South China Sea Annual Meeting 2023 (SCSAM)

55 / 2023-07-24 15:08:10

Vertical distribution of phytoplankton functional types in the South China Sea and its adjacent Taiwan Strait based on phytoplankton absorption spectrum and machine learning

phytoplankton functional types (PFTs), absorption coefficient of phytoplankton (aph), Support Vector Regression (SVR), South China Sea (SCS) and its adjacent Taiwan Strait (TWS).

There are various types of phytoplankton, or phytoplankton functional types (PFTs), in natural aquatic environments, which play different roles in biogeochemical cycles of different elements. In this study, based on a large (1747 points) dataset collected from 2004 to 2019 in the South China Sea (SCS) and its adjacent Taiwan Strait (TWS) that has concurrent measurements of the spectral absorption coefficient of phytoplankton (aph) and chlorophyll a concentration of 9 PFTs (PFTsChla), along with depth and time information, a Support Vector Regression (SVR) model was developed to retrieve these 9 PFTsChla in the water column. These PFTs include diatoms, dinoflagellates, haptophytes_8, haptophytes_6, chlorophytes, cryptophytes, prochlorococcus, and prasinophytes. Independent validation results indicate that the SVR model outperforms traditional PFTsChla retrieval algorithms, with an average of Mean Biases as -14.2%, average of Mean Absolute Unbiased Relative Differences as 60.3%, and average of coefficient of determinations as 0.56. The distribution of predicted PFTsChla in the water column was subsequently analyzed, and the predicted diatoms were found to contribute the most to particulate organic carbon. Last, the SVR model was found to be applicable to most PFTsChla retrieval in the adjacent East China Sea.