Multivariate Analysis of Temperatures and Winds Using Optimum Interpolation

Abstract The design of a statistical “optimum interpolation” analysis system for multivariate analysis of temperature and wind fields is described. The scheme uses three-dimensional correlation functions, defined as products of quasi-horizontal and vertical correlations. A numerical prediction is used to provide background fields, and corrections to them are obtained using optimum interpolation. Observations are assigned rms error levels, and for some observational types the errors are assumed to be vertically or laterally correlated. A procedure for using oceanic surface data in the upper air analysis is included. Some special design features, including data selection and error-checking procedures, are discussed. The mechanics of the analysis system are illustrated with a step-by-step example analysis. Several experimental analyses are compared in order to illustrate sensitivity of the analysis scheme to changes in design features and governing parameters.