Some Experiments with a Multivariate Statistical Objective Analysis Scheme

Abstract

A statistical scheme for simultaneous analysis of the wind and geopotential height fields has been developed based upon optimum interpolation. The matrix weights applied to the observation vectors depend upon covariances of observed-minus-forecast differences. The simplest possible forecasts (or first guesses)?climatology, persistence and damped persistence?are used. The geostrophic relationship and the height-height covariances computed from historical data are used to derive the other required covariances. The scheme has been tested at a single point based upon 500-mb winter U.S. radiosonde data. Results are promising; when either climatology or damped persistence is used as a first guess, the root-men-square (rms) differences between analyzed and observed values were about 13 m for the height and 4.0 m s?1 for the u-and v?components of the wind. When persistence is used as a fist guess, results are slightly worse. The multivariate approach is clearly superior to a univariate approach for height analyses. On the other hand, geopotential height data do not significantly improve wind analyses. The applicability of the scheme to objective analysis over large areas is briefly mentioned.