Oral Session 2

Active and passive optical remote sensing data fusion and ocean optical parameter retrieval method based on deep learning

Passive ocean optical remote sensing has yielded extensive, long-term observations covering a broad spatial range, capturing global ocean primary productivity trends and changes in marine ecological environments. However, its capability is confined to observing the integrated volume of the ocean surface layer. Conversely, active optical remote sensing, such as space-borne Lidar, can provide profile information on the optical properties of the ocean. However, it encounters two persistent challenges: Firstly, the Lidar signal is influenced by the Lidar system response function, impeding direct reflection of water body optical property profiles. Secondly, the ocean Lidar equation involves two unknowns: the 180° backscattering coefficient β(π) and the Lidar attenuation coefficient Klidar. To address both of these challenges, this study integrates MODIS, ICESat-2, and BGC-Argo data to develop a deep learning model for retrieving the profiles of water optical properties. This deep learning model employs a Generative Adversarial Network (GAN) architecture, based on Bayesian estimation, to extract attenuation characteristics of laser photons emitted by Lidar in seawater. The final output of the model is the 490 nm downward diffusion attenuation coefficient Kd(490) and the 532 nm backscattering coefficient bbp(532), verified against independent in-situ profile observation data from BGC-Argo. The validation results show that the profiles of Kd(490) and bbp(532) estimated by our model are consistent with the in-situ profile data measured by BGC-Argo. The application of this model can provide a novel approach for fusing active and passive ocean optical remote sensing data.

Xiaolong Li, Institute of Oceanology, Chinese Academy of Sciences, China

Xiaofeng Li, Institute of Oceanology, Chinese Academy of Sciences, China

Yi Yang, Institute of Oceanology, Chinese Academy of Sciences, China

Code of Conduct
Privacy Policy