And on the weather side, NOAA/GSL is working to catch up on these issues as well. We’re working with colleagues at NCAR and NVIDIA to port physics parameterizations to GPU (primarily OpenACC as that’s what wraps around MPAS dynamics). The Air Force is also interested. Suspect 2026 will be exciting.

That’s so much of my struggle. My best mitigation approach has been writing out rough ideas somewhere else - usually my whiteboard - that I can then play around with (write, erase, erase again, etc.) in something that isn’t the formal document. When I’m somewhat happy with it, I pull it into Word.

Could you try MPASSIT? It can be a little tough to compile on non-NOAA HPCs, but is another option for converting to lat/lon. Or maybe uxarray for plotting directly without postproc?

They’re splitting off from NBC so they need a new name - though I’d say this isn’t a great new name…

Do you get the same if you try to use the same fields to initialize a WRF domain? That would help assess if it is a data issue (since that part of the init process is the same) or an init_atmosphere issue with MPAS.

If you’re looking to do computationally intensive work on the computer, I’d choose the MacBook Pro as it has better CPU cooling. The linked Reddit comment is a decent summary of the nuanced differences between the two.

www.reddit.com/r/macbook/co...