Effects of the lack of stationarity on deep water wave statistics

A year-long, near-continuous record of wave heights from offshore of Southern California was examined to determine the effects of lack of stationarity on aspects of two important statistics, the significant wave height and the centroidal frequency. A nonparametric runs test was used to determine the distribution of stationary intervals of significant wave height. The mode of this distribution is at about 5-6 hour periods for this data set. The sampling variability of the height and frequency parameters was calculated as a function of sample size, using a nominal six hour interval of data to represent the true process. The resulting statistics were compared with a theoretical prediction for a stationary process, and the observed sampling variability was found not to decrease as rapidly with increasing sample size as the theory predicted. Doubling the typical sampling time of 17 min produced only modest improvement in estimates of height and period.