Weather Models

Convection-allowing; assimilates radar reflectivity every 15 min.

Retired November 3, 2026 and replaced by RRFS v1.0. Was the hourly 13 km parent of the HRRR, used for aviation and SPC mesoanalysis.

Retired November 3, 2026, together with its 3 km NAM-NEST, and replaced by RRFS v1.0.

European mesoscale; convection-allowing over Central Europe.

High-resolution operational model for France and surroundings.

Convection-allowing mesoscale model covering Canada and portions of the northern U.S. Excellent for orographic precipitation and lake-effect snow detail.

CMC's primary regional model covering North America. Serves as the mesoscale companion to GEM-Global and feeds into GEPS.

NOAA’s FV3-based, convection-allowing model with hourly data assimilation including radar. Operational from the 12Z run on November 3, 2026 (SCN 26-48), replacing the NAM, NAM-NEST and RAP. Runs every hour at 3 km; the 00, 06, 12 and 18Z cycles go out to 84 hours.

NOAA's primary operational wave model; global and regional nests cover U.S. coastal zones, Great Lakes, and the western North Atlantic.

Spectral wave model covering the Great Lakes and coastal Atlantic/Pacific marine zones. Provides significant wave height, peak period, and direction at 2.5 km grid spacing — primary source for CMC marine and Great Lakes wave guidance.

CMC's global spectral wave model providing open-ocean wave height, period, and direction on a 0.25° grid. Feeds into RDWPS boundary conditions and supports offshore marine forecasting.

Third-generation spectral wave model coupled to the IFS atmospheric model; drives ECMWF's marine bulletins and extreme wave products.

Wave component of the GFS model suite using WAVEWATCH III physics; fully coupled to the atmospheric GFS for two-way air–sea feedback.

Wave ensemble driven by GEFS atmospheric members; underpins probabilistic significant wave height and swell guidance beyond day 3.

Graphical marine forecast analyses and warnings for North Atlantic and North Pacific including gale, storm, and hurricane force wind warnings plus sea state products.

NOAA's primary operational hurricane model; replaced HWRF in 2023. Two-way coupled to HYCOM ocean model with triple-nested moving domains that follow the storm.

Second HAFS configuration with alternate initialization and physics suite; run alongside HAFS-A for ensemble-like spread and consensus blending at NHC.

NOAA supplemental hurricane model coupled to HYCOM; provides additional track and intensity guidance for NHC consensus blends alongside HAFS.

Legacy operational hurricane model formally retired from primary operations in 2023 and replaced by HAFS. Still referenced in forecast discussions and historical verification archives.

Widely used statistical-dynamical intensity model blending large-scale environmental predictors — vertical shear, SST, mid-level humidity, and divergence — with climatological regression. Backbone of NHC intensity consensus products.

Probabilistic rapid intensification guidance derived from SHIPS predictors; provides the likelihood of RI (≥35 kt in 24 hrs) at each forecast cycle. Critical input to NHC intensity discussions.

Statistical intensity model using a logistic growth equation to model tropical cyclone lifecycle. Uses the same large-scale predictors as SHIPS; contributes to the NHC intensity consensus.

Simple consensus of GFS, ECMWF, UKMET, GEM, HAFS, and other track models. Consistently among the top-performing guidance products in NHC annual verification reports.

Consensus of SHIPS, LGEM, HAFS-A, HAFS-B, HMON, and dynamical model intensity guidance. Used alongside SHIPS-RII to inform NHC official intensity forecasts.

GFS output processed through the NHC tropical cyclone vortex tracker. AVNO is the standard NHC designation for GFS tropical track guidance in forecast discussions and verification.

GEFS ensemble mean and member tracks for extended-range tropical probability cones. The 31-member spread underpins probabilistic NHC outlooks and long-range genesis discussions beyond day 5.

ECMWF IFS deterministic output tracked by the NHC tropical cyclone tracker. Frequently the top-ranked individual model in NHC annual verification archives for track error.

51-member ENS spaghetti track plots provide probabilistic track uncertainty envelopes; widely used for extended-range tropical cyclone landfall probability assessments beyond day 5.

UK Met Office Unified Model output tracked for tropical cyclone guidance; labeled UKX in NHC model summary tables. Consistently competitive on Atlantic and East Pacific track forecasts.

GEM global model output tracked by the NHC; labeled CMC in forecast discussions and model summary tables. Contributes to the TVCN consensus.

Statistical intensity model developed by the Naval Research Laboratory for western North Pacific typhoon guidance; uses the same predictor framework as SHIPS.

U.S. primary global model; publicly available at full resolution.

Consistently among the world's most accurate deterministic models.

Strong performance in mid-latitudes; forms basis for MOGREPS ensemble.

Icosahedral nonhydrostatic model; good ensemble-scale accuracy.

Variable-resolution; finer over Europe, coarser at antipode.

Japan Meteorological Agency global spectral model.

Korean Meteorological Administration; adapted from UK Met UM.

Australian Bureau of Meteorology; strong Southern Hemisphere performance.

Canadian operational global model; regional version at 2.5 km.

China Meteorological Administration global model.

U.S. global ensemble; underpins all NOAA probabilistic guidance.

Gold-standard medium-range ensemble; EFI/SOT products at 15 days.

Sub-seasonal guidance to 6 weeks; weekly plumes widely used.

Global Met Office ensemble; feeds into IMPROVER blended products.

European ensemble; good Atlantic storm-track spread.

Japan ensemble system; strong western Pacific performance.

Canadian global ensemble prediction system.

CMC's short-range regional ensemble; 20 members at convection-permitting scale over North America. Feeds probabilistic QPF and severe weather guidance into WPC and SPC equivalent Canadian products.

SPC and WPC rely on SREFs for convective and QPF probabilistic products.

The RRFS-based convection-allowing ensemble, operational from 12Z on November 3, 2026 (SCN 26-48). Its NOMADS files carry members m001–m005 each cycle; our REFS FOUS adds the previous cycle’s five members, 6 hours lagged, for a 10-member ensemble.

ECMWF's operational AI model; matches HRES on many metrics and runs in minutes.

AI-based ensemble counterpart to AIFS; 50 members generated via stochastic perturbations, matching ECMWF ENS spread-skill on most metrics at a fraction of the compute cost.

Graph Neural Network trained on ERA5; outperformed HRES on 90% of test metrics.

Transformer-based model; competitive medium-range accuracy with very fast inference.

Spherical Fourier Neural Operator; among the first neural models to rival operational NWP.

Probabilistic diffusion-based ensemble; outperforms ECMWF ENS on a majority of metrics.

Foundation model pretrained on diverse atmospheric datasets; strong generalization.

Hybrid ML+physics model coupling a neural network with a differentiable dynamical core.

Transformer architecture trained directly on ERA5; strong on medium-range temperature fields.

Diffusion-based ensemble emulator conditioned on deterministic model output.

Foundation model for weather and climate built on NASA MERRA-2 reanalysis data.

NOAA's AI-driven replacement for GFS, trained on ERA5 and GFS reanalysis; targets parity with IFS on standard benchmarks.

AI-driven ensemble successor to GEFS; uses diffusion-based perturbation methods for probabilistic spread calibration.

ECMWF's experimental sub-seasonal AI ensemble covering weeks 2–6; resolution and member count are not yet verified here.

Foundation model pretrained across heterogeneous climate datasets; designed for fine-tuning on downstream forecast tasks.

Large-language-model-inspired architecture for weather; integrates textual and gridded inputs for probabilistic guidance.

Basis for CPC seasonal outlooks; 6-10 day through seasonal guidance.

ECMWF seasonal system; widely regarded best seasonal forecast skill.

Sub-seasonal to seasonal; bridges medium-range and seasonal forecasting.

UK seasonal system; good NAO and stratosphere coupling skill.

Japan Climate Prediction System, version 3.

Australian seasonal system; strong ENSO-driven anomaly forecasting.

NOAA's dedicated sub-seasonal system replacing the extended GEFS; targets week 3–4 temperature and precipitation skill. Underpins CPC week 3–4 outlooks.

Pools CFSv2, GFDL-SPEAR, NCAR CESM, NASA GEOS-S2S, and others into a combined probabilistic seasonal outlook. Underpins NOAA's official seasonal temperature and precipitation forecasts.

Combines CanCM4i and GEM-NEMO seasonal models; contributes to WMO lead centre seasonal outlooks and the NMME archive.

Next-generation NOAA seasonal system under active development; intended to eventually replace CFSv2. Incorporates improved ocean–atmosphere coupling and updated land surface parameterizations.