Evaluation of a Scheme for Improving the Acceptance of Low-Latitude Data in Atmospheric Data Assimilation

Abstract

A scheme of iterated multivariate optimum interpolation and nonlinear normal-mode initialization is applied to a shallow-water model in an attempt to reduce the rejection of low-latitude height and divergent wind data. The amount of rejection and the performance of the iterative scheme are evaluated as a function of the adjustable analysis parameters in the Tropics. Results typically indicate a modest convergence of the scheme, with analysis- and initialization-produced changes to height and velocity potential showing some decrease. Allowing the normalized observation error to vary as a function of iteration significantly accelerates this decrease because the increasing ratio of observation error to background error results in smaller analysis increments. In all cases, however, the final initialized state is not greatly different from the first guess in the Tropics. The behavior of the scheme is relatively insensitive, at least qualitatively, to the value of any of the analysis parameters. Increasing (decreasing) the value of the geostrophic (divergent) parameter yields better-balanced analyses, meaning analyses with less rejection, as well as improved convergence of the iterative scheme. Varying other analysis parameters, as well as including topography and diffusion, has little qualitative impact. One finding of this study is that retaining some mass?wind coupling in the Tropics may be advantageous, since it tends to produce slightly improved balance in the analysis. The main conclusion, however, is that the combined analysis?initialization scheme appears only marginally useful at low latitudes, at least for current operational prediction systems and data sources.