POSTER Session 4

ONE-STEP RETRIEVAL OF WATER PARAMETERS FROM SATELLITE TOP-OF-ATMOSPHERE MEASUREMENTS

This study proposes a novel scheme to generate water color parameters based on digital numbers (DN) obtained at the satellite altitude. Traditionally the derivation of water quality parameters from satellite ocean color measurements takes complex multiple steps, where in addition to sensor’s radiometric calibration, algorithms for atmospheric correction and water properties retrieval have to be developed, all take large amounts of resources and efforts. Given the availability of numerous water-quality related products from mature ocean color satellite missions, a Combined Deep Learning Model (CDLM) is developed, which is capable of directly extracting water quality parameters from satellite’s top-of-atmosphere DN data. Taken products from the Sentinel-3 (reference) and HiSea-II (target) as examples, it is observed that the determination coefficient (R²) for water clarity (Zsd) exceeds 0.9, with a Mean Absolute Percentage Difference (MAPD) of approximately 10% for the training dataset. Upon application to a new dataset, the R² for Zsd exceeds 0.8, and the MAPD remains within 20%, indicating the model’s predictive accuracy and generalization capabilities. This approach has been successfully extended to multiple satellites, including HY1C/1D, GF1, and HJ2A/2B, with Zsd measurements showing good consistency across different satellites. Moreover, matchup Zsd data between satellite and insitu measurements revealed MAPD of 19.6% and Mean Absolute Error (MAE) of 0.23 m. These results indicate that with today’s wide range of data sources, it is feasible to obtain reliable water-quality parameters from satellite’s top-of-atmosphere DN data, which greatly simplifies the processing procedure for many satellites.

*Hanyang Qiao, Xiamen University (XMU), China, [email protected], https://orcid.org/0009-0004-7458-3226

Zhongping Lee, Xiamen University (XMU), China, [email protected]

Code of Conduct
Privacy Policy