POSTER Session 2

A DEEP-LEARNING ASSISTED QUASI-ANALYTICAL ALGORITHM (QAA-DL) FRAMEWORK FOR THE ESTIMATIONS OF INHERENT OPTICAL PROPERTIES OVER INLAND AND NEARSHORE COASTAL WATERS

Inherent optical properties (IOPs) are crucial parameters for assessing water quality, with widely applied estimation methods established for open oceans. However, the estimation of IOPs for inland and coastal waters remains a longstanding challenge due to their complex optical properties. To address this problem, we developed a deep-learning assisted quasi-analytical algorithm (QAA-DL) framework to estimate IOPs for inland and coastal waters. This approach utilizes spectral remote sensing reflectance (Rrs) as input parameters, and optimizes the QAA process considering to use the neural network scheme for highly turbid water areas. Moreover, we incorporated the soft-wired classification scheme to ensure smooth IOPs retrievals for slightly turbid waters, as observed in coastal regions. This classification mechanism considers the characteristics of both clear and highly turbid waters, thus reducing uncertainties of IOPs estimation. Validation analysis showed that the IOPs retrievals using QAA-DL algorithm agreed well with the worldwide in situ measurements. QAA-DL outperformed the common IOPs algorithms in not only the accuracy of the IOPs retrievals, but also the valid data coverage. Application of the QAA-DL method to Moderate Resolution Imaging Spectroradiometer (MODIS) imageries, it is shown consistency of the spatial pattern for the IOPs retrievals, indicating that the robustness of the model proposed in this study. The QAA-DL algorithm could be implemented in the ocean color missions to produce high-quality IOPs retrievals for global inland and coastal waters.

Dan Zhao, Southern University of Science and Technology, China, https://orcid.org/0000-0002-6694-6945

Lian Feng, Southern University of Science and Technology, China

Code of Conduct
Privacy Policy