Home

VIFOP User Guide: an example of set up and utilization

image

Contents

1. b LNEAR LR LBUB 100 km x x x X XA xx x X X DOOOCX XX X X DODOCUX XX X DOO K x x x x x 5 s Q x x XQOOOCOQOOOCUX OX X X X KOKOMO OK X X X X X X XX30000QO0OCX X X X XXOOOOOOOOOOOCUX XCX7X AA X X XOOOOOOOOOOOOOCXX X X XXX0000O0QOOCOC X XX X XX XX OXOXOXOGQOXOGOOX OX XC xx X KR KKK xxx X x3 x gt 4 CK KK ORO X X X X X X X X X XXXXOOQQOOOCX X X X XD XXX X x MR KK KKK RK X E x X XX xXx XX XOOXCX XX 50 100 X distance km Figure 3 a the red and the green x signs respectively represent the location of the LR grid and HR density points on a cartesian map The magenta squares mark the LR selected points to be taken into account in the interpolation of the LR data on the HR density point highlighted by the blue asterisc in the case of LBUB LR 50km b same but for LBUB LR 100km 0 2 0 18 0 16 0 14 0 1 0 08 Normalized weight 0 06 0 04 0 02 0E0 Normalized weight of LR data in the interpolation scheme X X ry Th mp KO y Xs FN 50E3 100E3 X distance km L ae ae Agauses 30 km Rgauss 50 km Rgauss 100 km Fi Es 1000 LT Lia lt bo 150E3 XX 200E3 Figure 4 Normalized weight of LR variables in fonction of the distance Each set of point correspond to a different R_GAU SS value X X XOXOXOOOOQOOQOCOX X X X X NY X DOXKOO0ODO0pOQOO X XX X X 200 From an academic sea
2. web browser interface associated to CVS allows to navigate into the sources and to retrieve any release of any file of the project To download all the files related to a VIFOP given release connect the Noveltis web site http www noveltis fr mf step wp7 then navigate through out WP7 products How to obtain After filling out your login name and password follow the two steps illustrated in figure 1 to select a VIFOP release and to download it either as a tarball 2 2 What is included Untar the downloaded file to create a vifop directory that can be renamed The material held in the vifop directory may not be modified by regular users The VI FOP subdirectories include the FORTRAN77 source files SOURCES the com mon declarative files COMMONS the namelists to set VIFOP input arguments NAMELIST the files defining the low and high resolution LR and HR domains the LR input fields LR and HR the mathematical libraries LIBRARY Sev eral scripts and makefiles are also included in vifop to install compile and run VIFOP 3 Running VIFOP The install vifop script creates a LOCAL directory at the same level as the installed VIFOP version currently the vifop directory In that LOCAL directory whose architecture is a copy of vifop plus a working and a compilation and run diagnostic directories WORKDIR and OUTPUT the user can handle the files for instance modifying or adding source files in the SOURCES directory compil
3. OUT restart0 out file e 3 Copy LR fields in HR variables for model tangent testing purpose The gaussian interpolation of LR data on the HR grid 1s driven by two param eters LBUB LR and R_GAUSS tunable in the namelist namelist int The flag LBUB LR is the radius around HR points defining area in which LR points are selected to be taken into account in the interpolation The influence of LBUB LR on the number of considered LR points is illustrated in fig 3 Higher LBUB LR values slower is the interpolation calculation Once selected the data at LR grid points Yzpr are interpolated into the HR variable Xyp following D i k l 2 York expl R GAUSS San D k l IR GAUSS where D k l is the distance between the HR point and the LR point k k are indexes in the 1D vector representation The R_GAUSS parameter allows to adjust the LR point weights as illustrated in figure 4 Xur l 1 3 OUT is currently the OUTPUT directory Y distance km 50 0 50 a LNEAR LR LBUB 50 km rx X X X X x 4 CK MICROIOC X OOOO X X gt E X i 1 E X X X X OOOOOOQOOCOK x x x x K K 5 14 x R x K K X X x x x 00000090OUx X X x X X X X0OOOO0O000OOOUX X X X X X X XXX0OOOOQUCXXX X X X X X X OOOQOOOOOOUCX X X X X XXXXX X X X XX0000pOQOO X X X X X X X X900OOOOOOOOCOXCX X X X X X0X0OOOQQOX X X X X xx xX x 100 150 X distance km 200 Y distance km 50 0
4. This block is repeated for the 4 grid point types 4 2 2 The Nearest LR stations around HR grid points block Indexes of Nearest LR stations around HR grid points are dumped in LNEAR HR The way to retrieve LR point indexes is the same as described in the previous sec tion but using LVEARBEG HR in place of LNEARBEG LR In addition it is printed the Distance to all LR stations from a HR point selected by ISEC HR OUT and JSEC HR OUT in namelist output This block is repeated for the 4 grid point types 4 23 5 The Interpolated 2D HR variables block V2D and V3D HR variables resulting of the V2D and V3D LR variable interpola tions are displayed in this block 4 3 Optimization The optimization leads the model control variables to a compromise between their values involved by the input data and the physical constrains they must verify The physical constrains are gathered in modulus the applied modulus can be selected in 10 E ME E p P gt NSSE NSSE NSSE OL 09 0S OF o o o o OF o Oc o o o o o ISN ASSN ISN ASSN ISN ASSN d e f Figure 5 a LR sea surface elevation anomaly field NSSE NSSE ranges from 2 to about 75 cm The domain size is 300x300 km wide The black rectan gle indicates the HR subdomain b NSSE interpolated on the HR grid with LBUB LR 50 km and R GAUSS 50 km c same as b but for LBUB_LR 100 km and R_GAUSS 100 km d 3 North South sections of graph a are symbol ized by
5. 13 4 3 4 4 3 5 4 3 6 4 3 7 4 3 8 4 3 9 4 3 10 The State Vector Covariance Matrix block Level 2 Users 13 The Constrain Matrix block Level 2 Users 14 The External mode variables block 14 The After opt analyzed 2D or 3D HR variable block 15 The Ext mode analysis block 15 Check VIFOP or The Results of the optimization Error on constrains blocks 02 2 000 4 15 Sensibility tests lt s onse cs s dw odo RE RRA EES 16 1 Introduction This document intends to show a practical installation compilation and run of VIF OP Variational Initialization and Forcing Platform The case study is imple mented on a small domain in order to be able to easily play around in tuning the VIF OP control parameters The first part of this document looks like a user guide for the installation compilation and run of VIFOP the second part describes some optimized field sensitivities to control parameters VIFOP is a research oriented toolbox Thus for users to add developments some sections consist of technical descriptions of the vector and matrix management in VIFOP they are labelled Level 2 Users A runtime check of VIFOP is proposed in section 2 Install the toolbox 2 1 To get the toolbox source files The source files along the various releases of VIFOP are managed by CVS Con current Versions System http www cvshome org a free source manager The
6. HR bck nc Frame 90 in File champ HR ana dssedt1 nc Frame 90 in File champ HR ana duvbdt1 nc UBAR USI UBAR USI UBAR USI Y km Y km Y km gilles Fri Nov 28 01 14 00 2003 gilles Fri Nov 28 02 03 18 2003 gilles Fri Nov 28 01 14 21 2003 X km X km X km Range of UBAR 0 to 95 5113 USI Range of UBAR 0 to 101 997 USI Range of UBAR 0 to 131 415 USI Range of X 1927 69 to 1846 7 km Range of X 1927 69 to 1846 7 km Range of X 1927 69 to 1846 7 km Range of Y 0 to 0 km Range of Y 0 to 0 km Range of Y 0 to 0 km Frame 90 in File champ HR bck nc Frame 90 in File champ HR ana dssedt1 nc Frame 90 in File champ HR ana duvbdt1 nc VBAR USI VBAR USI VBAR USI E a g gt gt gt X km X km X km Range of VBAR 0 to 188 638 USI Range of VBAR 0 to 188 638 USI Range of VBAR 0 to 326 696 USI Range of X 1927 69 to 1846 7 km Range of X 1927 69 to 1846 7 km Range of X 1927 69 to 1846 7 km Range of Y 0 to 0 km Range of Y 0 to 0 km Range of Y 0 to 0 km Frame 90 in File champ HR bck nc Frame 90 in File champ HR ana dssedt1 nc Frame 90 in File champ HR ana duvbdt1 nc Figure 7 Comparison of the optimization using the type 1 and 2 and 3 constrains a Interpolated sea surface elevation b same as a but for type 1 contrain opti mization c same as b but for type 3 Ald 4 contrain optimizations d same as a but for t e same as b but for u f same as c but for u g same as a but
7. back and forth from the 2D to the 1D representation The J2LX COM table is shown for the state vector variables in plots out NUM ZTA corresponds to the numbering of n RHBAR to p VBARX to u and VBARY to v Thus one can remark that 1 2 is the first value of X the second is p 1 2 and the third 2 2 because VBARX is not de fined in the first column and first row of the domain Similarly VBARY is not defined in the first 2 rows and in the first column 4 3 4 The State Vector Covariance Matrix block Level 2 Users This is the B matrix of equation 2 17 of Auclair et al 2000 The way the matrices are stored in memory is explained in the Matrix representation section of the VIFOP reference manual In this block are printed the variables required to retrieve any value of the state vector covariance matrix e VCOVX COM values of the state vector covariance matrix e LINCOVX COM beginning of each matrix line e JCOVX COM non zero element index in the state vector It is very difficult to print matrices that have dimensions up to 3 x 3 at each grid point Consequently only the matrix elements are printed type by type in VCOVX COM LINCOVX COM indicates the storage index for a new matrix line in VVOVX COM Thus the first element of VCOVX COM corresponds to the point 1 2 where LIN COVX COM is 1 and as LINCOVX COM is 3 at 2 2 one can deduce there are 2 lines in the matrix at 1 2 Moreover there are 3 lines at the other point
8. for y h same as b but for v i same as c but for v References Auclair F S Casitas and P Marsaleix 2000 Application of an inverse method to coastal modelling J Atmos Oceanic Technol 17 1368 1391 Blumberg A F and G L Mellor 1987 A description of a three dimensional costal ocean circulation model volume 4 of Three Dimensional Costal Ocean Models American Geophysical Union 20
9. km Range of Y 0 to 0 km Range of Y 0 to 0 km Frame 90 in File champ HR bck nc Frame 90 in File champ HR ana dssedt1 nc Frame 98 in File champ HR ana dssedt2 nc VBAR USI VBAR USI VBAR USI E a g gt gt gt X km X km X km Range of VBAR 0 to 188 638 USI Range of VBAR 0 to 188 638 USI Range of VBAR 0 to 348 407 USI Range of X 1927 69 to 1846 7 km Range of X 1927 69 to 1846 7 km Range of X 1927 69 to 1846 7 km Range of Y 0 to 0 km Range of Y 0 to 0 km Range of Y 0 to 0 km Frame 90 in File champ_HR_bck nc Frame 90 in File champ HR ana dssedt1 nc Frame 90 in File champ HR ana dssedt2 nc Figure 6 Optimization using the type 1 constrain n Q a Interpolated sea surface elevation b same as a but optimized c same as b but large a priori error on z and v that is 200 USD d same as a but for 7 e same as b but for Uu f same as c but for u g same as a but for v h same as b but for v 1 same as c but for v a b c NSSE USI NSSE USI NSSE USI Y km Y km Y km gilles Fri Nov 28 01 12 26 2003 gilles Fri Nov 28 01 13 20 2003 gilles Fri Nov 28 02 02 51 2003 X km X km X km Range of NSSE 0 to 34 2672 USI Range of NSSE 0 to 34 2672 USI Range of NSSE 0 to 56 5158 USI Range of X 1927 69 to 1846 7 km Range of X 1927 69 to 1846 7 km Range of X 1927 69 to 1846 7 km Range of Y 0 to 0 km Range of Y 0 to 0 km Range of Y 0 to 0 km Frame 90 in File champ
10. surface elevation tendency as a strong constrain 0 IFLTC DUBDT EXT Optimization of the depth average x velocity 0 IFLTC DVBDT EXT Optimization of the depth average y velocity 0 IFLTC OBC VBNORM EXT Open boundary conditions for normal depth averaged velocities 0 IFLTC OBC VBTGT EXT Open boundary conditions for tgt depth 0 IFLTC SMOOTH EXT Optimization of the smoothing constrains Such optimizations will affect u and v The difference between the first and the second experiment concerns the a priori error on u and v In the first experiment it is set to realistic values 20 e 2 IND COVX COM TX SSE EXT A priori error on the surface elevation 1 5e1 IND COVX COM TX RHBAR EXT A priori error on the depth average density RHb 20 IND COVX COM TX VBARX EXT A priori error on the x transport Ub 20 IND COVX COM TX VBARY EXT A priori error on the y transport Vb In the second experiment the a priori error on the x and y transports are mul tiplied by 10 The optimized variables after the first and the second experiments are compared to the interpolated fields in fig 6 We verify that this kind of optimiza tion doesn t affect n Hence in the block 165 of plots out no modification of the X ANALYSIS COM type 1 variable is reported Comparisons between the first and the second experiment raise up the role of the a priori error on the x and y transports This role can also be observed on the X ANALYSIS COM type 3 and 4 variable modific
11. thin white lines e same as d but sections are from graph b f same as d but section are from graph c 11 namelist flags The optimization is proceeded sequentially that is an optimization modulus computes a set of analyzed variables reintroduced as the first guess in the next modulus These modulus are e SSE The Sea Surface Elevation modulus reconstructs the sea surface eleva tion variable e HPG The Horizontal Pressure Gradient modulus constrains the flow to ver ify the geostrophic balance on Z levels e MPV The Potential Vorticity Modulus constrains the vorticity verifying the quasi geostrophic balance e EXT This External mode includes constrains on the barotropic mode a 0 Z 0 Z 0 with n the sea surface elevation u and v the horizontal west east and north south transports To illustrate the optimization modulus we will focus on the external mode optimization modulus the EXT modulus The optimization modulus purpose is to compute the X variable such it will be consistent with the model physics to a degree fixed by the constrain covari ance errors as shows the cost function equation 2 17 of Auclair et al 2000 R being the constrain covariance matrix and d Bo HoX Each modulus termed XXXX computes first Vp and Ho from a set of equations constrains for instance chosen in namelist ext for the EXT modulus by flags IFLTC XXXX EXT If both ZFLTC DSSEDT EXT IFLTC DUBDT EXT and IFLTC DVBDT EXT
12. DSSEDT EXT A priori error on the d sse dt constrain IND COVC COM TC DUBDT EXT A priori error on the d ub dt constrain IND COVC COM TC DVBDT EXT A priori error on the d vb dt constrain IND COVC COM TC SMOOTH EXT A priori error on the smoothing constrain MATRUNC COM Truncation threshold no threshold 0 AEn PIS k IE IS k IE k kk k State vector matrix covariance ko kkk 2 kk EHE k k kk k k IFLDIAG_COVX_COM Flag for a diagonal matrix IFLINTER COVX COM _ To estimate the consequences of the 1 0 0 4 3 10 Sensibility tests interpolation 1 2 3 IFLBATHY COVX COM To parametrize the difference between the HR and LR bathy IFLHOMO COVX COM _ To use an homogeneous component namelist flags must be O oO c C em IFLINTE IFLINIT SSE IFLINIT HPG IFLINIT EXT IFLINIT MPV IFLINIT TRA Flag for the interpolation 1 2 3 Flag for the initialization of the SSE Flag for the initialization of the truncation error Flag for the initialization of the external mode Flag for execution of the vorticity modulus Flag for execution of the extrapolation modulus Three experiments have been operated They intend to show the sensibility of the a priori error on the state vector variables and the constrain sensibilities Let us on i j first consider only the type 1 constrain MGI 0 The namelist ext flags are t 16 1 IFLTC DSSEDT EXT Optimization of the surface elevation tendency IFLTC DSSEDT ADJ EXT Optimization of the
13. VIFOP User Guide an example of set up and utilization Gilles Molini Francis Auclair P le d Oc anogaphie C ti re de Toulouse Laboratoire d A rologie 31400 TOULOUSE francis auclair Q aero obs mip fr gilles molinie aero obs mip fr 15th December 2003 Contents 1 Introduction 3 2 Install the toolbox 3 2 1 To get the toolbox source files 3 22 Whatis incded em 2boeen de eedt ed BOSS Se aS 3 3 Running VIFOP 3 Sk stall VER OP es uon ots IS Fe cee nh Os Se BOS Ak AO ee Be 5 3 2 Compiling and running VVFOP 02 5 3 3 Possible compilation or run time issues 5 4 Example case a walk across plots out 6 4l OmuddesenpLbos 3 2 ac e sebo the a teed Bn e Ee S 6 4 1 The HR grid characteristics block f 4 1 2 The LR grid characteristics block 8 4 1 3 The 2D and 3D LR variables blocks 8 AD InterpolaWOf 4151038 48 amp otek IAE Bk dd ade eo a 8 4 2 1 The Nearest LR stations around LR stations block Level D AUSELS veru rinsed ur era Dante ede ty diss Ey ea een Hout e ud 10 4 2 2 The Nearest LR stations around HR grid points block 10 4 2 3 The Interpolated 2D HR variables block 10 45 Opm 22 3 eue moa puR Ana eS eh EA SS ee eo a 10 4 3 1 The External mode variables block 12 4 3 2 The External mode Smagorinsky dif coef block 13 4 3 3 The Numbering block Level 2 Users
14. are set to 1 the Ho operator consists on the linear terms of equations 45 46 and 47 of Blumberg et Mellor 1987 corresponding respectively to ones Lj ont 0 hd 0 and y to the true tendencies on J Abi D a x tad after one time step model run the comprehensive selected constrains ee and non linear terms are coded in VIFOP n i j u i j and v i j are gathered in the state vector X 4 3 1 The External mode variables block In order to fill in the state vector X the external mode variables are retrieved from the V2D and V3D HR fields Thus printed in plots out are e EL Z Sea surface elevation computed at n points of the staggered C grid as defined in Blumberg et Mellor 1987 e VBAR_X tu x D atu points D H is the total depth e VBAR Y v xDatv points e UA POM u at u points e VA POM at V points e RHO depth averaged density at n points 12 4 3 2 The External mode Smagorinsky dif coef block The block consists of the Smagorinsky diffusion coeficients 4 3 3 The Numbering block Level 2 Users All the model variables control variables are dumped into the same 1D state vec tor X for each HR grid points see the VIFOP reference manual section X and Y vector representation At HR point 1 1 type 1 2 3 etc non zero variables are sequentially stored into X The process to fill in X continues at 1 2 The tables IJ2LX COM L21X COM and L2JX COM allows to go
15. ation block 165 of plost out that is one order of magnitude larger than in the second experiment The C ANALYSIS COM values block 170 re veal that the larger corrections occur around the seamount as C ANALYSIS COM is one order of magnitude lower in the second than in the first experiment around the seamount The third experiment considers both the type 3 and 4 constrains 3 0 and g 0 The second sensibility test consists of comparing results of the first and the third experiments The a priori error are let to realistic values This comparison is illustrated in fig 7 In that case we can see that both X state vector variables are optimized 17 a b NSSE USI NSSE USI NSSE USI E g E xz xz xz X km X km X km Range of NSSE 0 to 34 2672 USI Range of NSSE 0 to 34 2672 USI Range of NSSE 0 to 34 2672 USI Range of X 1927 69 to 1846 7 km Range of X 1927 69 to 1846 7 km Range of X 1927 69 to 1846 7 km Range of Y 0 to 0 km Range of Y 0 to 0 km Range of Y 0 to 0 km Frame 90 in File champ HR bck nc Frame 90 in File champ HR ana dssedt1 nc Frame 90 in File champ HR ana dssedt2 nc UBAR USI UBAR USI UBAR USI E E E lt lt gt gt gt X km X km X km Range of UBAR 0 to 95 5113 USI Range of UBAR 0 to 101 997 USI Range of UBAR 0 to 206 526 USI Range of X 1927 69 to 1846 7 km Range of X 1927 69 to 1846 7 km Range of X 1927 69 to 1846 7 km Range of Y 0 to 0
16. ex first index of the 2D arrays is indicated on the first line and evolves from column to column 1 15 the 7 index second index of the 2D arrays is printed on the first column and varies from line to line 1 21 The same way vertical and horizontal sections of the 3D variables are plotted The vertical section indexes can be selected by the ISEC HR OUT JSEC_HR_OUT ISEC_LR_OUT JSEC LR OUT variables in namelist output The process to check the installation success is indicated in section 4 3 9 4 1 Grid descriptions VIFOP deals with two grids the LR Low Resolution and the HR High Resolu tion grid I Printing a block is activated in namelist output through the IFL OUPUT flags Type 1 points are density points type 2 are T points type 3 are v and type 4 vorticity points 2000 ee ee PRU 4000 er ae 6000 leere tttm 0 72 1000 eS ee 2000 EMT S SO 3000 NENNT S 4000 MT ia 0 74 0 04 068 0 02 0 01 Figure 2 Upper graph HR upper graph and LR lower graph discretization of the seamount bathymetry 4 1 1 The HR grid characteristics block MSK_HR Mask of the HR domain IJ2L_HR 1D index for the 2D compressed vectors stacked representation see VIFOP reference manual section Vector representation H_HR Bathymetry Lat_HR Latitude of grid points Lon_HR Longitude of grid points XM_HR West East coordinate in the Mercator projection YM_HR S
17. ing libraries Go Bookmarks Tools Window Help EJ 4 http aeropc12 cgi bin cvsweb cgi VIF OP A Search add sources of the release 1 5 and remove files non existing in the version 1 3 add sources of the release 1 5 and remove files non existing in the version 1 3 remove empty directories REVTUTU before installating VERSION 1 5 include TRA extrapolation module netcdt data management and important revisio include TRA extrapolation module netc f data management and important revisio import vitop version 1 3 sources in the CVS repository include TRA extrapolation module netcdt data management and important revisio Ble le R le ie add sources of the release 1 5 and remove files non existing in the version 1 3 add sources of the release 1 5 and remove files non existing in the version 1 3 H add sources of the release 1 5 and remove files non existing in the version 1 3 import vifop version 1 3 sources in the CVS repository include TRA extrapolation module netc f data management and important revisio add sources of the release 1 5 and remove files non existing in the version 1 3 E Ll import vifop version 1 3 sources in the CVS repository include TRA extrapolation module netcd data management and important revisio include TRA extrapolation module netcdt data management and important revisio lecvs lecvs lecvs lecvs lecvs 1111 lecvs 13 lecvs 11 lecvs 11 lecvs
18. l first be performed in order to define the LR Low Resolution and HR High Resolution domain parameters the LR and HR variable lists and the specific parameters to the model configuration for which LR data are optimized Thus this first step needs to be done only if you run VIFOP for the first time or if any of the LR or HR domain has been changed or if the model used is different from this of the previous VIFOP run To go through the PREP PARAMETER step launch as a shell command gt run var init prep parameter Following the PREP PARAMETER step such a command line compiles and runs the optimization software variat init Compiling and running only variat init in the same domain and with the same libraries as previously corresponds to the command line run var init noprep parameter 3 3 Possible compilation or run time issues e The objects files may not be linked Possibly the mathematical libraries prep parameter and variat init are not compiled with the same compiler or options To fix this issue reinstall without removing the LOCAL directory and recompile the code e The optimization can not be proceed Verify that the correct mathematical libraries are set in the run var init script 4 Example case a walk across plots out The example case study proposed in the VIFOP package is called seamount The LR and HR bathymetries and the LR fields are included in the vifop LR and vifop HR directories with the seamount s
19. lecvs lecvs lecvs lecvs lecvs lecvs lecvs igdule path or alias MFOP Go Figure 1 VIFOP CVS management frontpage 1 select a VIFOP release 2 download it with different compilers or options The LOCAL directory can be renamed to suit users 3 1 Install VIFOP The installation procedure is described in the INSTALL TXT file of the VIFOP distribution The first step is to initialize the shell environment variable MAKE to the path of your gmake command The second step in installing VIFOP is to set the compiler name and options by editing the vifop script compil shell script script compil is the only file required to be edited in vifop Then run the in stall vifop script as a shell command It will be asked where to installed VIFOP it is assumed in the following that the installation is done in the LOCAL directory but no matter wherever it is actually done It compiles the mathematical libraries considering the chosen compiler name and options and will install the LOCAL directory at the same level as vifop 3 2 Compiling and running VIFOP Change directory to LOCAL and check variables initialized in run var init Insure the csckit mathematical library is selected export LIB USED csckit must be set inrun var init this is the default Note that using Matlab required Matlab to be installed on your computer Then you are ready to launch the compilation and execution of VIFOP sources files A PREP PARAMETER step wil
20. outh North coordinate in the Mercator projection Z_HR Surface Height of the sigma coordinate level near the surface Z_HR Bottom Height of the sigma coordinate level near the bottom Z HR Height of the sigma coordinate level 1 and j constant sections BEG_HR Index for the representation of 3D compressed vectors stacked in ID vectors see VIFOP reference manual section Vector representation e ROT HR grid rotation This block is printed out for the 4 point types of the HR C grids Sometimes stars are printed in place of the field value This is because the conversion of field values after scaling into four digit integers runs over a wide range 4 1 3 The LR grid characteristics block The printed arrays are similar to those of the HR grid characteristic block 4 1 3 The 2D and 3D LR variables blocks The 2D and 3D LR variables are concatenated in the generic arrays called V2D LR and V3D_LR The variable indexes in those generic arrays are set in the namelist lr files These two arrays are printed in the current block Two horizontal and vertical sections of the 3D variables are printed the sections are selected via namelist output as mentioned above 4 2 Interpolation The flag IFLINTE in namelist flags allows selecting the interpolation scheme IFLINTE can be set to e The gaussian interpolation scheme is processed e 2 No interpolation is processed the all ready interpolated HR fields are read in the
21. s issued from the optimization 4 3 9 Check VIFOP or The Results of the optimization Error on con strains blocks e RHS COM y of Auclair et al 2000 equation 2 16 e C ANALYSIS COM d of Auclair et al 2000 equation 2 16 If only the ZIFLTC DSSEDT ADJ EXT flag has been set to 1 in namelist ext strong constrain see the end of the current section for a comprehensive description of the incriminated namelists the optimization result of ented 0 must match the RHS_COM values Thus C ANALYSIS COM of type 2 must be lower 1074 in the current case using the g77 compiler without specific options This is a mean to check the routine is well functionning namelist ext must be IFLTC_DSSEDT_EXT Optimization of the surface elevation tendency 1 IFLTC DSSEDT ADJ EXT Optimization of the surface elevation tendency as a strong constrain 0 IFLTC DUBDT EXT Optimization of the depth average x velocity 0 IFLTC DVBDT EXT Optimization of the depth average y velocity 0 IFLTC OBC VBNORM EXT Open boundary conditions for normal depth averaged velocities 0 IFLTC OBC VBTGT EXT Open boundary conditions for tgt depth 0 IFLTC SMOOTH EXT Optimization of the smoothing constrains Wok k k k I ok oo KE EK y ternal mode constrain covariance dk rk ore oe k k kk k kkk k 15 e 2 2 e 2 2 e 2 l 0 IFLDIAG COVC COM IFLNLIN COVC COM Flag for a diagonal matrix To compute the matrix from non linear term estimate IND COVC COM TC
22. s of the first row and 4 at the remaining of the domain except column 1 This is consistent with the zero values for VBARX and VBARY at the first column and rows as reported in the previous section JCOVX COM indicates the non zero element indexes in the state vector The diagonal of JCOVX COM being printed type by type con sequently type 1 of JCOVX COM is identical to NUM_ZTA type 2 to RHBAR type 3 to VBAR X and type 4 to VBAR Y the variable type is its index in the state vector X 13 4 3 5 The Constrain Matrix block Level 2 Users This is the H matrix of equation 2 17 of Auclair et al 2000 e VCONS COM values of the constrain matrix e LINCONS COM beginning of each matrix line e JCONS COM non zero element index in the state vector Imagine we would like to retrieve the m equation terms at 6 3 You must get the index value say l IJ2LC COM 6 3 TC DUBDT COM and then 11 BEGC COM L TC DUBDT COM Thus we have the linearized tendency k LINCONS L1 1 1 mj Y VCONS k x X JCONS k k LINCONS L1 In plots out the terms of each selected equation of the optimization modulus are printed The corresponding blocks are marked by OUTPUT 150 0000 There is first a bunch of blocks associated to each type 1 equation term The equa tion type is marked in the block header KEK KK K E K K E K E K K E K E ECKE CE E K E K ECKE ICE EEG CR K E K E K K K K K K K K K K K K K K K K K K K K Constrain Matrix T
23. surface elevation field on the LR grid fig 5 shows how can the interpolated fields differ following the R_GAUSS and LBUB_LR parameter tuning 4 2 1 The Nearest LR stations around LR stations block Level 2 Users In order to speed up the interpolation scheme some useful stuffs are processed like the indexes of LR points next to each LR point These indexes are concatenated in the 1D array LNEAR LR The index list beginning for a given LR point L is provided by LNEARBEG LR L Currently LVEARBEG LR L is displayed in plots out but everybody can add the variable he would like to be printed by pro gramming in the file output 67 F in the section corresponding to ICHOICE_P 45 For example LVEARBEG LR LNEAR LR values currently printed can be used to verify which points are included in a circle of radius LBUB_ER around a specific LR point For instance for i25 and j 7 LNEARBEG LR indicates 2787 and at i 5 and j 8 LNEARBEG LR 2876 Then all the index given by LN EAR LR 2787 to LNEAR_LR 2876 1 indicate the closest LR points to the LR point 1 5 and j27 Their i and j indexes could be retrieved from the L2 LR and L2J LR or IJ2L LR arrays this last one being printed near the plots out begin Distance to all LR stations is also printed in plots out This distance is the key parameter to select points that fill out LVEAR LR It is printed for a particu lar point selected in namelist output by the ISEC LR OUT and JSEC LR OUT parameters
24. uffixes It represents a seamount at the center of a basin The data LR fields are coarsely discretized 10 x 11 x 7 points and must be analyzed on a tighter HR grid 15 x 21 x 11 points see fig 2 The initial fields are not physically consistent During the VIFOP runs several diagnostic files are written in the OUTPUT directory Among them plots out gives an ASCII representation of the VIFOP ar rays allowing to follow step by step the interpolation and optimization process In the remaining we scroll through plots out as processing the example case Along the VIFOP run several routines write some fields in plots out gathered together in blocks The block header gives the block generic title the writing routine and the type of the printed fields KKK EKE KE E KE E KE E E E E K E E E KE EKE E E KE E E E KE E E E E E CE E KE K CE KE K K K K K K K K K KK K KK K K K K K K K HR grid characteristics of type 1 ROUTINE INTER_INIT_HRGRID F KOKKE K K K K E K K E K E K E EK E K E K E K K K K K K K E KE E K CICER E K K CE K K K K K K K K K K K K K K K KK KK K K The field type indicated here corresponds to one of the four C grid points When the displayed field is a vector or a matrix the type indicates an equation or variable index this is detailed in the following The block index is given by the OUTPUT variable and the field name is indicated one line above Each value of the 2D arrays are printed out in a table like structure The 7 ind
25. ype 1 ROUTINE MODULE_EXT_CONS F KEK KKE K E K K E K EOK K E K KE K E E K K EK K K K K K K K K E K EK E K E K KE K EK K K K K K K KK K K K K K KK K K K K In each block VCONS_COM and JCONS_COM indicate respectively the value and the corresponding state vector variable index at the point JSEC_HR_OUT JSEC_HR_OUT namelist output as indicated by the line Plot of 92 th LIN I2 8 J 8 for Type 4 Thus the terms printed in the described block correspond to v in the equation on Q that is 8 8 which in the POM finite difference scheme is computed between the 8 8 and 8 9 north south flux points of the C grid The indexes reported in JCONS_COM be careful to the multiplying factor are included in VBARY block 115 4 3 6 The External mode variables block Here is the beginning of the red meat The analysed external mode variables are printed Hags IFLTC_DSSEDT_EXT IFLTC DSSEDT ADJ EXT IFLTC DUBDT EXT and IFLTC DVBDT EXT set in namelist ext select respectively constrain of type 1 qe jc 0 type 2 same as type 1 but applied exactly type3 gu 0 and type 4 PED 0 14 4 3 7 The After opt analyzed 2D or 3D HR variable block Display 2D and 3D HR variables recomputed from the analyzed external mode variables 4 3 8 The Ext mode analysis block Here begins a bunch of blocks providing analysis of the optimization The block entitled X_ANALYSIS_COM display the corrections on the X variable

Download Pdf Manuals

image

Related Search

Related Contents

Savingr  取扱説明書  bedienungsanleitung-5a-laderegler-pwm - Solarzellen  OARS Training Manual - Human Resource Services  MGW-CMAP 8010 Ver 2.3 Operation and Installation    MODE D`EMPLOI - Observatoire des industries chimiques  医政指発第033000 ー号  Procédure TEST_TEMPS  RADIO SWITCH  

Copyright © All rights reserved.
DMCA: DMCA_mwitty#outlook.com.