Product updates and fixes to delivery test script

[scottstanie]

Changes to output netcdf:

• Added "metadata" group, with both dolphin/disp-s1 software versions included
• Started to pull relevant metadata from the CSLC products, included zero_doppler_<start/stop>_time in the identification group to show the frame start/stop
• Added identification/bounding_polygon based what's inside the CSLC products
  • This does not close #3 because we don't fully trust this metadata to be accurate in all cases, and we want to make the polygon based on the real data.
• Added radar_wavelength

• added a /time dimension. Currently using the zero_doppler_start_time for this value.

  • While this appears to make it shaped (1, rows, cols), GDAL actually squeezes out the singleton dimension:

    io.load_gdal("netcdf:output/historical/20221119_20221201.unw.nc:unwrapped_phase").shape
    # (6194, 10017)

  • The benefit is that, by storing time as a dimension instead of just in the filename, the products are more analysis ready and can easily be loaded into a stack with, e.g. xarray:

    
    In [11]: xr.open_mfdataset(Path("output/historical/").glob("*nc"))
    Out[11]:
    <xarray.Dataset>
    Dimensions:                      (time: 2, y: 6194, x: 10017)
    Coordinates:

  • y (y) float64 4.404e+06 4.404e+06 ... 4.218e+06

  • x (x) float64 1.936e+04 1.94e+04 ... 3.198e+05

  • time (time) datetime64[ns] 2022-12-01T14:07:51.24... Data variables: spatial_ref (time) int64 0 0 unwrapped_phase (time, y, x) float32 dask.array<chunksize=(1, 6194, 10017), meta=np.ndarray> connected_component_labels (time, y, x) float32 dask.array<chunksize=(1, 6194, 10017), meta=np.ndarray> temporal_coherence (time, y, x) float32 dask.array<chunksize=(1, 6194, 10017), meta=np.ndarray> interferometric_correlation (time, y, x) float32 dask.array<chunksize=(1, 6194, 10017), meta=np.ndarray> ...

See this talk about Tidy Datacubes for more on the rationale

• Also added hdf5 exploration tools in disp_s1.plottting, copied from https://github.com/scottstanie/apertools/blob/master/apertools/h5explorer.py.

Closes #4 Closes #5

[scottstanie]

We need new tests for this, but they probably will need a full dummy data setup. I'm going to merge this for now so we can start testing testing (as running through the delivery test data works for me)