Using Signal Coherency to Improve Detection on Weather Radars

Typically, signal detection in operational weather radars is performed using thresholds on estimated signal-to-noise ratio and/or magnitude of the autocorrelation coefficient. In this paper, a novel approach to signal detection that combines the sums of the estimated power and autocorrelations is developed. The hypothesis that “signal is present” is accepted if the sum exceeds a predetermined threshold. Otherwise, data are considered ”non significant” and are not displayed or used as inputs for algorithms. The threshold for the coherency sum is determined based on a given maximum acceptable rate of false detections. The detection scheme is evaluated both in simulations and through implementation on time-series data collected with two research weather surveillance radars in Norman, Oklahoma. The comparison to the traditional techniques shows that the coherency-based method leads to increased detection rates in the areas of weak reflectivity.