Methods for automatized detection of rapid changes in lateral boundary condition fields for NWP limited area models

摘要

Three-hourly temporal resolution of lateral boundary data for limited areamodels (LAMs) can be too infrequent to resolve rapidly moving storms. Thisproblem is expected to be worse with increasing horizontal resolution. Inorder to detect intensive disturbances in surface pressure moving rapidlythrough the model domain, a filtered surface pressure field (MCUF) iscomputed operationally in the ARPEGE global model of Météo France.The field is distributed in the coupling files along with conventionalmeteorological fields used for lateral boundary conditions (LBCs) for theoperational forecast using limited area model ALADIN (Aire LimitéeAdaptation dynamique Développement InterNational) in the Meteorologicaland Hydrological Service of Croatia (DHMZ). Here an analysis is performed ofthe MCUF field for the LACE coupling domain for the period from23 January 2006, when it became available, until 15 November 2014. The MCUFfield is a good indicator of rapidly moving pressure disturbances (RMPDs).Its spatial and temporal distribution can be associated with the usualcyclone tracks and areas known to be supporting cyclogenesis. An alternativeset of coupling files from the IFS operational run in the European Centre forMedium-Range Weather Forecasts (ECMWF) is also available operationally inDHMZ with 3-hourly temporal resolution, but the MCUF field is not available.Here, several methods are tested that detect RMPDs in surface pressure aposteriori from the IFS model fields provided in the coupling files. MCUF iscomputed by running ALADIN on the coupling files from IFS. The error functionis computed using one-time-step integration of ALADIN on the coupling fileswithout initialization, initialized with digital filter initialization (DFI)or scale-selective DFI (SSDFI). Finally, the amplitude of changes in the meansea level pressure is computed from the fields in the coupling files. Theresults are compared to the MCUF field of ARPEGE and the results of samemethods applied to the coupling files from ARPEGE. Most methods give a signalfor the RMPDs, but DFI reduces the storms too much to be detected. The errorfunctions without filtering and amplitude have more noise, but the signal ofa RMPD is also stronger. The methods are tested for NWP LAM ALADIN, but couldbe applied to other LAMs and benefit the performance of climate LAMs.