NEMO/AGRIF Nesting tools User's Guide - Forge
Contents
1. 3zps_rat is the minimun percentage of the partial step regarding to corresponding level thickness Those both parameters should stay the same than in NEMO namelists 13 The jpizoom and jpjzoom variables are parent grid lower left position on a large model eg ORCA2 Namelist files The wind stress and forcing file sections enables the user to specify which parent fluxes files to interpolate on child grid Then in interp section user can choose what type of interpolation will be used for each variable among bilinear and bicubic It s also possible to choose a correction of child grid values in order to ensure conservation on cells at the boundaries When one doesn t have any wind stress files to interpolate in case of coupled run just set taux file NULL Namelist restart If restart file interpolation is required restart fle variable determines the location of the par ent file shlat defines the lateral friction shlat 0 free slip no shear along the coast shlat 2 no slip specified zero velocity at the coast The value shlat is used for the computation of the relative vorticity One can choose the format of the output restart file by default a file in netcdf format is created but dimg format file can be built when d mg value is true in this case the name of the output file is dimg output file and user must specify adatr7 which corresponds to number of elapsed days since the begining of the run given by oc2. NEMO AGRIF Nesting tools User s Guide January 30 2006 minor update Dec 6 2010 Contents 1 Introduction 2 Installation 3 Input file 4 Running gt Technical aspects Dl Coord mates Interpolation sasae aae ee dte a Re ae en 0 2 Bathymetty building ze ed aa Dad Interpolation method ub s assw odo Wrote ao CU ea da A mE REA 5 2 2 Connection with parent grid bathymetry 022 9 OMONE Metodo O bec a Reb d an 5 2 4 Linear connection method 5 2 5 Adding an external high resolution database Oso Forcing les creacion ohne aa are A R29 eS Oo Remapping Method beak ee rara 5 3 2 Conservation at the boundaries 5 3 3 Extrapolation Process 5 4 Summary of Nesting tools possibilities 5 4 1 What kind of external database for bathymetry 5 4 2 What kind of tracers flux and momentum fields 5 4 3 What kind of input files are supported Chapter 1 Introduction This guide presents a series of Fortran 95 procedures that could be useful for the pre processing of OPA NEMO ocean model when running an embbeded model The version 9 called NEMO of the code makes use of the AGRIF grid refinement procedure developed at the LMC IMAG AGRIF is a Fortran 95 package f
3. athymetry mapped on the fine grid points the smoothing option isn t relevant indeed the smoothing algorithm could modify the values of this improved bathymetry If no smoothing option is activated outside the connection zone the interior do main bathymetry stay the same than the one given by the external file The type of interpolation used is closely linked to the type of external bathymetry used Indeed median and arithmetic average are relevant when the resolution of external bathymetry is much higher than the one of fine grid for example when using etopo2 A bilinear interpolation is more accurate and fast when resolutions are quite similar 5 4 2 What kind of tracers flux and momentum fields This problem is quite similar to the one of the previous section One can choose to interpolate the flux tracers and momentum fields from parent grid files T he code also offer the possibility to use ORCA025 or ORCA05 forcing files to interpolate on child grid even if one is implement ing ORCA2 zoom This is a quite easy stuff to do just replace in the namelist file the coordinate file by the one corresponding to the new fields for example ORCAO5 coordinate file and then user have to mention ORCA05 forcing files names in FLX FILES and wind stress sections This is relevant only when forcing files are not masked and when the vertical level position for 3D fields is exactly the same this is the case between ORCA05 and ORCA2 configurations becau
4. ation The general operations are described below Let A 9 z be the geographical coordinates system 1 Interpolation of latitude and longitude coordinates on child grid at T U V and F points 2 Scale factors computation on T U V and F points e amp a 4 z B coso Mn ae e2 a z 3 cos o E ay It s possible to easily change directly in the code if user prefers to apply an interpolation process for e and ea computation 50 2 Bathymetry building This is a crucial part when running an embedded model Various options are offered for the creation of child grid bathymetry As mentioned in chapter 3 users can choose between full steps 18 and partial steps the code also offers the possibility to use external high resolution database in order to better resolve topographic variations This section presents some key components of bathymetry construction namely smoothing and interpolation method as well as connection with parent grid bathymetry 5 2 1 Interpolation method Interpolation from parent bathymetry The method used to interpolate bathymetry on child grid is bilinear interpolation based on SCRIP Spherical Coordinates Remapping Interpolation Package The first step is to compute remapping weights then interpolation process is similar to a matrix vector multiplication Wii U12 Wim Wa W22 W2m Ci C2 Cn 2 E or ssec Wnt Wn2 Wnm The 1xn matrix C contains the 2D coarse grid fie
5. d parent grid coordinates file meshmask file containing the mbathy variable Namelist bathymetry The bathymetry on the child grid can either be interpolated from parent grid values if flag new topo is set to false or be recomputed from an external database etopo2 or OPA format file currently in that case elevation database should contain the path to this fine bathymetry ee See important bathymetry gradients a smoothing algorithm is available if flag smoothing is set to true and is carried out according to a criteria called smoothing f actor whose values are typically between 0 6 and 0 8 Actually smoothing f actor is slope parameter factor grad h h maximum value for topography smoothing One can choose to avoid applying smoothing in the case of tuned bathymetry then smooth ing must be false In order to ensure the compatibility between parent grid and fine grid bathymetry a linear connection is made on the boundaries of fine domain on nb connection pts points Thanks to removeclosedseas parameter it is possible to remove closed seas from your fine grid domain This option can be useful when using for example etopo2 database which can add some extra seas in fine domain Figure 3 2 illustrates this problem with a zoom on the gulf of Mexico from NATL3 we see that when using etopo a part of Pacific Ocean is incorporated whereas large scale domain doesn t contain it If a high resolution bathymetry is provided as input o
6. e The current code of nesting tools is designed to include etopo2 global elevation data base sridded at 2 minute latitude longitude resolution as well as various OPA format bathymetry file The user can choose to add his own database for bathymetry interpolation In that way the code is flexible and is able to properly work if this external database is in netcdf format with the structure presented in figure 5 2 If for any reasons user doesn t want to change the structure of his file to match the structure required by the code it s of course possible to edit directly in the code the routine responsible for external bathymetry file reading 5 3 Forcing files creation 5 3 1 Remapping method As introduced in section 5 2 1 we use remapping to interpolate various fields between coarse domain grid and fine domain grid Remapping method are well adapted in our case because it accelerates interpolation process Indeed remapping matrix is calculated once for all and can be re used for all fields and also for all levels if fields to interpolate are not masked At this point we have the choice between bilinear and bicubic interpolation schemes We can quickly present those both methods Let f be the function to approximate on child grid point P A search algorithm on coarse grid finds in which box the point P resides If we use indices defined in figure 5 3 bilinear scheme gives 24 AOS ha al aia A Og sd whereas bicubic scheme can be writt
7. e bathymetry whose derivative is a will be smoothed until it fits the slope parameter called smoothing_factor in nesting tools namelist file Thus the stop criterion of the smoothing algorithm is when m lt smoothing_factor on all points The method used to carry out this smoothing makes use of hanning filter if we note Inh the logarithm of h hanning filter is defined by Inh i 5 Inh i 1 3 Inh i 1 7 lInh i j 1 Inh i 5 1 4 Inh i j oo m 20 Fine grid western boundary O 6G GGG G oc OO e oo oooooo ooo OO OOQO O OO O OO OOOCO o ooo oo o oo 000000 Ooo oo ooo o oo o 000000 oc ocQocQoocopooQpo DIR OO j coarse jmin 1 20 i_coarse imin 1 00000000000 Coarse grid points Ghost points Connection zone copy of parent grid values ooo Points involved in linear connection process Figure 5 1 Bathymetry connection process for full cells example with space refinement ratio 3 21 Then the following convolution kernel is applied to resulting bathymetry 00 0 0 0 11 21 00 K 5 9 4 1 0 Lee qu EM 0 1 5 10 we can note that the sum of the elements of this 5x5 kernel is equal to 1 so this kernel is normalized If the kernel has m rows and n columns mathematically we can write the convolution as Inhsmoothed i j Y Y lnh i k 1 4 1 1 K k 0 k li 1 Smoothing algorithm iterates those two ste
8. ean output file not needed for netcdf format file Then the type of interpolation applied to restart file variables is type interp and one can choose bicubic or bilinear interpolation When using dimg format compiler options like assume byterecl and convert big endian used by ifort compiler can be required 14 15 Chapter 4 Running Once namelist input file is propely filled in it s possible to run the code For basic NEMO run using AGRIF capabilities you have to execute following programs create_coordinates exe must be execute first 1 create coordinates exe namelist file 2 create bathy exe namelist file 3 create_data exe namelist file If your namelist file is called namelist input you don t need to specify namelist file name as argument Additional executable named create restart exe interpolates any given parent grid restart file on child grid If your grid hierarchy comprises n levels you just have to create a namelist file for each level and fill in data file section with files created at level n 1 in that case you should give value 1 for jpizoom and jpjzoom 16 17 Chapter 5 Technical aspects 5 1 Coordinates interpolation This part aims at describing the work carried out by create coordinates exe This executable creates coordinates file for child grid according to his position into coarser grid The type of interpolation used for this task is currently 4 order polynomial interpol
9. en ws SE i j i 1 7 wr LG 41 941 08 i j 1 wg gt j wise i 4 1 wu el 1 t wis i j 1 2 T 2 2 2 M wisg 1 J dE wis seh i PLIT ws 5535 1 ddr ep dose w16 5 35 1 J 1 For each field to interpolate user can choose between the both schemes You can note that bicubic remapping can create some local extrema whereas values interpolated thanks to bilinear scheme are limited by parent grid values 5 3 2 Conservation at the boundaries For two way interactive grid nesting user can choose to ensure conservation mainly for tracers This conservation is implemented on the two coarse grid cells at the boundary of child grid namely the zone where coarse grid points will be updated The process is quite simple a correction is applied to each fine grid points covered by the same coarse grid cell If we note a this correction p and p space refinement ratio in x and y direction for a field F the correction is given by px Py Q Psp coarse gt gt Diane a 9 i l j 1 5 3 3 Extrapolation process This part is useful when the field to interpolate on child grid is already masked e g when interpolating restart files fields Indeed the main drawback of SCRIP package is the fact that it doesn t manage extrapolations During search algorithm if one of the four points surrounding one given child grid point to interpolate is masked then no interpolation is done and child grid point receive the va
10. ents are planed to enable the interpolation of U and V type 4D variables this feature can be useful when using TOP ocean tracer and biochemistry model or other additional option of NEMO 28 Bibliography 11 G Madec OPA Ocean General Circulation Model Reference Manual http www lodye jussieu fr N EMO 2 Philip W Jones A User s Guide for SCRIP A Spherical Coordinate Remapping and Interpolation Package http cl mate lanl gov Software SC RI P SC RI Pusers pdf 3 Van Leer B 1976 Towards the Ultimate Conservative Difference Scheme IV A New Approach to Numerical Convection J Comput Phys 23 276 299 4 P Penven and P Marchesiello A Regional Oceanic Modeling System with mesh refinement pre and post processing tools http www brest ird fr Roms tools 29
11. ld to interpolate unrolled into a single vector The Ixm matrix F is the result of matrix multiplication so F contains the interpolated field on fine grid Bathymetry is mapped with weights calculated using SCRIP s bilinear inter polation scheme in a quadrilateral grid This scheme is quite simple to explain let s consider point P on destination grid we search for a cell formed by four points on source grid containing P and then the remapping weights must be assigned to each of those four points Interpolation from external high resolution bathymetry When using a high resolution bathymetry the relevance of bilinear interpolation is not obvious In this way an interpolation method based on an average of all high resolution points contained in a given child grid cell has been implemented This method is quite simple the value assigned to every child grid points is the average of all high resolution points present in its cell There is two different way to carry out this average indeed user have the possibility to choose between arithmetic and median average http climate lanl gov Software SCRIP 19 5 2 2 Connection with parent grid bathymetry The bathymetry thus interpolated will be connected to the parent topography before being smoothed The connection is made on three coarse grid cells at the boundaries of child grid as well as on both ghost cells This connection is not exactly the same between full cells and partial cell
12. lue zero even if this point is not masked So we have to preprocess 25 LEVEL N 1 LEVEL N Figure 5 4 Directions investigated around a point where extrapolation is required interpolation step to avoid such problems by correcting mask on coarse grid and giving value to points where mask has been changed The principle is to construct a box around the point to be extrapolated as shown in figure 5 4 and to search for all no masked points candidates for this extrapolation Then between all possible points we calculate the error made by replacing the value of the point to be extrapolated with this new possible value Using Taylor expansion the error is given by Of Of hz y hy f x y h h FE hay hy FC ha hy So at each no masked points around a point to be extrapolated we compute gradients in x and y directions as well as distances h and h Of course the point giving minimum error is chosen for extrapolation 26 5 4 Summary of Nesting tools possibilities 5 4 1 What kind of external database for bathymetry The choice of high resolution database for bathymetry is linked to the resolution of child grid The use of etopo2 is not completely relevant in case of an ORCA2 zoom it would be better to use a higher resolution ORCA bathymetry For example for an ORCA2 zoom with a space refinement ratio of 3 the best choice could consist in using ORCA05 or ORCA025 bathymetry If one already owns a tuned fine b
13. n the Agrif FixedGrid in when running the model rho and rhot are space and time refinement ratio time refinement ratio is used only when creating a restart file Namelist vertical grid Vertical grid section contains various parameters related to vertical grid structure in z coordinates N refers to the number of vertical levels and ppacr is the stretching factor for the grid while ppkth is the approximate layer number above which stretching will be maximum Values of those variables can be found in NEMO source code for example in file par ORC A R2 h90 for ORC A2 LI M configuration note that as AGRIF doesn t yet support vertical refinement the number N of vertical levels must be the same for each grid of the hierarchy ppdzmin is the vertical thickness of the top layer in meters and pphmaz is the maximum depth of the ocean If in NEMO h90 file ppdzmin and pphmaz are initialized with character string pp_not_used insteed of those parameters you have to specify pa0 pal and psur also defined in this configuration file Namelist partial cells Ihe NEMO code offers the option to include partial cells at the bottom in order to better resolve topographic variations If this option is chosen an additional bathymetry file in meters is required by NEMO In this case partial_steps must be set to true and you have to add the path to parent bathymetry file in meters e zps minis the minimun thickness of the partial step and e
14. ne can choose between median aver age type bathy interp 1 arithmetic average type bathy interp 0 interpolation methods as well as bilinear interpolation type bathy interp 2 Without any external bathymetry dataset a bilinear interpolation is carried out When user aims at using mesh refinement in two way grid nesting mode it s possible to update parent domain bathymetry in order to ensure the compatibility between coarse and fine grids masks for interior points of the nest T his action is carried out when parent bathy update variable is true then one can choose the name of the updated parent bathymetry file thanks to updated parent file variable 11 removeclosedseas true removeclosedseas false Figure 3 2 Example of the influence of removeclosedseas option in a zoom on the gulf of Mexico from NATL3 configuration Namelist nesting The third section is very important to implement an embbeded model Here is the specification of fine grid position on its parent grid and its space and time refinement factor imin imaz jmin and maz refer to the positions of the fine grid corners relatively to its parent starting with the value 1 in each space direction When the user writes these positions one has to take care that fine grid is contained in the parent grid imin jmin are the lower left corner coordinates and imax jmax are the upper right corner position Those values must be identical to the ones given i
15. or the inclusion of adaptive mesh refinement AMR features within a finite difference numerical model To implement an embbeded model using AGRIF the user needs to provide a whole set of data containing the surface forcing fields and the initial conditions for each child domain of the grid hierarchy Indeed for NEMO AGRIF run each grid must have his own input files namely bathymetry coordinates and forcing fields files This package provides procedures that have been designed to perform these tasks including bathymetry interpolation and smoothing as well as forcing data interpolation The next chapter describes system requirements and installation process while the following sections describe how to use nesting tools and the various capabilities provided http www lodyc jussieu fr NEMO http www Imc imag fr IDOPT AGRIF Chapter 2 Installation The distribution file is a gzipped tarfile so you must uncompress the file using gunzip and then extract package from the tar file using tar xvf nesting_tools tar The extraction process will create a directory called Nesting tools System requirement In order to compile nesting tools you need a Fortran 95 compiler and a netCDF library http my unidata ucar edu content software netcdf index html with F90 enabled This toolbox has been tested on Linux workstations with g95 pgf90 version 5 2 4 or later and ifort9 0 compilers as well as on Alpha platform with f95 com
16. per right point indice jmin 30 I j direction lower left point indice jmax 50 I j direction upper right point indice rho 3 I space refinement ratio rhot 3 time refinement ratio only for restart amp vertical grid l Vertical grid parameter for domzgr see par ORCA h90 or par_NATL h90 files ppkth 21 4333619793800 ppacr 3 ppdzmin O pphmax O psur 4762 96143546300 pad 255 58049070440 pal 245 58132232490 N 31 amp partial_cells partial_steps true ri ge bathy_meter nc 3zps min 50 ame e3zps rat 0 25 amp nemo_coarse_grid jpizoom 1 left bottom i j indices of the zoom jpjzoom 1 I in data domain indices amp wind_stress taux file taux 1m nc tauy file tauy_1m nc amp forcing files FLX FILES data 1m salinity nomask nc data 1m potential temperature nomask nc TISSone geothermal_heating nc amp interp VAR_INTERP votemper bilinear conservative vosaline bilinear conservative heatflow bilinear conservative restart restart file NULL shlat 0 dimg false dimg_output_file test_dimg adatrj 360 25 interp_type bilinear 10 Namelist coarse_grid_files ar 5 number of ocean levels T points in the ocean The first part of this namelist contains path to files needed for bathymetry and coordinates interpolation process barentbathytevel and parent coordinate file are names of parent grid an
17. piler In the src directory you must edit the makefile corresponding to your compiler in order to insert the appropriate path to netCDF library Once the makefile has been properly edited type make and let it work By default the makefile builds in bin directory four executables named e create coordinates exe computation of child grid coordinates e create bathy exe bathymetry interpolation smoothing and connection e create data exe interpolation of the parent surface fluxes and tracers on child grid e create restart exe interpolation of a parent restart file on a given child grid Chapter 3 Input file lhe different options of the Nesting Tools are passed to the program through an entry file which has the structure of a namelist file Namelist Input The namelist input file name can be given as an argument to various executables By default this file is called namelist input and contains a group of namelist described in Fig 3 1 Figure 3 1 Required input namelist amp coarse_grid_files parent_coordinate_file coordinates nc 2 parent meshmask file meshmask nc must contain mbathy amp bathymetry new topo true elevation database etopo2 nc elevation name topo smoothing true smoothing_factor 0 6 nb_connection_pts 3 removeclosedseas true type_bathy_interp 1 amp nesting imin 56 I i direction lower left point indice imax 76 I i direction up
18. ps as long as smoothing criterion isn t satisfy 5 2 4 Linear connection method In this case fine grid bathymetry and interpolated coarse grid bathymetry fields are blended across the nb connection pts mentioned in namelist input file outer rows and columns of the fine grid The blending is a simple linear weighting between the interpolated coarse grid values and the fine grid values from the input file The weighting scheme for a 5 point connection is given as e row column 1 100 interpolated coarse grid 0 fine grid e row column 2 75 interpolated coarse grid 25 fine grid e row column 3 50 interpolated coarse grid 50 fine grid e row column 4 25 interpolated coarse grid 75 fine grid e row column 5 096 interpolated coarse grid 100 fine grid This relaxation zone takes place between the three coarse grid at the boundaries and the interior domain 22 netcdf etopo_file dimensions lon x lat y 3 variables float lon lon float lat 138 float topo 1at lon netcdf OPA_format_file dimensions x 182 y 149 variables float nav lon y x float nav lat y x float Bathymetry y x Figure 5 2 sample input high resolution bathymetry file 29 child grid point O coarse grid point 1 j 1 i 1 j 1 i j i 1 j Figure 5 3 Coarse grid points box around a child grid point to interpolate 5 2 5 Adding an external high resolution databas
19. s Then the way to connect this connection zone to the rest of the interior high resolution domain is a linear connection Full cells In full cells configuration only bathymetry in levels is required So the connection is simply made by copying parent grid values on corresponding child grid cells as shown in figure 5 1 Then bathymetry in levels is converted into meters in order to implement smoothing process This smoothing is carried out by using interior domain points To ensure connection between interior domain and boundaries a linear connection is made The bathymetry is then reconverted in levels Partial cells In partial cells the process is quite similar But in this case we work directly on bathymetry in meters For the three coarse cells at the boundaries it s no need to directly copy parent value we just have to take care about the fact that child grid value remain at the same level than corresponding parent grid value The interpolation process 1D in each space directions is based on Van Leer slope limiters with an additional criterion to check that the interpolated points remain at the same level 5 2 3 Smoothing method In order to prevent horizontal pressure gradients errors interpolated bathymetry requires some smoothing This is not the values of bathymetry but the values of its logarithm that are smoothed This is a usual method which ensures that the smoothing won t be too strong in deep areas The logarithm of th
20. se no vertical interpolation is currently implemented The default version of the code doesn t carry out extrapolation during forcing files interpo lation because NEMO standard files are not masked If user wants to activate extrapolation during interpolation process on files mentioned by FLX FILES it s possible to activate this option by editing agrif interpolation f90 file in routine Interp_Extrap_var the logical variable Extrapolation must be initialize to true value this option is not available for wind stress files interpolation 27 namelist variable netcdf variable variable variable position file type name taux file x y time counter 3D taux or sozotaux BINMNEBNF TP tauy file x y time counter 3D tauy or sometauy BMNBNEF cq CREE deptht or z x y time counter different from nav lat x y z time counter nav lon nav lev x y deptht time counter deptht time steps time counter tmask FLX FILES x y time counter AD or 3D interpolate any name 0 optional dimensions Figure 5 5 Netcdf files structure supported by nesting tools 5 4 3 What kind of input files are supported lhe current version of Nesting tools have been fully tested with NATL3 and ORCA2 LIM configurations tests with NATL4 and ORCA025 are currently carried out In the case of tuned input files user can refer to figure 5 5 to check the current kind of netcdf file structure currently supported by nesting tools Developm
Download Pdf Manuals
Related Search