The spatial variation of geopotential and temperature on constant pressure surfaces in the Australian region has been described by fitting analytic functions to observed correlation coefficients. These coefficients were computed for 296 station pairs at six levels in summer and winter for the period 1962 to 1973. Similar computations, based upon a subset of this data base, were performed on correlation coefficients of 24-hour changes. The dependence of the two-point correlation coefficient upon the separation of the points, and upon the orientation of the line joining them, is shown to vary systematically between levels, between seasons, and between tropics and mid-latitudes. Characteristic differences in anisotropy are apparent between Australia, and Europe and North America. It is proposed that correlation coefficient functions specific to particular geographical regions provide an appropriate measure of spatial variability, against which to verify dynamical climate models which aim to simulate regional characteristics. It is also suggested that in the Australian region, both the anisotropy of the coefficients and the contrasts between tropics and mid-latitudes are sufficiently systematic to be taken into account in objective analysis.
R. S. Seaman (Mon,) studied this question.