Welcome to NexStorm

NexStorm is a personal project of mine. Being a bit of a weather hobbyist and also a electrical/computer engineer, I decided it would be a fun project to build, operate and maintain a cluster computer that runs my own weather models.

I initially built the cluster in 2020 with 2 FX8350s and a Ryzen 1700X system that I had laying around but quickly found out that I need faster hardware. The system now consists of 3 GTX 1070s with a Ryzen 7 2700X, a Ryzen 7 1700X, and a FX8350.

NexStorm AI Model (NSAIM)

The NSAIM is an Convolutional Neural Network based weather model that was trained using some of the NOAAs NAM output parameters and ground truth radar data that corresponds to that NAM forecast hour and date.

For operational runs, the Cluster Computer starts by downloading all 60 hours of NAM model data and then processes it into the needed inputs that the NSAIM was trained on (PWAT, SBCAPE, TMP, RH, VGRD, UGRD, DZDT). The NSAIM takes a 3D input and so it actually processes 12 millibar levels (450mb-1000mb) for all parameters except for PWAT and SBCAPE. After this is all processed, the NSAIM then runs using this 3D data as inputs to produce the predicted radar reflectivity.

Disclaimer: This is an experimental model and should not be used for safety related decisions. Refer to the NWS (weather.gov) for the most up to date information.

North American Model (NAM)

The NAM is a 3KM resolution model that predicts 60 hours into the future. The National Weather Service runs and maintains this model on their supercomputers named “Dogwood” and “Cactus”. Its outputs are used as inputs to the NSAIM. You can view the NAMs REFC output by clicking the change model button.