One-step retrieval of water-quality parameters from satellite top-of-atmosphere measurements

Abstract

Traditionally, deriving water-quality parameters from satellite ocean color measurements involves multiple complex steps, including in-orbit radiometric calibration, atmospheric correction, and water property retrieval, all of which require substantial resources and effort. Given the abundance of water-quality related products provided by established ocean color satellite missions, we propose a scheme (DN2WP) to estimate water-quality parameters directly from the digital number (DN) data obtained at satellite altitude. Using Secchi Disk Depth (Z_SD) as an example and the Z_SD data from Sentinel-3 as a reference, HiSea-II DN data were converted to Z_SD directly by DN2WP. In the training phase, the coefficient of determination (R²) for Z_SD reached 0.95, with a mean absolute percentage difference (MAPD) of ∼10 %. When applied to new datasets, the R² for Z_SD exceeded 0.8, and the MAPD remained within 20 %. The DN2WP scheme was further tested using the multispectral sensors of the CZI on board the HY-1C satellite and MII on the SDGSAT-1 satellite, respectively, and a high level of performance was maintained by achieving an R² of 0.83 (0.84) with MAPD as 19.8 % (15.8 %) for CZI (SDGSAT-1) on independent data. These results are significantly better than the consistency measure of Z_SD between such satellites and Sentinel-3A/3B obtained individually through the traditional multi-step approach. It indicates that DN2WP is an efficient approach to get water-quality parameters from top-of-atmosphere DN data measured by a satellite sensor, which also results in much better consistency across different satellite missions.