Lee, E.-J.; Chae, J.-Y., and Park, J.-H., 2020. Reconstruction of sea level data around the Korean coast using Artificial Neural Network methods. In: Malvárez, G. and Navas, F. (eds.), Global Coastal Issues of 2020. Journal of Coastal Research, Special Issue No. 95, pp. 1172–1176. Coconut Creek (Florida), ISSN 0749-0208.

The coastal sea level is an important element for better understanding and clarifying of oceanic physical phenomena in the coastal regions. However, interruptions of sea level measurements happen quite often, and hence it is crucial to find out an optimum method to fill in data gaps to retrieve continuous sea level time series. Traditionally, single-variable methodologies such as spline interpolation and ARIMA model have been utilized to fill the data gap. Sea level time series can be divided into regular astronomical tides and its residual, meteorological tides. The former can be predicted precisely, while the latter needs a specific methodology to reconstruct. In this study, we examine the feasibility of artificial neural network methods for reconstruction of the sea level time series through Long-Short Term Memory (LSTM) networks, one of Recurrent Neural Networks suitable for application to sequential data. Input data is composed of oceanic, atmospheric, and processed variables in consideration of air-sea interactions. Optimization is applied to five coastal tide station data to evaluate the effects of sea level change-related variables. Our artificial neural network methods show a remarkable performance in reconstructing continuous sea level time series with an error less than 8 cm and the Pearson correlation coefficient over 0.85, implying that 85% of sea level can be explained by oceanic and atmospheric variables.

This content is only available as a PDF.