Numerical analyses in the Australian region are heavily dependent upon the interpretation of cloud imagery over oceanic areas. Also, particularly at 1200 GMT, observations of wind outnumber those of geopotential. Recurrent analysis deficiencies have arisen from an unsatisfactory mix of conventional and interpretatively based input (the 'data-no data' problem), and the failure of geopotential analyses based primarily on winds to adequately reflect the observed winds (the 'wind only' problem). The frequency and severity of both these problems have been significantly reduced through application of the variational 'field information blending' technique (Holl and Mendenhall 1971). The first deficiency is ameliorated by placing a greater reliance on the gradients and Laplacian rather than on the absolute value of field variables dependent on interpretative input; the second by performing separate scalar analyses of thicknesses and wind components, and subsequently blending these analyses to produce fields of goepotential, stream function and velocity potential. A similar technique may be useful for exploiting the information in future observational systems which define gradients more reliably than absolute values.
Seaman et al. (Tue,) studied this question.