Improved estimation of soil organic matter content in the Yellow River delta using spectral data optimization and environmental variables integration

Abstract

Soil Organic Matter (SOM) is a critical component of soil fertility, playing a key role in crop growth and soil improvement. Visible and near-infrared (Vis-NIR) spectroscopy is an ideal method for rapid and accurate estimation of SOM content. This study utilizes fractional-order derivative (FOD) and optimal band combination algorithm (two-dimensional spectral index (2DSI) and three-dimensional spectral index (3DSI)) collaboratively to optimize soil Vis-NIR spectral data and combines environmental variables (EV) to establish an estimation model for SOM content. The results indicate that the use of FOD significantly enhances the correlation between spectra and SOM, especially within the range of 0.4 to 1.4 orders. Modeling outcomes demonstrate that combining FOD with either 2DSI or 3DSI yields superior results compared to models constructed using FOD alone. Specifically, the introduction of 2DSI and 3DSI increases R_v2 values by 3.66 % and 8.54 %, respectively, compared to using FOD alone. Furthermore, when optimized spectral data are combined with EV as input variables, the model’s estimation performance is markedly improved. The optimal model in this study is constructed using a combination of 2Ds + 3Ds + EVs, achieving R_v², RMSE_v, and RPD values of 0.92 g kg⁻¹, 0.40 g kg⁻¹, and 2.83, respectively. The research findings confirm the capabilities of FOD technology combined with the optimal band combination algorithm in optimizing spectral data, highlighting the advantages of multi-source data fusion modeling.