MOM1_NPD User Manual
Contents
1. MOM NPD User manual OGS Tech Rep 112 2000DOGA1 run the model time pe4131 1 lt lt EOF gt Soutdata out 1 2761 time pe4131 1 lt lt EOF amp contrl days 360 dgnstc 90 tsi 10 nmix 20 eb T restrt T init T ncon 20 travg 90 snaps 30 taver 4000 end amp eddy am 1 6e19 ah 8e19 fkpm 1 5 fkph 0 3 aidif 0 0 amp end amp tsteps dtts 2400 dtuv 2400 dtsf 2400 amp end amp parms acor 0 mxscan 400 sor 1 60 crit 5e8 amp end amp iland alonis 1 1 25 alonis 2 3 25 alonis 3 9 5 alonis 4 25 25 alons 5 33 5 alatis 1 39 0 alatis 2 39 5 alatis 3 41 alatis 4 35 25 alatis 5 35 0 end amp ECO kn 0 25 g 1 157e 5 kp 1 0 alpha 0 75 drate 5 55e 7 rprime 1 18e 6 rkphy 0 000 rkdet 0 000 conc 0 acl 0 growth max 6 83e 6 texp 0 0633 amp end EOF rm cloud_ rm ave rm data rm fort echo MOMdone The script can be activated typing cun npdo Namelists The namelists included in this version of MOM NPD in the version compiled with the type 2 procedure above reported are basically mentioned CZCS Dataset afterGDAAC MLD 01_013 CZCS Guide modified The color of the ocean isdetermined primarily by the abundance of phytoplankton and theirassociated photosynthetic pigments Asthe concentration of phytoplankton pigments increases ocean color shifts from blue to green Taking advantage of this change NASA develop2. largely determined by the amount of memory available Simila dy the amount of disk space required to run the MOM code depends upon the model configuration and code options selected Itis recommended that the MOM NPD code be managed ona UNIX based system although one can certainly manage the code successfully on other platforms Here examples are presented using UNIX commands The MOM NPD code hasbeen designed in some ways to take advantage of certain UNIX utilities Recommended Directory Structure for MOM NPD Files We suggest that users create a specific directory structure on their local machines to hold the MOM NPD files The following UNIX commands generate the desired directory structure user input in bold type cd HOME_MOM directory MOM NPD User manual OGS Tech Rep 112 2000DOGA1 mkdir DATAic directory with input data mkdir MOMor directory with original source mkdir MOMout directory with output list mkdir MOMrun directory with run script and symbolic links mkdir PERPEbc directory with perpetual year forcing functions MOM NPD Source Code Files Before beginning to work with the MOM NPD source code files we strongly recommend that a copy of the orginal files be saved asa backup and protected against being overwritten F and f files these are the Fortran source files Subroutines main programs and functions Files ending with F contain cpp directives while those
3. ending with f do not h files these are the so called include files groups of parameter statements common blocks statement functions etc that are inserted into F files include cpp directives Creating a MOM NPD Executable T Compiling the sorce code Pre processing Computers having C language compilers and or those running a UNIX based operating system can be expected to have cpp available asa stand alone program Many Fortran compilers also have built into them the cpp functionality needed to run the MOM NPD code even if not explicitly stated Lines in MOM NPD code files having a hash symbol sometimes called a pound sign ornumber sign in column one are directives recognized by cpp The MOM 1 code usesa small straight forward subset of the directives available in cpp fortwo general purposes 1 copying code from one file into other files 2 identifying and testing blocks of code associated with various model Options for conditional compilation The first type of cpp directive has the fom include filename h where filename h represents one of the approximately 40 MOM NPD files having a h suffix hereafter referred to as h files The contents of h files are copied to the location of include directives by cpp The most frequent use of h files in the MOM NPD code isto hold common blocks Using include directives ensures that the contentsand arrangement of common blocks are identical in all routines in which th
4. horizontal subgrid scale mixing after Sma gorinsky In this formulation the horizontal eddy diffusion coefficient is proportional to the honzontal grid length and to the local deformation field The coefficients are sensitive to the spatial scales of motion and therefore are relatively small in the open ocean where the scalesof motion are comparatively large and are relatively large in regions where the scales are comparatively small Vertical mixing schemes Vertical mixing applies to both momentum and tracers One and only one scheme must be enabled The default is for explicit solution Option implicitvmix will solve them implicitly constvmix uses constant eddy coeficients fkom amp fkph for linear second order mixing of momentum amp heat in the vertical In the explicit case no implic itvmix convective adjustment handles regions of gravitational insta bility If imp lic itvmix is selected then the convective section is bypassed and the large limits vdclim amp vvc lim are used to set the mixing coeffiecints which handles the insta bility ppvmix isa vertical mixing scheme based on the Pacanowski amp Philander richardson mixing scheme jpo vol 11 11 1981 This parameterization was designed primarily for equatorial models with vertical resolution of about 10 meters between levels in the upper ocean we were most interested in the structure of the undercurrent Previous versions of this code assume
5. MOM NPD User manual OGS Tech Rep 112 2000DOGA1 1 A Crise E Bouma sky Z Boumazka MOM1 NPD User Manual OGS Technical Report 112 2000 DOGA1 31 August 2000 MOM NPD User manual OGS Tech Rep 112 2000DOGA1 MOM NPD Model for the Mediterranean Sea MOM NPD was written asa collaborative effort by Alessandro Crise Guido Crispi using an ecological model orginally designed by Renzo Mosetti The hydrodynamical submodel was developed using GFDL MOM 1 scheme asimplemented during MTP and MTP II MERMAIDS projects in cooperation with Nadia Pinardi MOM 1 documentation Documentation for MOM 1 primarily consists of several README files MOM 1 documentation is less than ideal Some MOM users have referenced the information contained in the documentation files and within the MOM code itself as follows The MOM 1code itself written in Fortran 77 as well as supporting utilitiesand some documentation files are available free of cost via anonymous ftp from a server located at GFDL in Princeton NJ The NPD ecological submodel manual is described in this manual merged together with some relevant art of the MOM 1 manual in order to obtain a comprehensive user manual forthe coupled model How to use MOM_NPD Despiite its potential generality the code should be installed customoized compliled and run according some rules meant to help in particular the beginners To run the MOM NPD code the keyword based customization of the modularcode can be
6. acheived in two waysas we will see later in further detail type 1 C languiage preprocessor cpp or functional equivalent and a Fortran 77 compiler type 2 modem g77 f77 like compilers In both cases the source codes identified with F extenction that contain C like directives forthe optional inclusion of part of the code are precompiled with cpp in order to resolve but in type 1 procedure this must be done calling cpp orcc explicitely With type2 procedure a simple calll to the compiler is sufficient to automatically resolve the cpp directives befor the actual compilation Some non essential but very useful utilities provided with the model assume that the code will be managed ona UNIX T M system and that a C language compiler is available The MOM NPD code usesa small straight forward subset of the directives available in cpp fortwo general purposes 1 copying code from one file into other files 2 identifying and testing blocks of code associated with various model options for conditional compilation Computers having C language compilers and or those running a UNIX based operating system can be expected to have cpp available asa stand alone program type 1 precedure MOM NPD User manual OGS Tech Rep 112 2000DOGA1 Many Fortran compilers also have built into them the cpp functionality needed to run the MOM NPD code even if not explicitly stated Fortran 77 language and exceptions The vast majority of the MOM NPD cod
7. apply to either momentum ortracers but not both For all cases in which implicitvmix is not enabled convective adjustment is the default This mixes tracers vertically in regions of gravitational instability but NOT momentum The effectiveness of convective adjustment is controlled by ncon which is the number of passes through the convective section use grep i n ncon Ffh to see its usage If impllicitvmix is enabled the convective adjustment section is bypassed The assumption is that the vertical mixing schemes will handle it iso pyc mix isa scheme in which a mixing tensor is computed from the local isopycnal slope and the diffusion of tracers is then conducted along that direction It therefore influences both the lateral and vertical mixing of tracersin the model Its use is intended to partially mitigate a perceived shortcoming of z coordinate primitive equation ocean models in parametenzing oceanic mixing due to mesoscale motions Since such mixing is believed to take place along isopycnal surfaces the iso pyc mix option seeks to orient the mixing of tracers along isopycnal surfaces rather than purely honzontal surfaces Asdescribed by Cox Ocean Modelling issue 74 pg 1 5 J une 1987 in addition to specifying the mixing coefficient for along isopycnal diffusion ahisop a non zero background horizontal mixing coefficient ah is often needed to supress gridpoint noise Additionally a constraint is placed on the
8. biomass Original data t be inserted with newtonian relaxation derive from 12 1 4 degreee matrices with Chl mnthly mean CZC Sdata averaged in the pend 1979 1986 expressed in mg Chli m 3 These dataset is expressed in arbitrary units counts which are converted into pigments measurements through an empirical relationship Further informations of data processing can be obtainied from the above www address The conversion of Chl in Nitrogen equivalent Phyto according to Cloem etal 1995 needs to take into account Nitrogen concentration water temperature T and the angularintegrated Photosynthetic Available Radiation Chl TLN A 0 003 0 0154 e xp 0 050T e xp 0 00591 N N C n PhytoC where Chl and Carbon are expressed in mg Phyto is expressed in mgAtN m 3 and PhytoC Phyto 14 Redfield ratio where Redfield ratio C N ratio in phytoplankton 5 68 mgC mgN Inverting the above formula we have Phyto PhytoC 14 Redfield_ratio Chl 0 003 0 0154 e xp 0 050T e xp 0 00591 N N Cn 1 14 Redfield_ratio MOM NPD User manual OGS Tech Rep 112 2000DOGA1 Tland N can be obtained from the the prognostic fields of the models t i k j nm 1 is temperature C t i k j nm 4 is dissolved inorganic nitrogen matN m 3 and irradiance isexpressed in mol quanta m 2 day 1 Acknowledgments For the dynamical part these notes are inspired by Pacanowski R K Dixon and A Rosati 1991 1993 The G F D L Modular Ocean Model User
9. cant truncation due to the 5 point numenics In fact the accuracy of the streram funtion solution is limited only by crit and computer precision Its down side is that it may have stability problems with for topogra phies with sharp gradients and will take more time How much you ask It dependson your configuration Try itand see See congr9 F for more details hypergrid is a version of oldrelax solved using a checkerboard technique first on the black squares then on the red ones Numerically it s very much like oldrelax sor amp crit comments apply Lateral mixing schemes Lateral mixing appliesto both momentum and tracers One and only one scheme must be enabled consthmix is a linear second order mixing with constant eddy coefficients am for momentum amp ah for heat The idea is to choose them just large enough to suppress small scale numerical noise Typical vaules would be am 1 e7 cm 2 sec fora one degree resolution Usually ah is chosen smaller so as not to diffuse away frontal features biha monic isa linear fourth order mixing with constant MOM NPD User manual OGS Tech Rep 112 2000DOGA1 coefficients asin consthmix am for momentum amp ah for heat This is more scale selective than consthmix ie more severly dampsthe small scale features am should be roughly 10 19 to 10 20 cm 4 sec the minus sign is in the eqns fora 1 3 degree resolution nlhmix is the non linear
10. d the t grid was coincident with the u grid and gave good results In the present case we relaxed this assumption We tried thisa few waysand got lots of numerical noise particularly on the equator off Brazil The present configuration minimizes noise If noise develops off steep shelf breaks it can sometimes be suppressed by tuming ona bottom drag cdbot The background mixing bvdc should be kept lt lt 1 0 to avoid diffusing the thermocline away on long time scales In regions of gravitational instability vertical mixing limits vdclim amp vvclim are used In the explicit case no implic itvmix they must satisfy the cfl criterion and may be too small to remove MOM NPD User manual OGS Tech Rep 112 2000DOGA1 the instability However the convective adjustment is operative and tries to remove the instability Whether it doesornot dependson ncon which controlls the number of passes through the convective section If implicitvmix is used vdclim amp vvclim are set large and the convective adjustment section is bypassed This scheme also assumes the use of equal time stepson the density and momentum equations If a longertime step is desired try using the implicitvmix option tc vmix invokes the Mellor Yamada level 2 5 turbulence closure scheme Here the mixing coefficients are a function of the turbulence length scale I the turbulent kinetic energy q 2 the analytically derived stability factor
11. e multitasking crayio for now uses the getwa putwa routines If the disk units are set up on non SSD type disk then the efficiency cp time total time will be bad Thisshows the mismatch in speed between computation amp disk use SSD however is really the way to go lt can be looked atasan extended memory amp the efficiency is much better fio is for fortran direct accessi o The comments from crayio apply here also MOM NPD User manual OGS Tech Rep 112 2000DOGA1 Biological options bio introduces the code part responsible for the evolution of biological state variables this option hasto be used in conjunction with biological input files for what regards the initial condions in the basin plus the restoring data in the Gibraltar buffer zone bioinit but be called at the same time with bio option in order to initialize biological variables It hasbesn meant tomallow a re initialization of the biological code aftera spinup of the dynamical part 5 years of spinup forexample are standard for such model with the paramenters used in the sample run and only then the biological partcan be reinitialized to avoid possible problems at the beginning W hre the baroiclinic adjustment introduces hish sourous advective fluxes pretty dangerous for the correct num erical approximation of the biogeochemical tracer advection bio init must be called only once and then the program must be recomplied with simply the bio optio
12. e conforms to the American National Standards Institute ANSI Fortran 77 standards ANSI 1978 Two extensions found in most Fortran 77 compilers also appear in the code 1 Namelist input output Different compilers have different syntax specifications for na melis input output statements In extreme cases namelist may not be available at all since it is not part of the ANS Fortran 77 standards The user would then need to convertthe code to use standard Fortran read and write statements instead of na melist 2 Double quote string delimiters which appear in some write statements in two MO M NPD files are not part of ANSI Fortran 77 standards One can eliminate this by rewriting the statements to use numbered format statements Feedback from users indicates that very few if any Fortran code changes are needed to get MOM NPD to compile on most machines We have leamed that some Fortran compilers require that all data statements must appear in extemal block data routines In such cases users need to move data statements found in eight or so files into block data routines Memory and Disk A single memory requirement for the MOM NPD code can not be given because the model can be configured in a multitude of ways The MOM1 code s memory requirements can be sized to match the machine it isto be run on In fact users often find their choice of model grid dimensions input output options physics options and other model charactenistics are
13. ed the Coastal Zone Color Scanner CZCS which was launched on the Nimbus 7 satellite in October 1978 During its 7 1 2 year life time MOM NPD User manual OGS Tech Rep 112 2000DOGA1 October 1978 J une 1986 CZCS acquired nearly 68 000 im ages each covering up to 2 million square kilometers of oceansurface The Coastal Zone Color Scanner CZCS wasa multispectral line scanner devoted principally to measurements of ocean color Ithad six spectral bands channels There were four channels de voted to ocean color each of 20 nanometer band width and centered at 443 520 550 and 670 nanometers These are referred to as channels 1 through 4 respectively Channel 5 sensed reflectedsolar radiance and had a 100 nanometer bandwidth centered at 750 nanometers and a dynamic range which was more suited to land Channel 6 operated in the 10 5 to 12 5 micrometer region and sensed emitted thermal radiance for derivation of equivalent black body temperature The CZCS level 1 2 and 3 data productsare available from the Goddard Space Flight Center GSFC Distributed Active Archive Center DAAC For further details see http eosdata gsfc na sa gov DATASET DOCS czcs dataset html The Coastal Zone Color Scanner Level 3 images are available forthe Mediteranean in the GrADS machine independent format and can be prcocessed according the Goddard Space Flight Center prcedure CZCS data procedure for the conversion of chlorophyll into Nitrogen equivalent
14. en the F suffix The f suffix is given to files ready for the Fortran compiler those that had no cpp directives to begin with and those which already have been passed through cpp One can envision cpp as doing its work in the example case in two stages During the first stage include filename h directives in mom F are replaced by the contents of the h files themselves In the second stage cpp evaluates test directives based upon which code options have been defined by the D option to cpp 77 execution After the compile is complete the f files produced by cpp typically are no longer needed Modifications should be made only to copies of the original F f and h files Users should not attempt to make code modifications to the T files produced by cpp We are aware of two types of situations in which one may wish to retain the cpp output files fora short period of time Some usersfind portions of MOM 1 code F files that contain a high density of cpp directives somewhat difficult to read In such casesone may find it helpful to view the cpp output files since only the code relevant to the options selected will MOM NPD User manual OGS Tech Rep 112 2000DOGA1 appear Also some debugging tools may require the presence of the f output files In a type 2 procedure such compiling phase can be nduced as follows refering to the base installation included in the test example 77 02 0 MOMr
15. ey are included Various options available in MOM NPD are designed to be tumed on or off by MOM NPD User manual OGS Tech Rep 112 2000DOGA1 the user before compilation The placement of a second kind of cpp directive inside MOM code files allows blocks of code to be identified and tested for conditional compilation This second set of cpp directives will be referred to collectively as test directives Blocks of code affected by cpp test directives are identified as beginning with lines having the form ifdef option ifndef option or if expression where option represents the name of a MOM NPD option and expression represents potentially more complicated tests involving code options In the MOM NPD code test expressions are constructed using C language operators amp amp and or not Each test directive hasa matching endif directive marking the end of the block of code Test directives can be nested If then else constructscan be made through the use of else or elif expression else if lines When executing the cpp command users can specify which MOM code optionsare to be defined tumed on Forexample the command cpp P Dnamea Dnameb mom F mom f would cause cpp to take the file mom F as input process it and create an output file named mom f Two code options namea and nameb have been tumed on defined in thisexample We have adopted the convention that Fortran routines containing cpp directives are giv
16. g messes the solution up amp should only be done asa last resort Itis provided asa backward compatability feature for the COX code Note Both in MOM and in the COX code the amount of filtering depends on the number of ocean points within the longitudinal stip This actually leadsto possibly different truncations at the same latitude firfil is a simple finite impluse response filter based on the familiar 1 4 1 2 1 4 weights the response function isa cosine It comes in two flavors One to with symmetric boundary conditions ie for tracers amp one forasymmetric boundary conditons ie momentum It also works on longitudinal strips of ocean points amp may be applied an arbitrary number of times foran arbitrary number of latitudes The best that can be said for it is that the reason for using it isthe same as for fourfil it is quite fast amp we really haven t done any long running comparnsonsto comapre it to fourfil Miscellaneous keepterms allows the use of arrays instead of statement functions for component tems in the equations This may be desirable when using component terms over amp over for some purpose The trade off isthe MOM NPD User manual OGS Tech Rep 112 2000DOGA1 extra memory needed forthe arrays versus the extra time needed to recalculate the statement functions See size F to look at resource requirements nohilats stands for no high latitudes If the latitudinal domain of the m
17. ime levels are used for slabs on simulated disk to save memory and this is incompatible with nta sks gt 1 This condition may in principle be removed by setting ntlev 3 when enabeling the diskless option We haven t explored it yet So farwe have gotten to the point where all combinations of unitasked multitasked amp autotasked runsgive the same answers down MOM NPD User manual OGS Tech Rep 112 2000DOGA1 to the last hex digit To really utilize this feature on the CRAY YMP the SSD must be used forthe slabs unless large memory amp diskless is used We have not yet added the best i o scheme for the SSD so the wall clock times are higherthan they should be timing gives cp amp wall clock times for running on a CRAY YMP Additionally the time per grid point per time step is calculated Filtering Filtering is used to combat the effect of the convergence of meridians shrinking longitudinal grid spacing near the poles The problem is that the grid spacing severly limits the time step especially when using large density time steps due to the CFL criterian Filtering relaxes this constraint by truncating high wave numbers fourfil is a fourier filter which acts on longitudinal strips of ocean points The number of wave numbersto be allowed at each latitude isdecided upon amp wavenumbers above are truncated without using any reasonable window The best that can be said for it isthat it is time consumin
18. ions are enabled by using the D option form on the compiling statement as shown above in vi This form may also be used ona cpp statement see cray run to preprocess the code before sending separate it to the compiler Here isa terse summary of what s currently available Extemal mode Basically there are four Poisson solvers All of them require the rigidlid option to also be enabled MOM NPD User manual OGS Tech Rep 112 2000DOGA1 oldrelax isthe one from Mike Cox socean model If you use it you should twiddle with sor the overrelaxation constant to minimize the number of scans for your geometry crit controls the accuracy See relax F for more details Ccongrad5pt isa conjugate gradient technique sor is not needed here but crit still controls the accuracy This scheme is faster than oldrelax but may be a bit less stable in the presents of steep topographic gradients The surface pressure iscalculated only as a basisfor comparison with congrad9pt See congr5 F for more details congrad9pt is another conjugate gradient solver which uses a 9 point laplacian instead of the 5 point laplacians used by the other schemes It s advantage is that it isthe most accurate amp allows the surface pressure to be calculated directly Reducing cnt will give better results as shown by the closed line surface pressure integrals in the test case than congrad5pt because asin all other schemes there is Signifi
19. n s f Sindata azoini data azoini data Ter Assign output units In s f Soutdata rst 1 Fort 21 ln s f outdata hst 1 fort 22 ln s f Soutdata tke 1 tke ln s f outdata bio 1 bio ln s f dev null fort 70 link clouds files In s f Sforcing clouds cloud apr80 cloud apr80 In s f Sforcing clouds cloud aug80 cloud aug80 In s f Sforcing clouds cloud dec80 cloud dec80 ln s f forcing clouds cloud_feb80 cloud_feb80 In s f Sforcing clouds cloud jan80 cloud_jan80 In s f Sforcing clouds cloud jul80 cloud jul80 ln s f forcing clouds cloud_jun80 cloud_jun80 ln s f forcing clouds cloud_mar80 cloud_mar80 In s f Sforcing clouds cloud may80 cloud may80 ln s f forcing clouds cloud_nov80 cloud_nov80 In s f Sforcing clouds cloud oct80 Cloud oct80 In s f Sforcing clouds cloud sep80 cloud sep80 f rnn link NMC files ln s f forcing yr1980 apr80_ave apr80_ave In s f forcing yr1980 aug80_ave aug80_ave ln s f forcing yr1980 dec80_ave dec80_ave ln s f forcing yr1980 feb80_ave feb80_ave ln s f forcing yr1980 jan80_ave jan80_ave ln s f forcing yr1980 jul80_ave jul80_ave ln s f forcing yr1980 jun80_ave jun80_ave ln s f forcing yr1980 mar80_ave mar80_ave ln s f forcing yr1980 may80_ave may80_ave ln s f forcing yr1980 nov80_ave nov80_ave ln s f forcing yr1980 oct80_ave oct80_ave In s f forcing yr1980 sep80_ave sep80_ave
20. n to hold the restart information unaltered Running the executable To run the executable a number of step must be done in order to the script with your computational environment In the following a sample script is reported together with main comments the most probable modifications are highlited in italics The large part of the run script is general for MOM NPD fornulation although can be altered in case of digfferent forcing functions sa in the case of high frequency forcings Script file to run the MOM model on the SGI Origin 2000 architecture Tr run identification echo NPD Emil amp Zvetanka echo no clouds echo run identification 1 working directory and subdirectories run oga3 echouser NPD MOMrun src oga3 echouser NPD MOMor indata oga3 echouser NPD DATAic MOM NPD User manual OGS Tech Rep 112 2000DOGA1 outdata oga3 echouser NPD MOMout forcing oga3 echouser NPD PERPEbc pess symbolical links to initial and surface boundary conditions ln s f indata medtop431 dat fort 33 ln s f indata medts431 dat fort 53 In s f Sindata levsal fort 44 In s f Sindata initbio data initho data I
21. odel is limited to equatonal redions for instance then the metric nonlinearterms in the momentum equations may be dropped lt willsave some time resto rst stands for restore surface tracers to some prescribed values by a newtonian damping term In the test case the restoration is globally back to the annual mean conditions on a time scale of 50 days Without much work this option could easily be extended to damp certain regions while leaving others alone by simple making the damping time scale a function of latitude for instance testc fl tests whether any velocity exceeds the local cfl criterion by a factor of cflcrt see blkdta F It will print out the places where the cfl criterion is most nearly met on diagnostic time steps If the cflcrt factor is exceeded on any time step MOM will stop amp print out the surronding variables It s not meant to be left on all the time If MOM blows up it Can quickly point to where it s happening VO options diskless simulates disk usage using an amay in memory Since there is no wall clock time lost during disk transfers because they are memory to memory transfers this method gives the best efficiency cp time total time The down side is that the model may not be able to fit into available memory Asa memory saving feature diskless keeps only two time levels of prognostic variables instead of the usual three See size F to look at resource requirements Also se
22. s Sm amp Sh and the boundary conditions There are two options to compute the length scale eg solvesan additional equation for q 2I from which the length scale may be derived at each level alg obtains the length scale from an algebraic relationship This option may be sufficient for the boundary layer but not in cases where there would be multiple turbulent regimes The horizontal diffusion coefficient of turbulent kinetic energy is set using the variable aq ctcmix h namelist amp blkdta f For this scheme to be effective implic itvmix should be enabled and also the vertical resolution should be sufficient implicitvmix solves the vertical diffusion term implic itly for all vertical mixing schemes This allows for large mixing coefficients without the need to reduce the timestep With this option enabled convective adjustment is not done and the unstable density profile is mixed using large mixing coefficient limits vvclim amp vdclim For constvmix and po pvmix these limits are set in blkdta F and for tcvmix the coefficients are computed mplic itvmix also requires that aidif the implicit factor be set see cvmix h namelist and blkdta F O lt aidif lt 5 stable if ka ppa dt dz 2 lt 1 2 4 aidif 5 lt aidif lt 1 always stable where kappa isthe max vertical mixing coefficient Hybrid mixing schemes MOM NPD User manual OGS Tech Rep 112 2000DOGA1 These are schemes which
23. s Guide version 1 GFDLOcean Group Technical Report No 2 NOAA Geophysical Fluid Dynamics Laboratory Princeton NJ The original coupled model was developed under EU MAS2 C T93 0055 and MAS3 CT96 0051 MEDNET contracts The data assimilation scheme was implemented under NATO Linkage Grant 974997
24. steepest isopycnal slope slmxr that the scheme will consider when computing the components of the mixing tensor A typical value of slmxr is 100 0 which translates into a slope of 1 100 Steeper slopes would then be considered to be 1 100 for the purpose of computing the mixing tensor A vertical mixing scheme must be specified for use with iso pyc mix The zz component of the isopycnal mixing is added to the vertical diffusion coefficient produced by the vertical mixing scheme Since isopycmix only affects tracers one must also specify a lateral mixing scheme from the above list to be used for momentum This additional lateral mixing scheme hasno effect on tracer diffusion Grid options cyclic is for setting cyclic boundary conditons in a zonal direction If enabled anything exiting the eastem side of the grid comes back in through the westem side If not enabled then solid walls are assumed on the eastem and westem boundaries of the model MOM NPD User manual OGS Tech Rep 112 2000DOGA1 symmetry is for setting a symmetric condition across the northem boundary of the model This assumes the line of symmetry is at the equator which should be row jmt 1 on the velocity grid The condition is Tracers at row jmt on the t grid tracers at row jmt 1 on the t grid meridional velocity at row jmt 1 on the u grid is zero meridional velocity at row jmt on the Uu grid isthe negative of the meridional velocity at ro
25. un pe4131 npdo3 N Dskipland Drigidlid Dkeepterms Dislands N Doldrelax Dconstvmix Drestorst Dbiharmonic Ddiskless Dimga Ddprec Dbio Dab4 Dceh3 Dnewlight F The level of optimization can be raised to O3 but nome care must be taken because of the possible modifications of the results Code crashes have been experienced on old IBM AiX 3 2 5 workstations and IBM xif compiler with O3 optimization The compiling procedure must be run un MOMordirectory and generatsan executable code in MOMrun directory Thiscan be used succesfully also with g77 GNU public domain compiler Compiling directivesand Options excerpted from MOM1 user guide A very powertul feature of MOM isthe ability to define ifdef options and configure the model in a manner suitable forthe problem being investgated The options are quite extensive and range from various subgnd mixing parameterizations to computing performance enhancements The current implementation is such that if there are more than two options within a catagory there is no default value so care should be taken to define the desired option A subroutine checks wihin MOM searches the selcted options to see if any of them conflict For example if a user in selecting an VO scheme tumed on both diskless fully contained within memory and fio Fortran dirct access an obvious inconsistency exists and a message to this effect willbe printed and MOM will halt Opt
26. w jmt 2 on the u grid stream funcion at row jmt negative of the stream function at row jmt 1 on the t grid If not enabled solid walls exist at the northem and southem boundaries Optimizations skipland allows calculations to be done over blocks of ocean while skip ping land points The trade off here is the time saved by not doing the caluclation overland versus the time used in starting amp stoping a lot of vectors It s really a function of land mass geometry A global ocean would be a good candidate for this option but a idealized basin woulod not multita sking allows the slabs to be divided into tasks Each task contains approximately the same number of slabs and there should ideally be one task per processor This is all handled by simply spedcifying ntasks to be the number of processors we ve defaulted it to 8 in blkdta F Since all tasks are independent all taskscan be worked on simultaneously This reduces the total wall clock time forthe job but NOTthe total cpu time Note ntasks may be set gt 1 even when unitasking one processor but this isnot recommended since there is extra work being done at the interfaces between tasks Also on diagnostic tracer averaging and data saving time steps MOM reverts to one task of jmt 2 slabs so effectively multitasking is shut down for these time steps multita sking currently doesn t work with the diskless see below option because only two ntlev 2 t
Download Pdf Manuals
Related Search