Integrating Proximal and Remote Sensing with Machine Learning for Pasture Biomass Estimation

Area of Science:

• Agricultural Science
• Remote Sensing
• Machine Learning

Background:

• Accurate pasture biomass estimation is crucial for effective livestock management and rangeland sustainability.
• Integrating diverse data sources presents a challenge for precise biomass quantification.

Purpose of the Study:

• To develop and evaluate an integrated approach for pasture biomass estimation.
• To compare the performance of different machine learning models and remote sensing data.

Main Methods:

• Combined ground-based vegetation height (PaddockTrac) with satellite-derived vegetation indices (Landsat 7, Sentinel-2).
• Utilized Boruta algorithm for feature selection and evaluated Linear Regression, Decision Tree, Random Forest, and XGBoost models.
• Assessed prediction accuracy using R², Mean Absolute Error (MAE), and Root Mean Square Error (RMSE).

Main Results:

• XGBoost model demonstrated superior performance, achieving R² of 0.86 with Landsat 7 data.
• Ground-based height and vegetation indices were key predictors for biomass estimation.
• Sentinel-2 red-edge indices offered limited improvement in homogenous pasture environments.

Conclusions:

• The integrated approach combining proximal sensing, remote sensing, and machine learning provides a scalable and accurate method for pasture biomass estimation.
• XGBoost model is highly effective for biomass prediction in pasture ecosystems.
• Findings support advancements in precision agriculture and sustainable rangeland management through enhanced monitoring capabilities.