Improved modeling of Mollisol organic matter at 40% moisture using colorimetric and portable X‐ray fluorescence spectrometry techniques

Abstract Rapid and reliable assessment of soil organic matter (SOM) is essential for sustainable soil management and precision agriculture. However, existing color‐based prediction approaches often neglect the influence of soil moisture, rely on a single color space, and rarely integrate multi‐source spectral and elemental information, resulting in limited accuracy and field applicability. To address these gaps, this study systematically optimized soil moisture conditions and combined colorimetric and portable X‐ray fluorescence (pXRF) techniques to improve SOM prediction in Mollisols of Northeast China. A total of 240 surface soil samples were analyzed to extract color parameters across multiple color spaces and quantify chromogenic elements (Ca, Fe, and Mn). Four predictive models, including multiple linear regression, multilayer perceptron neural network, random forest (RF), and support vector regression, were developed under controlled moisture conditions (training/validation = 70%:30%). Results showed that soil color parameters exhibited a nonlinear response to moisture, with the strongest correlations between SOM and color components occurring at 40% moisture. The optimal model was the RF using CIE‐ L * u * v * color space (where CIE is International Commission on Illumination) and pXRF‐derived variables, achieving R 2 values of 0.96 (training) and 0.81 (validation), root mean square error values of 0.36% and 0.77%, and ratio of performance to deviation values of 5.06 and 2.31, respectively. This study is innovative in quantitatively determining the optimal soil moisture condition for color‐based SOM estimation and in integrating pXRF‐derived chromogenic element data with colorimetric information within a unified prediction framework. These methodological advances substantially enhance prediction accuracy and robustness, providing a practical and cost‐effective approach for in situ SOM monitoring and visible‐spectrum‐based soil fertility assessment.