Analysis accuracy achievable with specified observational networks

The theory of optimum interpolation (Gandin, 1963) is the basis of a program to compute the analysis accuracy likely to be achieved with a specified observational network. Input parameters include the locations of observing points, the random observational error, the magnitude and spatial correlation of systematic error, the variance and space and time autocorrelation functions of either the element being analysed, or its deviation from a first guess field. The output includes root-mean-square (RMS) analysis error, reduction of variance over climatology or a first guess field, and the probability of meeting a specified analysis accuracy. The program is applied to a few current Australian region and tropical Australian sector upper wind networks and input parameters are systematically varied. Results suggest that, in the absence of prognosis feedback, analyses based on optimum interpolation and derived from upper winds alone will frequently fail to meet ICAO aviation route forecast criteria, and that even where existing networks are relatively dense the greatest scope for improvement lies in increasing the density rather than the network's observational frequency or accuracy.