In operational meteorology, the term post-processing refers to one or more scientific software processes that capture the output from a numerical weather prediction (NWP) system and enhance its value in some way. In this case, the NWP model at the heart of the UFS will be NOAA’s operational Global Forecast System (GFS), based on the Finite-Volume Cubed-Sphere Dynamical Core (FV3; GFSFV3). Post-processing algorithms can be used to generate traditional meteorological variables (e.g., temperature, visibility, precipitation amount) and/or weather-dependent variables that are either not forecast or are poorly forecast directly by NWP models (e.g., road conditions, optimal evacuation path, crop disease susceptibility, renewable energy production). Often, these techniques generate or improve expressions of uncertainty (e.g., event probabilities, probability distributions). Post-processing can be said to include the following three broad areas:

  • Model Post-processing (ModPP)—A post-processing step that interprets NWP output in native model coordinates (e.g. sigma levels, spherical harmonic coefficients) and produces output in coordinates more familiar to human meteorologists (e.g., isobaric levels and regularly-spaced grids)
  • Diagnostic Post-processing (DiagPP)—A post-processing step that applies interpretive algorithms without training (e.g., the BUFKIT application, ensemble relative frequency) to NWP output
  • Statistical Post-processing (StatPP)—A post-processing step that uses statistical inference based on current NWP output, past forecasts, observations/analyses, and other data sets to create new or improved forecast quantities. Examples include Model Output Statistics (MOS) and multi-model blending such as the National Blend of Models (NBM).