Mapping Organic Carbon Content in Black Soil Using UAV Hyperspectral Remote Sensing and Deep Learning
-
摘要: 我国东北地区黑土作为重要且珍贵的农耕资源，受到长期开发的影响，退化问题日益加重. 通过卫星遥感技术反演黑土有机碳含量能够为保护利用黑土地提供技术支撑.针对卫星高光谱数据空间分辨率低以及黑土有机碳含量反演精度低的问题，本研究利用无人机高光谱数据及土壤地球化学数据，基于一维卷积神经网络思想，构建并对比了MDS-1DCNN、LLE-1DCNN、PLSR-1DCNN与KPCA-1DCNN四种模型土壤有机碳含量的反演效果.以黑龙江五大连池市典型黑土区为研究区，结果表明：LLE-1DCNN模型反演效果较好，在验证集上的R2为0.806，RMSE为0.572%，能够为黑土土壤有机碳含量反演提供技术支撑.Abstract: Black soil in northeastern China is an important agricultural resource but has been increasingly degraded due to long-term development. The use of satellite remote sensing technology to retrieve the organic carbon content in black soil offers technical support for the protection and sustainable use. However, satellite hyperspectral data suffer from low spatial resolution, and the retrieval accuracy for organic carbon content remains limited in fine-scaled study sites. To address these challenges, this study utilized UAV-based hyperspectral data and soil geochemical data instead. We proposed and compared four models based on one-dimensional convolutional neural networks (1DCNN)-MDS-1DCNN, LLE-1DCNN, PLSR-1DCNN, and KPCA-1DCNN, for organic carbon content retrieval using the Wudalianchi region in Heilongjiang Province as a case study. The results show that the LLE-1DCNN model outperforms the others, achieving an R2 of 0.806 and an RMSE of 0.572% on the validation set. This approach offers promising potential for accurately retrieving organic carbon content in black soil and supporting its conservation and management.
-
Key words:
- black soil /
- soil organic carbon /
- UAV hyperspectral /
- deep learning /
- remote sensing
-
图 2 野外土壤样本采样点分布图
Fig. 2. Distribution map of sampling points of soil samples in the field
图 3 原始数据、经过SG滤波的数据与经过4种预处理方法的无人机数据光谱曲线
Fig. 3. Original data、SG-filtered data and the spectral curves of UAV data after four preprocessing methods
图 5 耦合模型原理的矩阵解释
a. 输入耦合模型的原始数据；b. 输入耦合模型的降维后的特征数据
Fig. 5. Matrix Explanation of Coupling Model Principle
图 6 耦合模型预测值与真实值在验证集上的偏离程度图
Fig. 6. Deviation between coupled model predictions and actual values on the validation set
a. KPCA-1DCNN; b. MDS-1DCNN; d. LLE-1DCNN; c. PLSR-1DCNN
表 1 HRS主要参数表
Table 1. HRS main parameter table
参数 值 光谱范围（nm） 400~1 000 光谱波段（units） 224 光谱采样间隔（nm） 2.68 空间像素（px） 1 024 FOV 38 成像速度（fps） 330 位置精度（m） 0.02~0.05 
下载: 导出CSV
表 2 原始数据与经过SG滤波的不同预处理方法下的相关性统计表
Table 2. Correlation statistics table between original data and different preprocessing methods filtered by SG
预处理方法 PCC SRCC 相关性最大波段（nm） 相关性最大值 相关性最大波段（nm） 相关性最大值 RAW 748 0.440 1 023 0.423 SG-FD 722 0.678 1 307 0.628 SG-SD 1 173 0.559 2 434 0.541 SG-MSC 1 711 0.602 1 711 0.499 SG-SNV 1 677 0.595 1 677 0.503
下载: 导出CSV
表 3 1DCNN模型参数表
Table 3. 1DCNN model parameter table
模型结构 参数一 参数二 参数三 参数四 卷积核大小 3 2 3 2 卷积步长 1 1 1 1 卷积层数 4 4 5 5 激活函数 ReLU ReLU ReLU ReLU 池化层数 2 2 3 3 全连接层数 1 1 1 1
下载: 导出CSV
表 4 数据集SOC含量统计表
Table 4. Statistical table of SOC content in the dataset
样本类型 样本数量 最小值
（%）最大值
（%）平均值
（%）标准差
（%）变异系数（%） 训练集 72 1.91 8.83 4.54 1.36 29.95 验证集 18 2.11 8.60 4.53 1.37 30.24 总集 90 1.91 8.83 4.54 1.36 29.95
下载: 导出CSV
表 5 4种不同参数的1DCNN模型预测效果对比
Table 5. Comparison of prediction performance of the 1DCNN model with four different parameter sets
模型 训练集 验证集 R2 RMSE% R2 RMSE% 参数一 0.832 0.505 0.676 0.917 参数二 0.885 0.393 0.725 0.643 参数三 0.774 0.543 0.733 0.685 参数四 0.886 0.424 0.768 0.679
下载: 导出CSV
表 6 4种耦合模型预测效果对比
Table 6. Comparison of the prediction performance of four coupled models
模型 训练集 验证集 R2 RMSE(%) R2 RMSE(%) KPCA-1DCNN 0.915 0.438 0.64 0.859 MDS-1DCNN 0.872 0.515 0.699 0.775 PLSR-1DCNN 0.845 0.498 0.779 0.584 LLE-1DCNN 0.88 0.466 0.806 0.572
下载: 导出CSV
-
Angelopoulou, T., Tziolas, N., Balafoutis, A., et al., 2019. Remote Sensing Techniques for Soil Organic Carbon Estimation: A Review. Remote Sensing, 11(6): 676. https://doi.org/10.3390/rs11060676 Bendor, E., Chabrillat, S., Demattê, J. A. M., et al., 2009. Using Imaging Spectroscopy to Study Soil Properties. Remote Sensing of Environment, 113: S38-S55. https://doi.org/10.1016/j.rse.2008.09.019 Brunet, D., Barthès, B. G., Chotte, J. L., et al., 2007. Determination of Carbon and Nitrogen Contents in Alfisols, Oxisols and Ultisols from Africa and Brazil Using NIRS analysis: Effects of Sample Grinding and Set Heterogeneity. Geoderma, 139(1/2): 106-117. https://doi.org/10.1016/j.geoderma.2007.01.007 Chen, Y., Wang, J. L., Liu, G. J., et al., 2019. Hyperspectral Estimation Model of Forest Soil Organic Matter in Northwest Yunnan Province, China. Forests, 10(3): 217. https://doi.org/10.3390/f10030217 Deng, Y., Niu, Z. W., Feng, Q. Y., et al., 2023. A Novel Hyperspectral Prediction Model of Organic Matter in Red Soil Based on Improved Temporal Convolutional Network. Spectroscopy and Spectral Analysis, 43(9): 2942-2951(in Chinese with English abstract). Diwu, P. Y., Bian, X. H., Wang, Z. F., et al., 2019. Study on the Selection of Spectral Preprocessing Methods. Spectroscopy and Spectral Analysis, 39(9): 2800(in Chinese with English abstract). Gao, L. L., Zhu, X. C., Han, Z. Y., et al., 2019. Spectroscopy-Based Soil Organic Matter Estimation in Brown Forest Soil Areas of the Shandong Peninsula, China. Pedosphere, 29(6): 810-818. https://doi.org/10.1016/S1002-0160(17)60485-5 Huang, Y., Dend, G., 2013. Research on Development of Agricultural Geographic Information Ontology. Journal of Integrative Agriculture, 11(5): 865-877. https://doi.org/10.1016/S2095-3119(12)60077-X Ji, W. J., Li, S., Chen, S. C., et al., 2016. Prediction of Soil Attributes Using the Chinese Soil Spectral Library and Standardized Spectra Recorded at Field Conditions. Soil and Tillage Research, 155: 492-500. https://doi.org/10.1016/j.still.2015.06.004 Li, X. P., Zhang, F., Wang, X. P., 2019. Study on Differential-Based Multispectral Modeling of Soil Organic Matter in Ebinur Lake Wetland. Spectroscopy and Spectral Analysis, 39(2): 535-542(in Chinese with English abstract). Liaghat, S., Balasundram, S. K., 2020. A review: the Role of Remote Sensing in Precision Agriculture. American Journal of Agricultural & Biological Science, 5(1): 553-564. Liu, H. J., Zhang, B., Zhao, J., et al., 2007. Spectral Models for Prediction of Organic Matter in Black Soil. Acta Pedologica Sinica, 44(1): 27-32(in Chinese with English abstract). Liu, Y., Xu, Y., 2013. Application of Savitzky-Golay Smoothing Filter in the Pre-Processing of Near-Infrared Spectra for Rapid Analysis of Grape Juices. Food Chemistry, 139(1-4): 205-212. Padarian, J., Minasny, B., McBratney, A. B., 2019. Using Deep Learning to Predict Soil Properties from Regional Spectral Data. Geoderma Regional, 16: e00198. https://doi.org/10.1016/j.geodrs.2018.e00198 Pan, N., Wang, S., Liu, Y. X., et al., 2019. Advances in Soil Moisture Retrieval from Remote Sensing. Acta Ecologica Sinica, 39(13): 4615-4626(in Chinese with English abstract). Piikki, K., Wetterlind, J., Söderström, M., et al., 2021. Perspectives on Validation in Digital Soil Mapping of Continuous Attributes: A Review. Soil Use and Management, 37(1): 7-21. https://doi.org/10.1111/sum.12694 Rossel, R. A., Webster, R., 2012. Predicting Soil Properties from the Australian Soil Visible-Near Infrared Spectroscopic Database. European Journal of Soil Science, 63(6): 848-860. https://doi.org/10.1111/j.1365-2389. 2012. 01495. x doi: 10.1111/j.1365-2389.2012.01495.x Rostami, R., Fathollahi-Fard, A. M., 2022. A New Approach for Evaluating the Pearson Correlation Coefficient Using Machine Learning. Applied Sciences, 12(5): 2456. Shi, Z., Xu, D. Y., Teng, H. F., et al., 2018. Soil Information Acquisition Based on Remote Sensing and Proximal Soil sensing: Current Status and Prospect. Progress in Geography, 37(1): 79-92(in Chinese with English abstract). Tian, W. X., 2019. Hyperspectral Quantitative inversion of Black Soil Organic Matter Based on Statistical Model. Chengdu University of Technology, Chengdu(in Chinese with English abstract). Viscarra Rossel, R. A., Behrens, T., Ben-Dor, E., et al., 2016. A Global Spectral Library to Characterize the World's Soil. Earth-Science Reviews, 155: 198-230. https://doi.org/10.1016/j.earscirev.2016.01.012 Wang, D. M., Qin, K., Li, Z. Z., et al., 2018. Retrieval of Organic Matter Content in Black Soil Based on Airborne Hyperspectral Remote Sensing Data: Taking Jiansanjiang District in Heilongjiang Province as an Example. Earth Science, 43(6): 2184-2194(in Chinese with English abstract). Wang, H. F., Chen, Y. W., Zhang, Z. T., et al., 2019. Quantitatively Estimating Main Soil Water-Soluble Salt Ions Content Based on Visible-Near Infrared Wavelength Selected Using GC, SR and VIP. PeerJ, 7: e6310. https://doi.org/10.7717/peerj.6310 Wang, Y. D., Zhang, F., Hu, W. Y., et al., 2024. Reversing Organic Matter Contents in Black Soils in Northeast China Using Digital Image Technology. Soils, 56(5): 1051-1056(in Chinese with English abstract). Xiao, Y., Xin, H. B., Wang, B., et al., 2021. Hyperspectral Estimation of Black Soil Organic Matter Content Based on Wavelet Transform and Successive Projections Algorithm. Remote Sensing for Land & Resources, 33(2): 33-39(in Chinese with English abstract). Xie, Y. H., 2019. Study on Spatial Prediction of Soil Nutrient Distribution in Field Plots of Black Soil Region. Northeast Agricultural University, Harbin(in Chinese with English abstract). Yang, Y. C., Zhao, Y. J., Qin, K., et al., 2019. Prediction of Black Soil Nutrient Content Based on Airborne Hyperspectral Remote Sensing. Transactions of the Chinese Society of Agricultural Engineering, 35(20): 94-101(in Chinese with English abstract). Zhang, J. J., Xi, L., Yang, X. Y., et al., 2020. Construction of Hyperspectral Estimation Model for Organic Matter Content in Shajiang Black Soil. Transactions of the Chinese Society of Agricultural Engineering, 36(17): 135-141(in Chinese with English abstract). Zheng, M., Wang, X., Li, S. J., et al., 2022. Remote Sensing Inversion of Soil Organic Matter and Total Nitrogen in Black Soil Region. Scientia Geographica Sinica, 42(8): 1336-1347(in Chinese with English abstract). Zhu, A. X., Yang, L., Fan, N. Q., et al., 2018. The Review and Outlook of Digital Soil Mapping. Progress in Geography, 37(1): 66-78(in Chinese with English abstract). 邓昀, 牛照文, 冯琦尧, 等, 2023. 改进时间卷积网络的红壤有机质高光谱预测模型. 光谱学与光谱分析, 43(9): 2942-2951. 第五鹏瑶, 卞希慧, 王姿方, 等, 2019. 光谱预处理方法选择研究. 光谱学与光谱分析, 39(9): 2800. 李雪萍, 张飞, 王小平, 2019. 微分算法的艾比湖湿地自然保护区土壤有机质多光谱建模. 光谱学与光谱分析, 39(2): 535-542. 刘焕军, 张柏, 赵军, 等, 2007. 黑土有机质含量高光谱模型研究. 土壤学报, 44(1): 27-32. 潘宁, 王帅, 刘焱序, 等, 2019. 土壤水分遥感反演研究进展. 生态学报, 39(13): 4615-4626. 史舟, 徐冬云, 滕洪芬, 等, 2018. 土壤星地传感技术现状与发展趋势. 地理科学进展, 37(1): 79-92. 田尉霞, 2019. 基于统计模型的黑土有机质高光谱定量反演(博士学位论文). 成都: 成都理工大学. 汪大明, 秦凯, 李志忠, 等, 2018. 基于航空高光谱遥感数据的黑土地有机质含量反演: 以黑龙江省建三江地区为例. 地球科学, 43(6): 2184-2194. doi: 10.3799/dqkx.2018.612 王亚丹, 张凤, 胡文友, 等, 2024. 基于数字图像技术反演中国东北黑土有机质含量. 土壤, 56(5): 1051-1056. 肖艳, 辛洪波, 王斌, 等, 2021. 基于小波变换和连续投影算法的黑土有机质含量高光谱估测. 国土资源遥感, 33(2): 33-39. 谢雅慧, 2019. 黑土区田块土壤养分空间分布预测研究(博士学位论文). 哈尔滨: 东北农业大学. 杨越超, 赵英俊, 秦凯, 等, 2019. 黑土养分含量的航空高光谱遥感预测. 农业工程学报, 35(20): 94-101. 张娟娟, 席磊, 杨向阳, 等, 2020. 砂姜黑土有机质含量高光谱估测模型构建. 农业工程学报, 36(17): 135-141. 郑淼, 王翔, 李思佳, 等, 2022. 黑土区土壤有机质和全氮含量遥感反演研究. 地理科学, 42(8): 1336-1347. 朱阿兴, 杨琳, 樊乃卿, 等, 2018. 数字土壤制图研究综述与展望. 地理科学进展, 37(1): 66-78. -
点击查看大图
图(7) / 表(6)
计量
- 文章访问数: 1104
- HTML全文浏览量: 478
- PDF下载量: 76
- 被引次数: 0
