Premium
Improving initial condition perturbations for MOGREPS‐UK
Author(s) -
Tennant Warren
Publication year - 2015
Publication title -
quarterly journal of the royal meteorological society
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.744
H-Index - 143
eISSN - 1477-870X
pISSN - 0035-9009
DOI - 10.1002/qj.2524
Subject(s) - downscaling , data assimilation , ensemble forecasting , scale (ratio) , meteorology , computer science , convection , visibility , environmental science , artificial intelligence , geography , precipitation , cartography
Three different methods to initialise the Met Office limited‐area convection‐permitting ensemble prediction system (MOGREPS‐UK) are investigated. The first method, which forms the baseline for this study, is the current operational downscaling method where each 2.2 km high‐resolution ensemble member is nested within a corresponding 33 km resolution global ensemble member out to 36 h lead‐time. The second method is centring large‐scale perturbations, which are interpolated from the global ensemble, on an analysis interpolated from the operational UKV model, which is a 1.5 km high‐resolution convective‐scale model running a 3‐hourly data assimilation cycle. The third method is a perpetual cycling run where the 6 h forecast of each high‐resolution ensemble member is used to initialise the subsequent cycle, with only the lateral boundary conditions being updated to the new driving global ensemble model forecast. Each of the three methods have their own advantages, but the method that centres large‐scale perturbations on a high‐resolution analysis performs best as measured by standard objective verification measures and is shown to perform well in three case‐studies. Results from the perpetual cycling trial show that some forecast parameters, such as visibility, would appear to benefit from separate high‐resolution initial perturbations.