Development of a remote sensing algorithm to retrieve chromophoric dissolved organic matter in the Pearl River Estuary

Chromophoric dissolved organic matter (CDOM) plays crucial roles in the estuarine ecosystem and biogeochemical processes. However, existing algorithms for CDOM in the Pearl River estuary (PRE) require further optimization grounded on more observations. In this study, we developed a satellite remote sensing algorithm to retrieve CDOM in the PRE using a semi-analytical approach combined with mixed spectral decomposition. The algorithm was developed and validated using field data collected from 14 cruises conducted between 2003 and 2022. The power-law exponent for the particulate backscattering coefficient (bbp) was initially determined using neural networks and bio-optical model approximation in the near-infrared band. After refining the expressions of inherent optical properties (IOP), the IOP curves were normalized by setting the CDOM absorption coefficient at 443 nm ag(443), particulate backscattering coefficient at 443 nm bbp(443) and chlorophyll concentration to 1. Based on the normalized IOPs and the remote sensing reflectance, mixed spectral decomposition was employed to obtain the scaling factor of each normalized IOP, which served as the retrieval results of the model. The algorithm was implemented on Ocean and Land Colour Instrument (OLCI) data to estimate the ag(443). The proposed method exhibits good consistency between in situ measurements and retrieved ag(443), with R2 = 0.5776, RMSE = 0.2449 and MSE = 0.0600 for the 153 training samples, and R2 = 0.5765, RMSE = 0.1043, and MSE = 0.0109 for the 49 independent validation samples, exceeding the accuracy of the OLCI C2RCC-derived CDOM product (R2 = 0.0351, RMSE = 0.3556 and MSE = 0.1265) on the same validation samples in the PRE. Finally, we analysed the impacts of three distinct tropical cyclone pathways on CDOM dynamics in the region.