A new method to estimate the dissolved organic carbon concentration from remote sensing in the global open ocean

The Dissolved Organic Carbon (DOC) represents the largest organic carbon reservoir in the ocean. Therefore, describing its spatio-temporal distribution is crucial for better understanding the global carbon cycle. Recent studies have demonstrated the possibility to determine DOC in coastal waters from ocean color radiometry (OCR) based on its strong correlation with the absorption coefficient of Colored Dissolved Organic Matter (acdom(λ)). However, in the open ocean, the CDOM to DOC relationship is highly variable as they present different sources, sinks, and kinetics. Here we present a new approach to estimating DOC over the open ocean based on an Artificial Neural Network (ANN) algorithm. This model accounts for i) Optical Water Classes (OWC) ii) sea surface temperature (SST), mixed layer depth (MLD), acdom(443), and chlorophyll-a (Chl-a) concentration, and iii) different time lags depending on the input parameter. The satellite DOC estimated with this model is in good agreement with in situ measurements (MAPD = 7.04%), while the spatial patterns follow former observations and model outputs. A sensitivity analysis has shown that the main descriptors to assess satellite DOC at a given time for oligotrophic and mesotrophic open ocean waters are SST one week before, and acdom(443) two weeks before; with Chl-a one week before as an additional input parameter for more productive waters. This study allows for the first time the assessment of the contribution of the particulate organic carbon (POC) to the total organic carbon (TOC) over the global ocean. The POC/TOC ratio value varies between 1.31% and 9.07%, with a mean value of about 4.57 ± 1.87%.