Can metpy.parse_cf() be used on WRF files?

Question

I'm trying to follow the MetPy cross-section plotting example (https://unidata.github.io/MetPy/latest/examples/cross_section.html), but applying it to WRF output data. I read in the WRF file as an xarray dataset, and then apply metpy.parse_cf().squeeze() to it, as in the example. However, I get this set of errors:

Traceback (most recent call last):
  File "plot_wrf_cross_overlays.py", line 252, in <module>
    main(init_dt_first, init_dt_last, init_stride_h, plot_beg_lead_time, plot_end_lead_time, plot_stride, domain)
  File "plot_wrf_cross_overlays.py", line 236, in main
    DS_wrf_xr = DS_wrf_xr.metpy.parse_cf().squeeze()
  File "/glade/work/jaredlee/python/my_npl_clone_20200417_cheyenne_3.7.5/lib/python3.7/site-packages/metpy/xarray.py", line 521, in parse_cf
    for single_varname in varname])
  File "/glade/work/jaredlee/python/my_npl_clone_20200417_cheyenne_3.7.5/lib/python3.7/site-packages/metpy/xarray.py", line 521, in <listcomp>
    for single_varname in varname])
  File "/glade/work/jaredlee/python/my_npl_clone_20200417_cheyenne_3.7.5/lib/python3.7/site-packages/metpy/xarray.py", line 537, in parse_cf
    var = self._fixup_coords(var)
  File "/glade/work/jaredlee/python/my_npl_clone_20200417_cheyenne_3.7.5/lib/python3.7/site-packages/metpy/xarray.py", line 566, in _fixup_coords
    var = var.metpy.convert_coordinate_units(coord_name, 'meter')
  File "/glade/work/jaredlee/python/my_npl_clone_20200417_cheyenne_3.7.5/lib/python3.7/site-packages/metpy/xarray.py", line 159, in convert_coordinate_units
    data=self._data_array[coord].metpy.unit_array.m_as(units)
  File "/glade/work/jaredlee/python/my_npl_clone_20200417_cheyenne_3.7.5/lib/python3.7/site-packages/metpy/xarray.py", line 135, in unit_array
    return self._data_array.values * self.units
  File "/glade/work/jaredlee/python/my_npl_clone_20200417_cheyenne_3.7.5/lib/python3.7/site-packages/pint/quantity.py", line 1540, in __array_ufunc__
    return numpy_wrap("ufunc", ufunc, inputs, kwargs, types)
  File "/glade/work/jaredlee/python/my_npl_clone_20200417_cheyenne_3.7.5/lib/python3.7/site-packages/pint/numpy_func.py", line 894, in numpy_wrap
    return handled[name](*args, **kwargs)
  File "/glade/work/jaredlee/python/my_npl_clone_20200417_cheyenne_3.7.5/lib/python3.7/site-packages/pint/numpy_func.py", line 289, in implementation
    result_magnitude = func(*stripped_args, **stripped_kwargs)
numpy.core._exceptions.UFuncTypeError: ufunc 'multiply' cannot use operands with types dtype('<M8[ns]') and dtype('int64')

I get the same thing even when I drop the squeeze() function call at the end. Any ideas what's causing this, or how to fix this? This is a standard wrfout file from WRF v4.2. I know WRF unfortunately doesn't follow CF-compliant conventions, and that's probably what's causing this problem, but surely this is a problem that others have had, so I'm hoping someone can help point me in the right direction here so that I could still hopefully use MetPy for WRF cross-section plots.

Jon Thielen · Answer 1 · 2021-06-14T19:46:13.137

In direct answer to your question, no, .metpy.parse_cf() cannot be used with WRF files, as they are not CF-compliant with regards to grid mapping metadata. However, for datasets without CF-compliant grid mappings, MetPy provides the .metpy.assign_crs() method to use instead to register the needed projection information for calculations like cross_section to work. For example, for a WRF dataset on a Lambert Conformal Conic projection, you can use the following as a replacement for parse_cf():

DS_wrf_xr = DS_wrf_xr.metpy.assign_crs({
    'grid_mapping_name': 'lambert_conformal_conic',
    'earth_radius': 6370000,
    'standard_parallel': [DS_wrf_xr.TRUELAT1, DS_wrf_xr.TRUELAT2],
    'longitude_of_central_meridian': DS_wrf_xr.STAND_LON,
    'latitude_of_projection_origin': DS_wrf_xr.MOAD_CEN_LAT
})

This being said, taking a look at your problem as a whole, the nature of the error you encountered in attempting to use .metpy.parse_cf() is an unexpected bug, and so I would recommend raising an issue on MetPy's issue tracker. Additionally, documentation (or helper routines) for using MetPy with WRF's ...peculiar... metadata is admittedly still lacking (see https://github.com/Unidata/MetPy/issues/1004 and https://github.com/Unidata/MetPy/issues/1089 for relevant discussion). For example, an issue you are likely to encounter with WRF cross-sections is that WRF datasets lack dimension coordinates, which will have to be derived using .assign_y_x or custom routines.

Thanks for the suggestions, Jon. Your code snippet did indeed get me around .metpy.parse_cf(). However, then when I used the metpy.interpolate.cross_section() function, it gave me an error: `TypeError: Object of type float32 is not JSON serializable.` I don't know SO rules/etiquette as I'm new, so should that be a new SO post, or continued as a new answer below, so I could post full error messages and code snippets? — Jared Lee, Jun 17 '21 at 22:25
Based on its rules, SO is not always the best for extended discussion and troubleshooting...I'm not sure the best way of working around that here, but if you're okay with it, I'd recommend posting this question (along with the full error messages and code snippets) all on [MetPy's Discussions page](https://github.com/Unidata/MetPy/discussions) instead where there can be better interactive conversations with other users and MetPy devs. — Jon Thielen, Jun 18 '21 at 14:00

Can metpy.parse_cf() be used on WRF files?

1 Answers1