What is the proper way to preserve XArray and units when running a gaussian_filter?

[rolypolyman]

Let's say I have the following code that extracts some u-wind data from the models, along with other fields. I want to run a simple smoother like gaussian_filter() before feeding it into q_vector() because the source data is too noisy to use.

u_wind = data_wind_u.sel(isobaric=40000 * units.Pa).isel(time=time_index).metpy.quantify()

smoothed_u_wind = gaussian_filter(u_wind, sigma=sigma)

_.... process v_wind, temperature same as above ...._

u_q, v_q = mcalc.q_vector(smoothed_u_wind, smoothed_v_wind, smoothed_temperature, pressure, dx, dy)

The problem is I can't figure out a way to do this that preserves the Xarray structure, especially the units , which appear to be stored at the end of the array (not in attributes). Gaussian_filter() outputs a 2D array without the Xarray and I can't feed this into q_vector(). No problems if I feed u_wind directly into q_vector() without smoothing of the inputs, but I need the smoothing.

Maybe I am not approaching this the right way? Or maybe MetPy has its own filter?

[kgoebber]

HI @rolypolyman MetPy does have a couple of filtering options that we call smoothing.

https://unidata.github.io/MetPy/dev/examples/calculations/Smoothing.html#sphx-glr-examples-calculations-smoothing-py

The closest to the SciPy gaussian_filter would be MetPy's smooth_gaussian.

https://unidata.github.io/MetPy/dev/api/generated/metpy.calc.smooth_gaussian.html#metpy.calc.smooth_gaussian

If the function is fed an xarray DataArray, then it should give you back an xarray DataArray as is done in the MetPy xarray tutorial.

https://unidata.github.io/MetPy/dev/tutorials/xarray_tutorial.html#cf-compliant-dataset-example

[rolypolyman]

That did solve it. I will have to work through some more of the tutorials. I appreciate it!