Three-dimensional variational data assimilation for a limited area model Part II: Observation handling and assimilation experiments

A 3-dimensional variational data assimilation (3D-Var) scheme for the HIgh Resolution Limited Area Model (HIRLAM) forecasting system is described. The HIRLAM 3D-Var is based on the minimisation of a cost function that consist s of one term, J(b), which measures the distance between the resulting anal ysis and a background field, in general a short-range forecast, and another term. J(o), which measures the distance between the analysis and the obser vations. This paper is concerned with J(o) and the handling of observations , while the companion Paper by Gustafsson et al. (2001) is concerned with t he general 3D-Var formulation and with the J(b) term. Individual system com ponents. such as the screening of observations and the observation operator s, and other issues, such as the parallelisation strategy for the computer code, are described. The functionality of the observation quality control i s investigated and the 3D-Var system is validated through data assimilation and forecast experiments. Results from assimilation and forecast experimen ts indicate that the 3D-Var assimilation system performs significantly bett er than two currently used HIRLAM systems. which are based on statistical i nterpolation. The use of all significant level data from multilevel observa tion reports is shown to be one factor contributing to the superiority of t he 3D-Var system. Other contributing factors are most probably the formulat ion of the analysis as a single global problem, the use of non-separable st ructure functions and the variational quality control, which accounts for n on-Gaussian observation errors.