fitting a ROMS curvilinear grid

Hi, I’m trying to use xgcm for analyzing ROMS model output on a curvilinear grid, but I can’t seem to fit the grid

Here is how my dataset looks like:

import xarray as xr
from xgcm import Grid

model_path = 'http://barataria.tamu.edu:8080/thredds/dodsC/NcML/txla_hindcast_agg'
model = xr.open_dataset(model_path)

model
Out[2]:
<xarray.Dataset>
Dimensions:         (eta_psi: 190, eta_rho: 191, eta_u: 191, eta_v: 190, ocean_time: 210384, s_rho: 30, s_w: 31, tracer: 6, xi_psi: 670, xi_rho: 671, xi_u: 670, xi_v: 671)
Coordinates:
  * s_rho           (s_rho) float64 -0.9833 -0.95 -0.9167 -0.8833 -0.85 ...
  * s_w             (s_w) float64 -1.0 -0.9667 -0.9333 -0.9 -0.8667 -0.8333 ...
    lon_rho         (eta_rho, xi_rho) float64 ...
    lat_rho         (eta_rho, xi_rho) float64 ...
    lon_u           (eta_u, xi_u) float64 ...
    lat_u           (eta_u, xi_u) float64 ...
    lon_v           (eta_v, xi_v) float64 ...
    lat_v           (eta_v, xi_v) float64 ...
    lon_psi         (eta_psi, xi_psi) float64 ...
    lat_psi         (eta_psi, xi_psi) float64 ...
  * ocean_time      (ocean_time) datetime64[ns] 1993-01-01T01:00:00 ...
Dimensions without coordinates: eta_psi, eta_rho, eta_u, eta_v, tracer, xi_psi, xi_rho, xi_u, xi_v
Data variables:
    ntimes          int32 ...
...
    h               (eta_rho, xi_rho) float64 ...
    f               (eta_rho, xi_rho) float64 ...
    pm              (eta_rho, xi_rho) float64 ...
    pn              (eta_rho, xi_rho) float64 ...
    angle           (eta_rho, xi_rho) float64 ...
    mask_rho        (eta_rho, xi_rho) float64 ...
    mask_u          (eta_u, xi_u) float64 ...
    mask_v          (eta_v, xi_v) float64 ...
    mask_psi        (eta_psi, xi_psi) float64 ...
    zeta            (ocean_time, eta_rho, xi_rho) float32 ...
    u               (ocean_time, s_rho, eta_u, xi_u) float32 ...
    v               (ocean_time, s_rho, eta_v, xi_v) float32 ...
    w               (ocean_time, s_w, eta_rho, xi_rho) float32 ...
    temp            (ocean_time, s_rho, eta_rho, xi_rho) float32 ...
    salt            (ocean_time, s_rho, eta_rho, xi_rho) float32 ...
    dye_01          (ocean_time, s_rho, eta_rho, xi_rho) float32 ...
    dye_02          (ocean_time, s_rho, eta_rho, xi_rho) float32 ...
    dye_03          (ocean_time, s_rho, eta_rho, xi_rho) float32 ...
    dye_04          (ocean_time, s_rho, eta_rho, xi_rho) float32 ...
    AKs             (ocean_time, s_w, eta_rho, xi_rho) float32 ...
    tke             (ocean_time, s_w, eta_rho, xi_rho) float32 ...
    gls             (ocean_time, s_w, eta_rho, xi_rho) float32 ...
    Uwind           (ocean_time, eta_rho, xi_rho) float32 ...
    Vwind           (ocean_time, eta_rho, xi_rho) float32 ...
    shflux          (ocean_time, eta_rho, xi_rho) float32 ...
    ssflux          (ocean_time, eta_rho, xi_rho) float32 ...
    sustr           (ocean_time, eta_u, xi_u) float32 ...
    svstr           (ocean_time, eta_v, xi_v) float32 ...
Attributes:
    file:                      /scratch/user/d.kobashi/outputs_20yr_obc/1993/...
    format:                    netCDF-4/HDF5 file
    Conventions:               CF-1.4
...

As it’s a curvilinear grid the metadata seems to not be in a currently readable. So following the convention on the documentation site, I’ve tried to create my coord dictionary:

In [85]:

coords_roms={'xi':{'center':'xi_rho', 'inner':'xi_u', 'outer':'xi_psi'}, 
        'eta':{'center':'eta_rho', 'inner':'eta_v', 'outer':'eta_psi' }, 
        's':{'center':'s_rho', 'outer':'s_w'}}
grid = Grid(model, coords=coords_roms)
grid
Out[85]:
<xgcm.Grid>
xi Axis (periodic):
  * center   xi_rho (671) --> outer
  * inner    xi_u (670) --> center
  * outer    xi_psi (670) --> center
eta Axis (periodic):
  * center   eta_rho (191) --> outer
  * inner    eta_v (190) --> center
  * outer    eta_psi (190) --> center
s Axis (periodic):
  * center   s_rho (30) --> outer
  * outer    s_w (31) --> center

this seems to work but I’m losing xi_v in the xi axis (staggered on eta), and eta_u on the eta axis (staggered on xi)

so I expected something like this to work:

coords_simp={'xi':{'center':['xi_rho', 'xi_v'], 'inner':'xi_u', 'outer':'xi_psi'}, 
        'eta':{'center':['eta_rho', 'eta_u'], 'inner':'eta_v', 'outer':'eta_psi'}, 
        's':{'center':'s_rho', 'outer':'s_w'}}

but it doesn’t

grid = Grid(model, coords=coords_roms)
grid
---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
~/miniconda3/envs/owl/lib/python3.6/site-packages/IPython/core/formatters.py in __call__(self, obj)
    700                 type_pprinters=self.type_printers,
    701                 deferred_pprinters=self.deferred_printers)
--> 702             printer.pretty(obj)
    703             printer.flush()
    704             return stream.getvalue()

~/miniconda3/envs/owl/lib/python3.6/site-packages/IPython/lib/pretty.py in pretty(self, obj)
    398                         if cls is not object \
    399                                 and callable(cls.__dict__.get('__repr__')):
--> 400                             return _repr_pprint(obj, self, cycle)
    401 
    402             return _default_pprint(obj, self, cycle)

~/miniconda3/envs/owl/lib/python3.6/site-packages/IPython/lib/pretty.py in _repr_pprint(obj, p, cycle)
    693     """A pprint that just redirects to the normal repr function."""
    694     # Find newlines and replace them with p.break_()
--> 695     output = repr(obj)
    696     for idx,output_line in enumerate(output.splitlines()):
    697         if idx:

~/miniconda3/envs/owl/lib/python3.6/site-packages/xgcm-0.1.0-py3.6.egg/xgcm/grid.py in __repr__(self)
    783             is_periodic = 'periodic' if axis._periodic else 'not periodic'
    784             summary.append('%s Axis (%s):' % (name, is_periodic))
--> 785             summary += axis._coord_desc()
    786         return '\n'.join(summary)
    787 

~/miniconda3/envs/owl/lib/python3.6/site-packages/xgcm-0.1.0-py3.6.egg/xgcm/grid.py in _coord_desc(self)
    166         summary = []
    167         for name, coord in iteritems(self.coords):
--> 168             coord_info = ('  * %-8s %s (%g)' % (name, coord.name, len(coord)))
    169             if name in self._default_shifts:
    170                 coord_info += ' --> %s' % self._default_shifts[name]

~/miniconda3/envs/owl/lib/python3.6/site-packages/xarray/core/common.py in __getattr__(self, name)
    174                     return source[name]
    175         raise AttributeError("%r object has no attribute %r" %
--> 176                              (type(self).__name__, name))
    177 
    178     def __setattr__(self, name, value):

AttributeError: 'Dataset' object has no attribute 'name'

trying to create an “axis” attribute in my dataset doesn’t​ seem to work either:

In [92]:

model.xi_rho.attrs['axis']=model.xi_u.attrs['axis']=model.xi_v.attrs['axis']=model.lon_psi.attrs['axis']='xi'
model.xi_u.attrs['c_grid_axis_shift']= 0.5
model.xi_psi.attrs['c_grid_axis_shift']= -0.5
​
grid = Grid(model, periodic=['xi'])
grid
​
Out[92]:
<xgcm.Grid>

Any idea on how to solve this issue?

please, please, and thank you

[dcherian]

Regarding the z-coordinate, because ROMS output doesn’t generally have that as an output, we have to recreate it depending on the function used in the run.

should add this to cf-xarray: https://github.com/xarray-contrib/cf-xarray/issues/34

[rabernat]

xgcm doesn’t currently recognize the typical ROMS output grid, which follows the SGRID convention,

We would still like to solve this! See #109…just has not been implemented yet.

I don’t know then if there is a need to do something in the xgcm code, as we are happy with the current solution that we use in xroms.

As a long-term roadmap, we would love to improve things in xgcm such that fewer workarounds are needed downstream, e.g. in xroms. For the short-term, I’m glad things are in a useable state as is.