PhysicsNeMo Checkpoints: DLESyM-V1-ERA5

Description:

DLESyM-V1-ERA5 is an ensemble forecast model for global earth system modeling. This model includes an atmosphere and ocean component, using atmospheric variables as well as the sea-surface temperature on a HEALPix nside=64 (approximately 1 degree) resolution grid. This model package includes several individual trained checkpoints for the atmosphere and ocean components, which can be used to improve model variability in ensembles. The model architecture is a U-Net with padding operations modified to support using the HEALPix grid.

This model is for research and development only.

License/Terms of Use:

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

Deployment Geography:

Global

Use Case:

Industry, academic, and government research teams interested in subseasonal-to-seasonal weather forecasting, and climate modeling.

Release Date:

NGC 05/12/2025

Reference:

• Advancing Parsimonious Deep Learning Weather Prediction using the HEALPix Mesh
  
• A Deep Learning Earth System Model for Efficient Simulation of the Observed Climate
  

Model Architecture:

Architecture Type: DLESyM uses two UNet architectures adapted to the HEALPix grid, one for each of the atmosphere and ocean components.
Network Architecture: UNet

Input:

Input Type:

• Tensor (9 surface and pressure-level variables)

Input Format: PyTorch Tensor
Input Parameters:

• Six Dimensional (6D) (batch, lead time, variable, face, height, width)
  

Other Properties Related to Input:

• Input latitude/longitude grid: 0.25 degree 721 x 1440, regridded to HEALPix nside=64 grid in "XY" format with a north origin and clockwise order
  • See HEALPIX_PAD_XY in the earth2grid package for specific details
• Input state weather variables: z500, tau300-700, z1000, t2m, tcwv, t850, z250, ws10m, sst
  • tau300-700 (geopotential thickness) is defined as the difference between z300 and z700 geopotential levels.
  • ws10m (wind speed at 10m above surface) is defined as the square root of the sum of the squared zonal and meridional wind components, i.e. sqrt(u10m **2 + v10m **2).

For variable name information, review the HRRR Lexicon at Earth2Studio. Review the config.yaml provided in the model package for information on the input lead times required by the model.

Output:

Output Type: Tensor (9 surface and pressure-level variables)
Output Format: Pytorch Tensor
Output Parameters: Six Dimensional (6D) (batch, lead time, variable, face, height, width)
Other Properties Related to Output:

• Output latitude/longitude grid: 0.25 degree 721 x 1440, regridded to HEALPix nside=64 grid in "XY" format with a north origin and clockwise order.
  • See HEALPIX_PAD_XY in the earth2grid package for specific details
• Output state weather variables: z500, tau300-700, z1000, t2m, tcwv, t850, z250, ws10m, sst
  
  • tau300-700 (geopotential thickness) is defined as the difference between z300 and z700 geopotential levels.
  • ws10m (wind speed at 10m above surface) is defined as the square root of the sum of the squared zonal and meridional wind components, i.e. sqrt(u10m **2 + v10m **2).

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: Pytorch
Supported Hardware Microarchitecture Compatibility:

• NVIDIA Ampere
  
• NVIDIA Hopper
  
• NVIDIA Turing
  

Supported Operating System:

• Linux
  

Model Version:

Model Version: v1

Training, Testing, and Evaluation Datasets:

Total size (in number of data points): 110,960
Total number of datasets: 1
Dataset partition: training 90%, testing 5%, validation 5%

Training Dataset:

Link: ERA5

Data Collection Method by dataset

• Automatic/Sensors
  

Labeling Method by dataset

• Automatic/Sensors
  

Properties: ERA5 data for the period 1980-2015. 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. We re-grid to a healpix-64 grid that corresponds to an approximate 1 degree lat/lon grid.

Testing Dataset:

Link: ERA5

Data Collection Method by dataset

• Automatic/Sensors
  

Labeling Method by dataset

• Automatic/Sensors
  

Properties: ERA5 data for the period 2016-2017. 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. We re-grid to a healpix-64 grid that corresponds to an approximate 1 degree lat/lon grid.

Evaluation Dataset:

Link: ERA5

Data Collection Method by dataset

• Automatic/Sensors
  

Labeling Method by dataset

• Automatic/Sensors
  

Properties: ERA5 data for the period 2018-2019. 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. We re-grid to a healpix-64 grid that corresponds to an approximate 1 degree lat/lon grid.

Inference:

Acceleration 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 security vulnerabilities or NVIDIA AI Concerns here.