Ryan May

@lpilz Honestly, I have no idea what the problem is. If a dask user with a vested interest :wink: were to submit a PR with that change and it didn't...

@sgdecker I think it's good to capture it here. This particular function is high on our list to make sure it works with Dask for our current funded work on...

@sgdecker I'll be sure to reach out when we have something, then. That's literally one of the core workflows/use cases that we're trying to support with the funded work we...

Nothing really new to report. I'm curious if you've tried using dask with MetPy and run into any calculations in particular that cause you problems?

You might want to consider [xradar](https://github.com/openradar/xradar) for that, at least for level 2, since that's reading nexrad level 2 directly into xarray.

This feature in general is going to force us to consider how MetPy consists of a set of re-usable pieces. Just some notes for where I see this going: 1....

> To avoid the coordinate explosion, I suppose the alternative would be to set the unprovided levels (e.g., the 800-mb absolute vorticity in this example) to all NaN or some...

Dealing with PNG/JPEG compression (assuming it's not JPEG2k) should be doable using Pillow (which is already in our dependency chain thanks to Matplotlib. The hard part here really isn't decoding...

I kinda figured since that's why we were dealing with the JJ2000 library over on the netcdf-java side. Looks like Pillow **can** support jpeg2k, not sure how regularly it's included...

@kgoebber's work-around is good. Digging in, the problem is that `tke` is essentially performing a reduction along the given axis, and our `preprocess_and_wrap()` doesn't handle that. (Correct me if I'm...