Remove data_full_zdim from fieldfilebuffer

[VeckoTheGecko]

Removed data_full_zdim from fieldfilebuffer and the calling code.

• Chose the correct base branch (v4-dev)
• xref Create an inventory of features to drop in v4 #1844

[VeckoTheGecko]

@erikvansebille I had a questions about a failing test that I am encountering tests/test_grids.py::test_bgrid_interpolation[*-mom5], namely that it's not finding the sw_ocean variable in the datasets as requested by the dimension dict.

dimensions = {
    "U": {"lon": "xu_ocean", "lat": "yu_ocean", "depth": "sw_ocean", "time": "time"},
    "V": {"lon": "xu_ocean", "lat": "yu_ocean", "depth": "sw_ocean", "time": "time"},
    "W": {"lon": "xu_ocean", "lat": "yu_ocean", "depth": "sw_ocean", "time": "time"},
}

Looking at each of the datasets, it is true that there isn't a sw_ocean for the U and V files (but is an st_ocean variable).

(Pdb) xr.open_mfdataset(ufile)
<xarray.Dataset> Size: 4kB
Dimensions:      (st_ocean: 4, time: 3, nv: 2, yu_ocean: 8, xu_ocean: 11)
Coordinates:
  * st_ocean     (st_ocean) float64 32B 0.5413 1.681 2.94 4.332
  * time         (time) object 24B 1950-01-01 12:00:00 ... 1950-01-03 12:00:00
  * xu_ocean     (xu_ocean) float64 88B -247.4 -247.3 -247.2 ... -246.5 -246.4
  * yu_ocean     (yu_ocean) float64 64B -64.27 -64.22 -64.18 ... -64.0 -63.96
Dimensions without coordinates: nv
Data variables:
    time_bounds  (time, nv) timedelta64[ns] 48B dask.array<chunksize=(3, 2), meta=np.ndarray>
    u            (time, st_ocean, yu_ocean, xu_ocean) float32 4kB dask.array<chunksize=(1, 4, 8, 11), meta=np.ndarray>
Attributes:
    filename:   ocean_daily_3d_u.nc
    title:      ACCESS-OM2-01
    grid_type:  mosaic
    grid_tile:  1
    history:    Thu Oct  8 09:46:23 2020: ncks -d st_ocean,0,3 -d yu_ocean,39...
    NCO:        netCDF Operators version 4.8.1 (Homepage = http://nco.sf.net,...


(Pdb) xr.open_mfdataset(vfile)
<xarray.Dataset> Size: 4kB
Dimensions:      (st_ocean: 4, time: 3, nv: 2, yu_ocean: 8, xu_ocean: 11)
Coordinates:
  * st_ocean     (st_ocean) float64 32B 0.5413 1.681 2.94 4.332
  * time         (time) object 24B 1950-01-01 12:00:00 ... 1950-01-03 12:00:00
  * xu_ocean     (xu_ocean) float64 88B -247.4 -247.3 -247.2 ... -246.5 -246.4
  * yu_ocean     (yu_ocean) float64 64B -64.27 -64.22 -64.18 ... -64.0 -63.96
Dimensions without coordinates: nv
Data variables:
    time_bounds  (time, nv) timedelta64[ns] 48B dask.array<chunksize=(3, 2), meta=np.ndarray>
    v            (time, st_ocean, yu_ocean, xu_ocean) float32 4kB dask.array<chunksize=(1, 4, 8, 11), meta=np.ndarray>
Attributes:
    filename:   ocean_daily_3d_v.nc
    title:      ACCESS-OM2-01
    grid_type:  mosaic
    grid_tile:  1
    history:    Thu Oct  8 09:47:11 2020: ncks -d st_ocean,0,3 -d yu_ocean,39...
    NCO:        netCDF Operators version 4.8.1 (Homepage = http://nco.sf.net,...


(Pdb) xr.open_mfdataset(wfile)
<xarray.Dataset> Size: 4kB
Dimensions:      (sw_ocean: 4, time: 3, nv: 2, yt_ocean: 8, xt_ocean: 11)
Coordinates:
  * sw_ocean     (sw_ocean) float64 32B 1.083 2.279 3.601 5.062
  * time         (time) object 24B 1950-01-01 12:00:00 ... 1950-01-03 12:00:00
  * xt_ocean     (xt_ocean) float64 88B -247.4 -247.3 -247.2 ... -246.5 -246.4
  * yt_ocean     (yt_ocean) float64 64B -64.29 -64.24 -64.2 ... -64.03 -63.98
Dimensions without coordinates: nv
Data variables:
    time_bounds  (time, nv) timedelta64[ns] 48B dask.array<chunksize=(3, 2), meta=np.ndarray>
    wt           (time, sw_ocean, yt_ocean, xt_ocean) float32 4kB dask.array<chunksize=(1, 4, 8, 11), meta=np.ndarray>
Attributes:
    filename:   ocean_daily_3d_wt.nc
    title:      ACCESS-OM2-01
    grid_type:  mosaic
    grid_tile:  1
    history:    Thu Oct  8 09:47:47 2020: ncks -d sw_ocean,0,3 -d yt_ocean,39...
    NCO:        netCDF Operators version 4.8.1 (Homepage = http://nco.sf.net,...
(Pdb) 

Updating the dim dict to use st_ocean instead of sw_ocean for U and V gives:

        if "U" in dimensions and "W" in dimensions and dimensions["U"] != dimensions["W"]:
>           raise ValueError(
                "On a B-grid, the dimensions of velocities should be the (top) corners of the grid cells, so the same for U, V and W. "
                "See also https://docs.oceanparcels.org/en/latest/examples/documentation_indexing.html"
            )
E           ValueError: On a B-grid, the dimensions of velocities should be the (top) corners of the grid cells, so the same for U, V and W. See also https://docs.oceanparcels.org/en/latest/examples/documentation_indexing.html

parcels/fieldset.py:873: ValueError

Disabling for now, but just wondering if you can provide additional insight here?

[erikvansebille]

Yes, exactly! This is essentially the same argument that we use in the C-grid interpolation: the Parcels B-grid interpolation scheme expects the grid point on the w-level. And that is only available in the W field. This is exactly why we need to be very careful on how we construct the new grid in v4

[VeckoTheGecko]

I see. There is also a breakage in tutorial_nemo_3D.ipynb related to how the depth dimension is handled. Should we disable the notebook for now until we clarify and implement how we're going to ingest grids in v4 then we can re-enable?