Robust spectral optimization algorithms for retrieving inherent optical properties of global oceans

Area of Science:

• Oceanography
• Remote Sensing
• Optical Physics

Background:

• Ocean color remote sensing is crucial for global monitoring.
• Spectral optimization methods are key algorithms but sensitive to model and inherent optical property (IOP) errors.
• Existing methods struggle with accuracy due to inherent optical property (IOP) uncertainties.

Purpose of the Study:

• To improve the accuracy of ocean color remote sensing by developing error post-correction methods.
• To explore correlation patterns among inherent optical properties (IOPs) for enhanced inversion.
• To introduce novel correlation regularization models for spectral optimization algorithms.

Main Methods:

• Established cross-regression models for IOP magnitude parameters using in situ data.
• Developed global correlation regularization (G-CoReg) and locally weighted correlation regularization (L-CoReg) models.
• Reformulated semianalytical inversion as a spectral optimization task with correlation constraints.

Main Results:

• G-CoReg and L-CoReg algorithms showed superior accuracy over existing methods on NOMAD and SeaWiFS datasets.
• The new algorithms demonstrated improved robustness against atmospheric correction errors.
• Reduced accuracy degradation was observed with the proposed correlation regularization models.

Conclusions:

• The proposed G-CoReg and L-CoReg models significantly enhance ocean color remote sensing accuracy and robustness.
• Correlation regularization effectively mitigates errors from semianalytical models and atmospheric correction.
• These advancements offer more reliable ocean color data for scientific applications.