Augmenting Heritage Ocean-Color Aerosol Models for Enhanced Remote Sensing of Inland and Nearshore Coastal Waters

Frontiers in Remote Sensing

Jean-Claude Roger¹  Peng-wang Zhai²  Alexander Smirnov³  Robert Levy³  Martin Montes³  Brandon Smith³  P. Jeremy Werdell³  Nima Pahlevan³  David M. Giles³ 

[1] Department of Geographical Sciences, University of Maryland, College Park, MD, United States;Department of Physics, University of Maryland Baltimore County, Baltimore, MD, United States;NASA Goddard Space Flight Center, Greenbelt, MD, United States;Science Systems and Applications, Inc., (SSAI), Lanham, MD, United States;

关键词: absorbing aerosols;    aquatic remote sensing;    water quality;    atmospheric correction;    lakes;    rivers;   

DOI  :  10.3389/frsen.2022.860816

来源: DOAJ

【 摘 要 】

Satellite remote sensing of near-surface water composition in terrestrial and coastal regions is challenging largely due to uncertainties linked to a lack of representative continental aerosols in the atmospheric correction (AC) framework. A comprehensive family of absorbing aerosols is proposed by analyzing global AERONET measurements using the Partition Around Medoids (PAM) classifier. The input to the classifier is composed of Version 3, Level 2.0 daily average aerosol properties [i.e., single scattering albedo at λ = 0.44 μm, (SSA(0.44)) and the Angstrom exponents for extinction and absorption AEe(0.44–0.87) and AEa(0.44–0.87), respectively from observations from June 1993 to September 2019. The PAM classification based on low daily aerosol optical depth (AOD(0.44) ≤ 0.4) suggested 27 distinct aerosol clusters encompassing five major absorbing aerosol types (Dust (DU), Marine (MAR), Mixed (MIX), Urban/Industrial (U/I), and Biomass Burning (BB)). Seasonal patterns of dominant PAM-derived clusters at three AERONET sites (GSFC, Kanpur, and Banizoumbou) strongly influenced by U/I, DU, and BB types, respectively, showed a satisfactory agreement with variations of aerosol mixtures reported in the literature. These PAM-derived models augment the National Aeronautics and Space Administration's (NASA's) aerosol models (A2010) applied in its operational AC. To demonstrate the validity and complementary nature of our models, a coupled ocean-atmosphere radiative transfer code is employed to create a simulated dataset for developing two experimental machine-learning AC processors. These two processors differ only in their aerosol models used in training: 1) a processor trained with the A2010 aerosol models (ACI) and 2) a processor trained with both PAM and A2010 aerosol models (ACII). These processors are applied to Landsat-8 Operational Land Imager (OLI) matchups (N = 173) from selected AERONET sites equipped with ocean color radiometers (AERONET-OC). Our assessments showed improvements of up to 30% in retrieving remote sensing reflectance (Rrs) in the blue bands. In general, our empirically derived PAM aerosol models complement A2010 models (designed for regions strongly influenced by marine conditions) over continental and coastal waters where absorbing aerosols are present (e.g., urban environments, areas impacted by dust, or wildfire events). With the expected geographic expansion of in situ aquatic validation networks (e.g., AERONET-OC), the advantages of our models will be accentuated, particularly in the ultraviolet and short blue bands.

【 授权许可】

Unknown