PhysicsNeMo Checkpoints: StormScope-GOES-MRMS

Description:

StormScope is a mesoscale weather AI model that autoregressively predicts GOES (satellite) and MRMS (radar) variables. The model can make predictions at multiple spatiotemporal resolutions, and the HRRR model grid is used as the base 3 km grid over the continental US (CONUS).

The model can be conditioned on additional data, such as synoptic-scale 500 hPa geopotential (z500), if desired. Models predicting MRMS are conditioned on the most recent GOES state variables.

For inference see NVIDIA Earth2Studio.

This model is for research and development only.

License/Terms of Use:

Governing Terms: Use of this model is governed by the NVIDIA Open Model License.

Deployment Geography:

Global

Use Case:

High-resolution ensemble forecasting of GOES and MRMS observations over the continental US (CONUS)

Release Date:

Hugging Face: 01/26/2026 via https://huggingface.co/nvidia/stormscope-goes-mrms

Model Architecture

Architecture Type: StormScope uses a diffusion transformer architecture with 2D neighborhood attention.
Network Architecture: Diffusion Transformer (DiT)

This model provides the following checkpoints:

• 6 km grid, 1 hr DiT -- 194M Parameters
• 6 km grid, 10 min DiT -- 195M Parameters
• 3 km grid, 10 min DiT -- 260M Parameters

Input:

Input Type(s):

• Tensor (1-8 state variables from GOES/MRMS, and 1-8 optional conditioning variables.)
• DateTime (NumPy Array)
  

Input Format(s): PyTorch Tensor / NumPy array
Input Parameters:

• Five Dimensional (5D) (batch, lead time, variable, height, width)
  
• Input DateTime (1D)
  

Other Properties Related to Input:

• Input grid (height/width) dimensions are that of the HRRR model grid.
• Input lead time can be of size one, for a single input timestep, or of size 6 for a series of input timesteps.
• Input state GOES variables: abi01c, abi02c, abi03c, abi07c, abi08c, abi09c, abi10c, abi13c
• Input state MRMS variables: refc
• Conditioning variables (optional): z500, or the GOES variables listed above.

For variable naming information, review the GOES and MRMS Lexicons in Earth2Studio.

Output:

Output Type(s): Tensor (1-8 state variables from GOES/MRMS, and 1-8 optional conditioning variable.)
Output Format: Pytorch Tensors
Output Parameters: Five Dimensional (5D) (batch, lead time, variable, latitude, longitude)
Other Properties Related to Output:

• Output grid (height/width) dimensions are that of the HRRR model grid.
• Output lead time is of size 1 (model predicts one timestep)
• Output state GOES variables: abi01c, abi02c, abi03c, abi07c, abi08c, abi09c, abi10c, abi13c
• Output state MRMS variables: refc

Our AI models are designed and/or optimized to run on NVIDIA GPU-accelerated systems. By leveraging NVIDIA’s hardware (e.g. GPU cores) and software frameworks (e.g., CUDA libraries), the model achieves faster training and inference times compared to CPU-only solutions.

Software Integration

Runtime Engine(s): Not Applicable
Supported Hardware Microarchitecture Compatibility:

• NVIDIA Ampere
  
• NVIDIA Blackwell
  
• NVIDIA Hopper
  

Supported Operating System(s):

• Linux
  

The integration of foundation and fine-tuned models into AI systems requires additional testing using use-case-specific data to ensure safe and effective deployment. Following the V-model methodology, iterative testing and validation at both unit and system levels are essential to mitigate risks, meet technical and functional requirements, and ensure compliance with safety and ethical standards before deployment.

Model Version(s):

Model Version: v1

Training, Testing, and Evaluation Datasets:

Training Dataset:

Link: ERA5

Data Collection Method by dataset:

• Automatic/Sensors
  

Labeling Method by dataset:

• Automatic/Sensors
  

Data Modality:

• Gridded geophysical time series

Data Size:

• 131 GB subset used for model training

Properties: ERA5 data for the period January 2018 - December 2023. ERA5 provides hourly estimates of various atmospheric, land, and oceanic climate variables. The data covers the Earth on a 30km grid and resolves the atmosphere at 137 levels.

Link: GOES

Data Collection Method by dataset

• Automatic/Sensors
  

Labeling Method by dataset

• Automatic/Sensors
  

Data Modality:

• Gridded geophysical time series

Data Size:

• 18.5 TB subset used for model training

Properties: GOES imaging data for the period January 2018 - December 2023. GOES is a NOAA geostationary satellite product providing multispectral radiances at 3 km resolution over a disk covering the continental US.

Link: MRMS

Data Collection Method by dataset

• Automatic/Sensors
  

Labeling Method by dataset

• Automatic/Sensors
  

Data Modality:

• Gridded geophysical time series

Data Size:

• 2.3 TB subset used for model training

Properties: MRMS radar data for the period January 2018 - December 2023. MRMS is a NOAA precipitation data product incorporating weather radar, surface, and other observations at high resolution over the continental US.

Testing Dataset:

Link: ERA5

Data Collection Method by dataset:

• Automatic/Sensors
  

Labeling Method by dataset:

• Automatic/Sensors
  

Properties: ERA5 data for the period January 2024 - December 2024. ERA5 provides hourly estimates of various atmospheric, land, and oceanic climate variables. The data covers the Earth on a 30km grid and resolves the atmosphere at 137 levels.

Link: GOES

Data Collection Method by dataset

• Automatic/Sensors
  

Labeling Method by dataset

• Automatic/Sensors
  

Properties: GOES imaging data for the period January 2024 - December 2024. GOES is a NOAA geostationary satellite product providing multispectral radiances at 3 km resolution over a disk covering the continental US.

Link: MRMS

Data Collection Method by dataset

• Automatic/Sensors
  

Labeling Method by dataset

• Automatic/Sensors
  

Properties: MRMS radar data for the period January 2024 - December 2024. MRMS is a NOAA precipitation data product incorporating weather radar, surface, and other observations at high resolution over the continental US.

Evaluation Dataset:

Link: ERA5

Data Collection Method by dataset:

• Automatic/Sensors
  

Labeling Method by dataset:

• Automatic/Sensors
  

Properties: ERA5 data for the period January 2025 - December 2025. ERA5 provides hourly estimates of various atmospheric, land, and oceanic climate variables. The data covers the Earth on a 30km grid and resolves the atmosphere at 137 levels.

Link: GOES

Data Collection Method by dataset

• Automatic/Sensors
  

Labeling Method by dataset

• Automatic/Sensors
  

Properties: GOES imaging data for the period January 2025 - December 2025. GOES is a NOAA geostationary satellite product providing multispectral radiances at 3 km resolution over a disk covering the continental US.

Link: MRMS

Data Collection Method by dataset

• Automatic/Sensors
  

Labeling Method by dataset

• Automatic/Sensors
  

Properties: MRMS radar data for the period January 2025 - December 2025. MRMS is a NOAA precipitation data product incorporating weather radar, surface, and other observations at high resolution over the continental US.

Inference:

Engine: PyTorch
Test Hardware:

• A100
  
• H100
  
• L40S
  

Ethical Considerations:

NVIDIA believes Trustworthy AI is a shared responsibility and we have established policies and practices to enable development for a wide array of AI applications. When downloaded or used in accordance with our terms of service, developers should work with their internal model team to ensure this model meets requirements for the relevant industry and use case and addresses unforeseen product misuse.

Please report model quality, risk, security vulnerabilities or NVIDIA AI Concerns here.