Application of hyperspectral remote sensing technology in the quality evaluation of black soil

Black soil in Northeast China is a valuable agricultural resource, with the nutrient content and spatial distribution of elements being critical factors for evaluating cropland quality. This paper employs hyperspectral remote sensing data alongside field soil sampling to investigate the spatial distribution of key nutrient elements—total nitrogen (N), total phosphorus (P), and total potassium (K)—in black soil. By identifying key spectral bands and optimizing modeling approaches, we developed hyperspectral models capable of accurately extracting nutrient content. Our findings demonstrate strong correlations between soil nutrient concentrations and specific spectral transformation forms. The optimal spectral wavelengths for nitrogen (N), phosphorus (P), and potassium (K) vary based on the transformation methods and correlation metrics applied, underscoring the importance of tailored approaches for each nutrient. Specifically, soil nitrogen exhibits the highest correlation within the 510–570 nm spectral range, while phosphorus shows strong correlations at 576 nm, 1800 nm, and 2174 nm, and potassium correlates at 495 nm, 552 nm, 734 nm, and 1812 nm. In addition to traditional statistical methods, we applied a comparison between linear regression models and deep learning models, including multiple linear regression (MLR) and convolutional neural networks (CNN), which demonstrated exceptional performance. The CNN-based models achieved R² values of 0.994, 0.823, and 0.913 for nitrogen, phosphorus, and potassium, respectively, confirming their superior generalization capability across diverse soil types. Furthermore, the study focuses on the application of airborne hyperspectral data. The developed spectral transformation models and nutrient extraction methods were applied to high-resolution airborne hyperspectral imagery, enabling the generation of spatial distribution maps of nutrient element content with a 3.75-meter resolution. Compared to traditional geochemical sampling, this approach improves mapping efficiency, capturing localized field-scale nutrient variations more effectively and enhancing the understanding of nutrient element distribution in black soil. These airborne hyperspectral maps contribute to more precise agricultural practices, facilitating sustainable land management in Northeast China.