Project Overview
This project focused on the prediction of soil pH using advanced machine learning techniques, integrating Sentinel-2 satellite imagery with field-collected soil samples.
Conducted in collaboration with Beyonsus (Japan), the study aimed to develop an accurate, scalable, and cost-effective approach for soil quality assessment. This solution was designed to support precision agriculture and sustainable land management by enabling farmers and planners to monitor soil health remotely without extensive lab testing.
Key Methodologies & Impact
- Remote Sensing Integration: Utilized Sentinel-2 spectral bands to correlate vegetation indices and soil reflectance with pH levels.
- Machine Learning Models: Trained regression models (Random Forest / XGBoost) to predict pH values across large areas.
- Cost-Effective Scalability: Reduced the need for expensive and time-consuming physical soil testing.
- Precision Agriculture: Provided actionable data for fertilizer application and crop suitability analysis.
Figure 1: Machine Learning Model Development Workflow