Three-Dimensional Applications of a Successive Corrections Analysis Scheme to Mesoscale Observations

Abstract

Bratseth?s scheme of objective analysis is applied to a mesoscale resolution dataset of dropsonde observations collected in the course of the Fronts ?87 experiment. The scheme is a version of the successive corrections method that converges, in the limit of the iteration cycle, to the statistical interpolation analysis. Following earlier applications on two-dimensional synoptic fields its performance is assessed here in a three-dimensional context. The convergence properties of the scheme are examined and its sensitivity to the prescribed values of the influence scale L and of the smoothing parameter ?0 is tested. A modified version of the scheme, based on Pedder?s analytical formulation of the weight function, is subsequently applied to the data. In the modified version the weights are defined as convolutions of a background error correlation function and a continuous linear filter, so as to express the effect of filtering on the analyzed rather than the observed field. Results are compared with those obtained through conventional filtering in observation space. It is suggested that the improved scale selectivity of postanalysis filtering and the relative insensitivity of Bratseth?s method to ?0 make the combined scheme suitable for the analysis of mesoscale datasets, where background and observing error correlation statistics are often unavailable or unreliable. Finally, based on an investigation of the sensitivity of the temperature, humidity, and wind analyses to variations in the value of the filter scale Lg, it is argued that postanalysis filtering can be used to complement spectral studies of mesoscale data.