Gradient Boosting for the Spectral Super-Resolution of Ocean Color Sensor Data

Area of Science:

• Oceanography
• Remote Sensing
• Spectroscopy

Background:

• Hyperspectral data offers detailed spectral information (100+ bands) crucial for oceanographic analysis.
• Multispectral sensors provide limited spectral resolution (3-11 bands), undersampling ocean color.
• Reconstructing hyperspectral signatures from multispectral data is a key challenge in ocean remote sensing.

Purpose of the Study:

• To develop a gradient boosting framework for reconstructing hyperspectral signatures from multispectral ocean imagery.
• To enhance the spectral resolution of satellite-based ocean color data.
• To improve accessibility to high-fidelity hyperspectral data for scientific and environmental applications.

Main Methods:

• Utilized a gradient boosting framework for spectral reconstruction.
• Applied the method to remote sensing reflectance (Rrs) data from various ocean color sensors (SNPP VIIRS, OLCI, HICO, PACE OCI).
• Validated the approach using in situ Rrs data from NOAA calibration and validation cruises.

Main Results:

• Demonstrated the feasibility of transforming low-spectral-resolution imagery into high-fidelity hyperspectral products.
• Successfully reconstructed hyperspectral signatures in the visible spectrum (400-700 nm).
• Showcased the potential for hyperspectral enhancement using readily available multispectral platforms.

Conclusions:

• The gradient boosting framework effectively reconstructs hyperspectral signatures from multispectral ocean data.
• This advancement significantly benefits ocean color monitoring and broadens access to spectral data.
• The technique supports scientific and environmental applications by providing high-resolution spectral information.