Review on Hyperspectral Remote Sensing of Tidal Zones

Hyperspectral data, known for providing detailed spectral information on targets, is recognized as a valuable remote sensing dataset for investigating natural processes and statuses of both terrestrial and oceanic ecosystems. Tidal zones, situated between land and ocean, exhibit characteristic features where hyperspectral data are uniquely useful for remotely identifying the bio-geochemical variables related to the processes in tidal zones. This article reviews studies on tidal zones that employed hyperspectral data for retrieving the variables on (i) physical properties of flats (texture and soil moisture content), (ii) abundance of primary producers on the surface (algae and vegetation), and (iii) water area properties (turbidity and bathymetry). The primary focus of this review is on the diversity of algorithms employed within each application field, rather than offering an exhaustive list of previous studies. The algorithms discussed range from traditional methods that exploit the geometric shape of the reflectance spectrum to statistical approaches, such as variations of regression, and more advanced machine learning techniques. In addition, relatively recent physics-based approaches involving radiative transfer simulation have been introduced for the relevant applications. While a direct comparison of performance results across different studies is often unfeasible due to variations in environments, sensors, platforms, and target species, we summarize the estimation results, along with the metrics used, to provide initial insights into the use of hyperspectral data for specific applications. Along with the algorithmic focus, this article identified critical wavelength ranges for each application based on the surveyed literature and our field measurement data, as feature selection and dimension reduction are often preferred prior to the analysis to manage the high data complexity and avoid the curse of dimensionality in hyperspectral data.