Global Forecast System (GFS)

Best (4-layer) lifted index uses parcels from the lowest four model layers to assess stability at 500 hPa. More negative values indicate greater instability.

• Units: K.
• Often used alongside CAPE and CIN in convective forecasting.

Absolute vorticity combines relative vorticity and planetary vorticity from Earth's rotation. It highlights rotation and dynamical forcing in the flow.

• Units: 1/s.

Convective precipitation accumulation represents precipitation produced by the convection scheme during the output period. It isolates convective rainfall from larger-scale precipitation.

• Units: kg/m².
• Useful for identifying convectively driven rainfall contributions.
• GFS selector {"name":"ACPCP","level":"surface","info":""}: Canonical product: Convective Precipitation, accumulated from forecast initialization to the valid time, at positive forecast leads only. No lead-zero row or synthetic zero is supplied. Difference endpoints within the same run to obtain interval totals; do not difference across runs. acc-from-init is an explicit alias of this base product; acc-reset is a different accumulation window after lead 6. Canonical statistical indexing has been validated for retained forecast initializations from 2021-03-22 12:00 UTC through 2026-11-07 12:00 UTC. This covers the affected statistical selectors, not unrelated source-offset exceptions or future incoming runs. Earlier base-selector results can differ from previously downloaded history after correction. Qualified selector definitions do not establish indexed history. Only selector tuples listed in this catalog are published; their advertised availability must be checked separately from the base selector. Units: kg/m^2.
• GFS selector {"name":"ACPCP","level":"surface","info":"acc-from-init"}: Convective Precipitation, accumulated from forecast initialization to the valid time, at positive forecast leads only. No lead-zero row or synthetic zero is supplied. Difference endpoints within the same run to obtain interval totals; do not difference across runs. This is the same native product as the empty-info base. Its window coincides with acc-reset at leads 1-6 and differs thereafter. Units: kg/m^2.
• GFS selector {"name":"ACPCP","level":"surface","info":"acc-reset"}: Convective Precipitation, accumulated over the source reset window at positive forecast leads only; no lead-zero row. For positive forecast lead h, the interval starts at 6*floor((h-1)/6) hours after initialization and ends at h. Its duration varies from 1 to 6 hours: leads 120, 123 and 126 cover 114-120, 120-123 and 120-126 hours respectively. It is not necessarily the interval since the previous output. Its window coincides with acc-from-init and the empty-info base at leads 1-6 and differs thereafter. Difference endpoints only within the same run and reset block. Units: kg/m^2.

Surface albedo is the fraction of incoming shortwave radiation reflected by the surface. Higher albedo means less solar energy absorbed and a cooler surface.

• Units: percent.
• Important for snow and ice feedbacks and surface energy balance.

Accumulated precipitation represents the total water-equivalent amount of rain and snow during the output period. It is a primary field for hydrologic impacts and event totals.

• Units: kg/m^2.
• Use with the model time interval to interpret totals and compare to gauges.
• GFS selector {"name":"APCP","level":"surface","info":""}: Canonical product: Total Precipitation, accumulated from forecast initialization to the valid time, at positive forecast leads only. No lead-zero row or synthetic zero is supplied. Difference endpoints within the same run to obtain interval totals; do not difference across runs. acc-from-init is an explicit alias of this base product; acc-reset is a different accumulation window after lead 6. Canonical statistical indexing has been validated for retained forecast initializations from 2021-03-22 12:00 UTC through 2026-11-07 12:00 UTC. This covers the affected statistical selectors, not unrelated source-offset exceptions or future incoming runs. Earlier base-selector results can differ from previously downloaded history after correction. Qualified selector definitions do not establish indexed history. Only selector tuples listed in this catalog are published; their advertised availability must be checked separately from the base selector. Units: kg/m^2.
• GFS selector {"name":"APCP","level":"surface","info":"acc-from-init"}: Total Precipitation, accumulated from forecast initialization to the valid time, at positive forecast leads only. No lead-zero row or synthetic zero is supplied. Difference endpoints within the same run to obtain interval totals; do not difference across runs. This is the same native product as the empty-info base. Its window coincides with acc-reset at leads 1-6 and differs thereafter. Units: kg/m^2.
• GFS selector {"name":"APCP","level":"surface","info":"acc-reset"}: Total Precipitation, accumulated over the source reset window at positive forecast leads only; no lead-zero row. For positive forecast lead h, the interval starts at 6*floor((h-1)/6) hours after initialization and ends at h. Its duration varies from 1 to 6 hours: leads 120, 123 and 126 cover 114-120, 120-123 and 120-126 hours respectively. It is not necessarily the interval since the previous output. Its window coincides with acc-from-init and the empty-info base at leads 1-6 and differs thereafter. Difference endpoints only within the same run and reset block. Units: kg/m^2.

Apparent temperature combines air temperature, humidity, and wind to estimate how hot or cold it feels to humans. It is useful for heat and cold stress assessments.

• Units: K.

Convective Available Potential Energy (CAPE) quantifies buoyant energy available to rising air parcels. Larger values generally imply stronger potential updrafts and greater convective intensity.

• Units: J/kg.
• Often used with CIN and lifting mechanisms to assess thunderstorm potential.

Categorical Freezing Rain indicates precipitation type. Instantaneous fields are binary indicators (0=no, 1=yes). Time-averaged variants may contain fractional indicator values; do not interpret them as calibrated probabilities. Read the selector statistic and interval.

• Units: dimensionless. The 0/1 code table describes binary indicators; time-averaged indicator fractions are not binary codes or calibrated probabilities. Read the selector statistic and interval.
• Useful as a mask for icing or glaze hazards.
• GFS selector {"name":"CFRZR","level":"surface","info":""}: Canonical product: Categorical Freezing Rain, reset-window average at positive forecast leads. For positive forecast lead h, the interval starts at 6*floor((h-1)/6) hours after initialization and ends at h. Its duration varies from 1 to 6 hours: leads 120, 123 and 126 cover 114-120, 120-123 and 120-126 hours respectively. It is not necessarily the interval since the previous output. Initialization exception: lead zero selects the instantaneous analysis, not an interval average. This base therefore mixes initialization analysis with positive-lead averages; ave-reset excludes analysis and instant selects instantaneous values at every available lead. Positive-lead values are time-averaged indicator fractions, not binary flags or calibrated probabilities; the lead-zero analysis is a binary indicator. Canonical statistical indexing has been validated for retained forecast initializations from 2021-03-22 12:00 UTC through 2026-11-07 12:00 UTC. This covers the affected statistical selectors, not unrelated source-offset exceptions or future incoming runs. Earlier base-selector results can differ from previously downloaded history after correction. Qualified selector definitions do not establish indexed history. Only selector tuples listed in this catalog are published; their advertised availability must be checked separately from the base selector. Units: fraction.
• GFS selector {"name":"CFRZR","level":"surface","info":"ave-reset"}: Categorical Freezing Rain, average over the source reset window at positive forecast leads only; no lead-zero or analysis row. It is the same native product as the empty-info base at positive leads and is distinct from instant. For positive forecast lead h, the interval starts at 6*floor((h-1)/6) hours after initialization and ends at h. Its duration varies from 1 to 6 hours: leads 120, 123 and 126 cover 114-120, 120-123 and 120-126 hours respectively. It is not necessarily the interval since the previous output. Values are time-averaged indicator fractions, not binary flags or calibrated probabilities. Units: fraction.
• GFS selector {"name":"CFRZR","level":"surface","info":"instant"}: Categorical Freezing Rain, instantaneous at the forecast valid time. This is distinct from the positive-lead average selected by empty info. At lead zero it selects the same instantaneous analysis as the empty-info base. Binary indicator: 1 means yes and 0 means no. Units: yes=1/no=0.

Categorical Ice Pellets indicates precipitation type. Instantaneous fields are binary indicators (0=no, 1=yes). Time-averaged variants may contain fractional indicator values; do not interpret them as calibrated probabilities. Read the selector statistic and interval.

• Units: dimensionless. The 0/1 code table describes binary indicators; time-averaged indicator fractions are not binary codes or calibrated probabilities. Read the selector statistic and interval.
• Use with precipitation fields to gauge intensity.
• GFS selector {"name":"CICEP","level":"surface","info":""}: Canonical product: Categorical Ice Pellets, reset-window average at positive forecast leads. For positive forecast lead h, the interval starts at 6*floor((h-1)/6) hours after initialization and ends at h. Its duration varies from 1 to 6 hours: leads 120, 123 and 126 cover 114-120, 120-123 and 120-126 hours respectively. It is not necessarily the interval since the previous output. Initialization exception: lead zero selects the instantaneous analysis, not an interval average. This base therefore mixes initialization analysis with positive-lead averages; ave-reset excludes analysis and instant selects instantaneous values at every available lead. Positive-lead values are time-averaged indicator fractions, not binary flags or calibrated probabilities; the lead-zero analysis is a binary indicator. Canonical statistical indexing has been validated for retained forecast initializations from 2021-03-22 12:00 UTC through 2026-11-07 12:00 UTC. This covers the affected statistical selectors, not unrelated source-offset exceptions or future incoming runs. Earlier base-selector results can differ from previously downloaded history after correction. Qualified selector definitions do not establish indexed history. Only selector tuples listed in this catalog are published; their advertised availability must be checked separately from the base selector. Units: fraction.
• GFS selector {"name":"CICEP","level":"surface","info":"ave-reset"}: Categorical Ice Pellets, average over the source reset window at positive forecast leads only; no lead-zero or analysis row. It is the same native product as the empty-info base at positive leads and is distinct from instant. For positive forecast lead h, the interval starts at 6*floor((h-1)/6) hours after initialization and ends at h. Its duration varies from 1 to 6 hours: leads 120, 123 and 126 cover 114-120, 120-123 and 120-126 hours respectively. It is not necessarily the interval since the previous output. Values are time-averaged indicator fractions, not binary flags or calibrated probabilities. Units: fraction.
• GFS selector {"name":"CICEP","level":"surface","info":"instant"}: Categorical Ice Pellets, instantaneous at the forecast valid time. This is distinct from the positive-lead average selected by empty info. At lead zero it selects the same instantaneous analysis as the empty-info base. Binary indicator: 1 means yes and 0 means no. Units: yes=1/no=0.

Convective Inhibition (CIN) measures the energy barrier that must be overcome to initiate convection. More negative values indicate a stronger cap that can suppress storms.

• Units: J/kg.
• When CIN is weak and CAPE is high, convection is more likely.

Cloud water mixing ratio is the mass of liquid cloud droplets per mass of air. It describes in‑cloud liquid water content.

• Units: kg/kg.
• Cloud fraction influences surface radiation and temperature.

Plant canopy surface water is liquid water intercepted by vegetation. It represents interception storage and evaporation potential.

• Units: kg/m^2.

Percent frozen precipitation indicates the fraction of total precipitation that is falling in frozen form. It helps distinguish mixed-phase and winter precipitation impacts.

• Units: percent.
• Higher values indicate a larger frozen contribution to the total precipitation.

Convective precipitation rate is the intensity of precipitation produced by convection. It highlights storm‑driven rainfall rates.

• Units: kg/m²/s.
• Accumulations represent water-equivalent depth over the output interval; rates describe intensity.
• GFS selector {"name":"CPRAT","level":"surface","info":""}: Canonical product: Convective Precipitation Rate, reset-window average at positive forecast leads. For positive forecast lead h, the interval starts at 6*floor((h-1)/6) hours after initialization and ends at h. Its duration varies from 1 to 6 hours: leads 120, 123 and 126 cover 114-120, 120-123 and 120-126 hours respectively. It is not necessarily the interval since the previous output. No lead-zero source product is present. At every available lead, ave-reset is an explicit alias of this base product. Canonical statistical indexing has been validated for retained forecast initializations from 2021-03-22 12:00 UTC through 2026-11-07 12:00 UTC. This covers the affected statistical selectors, not unrelated source-offset exceptions or future incoming runs. Earlier base-selector results can differ from previously downloaded history after correction. Qualified selector definitions do not establish indexed history. Only selector tuples listed in this catalog are published; their advertised availability must be checked separately from the base selector. Units: kg/m^2/s.
• GFS selector {"name":"CPRAT","level":"surface","info":"ave-reset"}: Convective Precipitation Rate, average over the source reset window at positive forecast leads only; no lead-zero or analysis row. It is the same native product as the empty-info base at positive leads and is distinct from instant. For positive forecast lead h, the interval starts at 6*floor((h-1)/6) hours after initialization and ends at h. Its duration varies from 1 to 6 hours: leads 120, 123 and 126 cover 114-120, 120-123 and 120-126 hours respectively. It is not necessarily the interval since the previous output. Units: kg/m^2/s.
• GFS selector {"name":"CPRAT","level":"surface","info":"instant"}: Convective Precipitation Rate, instantaneous at the forecast valid time. This is distinct from the positive-lead average selected by empty info. No lead-zero source product is present. Units: kg/m^2/s.

Categorical Rain indicates precipitation type. Instantaneous fields are binary indicators (0=no, 1=yes). Time-averaged variants may contain fractional indicator values; do not interpret them as calibrated probabilities. Read the selector statistic and interval.

• Units: dimensionless. The 0/1 code table describes binary indicators; time-averaged indicator fractions are not binary codes or calibrated probabilities. Read the selector statistic and interval.
• Use with accumulation or rate fields for intensity.
• GFS selector {"name":"CRAIN","level":"surface","info":""}: Canonical product: Categorical Rain, reset-window average at positive forecast leads. For positive forecast lead h, the interval starts at 6*floor((h-1)/6) hours after initialization and ends at h. Its duration varies from 1 to 6 hours: leads 120, 123 and 126 cover 114-120, 120-123 and 120-126 hours respectively. It is not necessarily the interval since the previous output. Initialization exception: lead zero selects the instantaneous analysis, not an interval average. This base therefore mixes initialization analysis with positive-lead averages; ave-reset excludes analysis and instant selects instantaneous values at every available lead. Positive-lead values are time-averaged indicator fractions, not binary flags or calibrated probabilities; the lead-zero analysis is a binary indicator. Canonical statistical indexing has been validated for retained forecast initializations from 2021-03-22 12:00 UTC through 2026-11-07 12:00 UTC. This covers the affected statistical selectors, not unrelated source-offset exceptions or future incoming runs. Earlier base-selector results can differ from previously downloaded history after correction. Qualified selector definitions do not establish indexed history. Only selector tuples listed in this catalog are published; their advertised availability must be checked separately from the base selector. Units: fraction.
• GFS selector {"name":"CRAIN","level":"surface","info":"ave-reset"}: Categorical Rain, average over the source reset window at positive forecast leads only; no lead-zero or analysis row. It is the same native product as the empty-info base at positive leads and is distinct from instant. For positive forecast lead h, the interval starts at 6*floor((h-1)/6) hours after initialization and ends at h. Its duration varies from 1 to 6 hours: leads 120, 123 and 126 cover 114-120, 120-123 and 120-126 hours respectively. It is not necessarily the interval since the previous output. Values are time-averaged indicator fractions, not binary flags or calibrated probabilities. Units: fraction.
• GFS selector {"name":"CRAIN","level":"surface","info":"instant"}: Categorical Rain, instantaneous at the forecast valid time. This is distinct from the positive-lead average selected by empty info. At lead zero it selects the same instantaneous analysis as the empty-info base. Binary indicator: 1 means yes and 0 means no. Units: yes=1/no=0.

Categorical Snow indicates precipitation type. Instantaneous fields are binary indicators (0=no, 1=yes). Time-averaged variants may contain fractional indicator values; do not interpret them as calibrated probabilities. Read the selector statistic and interval.

• Units: dimensionless. The 0/1 code table describes binary indicators; time-averaged indicator fractions are not binary codes or calibrated probabilities. Read the selector statistic and interval.
• Use alongside snowfall/snow-water-equivalent fields.
• GFS selector {"name":"CSNOW","level":"surface","info":""}: Canonical product: Categorical Snow, reset-window average at positive forecast leads. For positive forecast lead h, the interval starts at 6*floor((h-1)/6) hours after initialization and ends at h. Its duration varies from 1 to 6 hours: leads 120, 123 and 126 cover 114-120, 120-123 and 120-126 hours respectively. It is not necessarily the interval since the previous output. Initialization exception: lead zero selects the instantaneous analysis, not an interval average. This base therefore mixes initialization analysis with positive-lead averages; ave-reset excludes analysis and instant selects instantaneous values at every available lead. Positive-lead values are time-averaged indicator fractions, not binary flags or calibrated probabilities; the lead-zero analysis is a binary indicator. Canonical statistical indexing has been validated for retained forecast initializations from 2021-03-22 12:00 UTC through 2026-11-07 12:00 UTC. This covers the affected statistical selectors, not unrelated source-offset exceptions or future incoming runs. Earlier base-selector results can differ from previously downloaded history after correction. Qualified selector definitions do not establish indexed history. Only selector tuples listed in this catalog are published; their advertised availability must be checked separately from the base selector. Units: fraction.
• GFS selector {"name":"CSNOW","level":"surface","info":"ave-reset"}: Categorical Snow, average over the source reset window at positive forecast leads only; no lead-zero or analysis row. It is the same native product as the empty-info base at positive leads and is distinct from instant. For positive forecast lead h, the interval starts at 6*floor((h-1)/6) hours after initialization and ends at h. Its duration varies from 1 to 6 hours: leads 120, 123 and 126 cover 114-120, 120-123 and 120-126 hours respectively. It is not necessarily the interval since the previous output. Values are time-averaged indicator fractions, not binary flags or calibrated probabilities. Units: fraction.
• GFS selector {"name":"CSNOW","level":"surface","info":"instant"}: Categorical Snow, instantaneous at the forecast valid time. This is distinct from the positive-lead average selected by empty info. At lead zero it selects the same instantaneous analysis as the empty-info base. Binary indicator: 1 means yes and 0 means no. Units: yes=1/no=0.

Cloud water is the total liquid water content (often column‑integrated). It is used to assess cloud water burden.

• Units: kg/m².
• Cloud fraction influences surface radiation and temperature.

Cloud work function is an energy measure used by convection schemes to diagnose convective potential. Larger values generally indicate a more favorable environment for deep convection.

• Units: J/kg.
• Cloud fraction influences surface radiation and temperature.

Downward longwave (thermal) flux represents incoming energy at the surface. It is a key driver of surface heating and cooling.

• Units: W/m^2.
• Shortwave affects surface heating; longwave affects radiative cooling.

Dew point temperature is the temperature at which the air becomes saturated. It is a direct measure of moisture content.

• Units: K.
• When dew point is close to air temperature, fog or low cloud potential increases.

Downward radiative flux represents incoming energy at the surface. It is a key driver of surface heating and cooling.

• Units: W/m^2.
• Shortwave affects surface heating; longwave affects radiative cooling.
• GFS selector {"name":"DSWRF","level":"surface","info":""}: Downward Shortwave Radiation Flux, average radiative flux over the source reset window, not an instantaneous flux. Positive forecast leads only; no lead-zero source product. For positive forecast lead h, the interval starts at 6*floor((h-1)/6) hours after initialization and ends at h. Its duration varies from 1 to 6 hours: leads 120, 123 and 126 cover 114-120, 120-123 and 120-126 hours respectively. It is not necessarily the interval since the previous output. Units: W/m^2.

Geometric vertical velocity (dz/dt) is the vertical speed of air in height coordinates. Positive values indicate rising motion and negative values sinking motion.

• Units: m/s.

Field capacity is the soil water content after gravitational drainage. It represents the upper limit of plant-available water.

• Units: fraction.

Friction velocity (u*) is a measure of near‑surface shear stress. It is commonly used in turbulence, dust, and air‑sea exchange calculations.

• Units: m/s.

Ground heat flux is the conductive heat exchange between the surface and the subsurface. It is a key term in the surface energy balance.

• Units: W/m^2.

Graupel accumulation represents the water-equivalent amount of graupel in the column. It helps diagnose convective microphysics.

• Units: kg/kg.

Wind gust is the peak short-duration wind speed. It captures brief, hazardous bursts that can exceed sustained winds.

• Units: m/s.

High cloud cover is the fraction of the grid cell covered by clouds in the upper troposphere. It affects longwave radiation and aviation.

• Units: percent.
• Cloud fraction influences surface radiation and temperature.
• GFS selector {"name":"HCDC","level":"high cloud layer","info":""}: Canonical product: High Cloud Cover, reset-window average at positive forecast leads. For positive forecast lead h, the interval starts at 6*floor((h-1)/6) hours after initialization and ends at h. Its duration varies from 1 to 6 hours: leads 120, 123 and 126 cover 114-120, 120-123 and 120-126 hours respectively. It is not necessarily the interval since the previous output. Initialization exception: lead zero selects the instantaneous analysis, not an interval average. This base therefore mixes initialization analysis with positive-lead averages; ave-reset excludes analysis and instant selects instantaneous values at every available lead. Canonical statistical indexing has been validated for retained forecast initializations from 2021-03-22 12:00 UTC through 2026-11-07 12:00 UTC. This covers the affected statistical selectors, not unrelated source-offset exceptions or future incoming runs. Earlier base-selector results can differ from previously downloaded history after correction. Qualified selector definitions do not establish indexed history. Only selector tuples listed in this catalog are published; their advertised availability must be checked separately from the base selector. Units: %.
• GFS selector {"name":"HCDC","level":"high cloud layer","info":"ave-reset"}: High Cloud Cover, average over the source reset window at positive forecast leads only; no lead-zero or analysis row. It is the same native product as the empty-info base at positive leads and is distinct from instant. For positive forecast lead h, the interval starts at 6*floor((h-1)/6) hours after initialization and ends at h. Its duration varies from 1 to 6 hours: leads 120, 123 and 126 cover 114-120, 120-123 and 120-126 hours respectively. It is not necessarily the interval since the previous output. Units: %.
• GFS selector {"name":"HCDC","level":"high cloud layer","info":"instant"}: High Cloud Cover, instantaneous at the forecast valid time. This is distinct from the positive-lead average selected by empty info. At lead zero it selects the same instantaneous analysis as the empty-info base. Units: %.

Geopotential height is the altitude of a pressure surface in the atmosphere. It is the primary field for identifying ridges, troughs, and large‑scale flow patterns.

• Units: gpm.

Haines Index combines atmospheric stability and dryness to estimate the potential for large fire growth. Higher values indicate a more favorable environment for extreme fire behavior.

• Units: Numeric.
• Typically used in wildfire forecasting and risk monitoring.

Storm-relative helicity measures the potential for streamwise vorticity to be ingested by a storm updraft. Higher values suggest a greater likelihood of rotating updrafts.

• Units: m^2/s^2.
• Commonly evaluated in the lowest 1 to 3 km layer.

Boundary layer height marks the depth of the turbulent surface-influenced layer. It affects mixing, dispersion, and surface fluxes.

• Units: m.

Icing altitude is the height at which in-flight icing conditions are expected. It is used for aviation routing and hazard awareness.

• Units: m.
• More negative CIN indicates a stronger cap that can suppress convection.

Ice cover is the proportion of the grid cell covered by ice. It is used as a surface-state mask for sea ice and related cryosphere diagnostics.

• Units: Proportion.

Ice growth rate represents the rate of sea-ice thickness increase. It reflects freezing conditions at the surface and in the upper ocean.

• Units: kg/m^2.

Ice thickness is the physical thickness of sea or lake ice. It is important for marine navigation and heat exchange.

• Units: m.

Ice temperature is the thermal state of sea or lake ice. It affects melting and ice strength.

• Units: K.

Ice water mixing ratio is the mass of ice-phase hydrometeors per mass of air. It indicates ice cloud content.

• Units: kg/kg.

Land cover is a land/sea mask where 0 indicates sea and 1 indicates land. It is used to separate land and ocean surface processes.

• Units: Proportion.

Low cloud cover is the fraction of the grid cell covered by low clouds. It strongly affects visibility, fog potential, and surface temperatures.

• Units: percent.
• Cloud fraction influences surface radiation and temperature.
• GFS selector {"name":"LCDC","level":"low cloud layer","info":""}: Canonical product: Low Cloud Cover, reset-window average at positive forecast leads. For positive forecast lead h, the interval starts at 6*floor((h-1)/6) hours after initialization and ends at h. Its duration varies from 1 to 6 hours: leads 120, 123 and 126 cover 114-120, 120-123 and 120-126 hours respectively. It is not necessarily the interval since the previous output. Initialization exception: lead zero selects the instantaneous analysis, not an interval average. This base therefore mixes initialization analysis with positive-lead averages; ave-reset excludes analysis and instant selects instantaneous values at every available lead. Canonical statistical indexing has been validated for retained forecast initializations from 2021-03-22 12:00 UTC through 2026-11-07 12:00 UTC. This covers the affected statistical selectors, not unrelated source-offset exceptions or future incoming runs. Earlier base-selector results can differ from previously downloaded history after correction. Qualified selector definitions do not establish indexed history. Only selector tuples listed in this catalog are published; their advertised availability must be checked separately from the base selector. Units: %.
• GFS selector {"name":"LCDC","level":"low cloud layer","info":"ave-reset"}: Low Cloud Cover, average over the source reset window at positive forecast leads only; no lead-zero or analysis row. It is the same native product as the empty-info base at positive leads and is distinct from instant. For positive forecast lead h, the interval starts at 6*floor((h-1)/6) hours after initialization and ends at h. Its duration varies from 1 to 6 hours: leads 120, 123 and 126 cover 114-120, 120-123 and 120-126 hours respectively. It is not necessarily the interval since the previous output. Units: %.
• GFS selector {"name":"LCDC","level":"low cloud layer","info":"instant"}: Low Cloud Cover, instantaneous at the forecast valid time. This is distinct from the positive-lead average selected by empty info. At lead zero it selects the same instantaneous analysis as the empty-info base. Units: %.

Lifted Index compares a lifted parcel's temperature to the environment at 500 hPa. More negative values indicate greater instability and thunderstorm potential.

• Units: K.
• Values near zero indicate neutral stability.

Latent heat net flux represents energy exchange due to evaporation/condensation at the surface. It links moisture and energy budgets.

• Units: W/m^2.

Mid-level cloud cover is the fraction of the grid cell covered by clouds in the middle troposphere. It influences both shortwave and longwave radiation.

• Units: percent.
• Cloud fraction influences surface radiation and temperature.
• GFS selector {"name":"MCDC","level":"middle cloud layer","info":""}: Canonical product: Middle Cloud Cover, reset-window average at positive forecast leads. For positive forecast lead h, the interval starts at 6*floor((h-1)/6) hours after initialization and ends at h. Its duration varies from 1 to 6 hours: leads 120, 123 and 126 cover 114-120, 120-123 and 120-126 hours respectively. It is not necessarily the interval since the previous output. Initialization exception: lead zero selects the instantaneous analysis, not an interval average. This base therefore mixes initialization analysis with positive-lead averages; ave-reset excludes analysis and instant selects instantaneous values at every available lead. Canonical statistical indexing has been validated for retained forecast initializations from 2021-03-22 12:00 UTC through 2026-11-07 12:00 UTC. This covers the affected statistical selectors, not unrelated source-offset exceptions or future incoming runs. Earlier base-selector results can differ from previously downloaded history after correction. Qualified selector definitions do not establish indexed history. Only selector tuples listed in this catalog are published; their advertised availability must be checked separately from the base selector. Units: %.
• GFS selector {"name":"MCDC","level":"middle cloud layer","info":"ave-reset"}: Middle Cloud Cover, average over the source reset window at positive forecast leads only; no lead-zero or analysis row. It is the same native product as the empty-info base at positive leads and is distinct from instant. For positive forecast lead h, the interval starts at 6*floor((h-1)/6) hours after initialization and ends at h. Its duration varies from 1 to 6 hours: leads 120, 123 and 126 cover 114-120, 120-123 and 120-126 hours respectively. It is not necessarily the interval since the previous output. Units: %.
• GFS selector {"name":"MCDC","level":"middle cloud layer","info":"instant"}: Middle Cloud Cover, instantaneous at the forecast valid time. This is distinct from the positive-lead average selected by empty info. At lead zero it selects the same instantaneous analysis as the empty-info base. Units: %.

Mean sea level pressure (ETA reduction) is MSLP computed using the ETA model reduction method. It is used for synoptic analysis.

• Units: Pa.

Ozone mixing ratio is the mass of ozone per mass of air. It is used in radiation, chemistry, and upper‑air diagnostics.

• Units: kg/kg.

Potential evaporation rate is the potential latent heat flux for evaporation under an assumed unlimited water supply.

• Units: W/m^2.

Pressure of the level from which the parcel was lifted indicates the parcel source level used for stability indices. It helps interpret lifted index and CAPE products.

• Units: Pa.

Potential temperature is the temperature a parcel would have if brought dry‑adiabatically to a reference pressure. It is used for stability.

• Units: K.

Precipitation rate is the total intensity of precipitation over the period. It combines convective and stratiform contributions.

• Units: kg/(m^2 s).
• Accumulations represent water-equivalent depth over the output interval; rates describe intensity.
• GFS selector {"name":"PRATE","level":"surface","info":""}: Canonical product: Precipitation Rate, reset-window average at positive forecast leads. For positive forecast lead h, the interval starts at 6*floor((h-1)/6) hours after initialization and ends at h. Its duration varies from 1 to 6 hours: leads 120, 123 and 126 cover 114-120, 120-123 and 120-126 hours respectively. It is not necessarily the interval since the previous output. Initialization exception: lead zero selects the instantaneous analysis, not an interval average. This base therefore mixes initialization analysis with positive-lead averages; ave-reset excludes analysis and instant selects instantaneous values at every available lead. Canonical statistical indexing has been validated for retained forecast initializations from 2021-03-22 12:00 UTC through 2026-11-07 12:00 UTC. This covers the affected statistical selectors, not unrelated source-offset exceptions or future incoming runs. Earlier base-selector results can differ from previously downloaded history after correction. Qualified selector definitions do not establish indexed history. Only selector tuples listed in this catalog are published; their advertised availability must be checked separately from the base selector. Units: kg/m^2/s.
• GFS selector {"name":"PRATE","level":"surface","info":"ave-reset"}: Precipitation Rate, average over the source reset window at positive forecast leads only; no lead-zero or analysis row. It is the same native product as the empty-info base at positive leads and is distinct from instant. For positive forecast lead h, the interval starts at 6*floor((h-1)/6) hours after initialization and ends at h. Its duration varies from 1 to 6 hours: leads 120, 123 and 126 cover 114-120, 120-123 and 120-126 hours respectively. It is not necessarily the interval since the previous output. Units: kg/m^2/s.
• GFS selector {"name":"PRATE","level":"surface","info":"instant"}: Precipitation Rate, instantaneous at the forecast valid time. This is distinct from the positive-lead average selected by empty info. At lead zero it selects the same instantaneous analysis as the empty-info base. Units: kg/m^2/s.

Pressure is the atmospheric pressure at the specified level. It is a fundamental mass-field variable for dynamics and thermodynamics.

• Units: Pa.

Pressure reduced to mean sea level is MSLP computed from surface pressure and temperature. It is used to locate highs and lows.

• Units: Pa.

Precipitable water is the total column water vapor from the surface to the top of the atmosphere. It is a key moisture indicator.

• Units: kg/m^2.

Composite reflectivity is the maximum reflectivity in the column. It is a standard radar‑style field for identifying convective cores and precipitation intensity.

• Units: dBZ.

Reflectivity is a radar‑style measure of hydrometeor concentration at a given level. Higher values generally correspond to heavier precipitation.

• Units: dBZ.

Relative humidity describes how close the air is to saturation at the specified level. High values favor cloud formation.

• Units: percent.
• Higher values indicate more atmospheric moisture and greater cloud or precip potential.

Rain water mixing ratio is the mass of rainwater per mass of air. It describes rain content within the column.

• Units: kg/kg.
• Accumulations represent water-equivalent depth over the output interval; rates describe intensity.

Surface roughness length is the effective height of surface obstacles that control drag. It strongly influences near-surface wind, turbulence, and exchange fluxes.

• Units: m.

Sensible heat net flux is turbulent heat exchange driven by temperature differences. It affects near‑surface temperature and stability.

• Units: W/m^2.

Snow mixing ratio is the mass of snow particles per mass of air. It describes suspended snow content relevant to microphysics and precipitation.

• Units: kg/kg.
• Accumulations represent water-equivalent depth over the output interval; rates describe intensity.

Snow depth is the physical thickness of the snowpack. It is distinct from snow water equivalent, which measures water content.

• Units: m.

Liquid volumetric soil moisture (non-frozen) is the liquid portion of soil water expressed as a proportion of soil volume, excluding soil ice.

• Units: Proportion.
• Soil state affects evapotranspiration and boundary-layer moisture.

Volumetric soil moisture is the fraction of soil volume occupied by water. It is commonly used in hydrology and drought monitoring.

• Units: Fraction.
• Soil state affects evapotranspiration and boundary-layer moisture.

Soil type class used by the land-surface model. It controls infiltration, drainage, and heat/moisture storage, and strongly affects runoff and evapotranspiration.

• Units: code.

Specific humidity is the mass of water vapor per mass of air at the specified level. It is central to moisture transport.

• Units: kg/kg.
• Higher values indicate more atmospheric moisture and greater cloud or precip potential.

Sunshine duration is the total time with direct sunlight at the surface. It is useful for solar energy and climatological applications.

• Units: seconds.

Total cloud cover is the fraction of the grid cell covered by clouds at any level. It influences radiation, temperature, and visibility.

• Units: percent.
• Cloud fraction influences surface radiation and temperature.
• GFS selector {"name":"TCDC","level":"entire atmosphere","info":""}: Canonical product: Total Cloud Cover, reset-window average at positive forecast leads. For positive forecast lead h, the interval starts at 6*floor((h-1)/6) hours after initialization and ends at h. Its duration varies from 1 to 6 hours: leads 120, 123 and 126 cover 114-120, 120-123 and 120-126 hours respectively. It is not necessarily the interval since the previous output. Initialization exception: lead zero selects the instantaneous analysis, not an interval average. This base therefore mixes initialization analysis with positive-lead averages; ave-reset excludes analysis and instant selects instantaneous values at every available lead. Canonical statistical indexing has been validated for retained forecast initializations from 2021-03-22 12:00 UTC through 2026-11-07 12:00 UTC. This covers the affected statistical selectors, not unrelated source-offset exceptions or future incoming runs. Earlier base-selector results can differ from previously downloaded history after correction. Qualified selector definitions do not establish indexed history. Only selector tuples listed in this catalog are published; their advertised availability must be checked separately from the base selector. Units: %.
• GFS selector {"name":"TCDC","level":"entire atmosphere","info":"ave-reset"}: Total Cloud Cover, average over the source reset window at positive forecast leads only; no lead-zero or analysis row. It is the same native product as the empty-info base at positive leads and is distinct from instant. For positive forecast lead h, the interval starts at 6*floor((h-1)/6) hours after initialization and ends at h. Its duration varies from 1 to 6 hours: leads 120, 123 and 126 cover 114-120, 120-123 and 120-126 hours respectively. It is not necessarily the interval since the previous output. Units: %.
• GFS selector {"name":"TCDC","level":"entire atmosphere","info":"instant"}: Total Cloud Cover, instantaneous at the forecast valid time. This is distinct from the positive-lead average selected by empty info. At lead zero it selects the same instantaneous analysis as the empty-info base. Units: %.

Maximum temperature is the highest temperature reached during the output period. It is commonly used for heat stress and daily extremes.

• Units: K.

Minimum temperature is the lowest temperature reached during the output period. It is commonly used for frost and cold-stress assessment.

• Units: K.

Air temperature is the thermal state of the atmosphere at the specified level. It influences density, stability, and energy exchange.

• Units: K.

Total column ozone is the integrated ozone amount through the atmosphere. It affects UV radiation reaching the surface and atmospheric chemistry.

• Units: Dobson.

Soil temperature describes the thermal state of the soil layer. It affects ground heat flux and freeze/thaw processes.

• Units: K.
• Soil state affects evapotranspiration and boundary-layer moisture.

Zonal gravity wave stress is the east-west component of momentum flux from unresolved gravity waves. It affects large-scale circulation and jet dynamics.

• Units: N/m².

Momentum flux, u‑component is the east‑west surface stress exerted by the atmosphere. It drives boundary-layer dynamics.

• Units: N/m².

U‑component of wind is the east‑west component of the horizontal wind. Combine with v for speed and direction.