############# GEMDM Model -- Version 3.2.0 ############### January 20 2005 Version 3.2.0 of the GEMDM dynamics is now ready and available in directories: "$ARMNLIB/modeles/GEMDM_shared/v_3.2.0/RCS_DYN" "$ARMNLIB/modeles/GEMDM_shared/v_3.2.0/RCS_4DVAR" It is connected to RPN/CMC physics version 4.2 in directory: "$ARMNLIB/modeles/PHY/v_4.2/RCS:4.2" ############################################################################ MAIN FEATURES: There has been many significant changes in this version which do not seem to change the meteorological results. This version will however not bit reproduce results of version v_3.1.2. It is connected to Physics version 4.2 so please refer to the Physics 4.2 release notes for details. Documentation on previous GEMDM releases can be found in: http://notos.cmc.ec.gc.ca/mrb/rpn/eng/gemdm/gemdm.html This note is available in the website under Version Rel. Date at v 3.2.0 Please note that there are additional files at this GEMDM website (v3.2.0) which are related to this version: "gem_settings.nml.txt" "outcfg.out.txt" "configexp.cfg.txt" #+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+# IMPORTANT NOTE The source code has been divided in 2 RCS with corresponding libraries. The non-linear forward code is stored in $gem/RCS_DYN whereas the 4DVAR code is stored in $gem/RCS_4DVAR. The makefile load target for the non-linear GEMDM is unchanged (make gemdm). An additional target has been added to the Makefile in order to load an executable containing 4DVAR capabilities - make gem4d. #+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+# #+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+ Changes that affect run time computation: #+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+ 1) New physics package (4.2) (Leduc/Bilodeau), 2) New optimized advection code by Valin+Tanguay. Control keys for this "turbo-advection" are in "gem_cfgs" namelist: - Adw_nosetint_L (default = .false.) when set to .true. this option activates an optimized version of subroutine adw_main_2_pos.ftn. The better performance will mainly be noticed under TL and AD configurations, - Adw_lag3d_L (default = .false.) when set to .true. this option activates the Lagrange cubic interpolation in the vertical. With this option both horizontal and vertical interpolations have been optimized, 3) System library "MASSVP4" has been upgraded since Sept.27 thus, comparison with previous operational executables will definitely not bit pattern validate, 4) Important correction in the computation of vertical velocity in pressure coordinate - impacts the physics behaviour (Gravel), 5) For LAM only, blending with lateral boundary conditions is enlarged by 1 point around global boundaries (Desgagne), 6) Bugfixes: - vertical sponge: correction to indexing - computation of dzmin in p_vmmphy.ftn #+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+ Major changes to the output package affecting final result and format of output. Output requests and configuration are also affected: (Lee & Desgagne) #+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+ 1) Output is no longer in SLAB files. The model directly produces FSTD RPN file format using grid type '#'. Moreover it is now possible to put more than 1 timeframe per file and to produce less files than the actual number of processor used for the run. This is made possible by the use of regular block partioning of the PEs themselves with "ptopo" namelist variables Ptopo_nblocx and Ptopo_nblocy. ex: Ptopo_npex = 4, Ptopo_npey = 8, Ptopo_nblocx = 2, Ptopo_nblocy = 4, will block partition the 32 PEs into 8 regular blocks of 2x2 PEs. Output of the whole domain would here produce 8 files. Of course this reduction in the number of output files is done at the cost of extra communications in order to collect the data. This can however be minimized through a wise choice of Ptopo_nbloc* that would actually minimized inter-node communications. The "delamineur" program is no longer used but the "bemol" program can still be used to re-assemble the global domain (d2z=1 in configexp.dot.cfg). The number of PEs must be divisible into the number of blocks on each axis i.e. mod(Ptopo_npex,Ptopo_nblocx)=0 and mod(Ptopo_npey,Ptopo_nblocy)=0. Be aware that an output of 4x8 versus 2x4 will look slightly different because of the compaction of FSTD files. The filtered fields are also slightly changed because they were previously scaled before being filtered in v_3.1.2 and now, it is the reverse. 2) Output requests and configuration have changed in two files. a) in gem_cfgs namelist All keys names Out_??? are changed to Out3_??? Out_xnbit and Out_filt keys are deleted and the control is now in "outcfg.out" file. There are 3 more new keys: (i) Out3_ip3 - to change value of IP3 written in output (ii) Out3_debug_L - control of stdio from FSTD functions (iii) Out3_etik_S - change "etiket" given in GEMNTR during GEMDM run (iv) Out3_closestep-requested number of steps to close output file See gem_settings.nml.txt. b) in outcfg.out file new output command to control special compaction on certain fields ie:xnbit([GZ,P0],bits,32) new output command to control special filter on certain fields ie: filtre([PN,GZ,TT], coef, 0.5, pass,2) user must now request physics fields using the command "sortie_p" and can use the variable names given in the first OR second column of the physics buses descriptions in the listing of the runs. ie: sortie_p([MF,MY,MG,ALVIS,LA,DLON,WFLUX,N2],grid,1,levels,1,steps,2) Variables requested under "sortie" are considered to be from dynamics. Definition of grid other than size of model or core is now defined with "reduc". Please read the file "outcfg.out.txt" for more details #+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+ Other Changes related to new schemes and/or packages: #+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+ 1) GEMDM offline mode (MEC). (P. Pellerin & Delage) E_schm_offline_L (default = .false.) 2) Patch for the surface prognostic variables when running without physics (Gravel) Schm_sfix_L (default = .false.) 3) Background vertical diffusion (Gravel) Vrtd_L (default = .false.) 4) Updated canonical cases (Gravel) 5) 4Dvar for operational implementation (Tanguay,Buehner,Zadra) 6) Re-introduction of variable diffusion operator (Toviessi & Qaddouri) Hzd_difva_L (default = .false.) 7) New keys to control physics v_4.2: P_pbl_impflx_L - implicit surface fluxes (default = .false.) P_pbl_z0tlat - latitude used to specify Z0T over water (default = 0.,0. ) P_cond_pcptype_S - PCPTYPE (default = 'NIL' ) P_cond_kticefrac_L - ice fraction (default = .true. ) #+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+# Sample run configurations can be found at $gem/run_configs Sample analysis given in these examples are not guaranteed to exist #+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+# ############################################################################