Modeling Nearshore Bed Topography with Principal Oscillation Patterns

A nontidal, natural, mildly sloping, multiple, mostly four bar nearshore zone at the Coastal Research Station Lubiatowo, Poland is analyzed. It is made of fine sand with the mean grain size $D_{50}=0.22\mathrm{mm}\text{,}$ which is typical of the south Baltic coast. The bars oscillate about their average locations offering good prototype conditions for testing a method that could reproduce and predict this type of bed dynamics. The dataset represents four adjacent, shore normal profiles, spaced every 100 m and sampled every 10 m between 100 and 900 m offshore of the geodetic base once a year between 1987 and 1998. In all, 324 grid points, each with 12 realizations were incorporated, allowing use of the data intensive principal oscillation pattern (POP) technique to come up with a data-driven model of bed dynamics. It revealed three patterns that reasonably reproduced long-term bed dynamics, and the modeling results are fairly accurate at the locations of bars 2, 3, and 4, where bed evolution is slow enough to be grasped by annual records. Developments around innermost bar 1 turned out to be too fast for annual measurements to retain sufficient information and achieve satisfactory results with the POP method. The POP model outperformed dominant empirical orthogonal functions (EOF) modes by 6% in terms of the explained variance, but the EOF method is highly recommended prior to the POP analysis, because it can identify clear-cut orthogonal patterns; whereas, the POP method produces interdependent patterns whose interpretation is difficult without the EOF modes.