TITLE:
Optical Classification of Water Types in Cook Inlet, Alaska
AUTHORS:
Hisatomo Waga, Mark A. Johnson
KEYWORDS:
Ocean Color Remote Sensing, Optical Classification, Machine Learning
JOURNAL NAME:
Advances in Remote Sensing,
Vol.14 No.3,
September
19,
2025
ABSTRACT: Ocean color is determined by the complex interactions of incident light with the optical properties of suspended and dissolved substances. Such interactions give water its characteristic color and reflect information about ocean constituents, which can be captured by satellite-borne ocean color sensors in orbit hundreds of kilometers above the Earth’s surface. Here, based on the spectral shape of remote sensing reflectance that is now readily available from diverse satellite ocean color sensors, the present study proposes a satellite approach for mapping surface water types in Cook Inlet, Alaska. The Iterative Self-Organizing Data Analysis Technique clustering identified 15 different water types in Cook Inlet, and then machine learning models were trained to accelerate the processing of high-resolution satellite images comprising thousands to millions of pixels each. A total of 31 classification algorithms were tested, and a neural network with medium preset showed the best performance with an accuracy of 99.9%. The main advantage of optical classification is its capability to identify water types with similar temperature or salinity properties, since the optical signatures are independent of such physical characteristics. Overall, the present study highlights that a combined approach with optical classification and the conventional temperature-salinity diagram improves our capability to differentiate water types, contributing to better monitoring of ocean dynamics that advances our baseline understanding of marine environments.