WRF VS ufs-mrweather-app

WRF-ARW Modeling System is in the public domain and there is a fork called WRFg that can utilize GPU acceleration. The price of used Nvidia tesla GPUs could allow an enthusiast group to band together and thrive however that doesn't solve the sheer complexity of using the model and obtaining enough accurate data points to even run the model.

The NAM, RAP, and HRRR are all different smaller-scale, higher-resolution models for the US and North America based on the Weather Research and Forecasting model. These are due to eventually be replaced by the RRFS, based on the UFS Short-Range Weather Application.

example: https://github.com/wrf-model/WRF

Hi, I'm going to nitpick and say that most physics modeling software is not written in C. A whole bunch of physics programs are in Fortran (like weather models) .

For some context, WRF (https://github.com/wrf-model/WRF) the current state-of-the-art weather modeling system (also developed by NCAR) can only run on CPU. There have been efforts to run it on GPU (https://wrfg.net)... though it doesn't look like it's been kept up to date.

It seems like FastEddy mostly replaces WRF-LES, which is used for high-resolution localized modeling.

Most of the weather models run on supercomputers, and require a LOT of input data. Here’s one you can try if you’re so inclined. https://github.com/wrf-model/WRF

If anyone is curious, the WRF (Weather Research and Forecasting) model has a repository on GitHub.

GFS (Global Forecasting System, the US forecasting model for the whole globe) is based on the UFS weather model in the context of the Medium-Range Weather Application (though the current version of the app is quite out-of-date as they're doing a behind-the-scenes overhaul of how it's organized).

You could theoretically do this yourself; there are many open-source atmospheric models out there and plenty of open-source visualization software. The UFS Medium-Range Weather App is a version of the global weather prediction model run by the US National Weather Service, and it can even run on a laptop if you set it up with low enough resolution. So all you would have to figure out is how to create some generic plain initial conditions rather than starting with the publicly available real data as input.