FCN3

GlobalMRF202580 GBNVIDIAPyTorch

FourCastNet 3 is NVIDIA's probabilistic machine-learning weather model, built on spherical (geometric) signal processing with a hidden-Markov ensemble formulation: each member evolves its own calibrated stochastic state, so the ensemble spread is learned rather than imposed by initial-condition perturbations. It forecasts 72 atmospheric variables globally at 0.25° resolution with a 6-hour step.

Skill

Pick a metric and variable; hover for exact values at each lead time.

Evaluation

16-member ensemble · 24 initial conditions · 14-day horizon · 72 variables

Scores are latitude-weighted (cos φ) and aggregated over the initial conditions. Evaluation is done against ERA5 fetched from ARCO.

Type	16-member ensemble
Initial conditions	24 (2025)
Initial condition source	ERA5 (ARCO)
Verification (ground truth)	ERA5 (ARCO)
Lead times	6 h to 14 days
Variables scored	72
Metrics	RMSE, Log spectral distance, RMSE (ensemble mean), CRPS, Spread, Spread / Skill

Variables

Scored output variables (72)

Name	Description	Unit	Group
msl	Mean sea level pressure	Pa	Surface
t2m	Temperature at 2m	K	Surface
tcwv	Total column water vapor / precipitable water	kg m⁻²	Surface
u100m	U-component of wind at 100 m	m s⁻¹	Surface
u10m	U-component (eastward, zonal) of wind at 10 m	m s⁻¹	Surface
v100m	V-component of wind at 100 m	m s⁻¹	Surface
v10m	V-component (northward, meridional) of wind at 10 m	m s⁻¹	Surface
z50	Geopotential at 50 hPa	m² s⁻²	Geopotential
z100	Geopotential at 100 hPa	m² s⁻²	Geopotential
z150	Geopotential at 150 hPa	m² s⁻²	Geopotential
z200	Geopotential at 200 hPa	m² s⁻²	Geopotential
z250	Geopotential at 250 hPa	m² s⁻²	Geopotential
z300	Geopotential at 300 hPa	m² s⁻²	Geopotential
z400	Geopotential at 400 hPa	m² s⁻²	Geopotential
z500	Geopotential at 500 hPa	m² s⁻²	Geopotential
z600	Geopotential at 600 hPa	m² s⁻²	Geopotential
z700	Geopotential at 700 hPa	m² s⁻²	Geopotential
z850	Geopotential at 850 hPa	m² s⁻²	Geopotential
z925	Geopotential at 925 hPa	m² s⁻²	Geopotential
z1000	Geopotential at 1000 hPa	m² s⁻²	Geopotential
t50	Temperature at 50 hPa	K	Temperature
t100	Temperature at 100 hPa	K	Temperature
t150	Temperature at 150 hPa	K	Temperature
t200	Temperature at 200 hPa	K	Temperature
t250	Temperature at 250 hPa	K	Temperature
t300	Temperature at 300 hPa	K	Temperature
t400	Temperature at 400 hPa	K	Temperature
t500	Temperature at 500 hPa	K	Temperature
t600	Temperature at 600 hPa	K	Temperature
t700	Temperature at 700 hPa	K	Temperature
t850	Temperature at 850 hPa	K	Temperature
t925	Temperature at 925 hPa	K	Temperature
t1000	Temperature at 1000 hPa	K	Temperature
u50	U-component of wind at 50 hPa	m s⁻¹	U wind
u100	U-component of wind at 100 hPa	m s⁻¹	U wind
u150	U-component of wind at 150 hPa	m s⁻¹	U wind
u200	U-component of wind at 200 hPa	m s⁻¹	U wind
u250	U-component of wind at 250 hPa	m s⁻¹	U wind
u300	U-component of wind at 300 hPa	m s⁻¹	U wind
u400	U-component of wind at 400 hPa	m s⁻¹	U wind
u500	U-component of wind at 500 hPa	m s⁻¹	U wind
u600	U-component of wind at 600 hPa	m s⁻¹	U wind
u700	U-component of wind at 700 hPa	m s⁻¹	U wind
u850	U-component of wind at 850 hPa	m s⁻¹	U wind
u925	U-component of wind at 925 hPa	m s⁻¹	U wind
u1000	U-component of wind at 1000 hPa	m s⁻¹	U wind
v50	V-component of wind at 50 hPa	m s⁻¹	V wind
v100	V-component of wind at 100 hPa	m s⁻¹	V wind
v150	V-component of wind at 150 hPa	m s⁻¹	V wind
v200	V-component of wind at 200 hPa	m s⁻¹	V wind
v250	V-component of wind at 250 hPa	m s⁻¹	V wind
v300	V-component of wind at 300 hPa	m s⁻¹	V wind
v400	V-component of wind at 400 hPa	m s⁻¹	V wind
v500	V-component of wind at 500 hPa	m s⁻¹	V wind
v600	V-component of wind at 600 hPa	m s⁻¹	V wind
v700	V-component of wind at 700 hPa	m s⁻¹	V wind
v850	V-component of wind at 850 hPa	m s⁻¹	V wind
v925	V-component of wind at 925 hPa	m s⁻¹	V wind
v1000	V-component of wind at 1000 hPa	m s⁻¹	V wind
q50	Specific humidity at 50 hPa	kg kg⁻¹	Specific humidity
q100	Specific humidity at 100 hPa	kg kg⁻¹	Specific humidity
q150	Specific humidity at 150 hPa	kg kg⁻¹	Specific humidity
q200	Specific humidity at 200 hPa	kg kg⁻¹	Specific humidity
q250	Specific humidity at 250 hPa	kg kg⁻¹	Specific humidity
q300	Specific humidity at 300 hPa	kg kg⁻¹	Specific humidity
q400	Specific humidity at 400 hPa	kg kg⁻¹	Specific humidity
q500	Specific humidity at 500 hPa	kg kg⁻¹	Specific humidity
q600	Specific humidity at 600 hPa	kg kg⁻¹	Specific humidity
q700	Specific humidity at 700 hPa	kg kg⁻¹	Specific humidity
q850	Specific humidity at 850 hPa	kg kg⁻¹	Specific humidity
q925	Specific humidity at 925 hPa	kg kg⁻¹	Specific humidity
q1000	Specific humidity at 1000 hPa	kg kg⁻¹	Specific humidity

All of the model's output variables that have ERA5 verification are scored.

Data

The numbers behind the plot are in eval_scores_fcn3.json, exported by the eval recipe scorecard (scorecard/export_scores.py --docs) -- one value per metric, variable and lead time, in the variable's own units.

Reproducibility

Run and environment details

Date scored	2026-08-20
Scores written	2026-08-20
GPUs	8 x NVIDIA H100 80GB HBM3 (single node)
PyTorch	2.12.0a0+0291f960b6.nv26.04.48445190
CUDA	13.2
Python	3.12.3
Repo commit	2e3d1fd9a3f3
Provenance source	run
Exported	2026-08-20
Locked dependencies	uv.lock @ 2e3d1fd9a3f3

Reference

Bonev, B., Kurth, T., Mahesh, A., Bisson, M., Kossaifi, J., Kashinath, K., ... & Keller, A. (2025). FourCastNet 3: A geometric approach to probabilistic machine-learning weather forecasting at scale. arXiv preprint arXiv:2507.12144.