Every forecast argument eventually becomes a model argument. This page covers the major weather models from the GFS to the HRRR, how ensembles and model consensus work, and the climate-scale patterns like ENSO and the MJO that tilt the odds weeks to months out. Model talk runs deep in the Xtreme Weather Discord (XWD), and this is the vocabulary for it.
A model run is a single execution of a numerical weather prediction model initialized at a specific time (00Z, 06Z, 12Z, 18Z). The GFS runs every 6 hours; the HRRR every hour. Each run produces updated forecasts as new observational data is assimilated. Forecasters compare runs to identify trends and model consistency.
Model consensus refers to agreement between multiple numerical weather models (GFS, ECMWF, NAM, HRRR, etc.) on a forecast scenario. High consensus increases confidence. When models disagree significantly ('model spread'), confidence is lower and broader uncertainty ranges are appropriate. Forecasters weight models based on skill for specific scenarios.
Ensemble forecasting runs many slightly different model versions (varied initial conditions and physics) to produce a range of possible outcomes. Ensemble spread represents forecast uncertainty — tight cluster = high confidence; wide spread = high uncertainty. The GEFS (GFS Ensemble, 31 members) and ECMWF ENS (51 members) are the primary global ensembles.
The GFS (Global Forecast System) is NOAA's primary global NWP model, running 4x daily out to 16 days. The basis for most U.S. medium-range forecasts. Historically less accurate than the European ECMWF model at 5–10 days — but improvements since 2019 (FV3 dynamical core) have narrowed the gap significantly.
The NAM (North American Mesoscale Model) runs 4x daily out to 84 hours at 12km resolution, with 3km nests over the CONUS. The NAM 3km is widely used for day-1 severe weather parameter forecasting — it provides better mesoscale detail than the GFS for fronts, boundaries, and convective parameters.
The HRRR (High-Resolution Rapid Refresh) is NOAA's convection-allowing model — 3km resolution, hourly runs out to 18 hours (extended 48hr runs at select times). It explicitly simulates individual thunderstorm cells rather than parameterizing convection. The gold standard for same-day storm timing, coverage, and mode forecasting.
The ECMWF (European Centre for Medium-Range Weather Forecasts) model is widely considered the world's most accurate global weather model — particularly at 5–14 day lead times. Operated by a European intergovernmental organization in Reading, UK. It consistently outperforms the GFS at extended ranges and famously predicted Hurricane Sandy's unusual westward track days before American models.
The NWS point-and-click forecast (available at forecast.weather.gov) provides hourly and 7-day weather forecasts for any location in the U.S. It draws from the National Digital Forecast Database (NDFD) — a gridded mosaic of human-edited model guidance from all 122 WFOs.
NWS text products are official written forecasts and statements issued by WFOs using standardized identifiers (VTEC codes, AWIPS headers). Categories include: Watches/Warnings/Advisories, discussions (AFD, MCD, PNS), summaries (HWO, ESF), and records (LSR, Storm Data). Available at mesonet.agron.iastate.edu and api.weather.gov.
The Climate Prediction Center (CPC) is the NWS national center for long-range outlooks — temperature and precipitation probability at 6–10 day, 8–14 day, monthly, and seasonal scales. Also monitors ENSO (El Niño/La Niña), the Arctic Oscillation (AO), and other teleconnection patterns that drive seasonal weather.
ENSO (El Niño–Southern Oscillation) is the large-scale coupled ocean-atmosphere pattern in the tropical Pacific that significantly influences global weather patterns. El Niño (warm Pacific) tends to suppress Atlantic hurricane activity, strengthen the southern jet, and bring wetter winters to the south. La Niña (cool Pacific) often does the opposite — fueling more Atlantic hurricanes and producing drier winters in the southern U.S.
The Madden-Julian Oscillation (MJO) is the dominant mode of intraseasonal tropical variability — an eastward-propagating pulse of enhanced and suppressed tropical convection circling the tropics every 30–60 days. Its phase position modulates tornado outbreaks, precipitation patterns, and even winter storms across North America on 2–5 week timescales.
The Atlantic Multidecadal Oscillation (AMO) is a long-term (~60-70 year) cycle of above- and below-normal SSTs in the North Atlantic. The warm phase of the AMO correlates with more active Atlantic hurricane seasons and drought in the Midwest/Great Plains. It has been in a warm phase since ~1995.
The Pacific Decadal Oscillation (PDO) is a long-term (~20-30 year) pattern of SST variability in the Pacific. The positive (warm) phase brings warm SSTs in the eastern Pacific; the negative (cool) phase does the opposite. The PDO modulates the effects of ENSO — a warm PDO amplifies El Niño's impacts on North America.
The Quasi-Biennial Oscillation (QBO) is a cycle of easterly and westerly winds in the tropical stratosphere, alternating with a period of ~28 months. The QBO influences the polar vortex, Atlantic hurricane activity, and winter weather patterns. The westerly (positive) QBO phase is associated with stronger polar vortex and fewer Midwest cold snaps.
A climate normal is a 30-year average of a weather variable (temperature, precipitation, snowfall) used as a baseline for comparison. The current normals period is 1991–2020, updated every decade. When forecasters say 'above normal' or 'below normal,' they are comparing to these climatological benchmarks.
The U.S. Drought Monitor is a weekly national map depicting drought conditions across five categories: D0 (Abnormally Dry) through D4 (Exceptional Drought). Produced jointly by the NOAA, USDA, and the National Drought Mitigation Center. It directly triggers drought declarations, crop insurance payments, and water restrictions.