Remote sensing estimates of global sea surface nitrate: Methodology and validation

Information about sea surface nitrate (SSN) concentrations is crucial for estimating oceanic new productivity and for carbon cycle studies. Due to the absence of optical properties in SSN and the intricate relationships with environmental factors affecting spatiotemporal dynamics, developing a more representative and widely applicable remote sensing inversion algorithm for SSN is challenging. Most methods for the remote estimation of SSN are based on data-driven neural networks or deep learning and lack mechanistic descriptions. Since fitting functions between the SSN and sea surface temperature (SST), mixed layer depth (MLD), and chlorophyll (Chl) content have been established for the open ocean, it is important to include the remote sensing indicator photosynthetically active radiation (PAR), which is critical in nitrate biogeochemical processes. In this study, we employed an algorithm for estimating the monthly average SSN on a global 1° by 1° resolution grid; this algorithm relies on the empirical relationship between the World Ocean Atlas 2018 (WOA18) monthly interpolated climatology of nitrate in each 1° × 1° grid and the estimated monthly SST and PAR datasets from Moderate Resolution Imaging Spectroradiometer (MODIS) and MLD from the Hybrid Coordinate Ocean Model (HYCOM). These results indicated that PAR potentially affects SSN. Furthermore, validation of the SSN model with measured nitrate data from different months and locations for the years 2018-2023 yielded a high prediction accuracy (N = 12,846, R² = 0.93, root mean square difference (RMSE) = 3.12 μmol/L, and mean absolute error (MAE) = 2.22 μmol/L). Further independent validation and sensitivity tests demonstrated the validity of the algorithm for retrieving SSN.