Machine learning for estimating phytoplankton size structure from satellite ocean color imagery in optically complex Pacific Arctic waters

In response to recent advances in satellite ocean color remote sensing, we have developed a chlorophyll a size distribution (CSD) model using machine learning (ML) approaches for optically complex Pacific Arctic waters. Previous CSD models have used principal component analysis (PCA) to retrieve spectral features from satellite-estimated phytoplankton absorption coefficient ( a ph ( λ )) by assuming a strong correlation between the spectral features and phytoplankton size structure determined from the exponent of the CSD ( η ). A weakness of such approach is that it relies on satellite retrievals of a ph ( λ ), which can be highly uncertain due to the optical effects of water constituents other than phytoplankton. In this study, we have developed a method based on ML to use remote sensing reflectance ( R rs ( λ )) for directly retrieving η , thus avoiding uncertainties due to the inversion of a ph ( λ ) from R rs ( λ ). Results show superior performance of the ML-based CSD models compared to the PCA-based model utilizing both R rs ( λ ) and a ph ( λ ) as predictors of η . For direct R rs ( λ )-based retrievals, a CSD model based on multivariable linear regression produced the best performance among all models considered. Nevertheless, models using in-situ a ph ( λ ) yielded better accuracy, reflecting a closer optical linkage between η and a ph ( λ ) than between η and R rs ( λ ). Our choice of an R rs ( λ )-based model for satellite application is therefore practical, motivated by the limitations and uncertainty of a ph ( λ ) inversions in optically complex waters. Another key finding is that more complex ML approaches do not always produce more effective models than standard linear regression. Indeed, multivariable linear regression outperformed other ML approaches for retrieving η directly from R rs ( λ ), whereas support vector machine performed the best among diverse ML approaches in the case of a ph ( λ ). Overall, this study found benefits in using R rs ( λ ) with ML to improve the retrieval accuracy of η for Pacific Arctic waters.