4. Running h-NUMO

4.1. Setting up the Workspace

Once you have successfully compiled h-NUMO, we recommend the following steps to running h-NUMO.

1. In the home/root directory create a directory called tests. Inside tests create another directory for a specific simulation (for example, let’s call it bump.

2. Inside of bump copy numa3d.in.bump from input_files and rename it numa3d.in.

3. Inside of bump copy one of the run_numo3d from run_scripts. Make sure that the other batch submission commands make sense.

4.1.1. Input file: numo3d.in

&gridnl
nelx  = 10,
nely  = 10,
nopx  = 4,
nopy  = 4,
xdims = 0.0,2e3,
ydims = 0.0,2e3,
nlayers = 2,

!P4est Input
refinement_levels_h = 0, !refinement levels in horizontal

lread_external_grid=F, ! Read external grid file

!Dimension of the domain: (x,y) in (min,max); z in (0,ztop)	
x_boundary = 4, 4 !Front and Back (x=-1 and x=+1)
y_boundary = 4, 4 !Left and Right (y=-1 and y=+1)
	   !These denote constant x,y, or z. 
	   !E.g. x_boundary(1) is the west and x_boundary(2) is the east
	   !BCs are: 0=do nothing, 4=freeslip, 5=noslip
	
/

&input
dt= 100,
dt_btp = 1.8,
time_initial = 0.0,
time_final   = 10800,
time_restart = 10.0,
time_scale = 1, !1=seconds, 3600=hours, 86400=days
!lrestart_file = .true.,
!irestart_file_number = 2

! mlswe parameter
ad_mlswe = 0.0, 
botfr = 0, ! bottom Friction flag: 0 = no friction, 1 = linear, 2 = quadratic
cd_mlswe = 0.0, !cd = 1.0e-7 for linear, cd = 1e-3 (default) for quadratic bottom friction
method_visc = 0,  !0=no viscosity
visc_mlswe = 0.0, ! horizontal viscosity
max_shear_dz = 0.0,
dg_integ_exact = T,
dump_data = T, ! write output files
beta = 0.0,  ! beta-plane Coriolis parameter (default = 0.0)
f0 = 0.0,    ! Coriolis parameter (default = 0.0)

!---- Test Cases ------------------------------------
test_case='bump',
	   !
	   ! 'bump' = 2 layers cosine test; 
	   ! 'lakeAtrest' = 2, ..., 20 layers lake at rest test; 
	   ! 'double-gyre' = 2 layers double-gyre test

ti_method_btp = 'rk35', ! Default = 'explicit' which is a two step method
kstages=5,

space_method = 'dg', !dg=discontinuous Galerkin
	    	
fname_root='mlswe',
format_vtk='ascii',
out_type='nc',
                  !txt=binary output
	       	  !nc=netcdf output; 
	       	  !vtk=Visual ToolKit (Paraview) output
	       	  !none=No Output

lprint_diagnostics=T, 
			   !T=true=print diagnostics
			   !F=false=do NOT print diagnostics

lcheck_conserved=T
/

Note

Check the sections below for more details.

4.1.2. gridnl

This namelist contains the following variables:

nelx : number of elements in the x-direction.
nely : number of elements in the y-direction.
nop : is the polynomial order inside each element.
xdims : for some tests, xdims gives the minimum and maximum values (in meters)
		in that direction.
ydims : for some tests, ydims gives the minimum and maximum values (in meters)
		in that direction.
x_boundary : This gives the boundary condition at the front and back of the 
			 x-direction (more on boundary conditions, below).
y_boundary : This gives the boundary condition at the front and back of the 
			 y-direction (more on boundary conditions, below).

4.1.3. input

The input variables:

dt : the time-step in seconds.
dt_btp : the time-step in seconds.
time_initial : initial time (can be nonzero if it is a restart).
time_final : final time (can be in seconds, minutes, hours, or days, 
			depending on the test case).

time_restart : time at which restart files are produced. The units of 
			   time depend on the test case as mentioned previously. 
			   The restart only works for 'txt' outputs at the moment.

cd_mlswe : is the bottom friction value.

method_visc : is for the use of viscosity term. If it is 0, that means
			  no use, otherwise, the viscosity solver is used.

visc_mlswe : is the value of the viscosity in the Laplacian term.

max_shear_dz : is used to discretize the vertical diffusion or 
			   friction between layer terms.

dg_integ_exact : a logical flag that turns off and on whether you
				 want to use 2N-1 quadrature points or 2N+1 for DG 
				 integrations.

dump_data : a logical flag that turns off and on the dumping 
			of the simulation data to a file

test_case : is the specific test case.  More on this in ``tests section``.

ti_method :  is the time-integration method (only the 
			 predictor-corrector is available now).

ti_method_btp : is the time-integration method for the barotropic equations

kstages : number of Runge-Kutta stages if ti_method='rk'. 
		  If ti_method='rk35' then the value of kstages 
		  does not matter.

fname_root : the root name of the output data. All data files 
             for this run will have the prefix defined by fname_root.

out_type : defines the type of output file.

lprint_diagnostics : a logical flag that turns off and on the printing. 
					 For scalability studies we turn off the printing
					 to the screen.

lcheck_conserved: a logical flag that turns off and on the 
				  compuation of mass conservation.

4.2. h-NUMO Output

1. Inside of bump directory run h-NUMO using your run script.

   • sbatch run_script.sh

2. At the end of the run, you will have a collection of output files. In the input file, you can dump the simulation data as a vtk file for Paraview visualizations or simply as a txt file for Matlab visualizations. For the Matlab visualization, turn matlab_viz to true; otherwise, it will output vtk files.