Rectilinear example

Example files

On a rectilinear grid, the grid cells are always rectangular and the grid axes are parallel to the coordinate axes. Rectilinear grids are often used in NEMO test configurations such as GYRE_PISCES.

Note

The ORCA grid is not rectilinear. If you are working with global model data, please refer to the ORCA example.

Example rectilinear data can be downloaded from the following archive

Link to Zenodo archive

You will need the following files

mesh_mask.nc - Grid information
GYRE_1m_04110101_04201230_grid_T.nc - T point data containing temperature and salinity information

Loading grid data

First, we will use the load_hgridata and load_zgriddata routines to load the necessary grid data from the model grid file (typically a mesh_mask or domaincfg file).

Loading the horizontal grid data is simple enough

from neutralNEMO.grid import load_hgriddata

hgd = load_hgriddata( "mesh_mask.nc"  )

hgd is a dictionary containing cell widths for the grid. The netcdf variable name for all of these terms can vary between datasets and can be adjusted accordingly using keyword arguments. For example:

hgd = load_hgriddata( "mesh_mask.nc"  , e1u_varname="e1u_0")

When loading the vertical grid data, we load cell thicknesses alongside the two-dimensional (y,x) and three-dimensional (z,y,x) T point mask. In this case, the necessary masks are contained in the mesh_mask.nc file.

from neutralNEMO.grid import load_zgriddata

zgd = load_zgriddata( "mesh_mask.nc", vert_dim="nav_lev" )

zgd is a dictionary containing depth information, the two-dimensional T mask, and the three-dimensional T mask. Note that the vertical dimension name of the netcdf file had to be specified, because it was a non-default name.

In some cases, such as the ORCA example, the masks are not provided explicitly in the grid file but are applied to the data. In this case, the infer_mask2d and/or infer_mask3d keyword arguments will be needed.

Variable names, coordinate names, and masking conventions can vary between NEMO datasets. To help with this, there are plenty of options to correctly load the grid data in the Public Function Reference.

Create a grid object

We then need to create a neutralocean grid object. This can be simply done using the build_nemo_hgrid routine

from neutralNEMO.grid import build_nemo_hgrid

neutral_grid = build_nemo_hgrid(hgd, iperio=False, jperio=False, gridtype="rectilinear")

The last step before calculating our neutral surface is the loading of the temperature and salinity data.

Load T-S data

from neutralNEMO.surf import load_tsdata

tsd = load_tsdata("GYRE_1m_04110101_04201230_grid_T.nc", zgd,
                  to_varname="votemper", so_varname="vosaline")

As shown above, netcdf variable names can be specified as a keyword argument. In this case, votemper is the variable name for the potential temperature and vosaline is the practical salinity.

Define the equation of state

We also need to define the equation of state used in the NEMO simulation. The default equations of state used in NEMO are described in neutralNEMO.eos and can be called here. In this case, we configure the simplified equation of state (S-EOS) to describe a linear equation of state using the same parameters found in the NEMO namelist.

from neutralNEMO.eos import NEMO_eos

eos = NEMO_eos( 'seos' , rn_a0 = 1.655e-1, rn_b0 = 7.655e-1,
                 rn_nu = 0., rn_lambda1 = 0., rn_lambda2 = 0.,
                 rn_mu1 = 0., rn_mu2 = 0. )

If we wished to use the TEOS10 equation of state (and it matched the NEMO simulation) we would simply do the following instead

eos = NEMO_eos( 'teos10' )

We also have the option to use EOS80. An example of this can be found in the ORCA example.

If you have used a non-standard equation of state then you will need to define the equation and its gradients with respect to T and S. Details on how to do this can be found in the neutralocean documentation .

Calculate time-independent surfaces

Now we have our temperature and grid information we can calculate our neutral surfaces for given moments in time.

from neutralNEMO.surf import find_omega_surfs

zpins = [100., 200.]   # List of depths to pin each surface to
ipins = [10, 10]       # List of i-indices to pin each surface to
jpins = [5, 5]         # List of j-indices to pin each surface to
tpins = [0, 0]         # List of time indices to pin each surface to

surf_dataset = find_omega_surfs( tsd, neutral_grid , zgd, zpins, ipins, jpins, tpins,
                                      eos=eos, eos_type='insitu', ITER_MAX=10)

#Save as netcdf (optional)
surf_dataset.to_netcdf("my_surfs.nc")

In the above example, two neutral surfaces are calculated. The first surface is pinned to 100 m depth at (i=10, j=5) in the first time step. The second surface is the same but pinned to 200 m depth. The depth, temperature, and salinity of the surfaces are outputted as an xarray DataSet and can be easilly saved to netcdf.

ITER_MAX=10 sets the maximum number of iterations carried out by the neutralocean algorithm.

If you are interested in calculating the associated Veronis density for the surfaces you calculate then calc_veronis=True enables the calculation of the Veronis density as a label for the density surfaces.

ver_ipins = 10
ver_jpins = 10

surf_dataset = find_omega_surfs( tsd, neutral_grid , zgd, zpins, ipins, jpins, tpins,
                                      eos=eos, eos_type='insitu', ITER_MAX=10, calc_veronis=True,
                                      ver_ipins=ver_ipins, ver_jpins=ver_jpins)

If you want the calculated Veronis densities to be consistent between surfaces at a given time step, the Veronis density should be calculate at the same point for all surfaces. In this case, the Veronis density is calculated at (i=15,j=150).

Calculate time-dependent surfaces

If we want to calculate time-dependent surfaces we need to use a different function find_evolving_omega_surfs.

from neutralNEMO.surf import find_evolving_omega_surfs

zpins = [100., 200.]   # List of depths to pin each surface to
ipins = [10, 10]       # List of i-indices to pin each surface to
jpins = [5, 5]         # List of j-indices to pin each surface to
tpins = [0, 0]         # List of time indices to pin each surface to

surf_dataset, surf_pin = find_evolving_omega_surfs( tsd, neutral_grid , zgd, zpins, ipins, jpins, tpins,
                                                         eos=eos, eos_type='insitu', ITER_MAX=10 )

#Save as netcdf (optional)
surf_dataset.to_netcdf("my_evolving_surfs.nc")

You will notice that we still need time indices to pin the surface. As before the first surface is pinned to 100 m depth at (i=10, j=5) in the first time step but is free to evolve at any other time. The values in tpins can be any valid time index for the data found in tsd.

Two datasets are calculated in the above code: surf_dataset contains the temperature, salinity, and depth of the time-dependent surfaces. surf_pin contains the temperature, salinity, and depth of the surfaces at the time of pinning.

find_evolving_omega_surfs can accept any keyword argument for find_omega_surfs.