Optimal Spectral Wavelengths for Discriminating Orchard Species Using Multivariate Statistical Techniques.

Area of Science:

• Agricultural Science
• Remote Sensing
• Spectroscopy

Background:

• Sustainable orchard management relies on accurate tree species identification.
• Precision agriculture programs benefit from detailed tree inventory data.
• Hyperspectral imagery offers a promising tool for orchard tree species mapping.

Purpose of the Study:

• To identify optimal wavelengths for discriminating between dominant orchard tree species using hyperspectral data.
• To develop and compare two multivariable methods for band selection in hyperspectral analysis.
• To assess the spectral separability of almond, walnut, and grape species.

Main Methods:

• Field spectroscopy was conducted on 165 leaf samples across the 350-2500 nm range.
• Two multivariable approaches were employed: ANOVA-RFC-PCA and Partial Least Squares (PLS).
• Discriminant analysis (DA) was used to evaluate species separation and identify optimal wavelengths.

Main Results:

• Distinct spectral behaviors were observed in the visible, red edge, and near-infrared ranges.
• The ANOVA-RFC-PCA method reduced wavelengths to five key values (363, 423, 721, 1064, 1388 nm).
• The PLS-DA model achieved 100% accuracy with optimal wavelengths at 397, 515, 647, 1386, and 1919 nm.

Conclusions:

• Hyperspectral data combined with optimized band selection effectively distinguishes orchard tree species.
• The identified optimal wavelengths provide a basis for developing accurate tree species mapping algorithms.
• This research supports precision agriculture and sustainable orchard management practices.