How to return an xarray with only valid values using `cross_section`?

[guoqing-noaa]

See below, uxvar.cross_section along a constant latitude will return an xarray with longitude from -180 to 180.
If you one try to check the values, it will return most nan.
Is it possible to return an xarray with only valid values? Thanks!

[philipc2]

Hi @guoqing-noaa

When we perform a cross-section, any sampled point that does not fall within a cell (i.e. outside a regional grid or within a continent in an ocean grid) is marked as nan

Since the output is now an xarray.DataArray (no longer attached to a Grid), we should be able to use the following:

# sampled cross section returns an Xarray DataArray 
da = uxds['var'].cross_section(lat=42.63, steps=100)

# Drop nan values along steps dimension
non_nan_da = da.dropna("steps")

[guoqing-noaa]

@philipc2 Thanks for the discussion.
Since we are working on a regional domain, if cross_section along a constant latitude/longitude, one would expect the returned xarray only covers regional domain.

We can live with this. But all nan values from the print out surprised me at first as I had thought I might did something wrong.

[erogluorhan]

Hi @guoqing-noaa , thanks for reporting this!

I agree there might be something to re-consider on this one. For a constant longitude cross-section, we are sampling latitudes from -90 to 90; while for a constant latitude cross-section, we are sampling longitudes from -180 to 180, and this is leading to all those NaNs when it is a regional grid, which might be confusing inspecting the output.

I don't have an idea or a strong preference though if it should be kept as is and this behavior should be documented clearly to reduce confusion, or whether we should consider sampling only for what lon/lat range is available from the data (rather than the global range for both).