Advancing Aerosol and Ocean Color Remote Sensing over Coastal Waters using Multiangle Polarimetry and Machine Learning

Abstract

Satellite remote sensing of aerosol and ocean color is crucial for advancing our understanding of air quality, ocean biogeochemical processes, and their impacts on climate change. A key challenge in ocean color remote sensing over coastal waters, complex with the presence of various constituents and absorbing aerosols, is atmospheric correction (AC), limited by weak aerosol information in traditional radiometric measurements. Measurements from multi-angle polarimeters (MAPs) can be used to overcome these challenges as they contain rich information on aerosol and hydrosol properties. NASA’s Plankton, Aerosol, Cloud, ocean Ecosystem (PACE) mission, launched in Feb 2024, provides global MAP data from UMBC’s HyperAngular Rainbow Polarimeter (HARP2) and SRON’s Spectropolarimeter for Planetary Exploration One (SPEXone). However, operational processing of such datasets using radiative transfer (RT) based retrieval algorithms remains impractical due to their high computational cost. This makes machine learning based RT emulators crucial for retrieval algorithms. The first part of this dissertation develops and validates a joint retrieval algorithm called FastMAPOL/component (Fast Multi-Angular Polarimetric Ocean coLor/component). The algorithm uses neural network (NN) based forward models to expedite the retrieval process and can be configured to process different combinations of PACE MAPs. It incorporates bio-optical models of coastal waters and component representation of aerosols to better capture the coastal marine environment. It also includes modules for AC and Bidirectional Reflectance Distribution Function (BRDF) correction. Validation using different combinations of PACE MAP measurements confirm the algorithm’s robustness and highlights the information content of each configuration. The second part evaluates FastMAPOL/component using real SPEXone measurements. Retrieved aerosol and ocean optical properties are validated against measurements from AERONET-OC stations worldwide. Aerosol component retrievals are qualitatively validated over dust and smoke scenes, and wind speed retrievals are evaluated against reanalysis estimates. The third part presents NN models for Instantaneous Photosynthetically Available Radiation (IPAR) at the ocean surface and its subsurface profile. These IPAR models are integrated into FastMAPOL/component for direct IPAR retrieval from satellites. The final part concludes with future perspectives, highlighting FastMAPOL/component’s potential for retrieving aerosols, ocean color, and biogeochemical parameters like IPAR in complex aerosol and ocean environment vital to carbon fixation and climate studies.