College of Biosystems Engineering and Food Science, Zhejiang University, Hangzhou 310058, China.
Key Laboratory of Spectroscopy Sensing, Ministry of Agriculture, Hangzhou 310058, China.
Molecules. 2019 Jul 7;24(13):2486. doi: 10.3390/molecules24132486.
Soil nitrogen is the key parameter supporting plant growth and development; it is also the material basis of plant growth. An accurate grasp of soil nitrogen information is the premise of scientific fertilization in precision agriculture, where near-infrared (NIR) spectroscopy is widely used for rapid detection of soil nutrients. In this study, the variation law of soil NIR reflectivity spectra with soil particle sizes was studied. Moreover, in order to precisely study the effect of particle size on soil nitrogen detection by NIR, four different spectra preprocessing methods and five different chemometric modeling methods were used to analyze the soil NIR spectra. The results showed that the smaller the soil particle sizes, the stronger the soil NIR reflectivity spectra. Besides, when the soil particle sizes ranged 0.18-0.28 mm, the soil nitrogen prediction accuracy was the best based on the partial least squares (PLS) model with the highest Rp of 0.983, the residual predictive deviation (RPD) of 6.706. The detection accuracy was not ideal when the soil particle sizes were too big (1-2 mm) or too small (0-0.18 mm). In addition, the relationship between the mixing spectra of six different soil particle sizes and the soil nitrogen detection accuracy was studied. It was indicated that the larger the gap between soil particle sizes, the worse the accuracy of soil nitrogen detection. In conclusion, soil nitrogen detection precision was affected by soil particle sizes to a large extent. It is of great significance to optimize the pre-treatments of soil samples to realize rapid and accurate detection by NIR spectroscopy.
土壤氮是支持植物生长和发育的关键参数；它也是植物生长的物质基础。准确掌握土壤氮信息是精准农业科学施肥的前提，近红外（NIR）光谱广泛用于快速检测土壤养分。本研究探讨了土壤近红外反射率光谱随土壤粒径的变化规律。此外，为了精确研究粒径对近红外土壤氮检测的影响，采用了四种不同的光谱预处理方法和五种不同的化学计量建模方法对土壤近红外光谱进行了分析。结果表明，土壤粒径越小，土壤近红外反射率光谱越强。此外，当土壤粒径在 0.18-0.28mm 范围内时，基于偏最小二乘法（PLS）模型的土壤氮预测精度最高，Rp 为 0.983，RPD 为 6.706。当土壤粒径过大（1-2mm）或过小时（0-0.18mm），检测精度不理想。此外，还研究了六种不同土壤粒径混合光谱与土壤氮检测精度之间的关系。结果表明，土壤粒径之间的差距越大，土壤氮检测的精度越差。综上所述，土壤粒径在很大程度上影响土壤氮的检测精度。优化土壤样品的预处理对于实现近红外光谱的快速准确检测具有重要意义。
