Determining the Spectral Requirements for Cyanobacteria Detection for the CyanoSat Hyperspectral Imager with Machine...

Area of Science:

• Remote Sensing
• Ocean Optics
• Machine Learning

Background:

• Cyanobacteria blooms pose risks to aquatic ecosystems and human health.
• Accurate monitoring of cyanobacterial pigments is crucial for water quality assessment.
• Existing remote sensing technologies require optimization for efficient cyanobacteria detection.

Purpose of the Study:

• To determine the optimal spectral configuration for the CyanoSat imager for cyanobacterial pigment retrieval.
• To evaluate the impact of spectral band number and positioning on pigment estimation accuracy.
• To guide the selection of spectral bands for future water color radiometry sensors.

Main Methods:

• Utilized a simulated dataset for machine learning (ML) model training and validation.
• Assessed various spectral configurations, from minimal bands to high spectral resolution.
• Compared the performance of different configurations in discriminating and retrieving cyanobacterial pigments like phycocyanin (PC) and chlorophyll-a (Chl-a).

Main Results:

• A minimum of three spectral bands can determine PC and Chl-a, but not cyanobacteria composition.
• Approximately nine ideally positioned spectral bands achieved accuracy comparable to 300 channels for CAR and pigment concentrations.
• Narrower spectral band full-width half-maximum (FWHM) did not enhance performance over a 12 nm configuration.

Conclusions:

• Continuous spectral sampling is not essential for cyanobacterial detection.
• A multi-spectral configuration with strategically placed, narrow bands is sufficient.
• The identified spectral configurations can inform the design of next-generation ocean and water color sensors.