DATA-DRIVEN ANALYSIS AND MODELING OF SHORELINE EVOLUTION TRENDS

摘要

Monthly records of shoreline positions, taken between Sep. 1983 and Sep. 1999 at IBW PAN Coastal Research Station at Lubiatowo, Poland, were examined with advanced statistical methods - Singular Spectrum Analysis (SSA) and Principal Oscillation Patterns (POP). The SSA allowed the extraction of shoreline change patterns. The most significant one reflects long-term trends of shoreline change, having a clearly visible periodical structure with cycles ranging between 8-10, through 16, up to 25-30yrs. Detailed periodic structure of these trends was examined with an FFT routine. The POP was applied to them and found that trend variations can be understood as sum of two cyclic patterns with periods very similar to those found with the SSA (10 and 18yrs) and a pattern reflecting simple exponential decay of a given shoreline configuration. These patterns may serve as a data-driven model for prediction of future trends of shoreline evolution.