Is power averaging the best estimator for undersea acoustic data?

Abstract

The power average is the statistic which is commonly used in all types of acoustic data processing. It is an excellent estimator for "well behaved" Gaussian distributions when the central or average value is desired. However, when there are "outliers" and errors or when the distributions are not Gaussian there are other statistics which are better estimators. In the case of ambient noise, the distributions are seldom Gaussian or free from "outliers" and errors. For this case the median and the geometric mean power level (dB average) are less sensitive to interference in the environment and to some of the minor system faults and are more powerful statistics.