Abstract
Results of a thorough study of the correlation structure of observation-minus-forecast increments for mandatory pressure level radiosonde observations of zonal and meridional wind components and geopotential, differenced with NMC's 6-hour global forecasts, are reported. Our work focused on the selection of a representation for spatial lag-correlations to be used in updating the multivariate statistical objective analysis algorithm of the global data assimilation system, with attention given to regional and seasonal dependence of the correlation structure, and on the degree to which the increments are in the same geostrophic balance as the signal and forecast fields individually.
We compare the performance of several candidates for representing autocorrelations of geopotential increments, on the one hand, and the auto- and cross-correlations of the wind component increments, on the other, for five mandatory pressure levels, for four regions of the Northern Hemisphere and for the Southern Hemisphere. A third-order auto-regressive correlation function is identified as optimal for NMC's global objective analysis. The parameters are shown to vary with level and season. Furthermore, the geopotential and wind correlation fits have identified important differences in corresponding parameter values. A single algorithm which covers the primary candidates in one fitting operation, for future semi-automatic updating, has been developed in the course of this work. Results of its use are presented and discussed.
