A Method for Predicting Alfalfa Biomass Based on Multimodal Data and Ensemble Learning Model

Area of Science:

• Agricultural Science
• Remote Sensing
• Data Science

Background:

• Traditional alfalfa biomass prediction methods struggle with accuracy in complex field conditions.
• Accurate biomass estimation is vital for effective pasture management and sustainable livestock production.

Purpose of the Study:

• To develop a highly accurate alfalfa biomass prediction method by integrating multispectral and LiDAR data with ensemble learning.
• To overcome the limitations of traditional methods in complex planting environments.

Main Methods:

• Extracted spectral and 3D structural features from UAV-based multispectral and airborne LiDAR data.
• Constructed an ensemble model using random forest, extra trees, and histogram gradient boosting.
• Performed feature selection to create a high-quality modeling dataset.

Main Results:

• The ensemble model achieved a coefficient of determination (R²) of 0.813, with RMSE of 0.178 kg m⁻² and MAE of 0.146 kg m⁻².
• Data fusion significantly outperformed models using only spectral indices (R² = 0.773) or LiDAR traits (R² = 0.576).
• Highest accuracy (R² = 0.917) was observed from bud emergence to early flowering stages, though high biomass intervals showed underestimation.

Conclusions:

• Multimodal data fusion and ensemble learning offer a robust approach for high-precision alfalfa biomass prediction.
• This method provides reliable technical support for pasture resource monitoring and precision agriculture.
• Further refinement is needed to address underestimation in high biomass scenarios.