--- license: apache-2.0 --- # Earth-2 Checkpoints: FourCastNet 3 ## Description: FourCastNet 3 advances global weather modeling by implementing a scalable, geometric machine learning (ML) approach to probabilistic ensemble forecasting. The approach is designed to respect spherical geometry and to accurately model the spatially correlated probabilistic nature of the problem, resulting in stable spectra and realistic dynamics across multiple scales. FourCastNet 3 delivers forecasting accuracy that surpasses leading conventional ensemble models and rivals the best diffusion-based methods, while producing forecasts 8 to 60 times faster than these approaches. In contrast to other ML approaches, FourCastNet 3 demonstrates excellent probabilistic calibration and retains realistic spectra, even at extended lead times of up to 60 days. All of these advances are realized using a purely convolutional neural network architecture specifically tailored for spherical geometry. Scalable and efficient large-scale training on 1024 GPUs and more is enabled by a novel training paradigm for combined model- and data-parallelism, inspired by domain decomposition methods in classical numerical models. Additionally, FourCastNet 3 enables rapid inference on a single GPU, producing a 60-day global forecast at 0.25°, 6-hourly resolution in under 4 minutes. Its computational efficiency, medium-range probabilistic skill, spectral fidelity, and rollout stability at subseasonal timescales make it a strong candidate for improving meteorological forecasting and early warning systems through large ensemble predictions. ![FCN3 15 member ensemble](https://raw.githubusercontent.com/NVIDIA/makani/main/images/fcn3_ens15.gif) This model is ready for commercial/non-commercial use. ### License/Terms of Use: [Apache 2.0 license](https://www.apache.org/licenses/LICENSE-2.0) ### Deployment Geography: Global ### Use Case: Industry, academic, and government research teams interested in medium-range and subseasonal-to-seasonal weather forecasting, and climate modeling. ### Release Date: NGC 07/18/2025 ## Reference: **Papers**: - [FourCastNet 3: A geometric approach to probabilistic machine-learning weather forecasting at scale](https://arxiv.org/abs/2507.12144v2) - [Neural Operators with Localized Integral and Differential Kernels](https://arxiv.org/abs/2402.16845) - [Huge Ensembles Part I: Design of Ensemble Weather Forecasts using Spherical Fourier Neural Operators](https://arxiv.org/abs/2408.03100) - [Huge Ensembles Part II: Properties of a Huge Ensemble of Hindcasts Generated with Spherical Fourier Neural Operators](https://arxiv.org/abs/2408.01581) - [Spherical Fourier Neural Operators: Learning Stable Dynamics on the Sphere](https://arxiv.org/abs/2306.03838) **Code**: - [Makani](https://github.com/NVIDIA/makani) - [PhysicsNeMo](https://github.com/NVIDIA/physicsnemo) - [Earth2Studio](https://github.com/NVIDIA/earth2studio) - [torch-harmonics](https://github.com/NVIDIA/torch-harmonics) ## Model Architecture: **Architecture Type:** Spherical Neural Operator. A fully convolutional architecture based on group convolutions defined on the sphere. Leverages both local and global convolutions. For details regarding the architecture refer to the [FourCastNet 3 paper](https://arxiv.org/abs/2507.12144v1).
**Network Architecture:** N/A
**Number of model parameters:** 710,867,670 **Model datatype:** We recommend that the model is run in AMP with bf16, however, the inputs and outputs are typically float32. ## Input: **Input Type:** - Tensor (72 surface and pressure-level variables) **Input Format:** PyTorch Tensor
**Input Parameters:** - Six Dimensional (6D) (batch, time, lead time, variable, latitude, longitude)
**Other Properties Related to Input:** - Input equi-rectangular latitude/longitude grid: 0.25 degree 721 x 1440 - Input state weather variables: `u10m`, `v10m`, `u100m`, `v100m`, `t2m`, `msl`, `tcwv`, `u50`, `u100`, `u150`, `u200`, `u250`, `u300`, `u400`, `u500`, `u600`, `u700`, `u850`, `u925`, `u1000`, `v50`, `v100`, `v150`, `v200`, `v250`, `v300`, `v400`, `v500`, `v600`, `v700`, `v850`, `v925`, `v1000`, `z50`, `z100`, `z150`, `z200`, `z250`, `z300`, `z400`, `z500`, `z600`, `z700`, `z850`, `z925`, `z1000`, `t50`, `t100`, `t150`, `t200`, `t250`, `t300`, `t400`, `t500`, `t600`, `t700`, `t850`, `t925`, `t1000`, `q50`, `q100`, `q150`, `q200`, `q250`, `q300`, `q400`, `q500`, `q600`, `q700`, `q850`, `q925`, `q1000` - Time: datetime64 For variable name information, review the Lexicon at [Earth2Studio](https://github.com/NVIDIA/earth2studio). ## Output: **Output Type:** Tensor (72 surface and pressure-level variables)
**Output Format:** Pytorch Tensor
**Output Parameters:** Six Dimensional (6D) (batch, time, lead time, variable, latitude, longitude)
**Other Properties Related to Output:** - Output latitude/longitude grid: 0.25 degree 721 x 1440, same as input. - Output state weather variables: same as above. 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 95%, testing 2.5%, validation 2.5%
## Training Dataset: **Link:** [ERA5](https://cds.climate.copernicus.eu/)
**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.
## Testing Dataset: **Link:** [ERA5](https://cds.climate.copernicus.eu/)
**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.
## Evaluation Dataset: **Link:** [ERA5](https://cds.climate.copernicus.eu/)
**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.
## 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. For more detailed information on ethical considerations for this model, please see the Model Card++ Explainability, Bias, Safety & Security, and Privacy Subcards. Please report model quality, risk, security vulnerabilities or NVIDIA AI Concerns [here](https://www.nvidia.com/en-us/support/submit-security-vulnerability/).