Sampling and Aliasing During the NORPAX Hawaii-to-Tahiti Shuttle Experiment

We examine leakage and aliasing problems in a seasonal multiple linear regression model of data taken during the NORPAX Hawaii-to-Tahiti Shuttle Experiment. The model includes a mean, trend, annual harmonic, and semiannual harmonic. Inclusion of the trend can exacerbate leakage, particularly for record lengths of 13 months or less. This is not a major problem for the 15-month shuttle record. The irregularity of the shuttle shipboard time series reduces the potential aliasing of single unresolved high frequencies. Confidence limits for regression model coefficients are determined empirically from Monte Carlo subsampling experiments on moored current meter and sea level data taken during the shuttle. These limits are generally narrower than the standard regression error estimates because there is lack of fit in the regression model. Confidence limits used here pertain only to the determination of model coefficients for a given data set; they say nothing about whether the annual cycle during the shuttle period is typical of any other time interval. The shuttle sampling at each of three longitudes (150 ø , 153 ø , and 158øW) was in general not adequate to reveal significant differences in low-frequency signals from one longitude to another. If data from the three longitudes are analyzed together without regard for longitude, the multiple linear regression parameters for temperature and zonal velocity component are fairly well determined. The meridional velocity component is so dominated by higher-frequency variability that the model parameters do not differ significantly from zero.