Estimating spatial correlations from spatial-temporal meteorological data

Abstract

In this paper, it is shown that temporal correlations can seriously bias fitted covariance functions that are used in optimal interpolation and optimal spatial-averaging methods. A result of this bias is that the fitting of correlation functions to temporal correlations can result in overestimates of spatial correlations. In contrast, the fitting of structure function models from data for fixed time periods does not suffer from the temporal biases of correlation function fitting and should be the preferred method for estimating spatial correlations. Structure function modeling must, however, accommodate anisotropy in the data. -from Author