Numerical models represent the heart of what is known as Numerical Weather Prediction - NWP. In recent years, NWP models have reached a constantly growing level of reliability becoming widespread tools for the forecast activity of operational meteorological centers. The forecast atmospheric fields (pressure, temperature, wind, humidity, etc.) are obtained by the integration of a set of equations describing the atmospheric physical processes and how meteorological variables change with time. Numerical methods are used to find numerical approximations to the solutions of the prognostic equations of NWP models. Grid points models use a discrete representation of the atmospheric variables on a regular spatial grid. Numerical weather prediction is a very long and complex process, that includes four fundamental phases:
  1. data collection;
  2. analysis;
  3. forecast;
  4. verification.
Operational centers, such as the Operational Center for METeorology (COMET), are able to provide information with a detail of few kilometers using NWP models of such complexity to require super calculators. The Italian Air Force Met. Service is provided with a "HP cluster linux" super calculator with performances of the order of 200 teraflops (floating point operations).



The operational short-range numerical forecasting system is composed of :
  • Ensemble Kalman Filter (EnKF) Data Assimilation System based on the COMET-LETKF algorithm and the High-resolution non-hydrostatic model COSMO integrated over the Mediterranean-European region (7km, 49 vertical levels);
  • High-resolution non-hydrostatic model COSMO-ME integrated over the Mediterranean-European region (5km, 45 vertical levels);
  • Very high-resolution non-hydrostatic model COSMO-IT integrated over the Italian region (2.2km, 65 vertical levels).
Along with the standard LETKF analysis, a deterministic analysis is computed using the standard Kalman gain and a control forecast, instead of the background ensemble mean. COSMO-ME is initialised by this deterministic analysis and driven by IFS-ECMWF boundary conditions. COSMO-IT is nested into COSMO-ME and initialised by an observation nudging cycle.
Along with the standard LETKF analysis, a deterministic analysis is computed using the standard Kalman gain and a control forecast, instead of the background ensemble mean. COSMO-ME is initialised by the deterministic LETKF analysis and driven by IFS-ECMWF boundary conditions. COSMO-IT is driven by IFS-ECMWF boundary conditions and initialised by the high-resolution deterministic LETKF analysis.