Initialization of the HIRLAM Model Using a Digital Filter

View More View Less
  • 1 Department of Meteorology, Stockholm University, Stockholm, Sweden
© Get Permissions
Restricted access

Abstract

Spurious high-frequency oscillations occur in forecasts made with the primitive equations if the initial fields of mass and wind are not in an appropriate state of balance with each other. These oscillations are due to gravity-inertia waves of unrealistically large amplitude; the primary purpose of initialization is the removal or reduction of this high-frequency noise by a delicate adjustment of the analyzed data. In this paper a simple method of eliminating spurious oscillations is presented. The method uses a digital filter applied to time series of the model variables generated by short-range forward and backward integrations from the initial time.

The digital filtering technique is applied to initialize data for the High-Resolution Limited-Area Model (HIRLAM). The method is shown to have the three characteristics essential to any satisfactory initialization scheme: (i) high-frequency noise is effectively removed from the forecast; (ii) changes made to the analyzed fields are acceptably small; (iii) the forecast is not degraded by application of the initialization.

The digital filtering initialization (DFI) technique is compared to the standard nonlinear normal-mode initialization (NMI) used with the HIRLAM model. Both methods yield comparable results, though the filtering appears more effective in suppressing noise in the early forecast hours. The computation time required for initialization is about the same for DFI and NMI. The outstanding appeal of the digital filtering technique is its great simplicity in conception and application.

Abstract

Spurious high-frequency oscillations occur in forecasts made with the primitive equations if the initial fields of mass and wind are not in an appropriate state of balance with each other. These oscillations are due to gravity-inertia waves of unrealistically large amplitude; the primary purpose of initialization is the removal or reduction of this high-frequency noise by a delicate adjustment of the analyzed data. In this paper a simple method of eliminating spurious oscillations is presented. The method uses a digital filter applied to time series of the model variables generated by short-range forward and backward integrations from the initial time.

The digital filtering technique is applied to initialize data for the High-Resolution Limited-Area Model (HIRLAM). The method is shown to have the three characteristics essential to any satisfactory initialization scheme: (i) high-frequency noise is effectively removed from the forecast; (ii) changes made to the analyzed fields are acceptably small; (iii) the forecast is not degraded by application of the initialization.

The digital filtering initialization (DFI) technique is compared to the standard nonlinear normal-mode initialization (NMI) used with the HIRLAM model. Both methods yield comparable results, though the filtering appears more effective in suppressing noise in the early forecast hours. The computation time required for initialization is about the same for DFI and NMI. The outstanding appeal of the digital filtering technique is its great simplicity in conception and application.

Save