Soil organic matter (OM) is a crucial indicator of soil productivity and sustainable management; however, traditional OM quantification is time-consuming and labor-intensive.This study aimed to develop land-use-specific OM predictive models using smartphone-based soil color data and machine learning (ML) algorithms.A total of 1,459 soil samples were collected from upland, paddy, and orchard fields in the Chungcheong province, South Korea.Soil images were captured using four different smartphones (e.g., Samsung and Apple) under controlled conditions, and color variables were converted into Commission Internationale de I'Elcairage (CIE)-guided color spaces (e.g., CIE-L * a * b * and CIE-L * c * h * ) to minimize devicespecific errors.Three ML algorithms, including multiple linear regression (MLR), support vector machine (SVM), and random forest (RF), were evaluated.Pearson correlation analysis identified L * , b * , and h * parameters as key predictors, confirming that soil color darkens as OM content increases.Among the ML algorithms, RF model demonstrated the most stable and superior regression performance, achieving a coefficient of determination (R 2 ) of up to 0.759 for paddy soils, while maintaining consistency across both training and test datasets.These findings suggest that integrating smartphone-based colorimetry with RF algorithms provides a potential alternative to traditional OM analysis, enabling efficient real-time field monitoring for sustainable agricultural land management.
Kang et al. (Sun,) studied this question.