A new implementation of the adaptive mesh transform in the Met Office 3D-Var System

This paper looks at the application of two new features of the adaptive mesh transform in the Met Office variational data assimilation which modifies the background error correlations to better represent the boundary layer structure. The new features are an improved normalization procedure and the recalculation of the adaptive mesh within the nonlinear analysis procedure. Relative to the initial implementation the results show improvements in the forecast skill of the UK4 model for both summer and winter and a reduction of the analysis root mean square errors with respect to radiosonde, aircraft, SEVIRI and ground GPS observations for both periods. Forecast errors at T + 2, 3, 4 h are also reduced with respect to aircraft, surface, SEVIRI, ground GPS observations and with respect to radiosonde in the boundary layer for relative humidity and for potential temperature around temperature inversions. These improvements are attributed to the variation of the background vertical correlations and the implied background error variances in the assimilation due to the greater movement of the mesh away from the lower boundary, in particular in areas where temperature inversions or stratocumulus layers are present. Copyright (c) 2012 British Crown copyright, the Met Office. Published by JohnWiley & Sons Ltd.