Tests of distribution-free non-Gaussian noise removal in spectral analysis of meteorological radar echo

The authors' previous paper (1992) presented a distribution-free technique for removing the noise level when performing spectral analysis of meteorological echo. The method applies the Kolmogorov-Smirnov (K-S) test to the periodogram obtained from the discrete Fourier transform, (DFT) of a sequence of pulse-to-pulse complex envelopes. The K-S test gives results that are independent of the noise distribution and depends only upon the pulse-to-pulse noise samples being uncorrelated, a condition almost always satisfied in radar. The cited paper tested the technique against only Gaussian noise. The present paper reports the test of the technique against severely non-Gaussian noise, viz., lognormal noise. The complex envelope of lognormal noise is characterized by an amplitude with a lognormal distribution and a uniformly distributed phase. In the authors' test, lognormal noise was generated in a fairly straightforward way to yield a specified power, so that the K-S technique could be applied to the same signal-to-noise ratios as in the previous paper. The new results show that the technique is truly distribution free. The authors also made comparisons with the pulse-pair method and the Hildebrand-Sekhon method.< >