FIELDS BY CORRELATION ASSEMBLY—A NUMERICAL ANALYSIS TECHNIQUE

Abstract

This paper describes a numerical technique to analyze meteorological or oceanographic variables. Associated distributions of reliability are also produced. As used here, reliability is defined as 1/(2σ2), where σ is the standard error. The methods have been adapted to the analysis of 500-mb. height, sea level pressure, and sea surface temperature. Sample case studies using the first two versions are presented. As applied to 500-mb. height, for example, the technique in its current version blends information on Z and ?Z (from winds). Information on ?2Z (e.g., from satellite data) could also be included. All data are combined in proportion to their reliabilities. Wind components are first geostrophically converted into height differences between adjacent grid points. These estimates are then assembled with the first-guess gradients over a 5?5 set of grid points (omitting the corner ones) surrounding the wind report. Height estimates are next extrapolated to their nearest grid points using the analyzed gradient fields. These modified estimates are then assembled with the first-guess heights. The assembled height and gradient fields are then blended to form the final height and reliability fields. Provision is also made to check height and wind reports for detectable gross errors. In the sea level pressure version, ship winds are used but not those from land stations. In the analysis of sea surface temperature, the only information on gradient is that contained in the first-guess field. Otherwise, the procedures for these two versions are similar to those employed to analyze 500-mb. height.