A novel approach to spectral moisture interference correction for nitrogen and soil organic matter inversion in native black soils: Bayesian-optimized dynamic moisture mitigation

作者：Jiaze Tang, Qisong Wang, Dan Liu, Junbao Li, Ruifeng Zhang, Meiyan Zhang, Jinwei Sun · 发表于：Ecological Informatics · 年份：2025 · DOI：10.1016/j.ecoinf.2025.103240 · 被引用次数：7 · 研究领域：Soil Geostatistics and Mapping、Soil Moisture and Remote Sensing、Remote Sensing in Agriculture

In recent years, portable near-infrared spectrometers have emerged as viable alternatives to conventional chemical methods for measuring total nitrogen (TN) and soil organic matter (SOM). Advances in unmanned aerial vehicle technology have enabled low-altitude aerial surveys, facilitating the quantification of TN and SOM in agricultural soils—an approach beneficial for applications such as fertilizer management. However, most studies rely on laboratory-based analyses using high-precision and nonimaging spectrometers that test dried and processed soil samples. This preference stems from the significant impact of moisture on soil reflectance spectra, particularly in moisture-rich black soils. To address this challenge, this study investigated the in situ quantitative inversion of TN and SOM contents in moist black soil using a high-throughput hyperspectral imaging system. We introduced the Bayesian-optimized dynamic moisture mitigation (BO-DMM) method—an approach that effectively corrected moisture-induced spectral distortions. The BO-DMM method reduced moisture interference, calibrating the spectral angle of moist soil spectra to shrink by 50 % toward that of dry soil spectra. To further assess the effectiveness of the BO-DMM method, we integrated it with different machine learning models to test soil properties and predict the TN and SOM contents. The results indicated that BO-DMM significantly enhanced the prediction accuracy of different soil properties across different mod...