[Estimation Models for Jujube Leaf Pigment Concentration with Hyperspectrum Data at Canopy Scale].

Area of Science:

• Agricultural remote sensing
• Plant physiology
• Spectroscopy

Background:

• Plant canopy pigment concentration is vital for agricultural remote sensing and linked to leaf nitrogen content.
• Accurate, non-destructive methods for measuring pigment concentration are needed.

Purpose of the Study:

• Compare PLSR and SVM for predicting chlorophyll, chlorophyll-a, b, and carotenoid concentrations in jujube canopies.
• Develop quantitative models using hyperspectral data for non-destructive pigment estimation.

Main Methods:

• Correlation analysis between hyperspectral data and pigment concentrations.
• Model development using Partial Least Squares Regression (PLSR) and Support Vector Machine (SVM).
• Independent sample testing to evaluate prediction performance.

Main Results:

• Hyperspectral data strongly correlated with jujube canopy pigment concentration, especially chlorophyll and chlorophyll-a.
• Both PLSR and SVM effectively estimated pigment concentrations, with SVM outperforming PLSR for chlorophyll and carotenoid.
• PLSR showed better precision for chlorophyll-a and chlorophyll-b prediction.
• Prediction precisions for chlorophyll, chlorophyll-a, and carotenoid were superior to chlorophyll-b.

Conclusions:

• Hyperspectral data combined with PLSR and SVM offers a viable method for non-destructive estimation of jujube canopy pigment concentration.
• Model selection (PLSR vs. SVM) depends on the specific pigment being analyzed.
• Accurate estimation of chlorophyll, chlorophyll-a, and carotenoid is achievable with high determination coefficients and low errors.