE!) IF ( section(is,s) >= nxl .AND. section(is,s) <= nxr ) & THEN local_2d = local_pf(section(is,s),:,nzb:nzt+1) ENDIF ENDIF #if defined( __parallel ) IF ( netcdf_data_format > 4 ) THEN ! !-- Output in netCDF4/HDF5 format. !-- Output only on those PEs where the respective cross !-- sections reside. Cross sections averaged along x are !-- output on the respective first PE along x (myidx=0). IF ( ( section(is,s) >= nxl .AND. & section(is,s) <= nxr ) .OR. & ( section(is,s) == -1 .AND. myidx == 0 ) ) THEN #if defined( __netcdf ) ! !-- For parallel output, all cross sections are first !-- stored here on a local array and will be written to the !-- output file afterwards to increase the performance. DO j = nysg, nyng DO k = nzb, nzt+1 local_2d_sections_l(is,j,k) = local_2d(j,k) ENDDO ENDDO #endif ENDIF ELSE IF ( data_output_2d_on_each_pe ) THEN ! !-- Output of partial arrays on each PE. If the cross !-- section does not reside on the PE, output special !-- index values. #if defined( __netcdf ) IF ( myid == 0 ) THEN WRITE ( 23 ) time_since_reference_point, & do2d_yz_time_count(av), av ENDIF #endif DO i = 0, io_blocks-1 IF ( i == io_group ) THEN IF ( ( section(is,s) >= nxl .AND. & section(is,s) <= nxr ) .OR. & ( section(is,s) == -1 .AND. & nxl-1 == -1 ) ) & THEN WRITE (23) nysg, nyng, nzb, nzt+1 WRITE (23) local_2d ELSE WRITE (23) -1, -1, -1, -1 ENDIF ENDIF #if defined( __parallel ) CALL MPI_BARRIER( comm2d, ierr ) #endif ENDDO ELSE ! !-- PE0 receives partial arrays from all processors of the !-- respective cross section and outputs them. Here a !-- barrier has to be set, because otherwise !-- "-MPI- FATAL: Remote protocol queue full" may occur. CALL MPI_BARRIER( comm2d, ierr ) ngp = ( nyng-nysg+1 ) * ( nzt-nzb+2 ) IF ( myid == 0 ) THEN ! !-- Local array can be relocated directly. IF ( ( section(is,s) >= nxl .AND. & section(is,s) <= nxr ) .OR. & ( section(is,s) == -1 .AND. nxl-1 == -1 ) ) & THEN total_2d(nysg:nyng,nzb:nzt+1) = local_2d ENDIF ! !-- Receive data from all other PEs. DO n = 1, numprocs-1 ! !-- Receive index limits first, then array. !-- Index limits are received in arbitrary order from !-- the PEs. CALL MPI_RECV( ind(1), 4, MPI_INTEGER, & MPI_ANY_SOURCE, 0, comm2d, & status, ierr ) ! !-- Not all PEs have data for YZ-cross-section. IF ( ind(1) /= -9999 ) THEN sender = status(MPI_SOURCE) DEALLOCATE( local_2d ) ALLOCATE( local_2d(ind(1):ind(2), & ind(3):ind(4)) ) CALL MPI_RECV( local_2d(ind(1),ind(3)), ngp, & MPI_REAL, sender, 1, comm2d, & status, ierr ) total_2d(ind(1):ind(2),ind(3):ind(4)) = & local_2d ENDIF ENDDO ! !-- Relocate the local array for the next loop increment DEALLOCATE( local_2d ) ALLOCATE( local_2d(nysg:nyng,nzb:nzt+1) ) #if defined( __netcdf ) nc_stat = NF90_PUT_VAR( id_set_yz(av), & id_var_do2d(av,if), & total_2d(0:ny+1,nzb:nzt+1),& start = (/ is, 1, 1, do2d_yz_time_count(av) /), & count = (/ 1, ny+2, nz+2, 1 /) ) CALL handle_netcdf_error( 'data_output_2d', 61 ) #endif ELSE ! !-- If the cross section resides on the PE, send the !-- local index limits, otherwise send -9999 to PE0. IF ( ( section(is,s) >= nxl .AND. & section(is,s) <= nxr ) .OR. & ( section(is,s) == -1 .AND. nxl-1 == -1 ) ) & THEN ind(1) = nysg; ind(2) = nyng ind(3) = nzb; ind(4) = nzt+1 ELSE ind(1) = -9999; ind(2) = -9999 ind(3) = -9999; ind(4) = -9999 ENDIF CALL MPI_SEND( ind(1), 4, MPI_INTEGER, 0, 0, & comm2d, ierr ) ! !-- If applicable, send data to PE0. IF ( ind(1) /= -9999 ) THEN CALL MPI_SEND( local_2d(nysg,nzb), ngp, & MPI_REAL, 0, 1, comm2d, ierr ) ENDIF ENDIF ! !-- A barrier has to be set, because otherwise some PEs may !-- proceed too fast so that PE0 may receive wrong data on !-- tag 0 CALL MPI_BARRIER( comm2d, ierr ) ENDIF ENDIF #else #if defined( __netcdf ) nc_stat = NF90_PUT_VAR( id_set_yz(av), & id_var_do2d(av,if), & local_2d(nys:nyn+1,nzb:nzt+1), & start = (/ is, 1, 1, do2d_xz_time_count(av) /), & count = (/ 1, ny+2, nz+2, 1 /) ) CALL handle_netcdf_error( 'data_output_2d', 452 ) #endif #endif do2d_yz_n = do2d_yz_n + 1 END SELECT is = is + 1 ENDDO loop1 ! !-- For parallel output, all data were collected before on a local array !-- and are written now to the netcdf file. This must be done to increase !-- the performance of the parallel output. #if defined( __netcdf ) IF ( netcdf_data_format > 4 ) THEN SELECT CASE ( mode ) CASE ( 'xy' ) IF ( two_d ) THEN iis = 1 ELSE iis = is-1 ENDIF ! !-- Do not output redundant ghost point data except for the !-- boundaries of the total domain. IF ( nxr == nx .AND. nyn /= ny ) THEN nc_stat = NF90_PUT_VAR( id_set_xy(av), & id_var_do2d(av,if), & local_2d_sections(nxl:nxr+1, & nys:nyn,1:ns), & start = (/ nxl+1, nys+1, 1, & do2d_xy_time_count(av) /), & count = (/ nxr-nxl+2, & nyn-nys+1, ns, 1 & /) ) ELSEIF ( nxr /= nx .AND. nyn == ny ) THEN nc_stat = NF90_PUT_VAR( id_set_xy(av), & id_var_do2d(av,if), & local_2d_sections(nxl:nxr, & nys:nyn+1,1:ns), & start = (/ nxl+1, nys+1, 1, & do2d_xy_time_count(av) /), & count = (/ nxr-nxl+1, & nyn-nys+2, ns, 1 & /) ) ELSEIF ( nxr == nx .AND. nyn == ny ) THEN nc_stat = NF90_PUT_VAR( id_set_xy(av), & id_var_do2d(av,if), & local_2d_sections(nxl:nxr+1, & nys:nyn+1,1:ns), & start = (/ nxl+1, nys+1, 1, & do2d_xy_time_count(av) /), & count = (/ nxr-nxl+2, & nyn-nys+2, ns, 1 & /) ) ELSE nc_stat = NF90_PUT_VAR( id_set_xy(av), & id_var_do2d(av,if), & local_2d_sections(nxl:nxr, & nys:nyn,1:ns), & start = (/ nxl+1, nys+1, 1, & do2d_xy_time_count(av) /), & count = (/ nxr-nxl+1, & nyn-nys+1, ns, 1 & /) ) ENDIF CALL handle_netcdf_error( 'data_output_2d', 55 ) CASE ( 'xz' ) ! !-- First, all PEs get the information of all cross-sections. !-- Then the data are written to the output file by all PEs !-- while NF90_COLLECTIVE is set in subroutine !-- define_netcdf_header. Although redundant information are !-- written to the output file in that case, the performance !-- is significantly better compared to the case where only !-- the first row of PEs in x-direction (myidx = 0) is given !-- the output while NF90_INDEPENDENT is set. IF ( npey /= 1 ) THEN #if defined( __parallel ) ! !-- Distribute data over all PEs along y ngp = ( nxrg-nxlg+1 ) * ( nzt-nzb+2 ) * ns IF ( collective_wait ) CALL MPI_BARRIER( comm2d, ierr ) CALL MPI_ALLREDUCE( local_2d_sections_l(nxlg,1,nzb), & local_2d_sections(nxlg,1,nzb), & ngp, MPI_REAL, MPI_SUM, comm1dy, & ierr ) #else local_2d_sections = local_2d_sections_l #endif ENDIF ! !-- Do not output redundant ghost point data except for the !-- boundaries of the total domain. IF ( nxr == nx ) THEN nc_stat = NF90_PUT_VAR( id_set_xz(av), & id_var_do2d(av,if), & local_2d_sections(nxl:nxr+1,1:ns, & nzb:nzt+1), & start = (/ nxl+1, 1, 1, & do2d_xz_time_count(av) /), & count = (/ nxr-nxl+2, ns, nzt+2, & 1 /) ) ELSE nc_stat = NF90_PUT_VAR( id_set_xz(av), & id_var_do2d(av,if), & local_2d_sections(nxl:nxr,1:ns, & nzb:nzt+1), & start = (/ nxl+1, 1, 1, & do2d_xz_time_count(av) /), & count = (/ nxr-nxl+1, ns, nzt+2, & 1 /) ) ENDIF CALL handle_netcdf_error( 'data_output_2d', 57 ) CASE ( 'yz' ) ! !-- First, all PEs get the information of all cross-sections. !-- Then the data are written to the output file by all PEs !-- while NF90_COLLECTIVE is set in subroutine !-- define_netcdf_header. Although redundant information are !-- written to the output file in that case, the performance !-- is significantly better compared to the case where only !-- the first row of PEs in y-direction (myidy = 0) is given !-- the output while NF90_INDEPENDENT is set. IF ( npex /= 1 ) THEN #if defined( __parallel ) ! !-- Distribute data over all PEs along x ngp = ( nyng-nysg+1 ) * ( nzt-nzb + 2 ) * ns IF ( collective_wait ) CALL MPI_BARRIER( comm2d, ierr ) CALL MPI_ALLREDUCE( local_2d_sections_l(1,nysg,nzb), & local_2d_sections(1,nysg,nzb), & ngp, MPI_REAL, MPI_SUM, comm1dx, & ierr ) #else local_2d_sections = local_2d_sections_l #endif ENDIF ! !-- Do not output redundant ghost point data except for the !-- boundaries of the total domain. IF ( nyn == ny ) THEN nc_stat = NF90_PUT_VAR( id_set_yz(av), & id_var_do2d(av,if), & local_2d_sections(1:ns, & nys:nyn+1,nzb:nzt+1), & start = (/ 1, nys+1, 1, & do2d_yz_time_count(av) /), & count = (/ ns, nyn-nys+2, & nzt+2, 1 /) ) ELSE nc_stat = NF90_PUT_VAR( id_set_yz(av), & id_var_do2d(av,if), & local_2d_sections(1:ns,nys:nyn, & nzb:nzt+1), & start = (/ 1, nys+1, 1, & do2d_yz_time_count(av) /), & count = (/ ns, nyn-nys+1, & nzt+2, 1 /) ) ENDIF CALL handle_netcdf_error( 'data_output_2d', 60 ) CASE DEFAULT message_string = 'unknown cross-section: ' // TRIM( mode ) CALL message( 'data_output_2d', 'PA0180', 1, 2, 0, 6, 0 ) END SELECT ENDIF #endif ENDIF if = if + 1 l = MAX( 2, LEN_TRIM( do2d(av,if) ) ) do2d_mode = do2d(av,if)(l-1:l) ENDDO ! !-- Deallocate temporary arrays. IF ( ALLOCATED( level_z ) ) DEALLOCATE( level_z ) IF ( netcdf_data_format > 4 ) THEN DEALLOCATE( local_pf, local_2d, local_2d_sections ) IF( mode == 'xz' .OR. mode == 'yz' ) DEALLOCATE( local_2d_sections_l ) ENDIF #if defined( __parallel ) IF ( .NOT. data_output_2d_on_each_pe .AND. myid == 0 ) THEN DEALLOCATE( total_2d ) ENDIF #endif ! !-- Close plot output file. file_id = 20 + s IF ( data_output_2d_on_each_pe ) THEN DO i = 0, io_blocks-1 IF ( i == io_group ) THEN CALL close_file( file_id ) ENDIF #if defined( __parallel ) CALL MPI_BARRIER( comm2d, ierr ) #endif ENDDO ELSE IF ( myid == 0 ) CALL close_file( file_id ) ENDIF CALL cpu_log( log_point(3), 'data_output_2d', 'stop' ) END SUBROUTINE data_output_2d