Highly Efficient JR Optimization Technique for Solving Prediction Problem of Soil Organic Carbon on Large Scale

Area of Science:

• Environmental Science
• Data Science
• Agricultural Science

Background:

• Accurate soil organic carbon (SOC) prediction is crucial for sustainable agriculture and environmental management.
• Integrating diverse data sources like remote sensing and ground-based measurements can enhance SOC prediction accuracy.
• Existing optimization algorithms may lack efficiency and optimal performance for complex environmental datasets.

Purpose of the Study:

• To develop and evaluate a novel optimization algorithm for soil organic carbon (SOC) prediction using multi-source environmental data.
• To create an integrated dataset (GeoBlendMDWC) combining remote sensing (DEM, MODIS) and ground data (WOSIS, CHELSA) for SOC analysis.
• To assess the efficiency and accuracy of the new algorithm against established methods (Jaya, GridSearchCV) for machine learning and deep learning models.

Main Methods:

• A novel data fusion approach integrating Digital Elevation Model (DEM), MODIS imagery, WOSIS soil data, and CHELSA climate data was employed.
• Machine learning (XGBoost, LightGBM, GBR, Random Forest, Decision Tree) and deep learning (MLP) models were utilized for SOC prediction.
• A new optimization algorithm was developed and compared against Jaya and GridSearchCV for performance and speed.

Main Results:

• The proposed optimization algorithm demonstrated superior performance and execution time compared to Jaya and GridSearchCV.
• Achieved an R² of 90.16 for machine learning models, a significant improvement over the base XGBoost R² of 79.08.
• Outperformed Jaya in deep learning optimization with an R² of 61.34 (vs. 30.04) and was 20-30 times faster than GridSearchCV.

Conclusions:

• The novel optimization algorithm offers a highly accurate and efficient solution for soil organic carbon (SOC) prediction.
• The GeoBlendMDWC dataset and the new algorithm provide a powerful tool for real-time environmental monitoring and agricultural applications.
• This methodology significantly advances the precision of SOC mapping, supporting better land management and farming practices.