Non-destructive estimation of maize carotenoids using reflectance-based spectral indices

Area of Science:

• Agricultural Science
• Plant Physiology
• Remote Sensing

Background:

• Carotenoids play crucial roles in light harvesting and photoprotection in plants.
• Leaf spectral reflectance is influenced by pigment concentrations, including carotenoids.
• Existing spectral indices have limitations in accurately estimating carotenoid content.

Purpose of the Study:

• To investigate the relationship between maize leaf carotenoid content and spectral reflectance.
• To evaluate existing carotenoid estimation indices.
• To develop novel spectral indices and machine learning models for improved carotenoid prediction.

Main Methods:

• Spectral analysis of maize leaves across visible wavelengths (500-650 nm).
• Development of nine new spectral indices using principal component analysis.
• Application and evaluation of machine learning models (REPTree) for carotenoid estimation.

Main Results:

• A strong positive correlation was found between carotenoid and chlorophyll content.
• New indices CAR7, CAR8, and CAR9 showed superior predictive ability compared to existing indices.
• Machine learning models, particularly REPTree, significantly enhanced estimation accuracy (R2 = 0.79, NRMSE = 13.84% on test data).

Conclusions:

• Spectral indices and machine learning models enable accurate, non-destructive estimation of maize carotenoid content.
• These methods offer potential for practical applications in crop monitoring and stress assessment.
• The study highlights the importance of developing targeted spectral approaches for plant pigment analysis.