Soil Organic Matter Estimation Model Integrating Spectral and Profile Features.

Area of Science:

• Soil Science
• Remote Sensing
• Machine Learning

Background:

• Accurate soil organic matter (SOM) measurement is essential for soil quality assessment and management.
• Traditional methods for SOM analysis can be time-consuming and labor-intensive.
• Developing efficient and accurate SOM prediction models is a key research area.

Purpose of the Study:

• To develop an innovative hybrid model for predicting soil organic matter (SOM) by integrating spectral and profile features.
• To evaluate the performance of different machine learning models in conjunction with these integrated features.
• To identify the optimal feature extraction and modeling strategy for accurate SOM prediction.

Main Methods:

• Feature extraction using Principal Component Analysis (PCA), Lasso, and Sequential Circulant Averaging (SCARS) on spectral data.
• Integration of extracted spectral features with soil profile data.
• Application and comparison of machine learning models including Random Forest, ExtraTrees, and XGBoost for SOM prediction.
• Validation using coefficient of determination (R²) and Root Mean Square Error (RMSE).

Main Results:

• The hybrid approach significantly improved SOM prediction accuracy across tested models, with R² boosted by up to 26%.
• The ExtraTrees model combined with PCA-extracted spectral features and profile data achieved the highest accuracy (R² = 0.931, RMSE = 0.068).
• Compared to single-feature models, the integrated approach demonstrated substantial improvements in R² (17% for PCA spectral features, 26% for profile features).

Conclusions:

• Feature integration, combining spectral and profile data, offers a robust strategy for enhancing SOM prediction accuracy.
• The ExtraTrees model, utilizing PCA-derived spectral features and profile information, proves to be a highly accurate and stable tool for large-scale SOM assessment.
• This approach provides a valuable alternative to traditional methods for monitoring and managing soil organic matter.