Detecting the Metal Elements and Soil Organic Matter in Farmland by Dual-Modality Spectral Technologies
WANG Jia-ying1, ZHU Yu-ting1, BAI Hao1, CHEN Ke-ming1, ZHAO Yan-ru1, 2, 3, WU Ting-ting1, 2, 3, MA Guo-ming4, YU Ke-qiang1, 2, 3*
1. College of Mechanical and Electronic Engineering, Northwest A&F University, Yangling 712100, China
2. Key Laboratory of Agricultural Internet of Things, Ministry of Agriculture, Yangling 712100, China
3. Shaanxi Key Laboratory of Agricultural Information Perception and Intelligent Service, Yangling 712100, China
4. Ningxia Runfeng Seed Industry Co., Ltd., Yinchuan 750000, China
摘要： 精准评估土壤质量是保障育种质量先决条件之一，对评估种子品质和精准施肥具有指导意义。土壤成分含量是土壤质量评估的重要指标，光谱技术已经被证实可以快速、绿色地进行土壤成分检测。然而单一模态光谱技术难以满足种田多种土壤成分含量检测的需求。故运用原子激光诱导击穿光谱（LIBS）和分子可见-近红外光谱（VIS-NIR）技术结合化学计量学方法，对宁夏润丰种业育种玉米田采集的288份土壤样本进行分析，建立金属元素和土壤有机质（SOM）含量的预测模型，并实现金属元素和SOM含量空间可视化分布。首先，利用共线双脉冲LIBS系统采集土壤样本的LIBS数据，利用air-PLS对光谱数据进行基线矫正以减少试验误差。选择的金属元素特征谱线查找于美国国家标准与技术研究院（National Institute of Standards and Technology，NIST）的标准原子光谱数据库。基于国家标准土样的LIBS光谱与其金属元素含量真实值，建立4种金属元素（Na、K、Mg、Mn）的偏最小二乘回归模型（PLSR），其中Mn含量的预测效果最好，R2p达到0.813，RMSEP为0.155 g·kg-1。另一方面，采集可见-近红外光谱数据后，利用SG卷积平滑(SGCS)、一阶导数变换、多元散射矫正（MSC）对光谱数据进行预处理，并分别建立SOM含量的PLSR预测模型对三种预处理方法进行评价，经MSC预处理后所建立的PLSR模型效果最好；随后利用蒙特卡洛交叉验证法（MCCV）剔除SOM含量异常样本。利用竞争自适应重加权采样法（CARS）和连续投影算法（SPA）选择特征波长，分别建立SOM含量的PLSR预测模型对两种算法进行评价；得出利用CARS算法选择的特征波长建立的预测模型性能有所提高。用CARS算法选择的特征波长与SOM含量真实值，分别建立PLSR和反向传播人工神经网络（BPNN）预测模型，其中PLSR模型的效果最好，R2p达到0.864，RMSEP为0.612 g·kg-1，RPDv为2.733。最后，利用国家标准土样所建立的PLSR模型预测玉米种田四种金属元素含量，建立PLSR模型预测值和BPNN模型预测值的空间分布图。研究结果表明，LIBS技术和可见-近红外光谱定量分析技术可以对种田土壤金属元素和SOM含量检测，为土壤成分含量的检测和空间可视化分布提供了参考价值并对土壤科学合理地施肥具有指导意义。
关键词：土壤有机质；金属元素；激光诱导击穿光谱；可见-近红外光谱；化学计量学方法
Abstract：Accurate evaluation of soil quality is one of the prerequisites for ensuring breeding quality, which is of guiding significance for evaluating seed quality and precise fertilization. Soil composition content is an important indicator of soil quality assessment; spectral technology has been proven to detect soil composition quickly and greenly. However, due to the limitations of different spectral excitation principles, single-spectral technology cannot meet the needs of multiple soil composition content detection in breeding fields. This study used laser-induced breakdown spectroscopy (LIBS) and visible-near-infrared spectroscopy (VIS-NIR) combined with intelligent algorithms to analyze 288 soil samples collected from the breeding corn field of Ningxia Runfeng Seed Industry. The prediction models of metal elements and soil organic matter (SOM) content were established, and the spatial visualization distribution of metal elements and SOM content was realized. The specific research is: (1) Detection of metal elements in the maize breeding field. After collecting LIBS spectral data using a collinear double pulse LIBS system, air-PLS was used to correct the baseline of the spectral data and reduce the experimental error. The selected characteristic spectral lines of metal elements were searched in the standard atomic spectrum database of the National Institute of Standards and Technology (NIST). Combined with the LIBS spectrum of national standard soil samples and the true value of metal element content, a partial least squares regression (PLSR) model of four metal elements (Na, K, Mg, Mn) was established in national standard soil samples. Among them, the prediction effect of Mn content was the best, R2p reached 0.813,RMSEP was 0.155 g·kg-1; (2) Detection of SOM in maize breeding field. After collecting visible-near infrared spectral data, the spectral data were preprocessed by Savitzky-Golay Convolution Smoothing (SGCS), first derivative transformation, and Multivariate Scattering Correction (MSC), and the PLSR prediction model of SOM content was established to evaluate the three pretreatment methods. The PLSR model established after MSC pretreatment was the best. Subsequently, Monte Carlo cross-validation (MCCV) was used to eliminate the samples of abnormal SOM content. Competitive Adaptive Reweighed Sampling (CARS) and Successive Projections Algorithm (SPA) were used to select the characteristic wavelengths, and the PLSR prediction models of SOM content were established to evaluate the two algorithms. It was concluded that the prediction model's performance, established by the characteristic wavelengths selected by the CARS algorithm, was improved. And the characteristic wavelengths selected by the CARS algorithm and the true value of SOM content were combined to establish PLSR and back propagation neural network (BPNN) prediction models. The PLSR model had the best effect, with R2p of 0.864, RMSEP of 0.612 g·kg-1, and RPDv of 2.733. (3) Spatial visualization distribution of metal elements and SOM content in maize breeding field. The PLSR model established by national standard soil samples was used to predict the content of four metal elements in the maize breeding field, and the spatial distribution map of predicted value content was established. Finally, the SOM content spatial distribution map of the real value, PLSR model predicted value, and BPNN model predicted value was established. The results show that LIBS technology and visible-near infrared spectroscopy quantitative analysis technology can detect the content of metal elements and SOM in the soil of the breeding field, which provides a reference value for the detection and spatial visualization distribution of soil component content.
王嘉滢，朱雨婷，白 浩，陈柯铭，赵艳茹，吴婷婷，马国明，余克强. 双模态光谱技术对种田土壤金属元素和有机质检测分析[J]. 光谱学与光谱分析, 2025, 45(08): 2317-2325.
WANG Jia-ying, ZHU Yu-ting, BAI Hao, CHEN Ke-ming, ZHAO Yan-ru, WU Ting-ting, MA Guo-ming, YU Ke-qiang. Detecting the Metal Elements and Soil Organic Matter in Farmland by Dual-Modality Spectral Technologies. SPECTROSCOPY AND SPECTRAL ANALYSIS, 2025, 45(08): 2317-2325.
[1] LU Ying(卢 莹). Molecular Plant Breeding(分子植物育种), 2024, 22(14): 4810.
[2] QI Xue-li, CHEN Yan-yan, WANG Yong-xia, et al(齐学礼, 陈艳艳, 王永霞, 等). Current Research Status and Development Suggestions for Advanced Breeding Technologies in China(中国作物育种先进技术的研发现状与发展建议). Molecular Plant Breeding(分子植物育种), 2024, (1): https://link.cnki.net/urlid/46.1068.S.20240102.1707.020.
[3] Hauer-Jakli M, Tränkner M. Frontiers in Plant Science, 2019, 10: 766.
[4] WANG Ren-jie, ZHU Fan, LIANG Hui-zi, et al（王仁杰, 朱 凡, 梁惠子, 等）. Acta Ecologica Sinica(生态学报), 2020, 40(6): 2019.
[5] Wood S A, Tirfessa D, Baudron F. Agriculture, Ecosystems & Environment, 2018, 266: 100.
[6] LI Min-zan, ZHENG Li-hua, AN Xiao-fei, et al（李民赞, 郑立华, 安晓飞, 等）. Transactions of the Chinese Society for Agricultural Machinery(农业机械学报), 2013, 44(3): 73.
[7] Ren J, Zhao Y, Yu K. Computers and Electronics in Agriculture, 2022, 197: 106986.
[8] LI Jin-chang, HE Hong-yuan, ZHAO Xue-jun, et al（李锦昌, 何洪源, 赵雪珺, 等）. Applied Chemical Industry(应用化工), 2022, 51(11): 3369.
[9] Tavares T R, Mouazen A M, Nunes L C, et al. Soil and Tillage Research, 2022, 216: 105250.
[10] Chen G, Yang G, Ling Z, et al. Analytical Methods, 2021, 13(12): 1502.
[11] Li C, Zhao J, Li Y, et al. Forests, 2021, 12(12): 1809.
[12] ZHANG Hai-liang, XIE Chao-yong, TIAN Peng, et al（章海亮, 谢潮勇, 田 彭, 等）. Spectroscopy and Spectral Analysis(光谱学与光谱分析), 2023, 43(7): 2226.
[13] CHENG Jie-hong, CHEN Zheng-guang, ZHANG Qing-hua（程介虹, 陈争光, 张庆华）. Journal of Agricultural Science and Technology(中国农业科技导报), 2020, 22(1): 162.
[14] Zhou W, Xiao J, Li H, et al. Journal of Soils and Sediments, 2023, 23(6): 2506.
[15] WANG Si-nan, LI Rui-ping, WU Ying-jie, et al（王思楠, 李瑞平, 吴英杰, 等）. Transactions of the Chinese Society for Agricultural Machinery(农业机械学报), 2022, 53(5): 332.
[16] AN Bai-song, WANG Xue-mei, HUANG Xiao-yu, et al（安柏耸, 王雪梅, 黄晓宇, 等）. Earth and Environment(地球与环境), 2023, 51(2): 246.
[17] Ren J, Yang Z H, Zhao Y R, et al. Optics Express, 2022, 30(21): 37711.
