Isolating time scales of drivers and shoreline response to predict shoreline changes.

Predicting shoreline evolution is one of the key challenges in coastal studies with far-reaching economic and societal implications. Changes in the shoreline position occur over different temporal scales (from seconds to decades and longer) which overlap and interact. External drivers of shoreline change, for example wave characteristics or Sea Level Pressure Fields (SLPF), can also contain oscillations at different temporal scales which may be reflected in shoreline evolution.

Here we present a new methodology to predict shoreline position based on Complete Ensemble Empirical Mode Decomposition (CEEMD). CEEMD decomposes a time series into a finite set of “intrinsic mode functions” (IMFs) characterized by different temporal scales. This method overcomes limitations of Fourier-based methods for time series analysis (e.g. FFT and wavelet techniques) since the method was designed to analyse non-linear and non-stationary phenomena, identifying patterns hidden in the data (Huang et al., 1998; Torres, 2011). The different temporal oscillations (IMFs) found in the drivers (SLPF and/or waves) are linked to shoreline changes at distinctive temporal scales (e.g. annual variability in the shoreline position can be predicted using annual variability in the drivers). The overall shoreline position is computed as the sum of the change at each individual scale (e.g., monthly, annual, bi-annual and trends). We applied the new methodology to some of the longest datasets of daily shoreline evolution, currently available, Tairua beach, New Zealand (18 years) and Narrabeen, Australia (11 years). Shoreline prediction through the new methodology displayed improvements compared to traditional equilibrium models like Yates et al. (2009) and ShoreFor (Davidson et al., 2013) showing that shoreline response can be decomposed and predicted using an approach that isolates the main temporal scales.