An Observational Study of Tropical Large-Scale Fields Part II: A Modified Scheme of the Optimum Interpolation Method for Climatological Data Analysis in the Tropics

Abstract

A modified scheme of OIM - a zero-guess scheme - for direct estimation of the total value of meteorological fields at grid-points, independently from dynamical assimilation models, is developed and successfully tested over the sparse-network tropical region.
In its formulation, the random empirical error' (observational error + small scale noises) is shown to have a remarkable influence not only on other statistical parameters used in the scheme, but also on the weighting factors, the interpolation error, and therefore the grid-point values themselves. Effective stations for each grid-point are selected using a search algorithm designed to fit the spatial structure of the statistical properties of the fields. A simple and steadily computable expression of the overall error at a grid-point is given, and the positive effects of low-pass filtering the data on this error is demonstrated.
The 5-year mean seasonal statistics obtained with these grid-point data show a significant consistency with both the observation and the existing statistics for which data from various sources were used in order to supplement the sparseness of the tropical observational network.
The generalization of this new scheme in order to use data from various sources, its applicability to climatological studies in the extratropics or its possible use in NWP assimilation schemes are to be investigated, keeping in mind that realistic estimates of statistical parameters should be used for the analysis.