Using PET with ROMS model

[vicemagui]

I have a time series of surface 'zeta', 'u', and 'v' fields modeled with ROMS. The model outputs can be found here: https://thredds.met.no/thredds/dodsC/fou-hi/norkystv3_800m_m00_be. I have downloaded PET dev version in a virtual environment, as the 'pip' and 'mamba install' options failed to be downloaded.

You can open the dataset using xarray, like:

import xarray as xr

path_to_file = 'https://thredds.met.no/thredds/dodsC/fou-hi/norkystv3_800m_m00_be'
df = xr.open_dataset(path_to_file, engine='netcdf4').sel(time=slice('2026-01-01', '2026-01-02'))

I have tried to apply the eddy identification using both the unregular and regular grid options. For the former, I get the following error:

[IN]: g = UnRegularGridDataset(
    dataset,
    "lon",
    "lat",
)

---- > NumbaTypeError: Unsupported array type: numpy.ma.MaskedArray.

Conversely, if I use the option 'RegularGrid'

g = RegularGridDataset(
    dataset,
    "X",
    "Y",
)

g.add_uv("zeta", "u_eastward", "v_northward")

date = datetime(2026, 1, 1, 0)

a_adt, c_adt = g.eddy_identification("zeta", "u_eastward", "v_northward", date, 0.002)

and the output of that:

We assume pixel position of grid is centered for [./test_nk_nomask.nc](http://sm-nx10076537-bc-compute.int.met.no:8925/lab/tree/test_nk_nomask.nc)
[/home/victorm/mambaforge/envs/pet/lib/python3.10/site-packages/numpy/ma/core.py:1044](http://sm-nx10076537-bc-compute.int.met.no:8925/home/victorm/mambaforge/envs/pet/lib/python3.10/site-packages/numpy/ma/core.py#line=1043): RuntimeWarning: invalid value encountered in cast
  np.copyto(result, da, casting='unsafe', where=m)
[/home/victorm/mambaforge/envs/pet/lib/python3.10/site-packages/py_eddy_tracker/dataset/grid.py:1915](http://sm-nx10076537-bc-compute.int.met.no:8925/home/victorm/mambaforge/envs/pet/lib/python3.10/site-packages/py_eddy_tracker/dataset/grid.py#line=1914): RuntimeWarning: invalid value encountered in sqrt
  self._speed_ev = sqrt(u * u + v * v)
[/home/victorm/mambaforge/envs/pet/lib/python3.10/site-packages/numpy/lib/function_base.py:4824](http://sm-nx10076537-bc-compute.int.met.no:8925/home/victorm/mambaforge/envs/pet/lib/python3.10/site-packages/numpy/lib/function_base.py#line=4823): UserWarning: Warning: 'partition' will ignore the 'mask' of the MaskedArray.
  arr.partition(

---------------------------------------------------------------------------
TypingError                               Traceback (most recent call last)
Cell In[21], line 16
     12 g.add_uv("zeta", "u_eastward", "v_northward")
     14 date = datetime(2026, 1, 1, 0)
---> 16 a_adt, c_adt = g.eddy_identification("zeta", "u_eastward", "v_northward", date, 0.002)

File [~/mambaforge/envs/pet/lib/python3.10/site-packages/py_eddy_tracker/dataset/grid.py:780](http://sm-nx10076537-bc-compute.int.met.no:8925/home/victorm/mambaforge/envs/pet/lib/python3.10/site-packages/py_eddy_tracker/dataset/grid.py#line=779), in GridDataset.eddy_identification(self, grid_height, uname, vname, date, step, shape_error, presampling_multiplier, sampling, sampling_method, pixel_limit, precision, force_height_unit, force_speed_unit, **kwargs)
    777     continue
    778 # FIXME : center could be outside the contour due to the fit
    779 # FIXME : warning : the fit is made on raw sampling
--> 780 _, _, _, aerr = contour.fit_circle()
    782 # Filter for shape
    783 if aerr < 0 or aerr > shape_error or isnan(aerr):

File [~/mambaforge/envs/pet/lib/python3.10/site-packages/py_eddy_tracker/dataset/grid.py:143](http://sm-nx10076537-bc-compute.int.met.no:8925/home/victorm/mambaforge/envs/pet/lib/python3.10/site-packages/py_eddy_tracker/dataset/grid.py#line=142), in fit_circle_path(self, method)
    141 if method not in self._circle_params.keys():
    142     if method == "fit":
--> 143         self._circle_params["fit"] = _fit_circle_path(self.vertices)
    144     if method == "equal_area":
    145         self._circle_params["equal_area"] = _circle_from_equal_area(self.vertices)

File [~/mambaforge/envs/pet/lib/python3.10/site-packages/numba/core/dispatcher.py:422](http://sm-nx10076537-bc-compute.int.met.no:8925/home/victorm/mambaforge/envs/pet/lib/python3.10/site-packages/numba/core/dispatcher.py#line=421), in _DispatcherBase._compile_for_args(self, *args, **kws)
    418         msg = (f"{str(e).rstrip()} \n\nThis error may have been caused "
    419                f"by the following argument(s):\n{args_str}\n")
    420         e.patch_message(msg)
--> 422     error_rewrite(e, 'typing')
    423 except errors.UnsupportedError as e:
    424     # Something unsupported is present in the user code, add help info
    425     error_rewrite(e, 'unsupported_error')

File [~/mambaforge/envs/pet/lib/python3.10/site-packages/numba/core/dispatcher.py:363](http://sm-nx10076537-bc-compute.int.met.no:8925/home/victorm/mambaforge/envs/pet/lib/python3.10/site-packages/numba/core/dispatcher.py#line=362), in _DispatcherBase._compile_for_args.<locals>.error_rewrite(e, issue_type)
    361     raise e
    362 else:
--> 363     raise e.with_traceback(None)

TypingError: Failed in nopython mode pipeline (step: nopython frontend)
Untyped global name 'fit_circle': Cannot determine Numba type of <class 'function'>


File "../mambaforge/envs/pet/lib/python3.10/site-packages/py_eddy_tracker/dataset/grid.py", line 180:
def _fit_circle_path(vertice):
    <source elided>
        return 0, -90, nan, nan
    centlon, centlat, eddy_radius, err = fit_circle(c_x, c_y)
    ^

During: Pass nopython_type_inference

Note: X and Y here represent the x- and y- coordinates in Cartesian system. Therefore, they are not lat and lon values. Given that the model resolution is 800 m, points have assigned values of 0, 800, 1600, ...., N $\times$ 800, where N is the number of grid points.

Has anyone experienced this before?

Thanks in advance.

[vicemagui]

I have created this quick Zenodo draft folder containing the dataset and a simple Jupyter notebook.