Accurate deep-learning estimation of chlorophyll-a concentration from the spectral particulate beam-attenuation coefficient

Different techniques exist for determining chlorophyll-a concentration as a proxy of phytoplankton abundance. In this study, a novel method based on the spectral particulate beam-attenuation coefficient (c p ) was developed to estimate chlorophyll-a concentrations in oceanic waters. A multi-layer perceptron deep neural network was trained to exploit the spectral features present in c p around the chlorophyll-a absorption peak in the red spectral region. Results show that the model was successful at accurately retrieving chlorophyll-a concentrations using c p in three red spectral bands, irrespective of time or location and over a wide range of chlorophyll-a concentrations.