An investigation of the multi-scale temporal variability of beach profiles at Duck using wavelet packet transforms

In an earlier paper a particular discrete wavelet transform (DWT) was used to study the complex variation of beach profile changes. However, use of the DWT requires that the sequence of spatial and temporal resolution is fixed as a dyadic sequence, which means that the variability over longer intervals is not characterised well. Here we introduce the discrete wavelet packet transform (DWPT) that uses an adaptive scaling to partition the data variance, according to an entropy cost function. The advantages of this approach are demonstrated by its application to the study of temporal variability of a 22 year record of beach profile data from the Field Research Facility (FRF) at Duck, North Carolina, USA. Time series of beach elevations at three locations across a particular profile are investigated in detail. We conclude that the DWPT provides a superior analysis of non-stationary time series to that of the DWT, with improved resolution of the scale intervals of the variability. The beach elevation around the shoreline is shown to respond at both sub-annual and interannual scales, but variability at an annual scale is weak. Moving seaward into deeper water, the variance is partitioned into fewer and longer scales. It is confirmed that elevation changes around the inner bar at Duck exhibit a strong interannual variation consistent with Plant et al. (Plant, N.G., Holman, R.A. and Freilich, M.H., 1999. A simple model for interannual sandbar behaviour. Journal of Geophysical Research 104(C7), 15755–15776). Around 23% of the variance around the inner bar is explained at the temporal scale of 64–128 months, which is consistent with the bar behaviour of 6 years found by Ruessink et al. (Ruessink, B. G., Wijnberg, K. M., Holman, R. A., Kuriyama, Y. and Van Enckevort, I. M. J., 2003. Intersite comparison of interannual nearshore bar behaviour. Journal of Geophysical Research, 108 (C8): 1–12). A significant new finding is, however, that about 26% of the variance is attributable to temporal scales of 16–21.3 months. Reconstruction of the wavelet packet components for individual temporal scales is shown to provide a means for identifying the impact and scale of non-stationary events, such as storms, on the beach response. This provides further information that can be used to interpret the morphological changes in terms of the forcing processes and also serves to inform morphodynamic modelling.