Analysis of field data of coastal morphological evolution over yearly and decadal time scales. Part 2 Non-linear techniques

TLDR

Non‑linear time‑series analysis techniques have recently been developed and applied across environmental sciences, and their relevance to coastal morphological data—distinguishing linear from non‑linear approaches—has prompted this review. This paper reviews and evaluates the usefulness of these techniques for coastal morphological data, illustrating their application in coastal morphology and related fields. The review covers time‑delay embedding, singular spectrum analysis, forecasting signatures, fractal analysis, and neural networks, and notes that remote‑sensing data could satisfy the longer, higher‑resolution series required. The authors conclude that non‑linear techniques are essential for understanding and forecasting coastal morphology on yearly and decadal scales, yet they typically demand longer, higher‑resolution data than are currently available.

Abstract

A number of techniques for non-linear analysis of time series data have been developed in recent years and applied in many environmental sciences. In this paper, some of these techniques are reviewed and their usefulness for coastal morphological data assessed, with examples in coastal morphology and related fields where these are available. The methods reviewed are time-delay embedding techniques, singular spectrum analysis, forecasting signatures, fractal analysis and neural networks. It is expected that readers from diverse backgrounds such as statistics, environmental modeling, and data measurement would be interested in this subject. Accordingly, introductions have been provided for some concepts and background material. These include the general purposes of data analysis, the approaches traditionally used by statisticians and physical scientists, the general nature of coastal morphological data, and the distinction between linear and non-linear analysis methods. It is concluded that the information potentially provided by these techniques is essential in understanding the generic behavior of coastal morphology on yearly and decadal timescales, and hence in constructing models and making forecasts of future morphological evolution. However, for these purposes, non-linear data analysis techniques usually require longer data series, and higher spatial and temporal resolution, than is available in present coastal morphological data sets. Data from remote sensing sources have the potential to meet these requirements.