UAV-Based Estimation of Tea Leaf Area Index in Mountainous Terrain: Integrating Topographic Correction and...

Area of Science:

• Agricultural Remote Sensing
• Plant Physiology
• Geospatial Analysis

Background:

• Leaf Area Index (LAI) is crucial for tea plant growth monitoring.
• Topographic relief and canopy heterogeneity in mountainous regions hinder accurate UAV-based LAI estimation.
• Existing methods struggle with radiometric consistency and topographic effects in complex terrains.

Purpose of the Study:

• To develop an integrated framework for improved UAV-based multispectral LAI estimation in rugged tea plantations.
• To evaluate the effectiveness of different topographic correction methods and machine learning models.
• To understand the retrieval mechanisms influencing LAI estimation accuracy.

Main Methods:

• Utilized UAV multispectral data from a tea-growing region in Fujian Province, China.
• Compared three topographic correction methods (SCS, SCS+C, Minnaert+SCS) with four machine learning models (LR, DT, RF, XGBoost).
• Employed SHapley Additive exPlanations (SHAP) for mechanism interpretability.

Main Results:

• The SCS+C topographic correction method significantly reduced topographic dependence and improved radiometric consistency.
• The XGBoost model combined with SCS+C correction achieved the highest retrieval accuracy (R² = 0.8930, RMSE = 0.6676).
• SHAP analysis identified red-band textural features (Correlation_R) as more important than vegetation indices, indicating a structure-dominated retrieval.

Conclusions:

• The integrated framework enhances tea LAI retrieval precision in complex mountainous terrains.
• The SCS+C correction and XGBoost model offer a robust approach for accurate LAI estimation.
• This study provides methodological insights for digital agriculture and precision management in challenging landscapes.