高光谱敏感波段筛选与机器学习协同提升土壤重金属预测精度

doi:10.16258/j.cnki.1674-5906.2025.06.012

生态环境学报 ›› 2025, Vol. 34 ›› Issue (6): 950-960.DOI: 10.16258/j.cnki.1674-5906.2025.06.012

• 研究论文【环境科学】 • 上一篇下一篇

高光谱敏感波段筛选与机器学习协同提升土壤重金属预测精度

孟畅¹^,²(), 红梅¹^,²^,^*(), 李斐¹^,²^,^*()

1.内蒙古自治区土壤质量与养分资源重点实验室，内蒙古呼和浩特 010018
2.内蒙古土壤质量与养分资源重点实验室/农业生态安全与绿色发展内蒙古高等学校重点实验室，内蒙古呼和浩特 010018

收稿日期:2024-11-14 出版日期:2025-06-18 发布日期:2025-06-11
通讯作者: * 李斐, E-mail: Lifei@imau.edu.cn；红梅, E-mail: nmczhm1970@126.com
作者简介:孟畅（1997年生），女，硕士研究生，主要研究方向为农业遥感。E-mail: 2021202040012@emails.imau.edu.cn
基金资助:
国家自然科学基金项目(M2042002);国家重点研发计划项目(2022YFD1500901);内蒙古自治区自然科学基金项目(2021ZD10)

Collaborative Enhancement of Soil Heavy Metal Prediction Accuracy Using Hyperspectral Sensitive Band Selection and Machine Learning

MENG Chang¹^,²(), HONG Mei¹^,²^,^*(), LI Fei¹^,²^,^*()

1. Inner Mongolia Key Laboratory of Soil Quality and Nutrient Resources, Hohhot 010018, P. R. China
2. Inner Mongolia Key Laboratory of Soil Quality and Nutrient Resources/Agricultural Ecological Safety and Green Development of Inner Mongolia University Key Experiment, Hohhot 010018, P. R. China

Received:2024-11-14 Online:2025-06-18 Published:2025-06-11

摘要/Abstract

摘要：

为探究土壤重金属有效波段提取方法，明确敏感波段耦合机器学习模型对土壤重金属浓度的估测潜力，以内蒙古多个废弃尾矿区周边典型污染场地为研究对象，通过高光谱遥感数据预测土壤Cu、Zn、Pb和Cr重金属的浓度。基于16种敏感波段提取方法（按过滤法、包裹法、嵌入法分类）并结合决策树（DT）、随机森林（RF）和梯度决策树（GBDT）模型，进行重金属浓度反演。结果表明，相比过滤法和嵌入法，包裹法提取的敏感波段对重金属浓度的解释性最高，敏感波段主要集中在450－750 nm和1829－2493 nm。在6种包裹法中，竞争自适应重加权抽样法（CARS）和可变迭代空间收缩法（VISSA）分别为Cu和Cr提供了关键光谱信息，而连续投影算法（SPA）则对Zn和Pb具有较高敏感度。相比DT和RF模型，GBDT在聚焦敏感波段时表现出更强大的拟合性能，耦合CARS、VISSA和SPA方法能更准确地估测土壤重金属浓度。利用独立矿区验证时，CARS、VISSA和SPA组合GBDT模型对土壤重金属浓度仍具有稳定的估测性能，Cu、Zn、Pb和Cr的决定系数（R²）分别为0.91、0.89、0.87和0.84。该研究构建的土壤重金属监测模型能有效增强土壤光谱信息可解释性，为矿区土壤重金属的快速监测提供了具有实际应用潜力的新方法。

关键词: 高光谱反演, 特征波段提取, 尾矿区, 土壤重金属, 决策树算法, 独立验证

Abstract:

This study aimed to investigate effective methods for extracting sensitive spectral bands and to assess potentially sensitive bands using machine learning models to estimate soil heavy metal concentrations. Hyperspectral remote sensing data were used to predict the concentrations of Cu, Zn, Pb, and Cr in soils from polluted sites near abandoned tailings in Inner Mongolia. Sixteen sensitive band extraction methods are classified into filtering, wrapper, and embedded approaches. These were coupled with the decision tree (DT), random forest (RF), and gradient boosting decision tree (GBDT) models for concentration inversion. Compared with the filtering and embedded methods, the results demonstrated that the wrapper method provided the highest explanatory power for heavy metal concentrations with sensitive bands, primarily in the 450-750 nm and 1829-2493 nm ranges. Among the six wrapper methods, Competitive Adaptive Reweighted Sampling (CARS) and Variable Iterative Space Shrinking Approach (VISSA) offered critical spectral information for Cu and Cr. At the same time, the successive projections algorithm (SPA) exhibited high sensitivity for Zn and Pb. Compared to the DT and RF models, the GBDT showed a superior fitting performance. In particular, in the sensitive bands, the combination of GBDT with CARS, VISSA, and SPA yielded more accurate estimates of the soil heavy metal concentrations. The results about independent validation from mining areas confirms that the combination of CARS, VISSA, and SPA with the GBDT model maintained stable estimation performance, with coefficients of determination (R²) of 0.91, 0.89, 0.87, and 0.84 for Cu, Zn, Pb, and Cr, respectively. The soil heavy metal monitoring model developed in this study enhances the interpretability of soil spectral data, offering a novel method with significant practical potential for rapid monitoring of heavy metals in mining-affected soils.

Key words: hyperspectral inversion, feature band extraction, tailings area, soil heavy metals, decision tree algorithm, independent validation

中图分类号:

X132
S151.9

孟畅, 红梅, 李斐. 高光谱敏感波段筛选与机器学习协同提升土壤重金属预测精度[J]. 生态环境学报, 2025, 34(6): 950-960.

MENG Chang, HONG Mei, LI Fei. Collaborative Enhancement of Soil Heavy Metal Prediction Accuracy Using Hyperspectral Sensitive Band Selection and Machine Learning[J]. Ecology and Environmental Sciences, 2025, 34(6): 950-960.

图/表 10

图1 研究区概况图

Figure 1 Overview of the studied area

表1 波段挑选方法

Table 1 Band selection method

类别	名称	缩略词	参考
过滤法	相关分析	CA	Liu et al.，2022
	互信息系数	MI	Zhou et al.，2022
	相关特征	RELIEF	Li et al.，2020
	最大信息系数	MIC	Liu et al.，2022
	最小冗余	MRMR	Gu et al.，2022
包裹法	K选择	SKB	Liu et al.，2005
	可变迭代空间收缩法	VISSA	Zhang et al.，2022
	连续投影算法	SPA	Mei et al.，2023
	遗传算法	GA	Rostami et al.，2021
	竞争性自适应重加权抽样	CARS	Mei et al.，2023
	无信息变量消除	UVE	Song et al.，2020
嵌入法	随即森林重要度	RFI	Yang et al.，2021
	逐步多元线性回归	SMLR	Liu et al.，2022
	偏最小二乘回归-VIP	PLSR-VIP	Yang et al.，2023
	岭回归	RR	Malik et al.，2022
	套索回归	LR	Tibshirani，1996

图2 模型构建流程图

Figure 2 Model construction flowchart

表2 研究区重金属质量分数描述性统计

Table 2 Descriptive statistics of heavy metal mass fraction in the studied area

区域与元素		变异系数/%	重金属质量分数/(mg∙kg⁻¹)
区域与元素		变异系数/%	最大值	最小值	平均值	偏斜率	内蒙古土壤背景值	中国土壤背景值
区域a	Cu	1.39	1617.45	4.82	218.47	2.28	22.91	100.00
	Zn	1.34	354.66	24.81	32.37	1.72	48.60	300.00
	Cr	1.29	193.08	12.61	19.97	1.20	68.20	350.00
	Pb	1.39	559.92	1.75	64.92	2.11	34.20	250.00
区域b	Cu	0.75	432.00	3.77	126.68	1.33	－	－
	Zn	0.30	123.56	8.46	68.87	0.61	－	－
	Cr	0.14	87.92	36.43	66.48	0.76	－	－
	Pb	1.21	715.84	4.62	77.74	2.49	－	－

图3 原始和预处理土壤光谱的反射率曲线

Figure 3 Reflectance curves of raw and pre-treated soil spectra

图4 16种特征波段挑选方法选择的土壤重金属特征波段

Figure 4 Characteristic bands of soil heavy metals selected by 16 characteristic bands selection methods

图5 敏感波段结合机器学习模型估测土壤中重金属浓度的交叉验证结果

Figure 5 Cross-validation results of sensitive bands combined with machine learning models to estimate heavy metal concentrations in soil

图6 最佳敏感波段方法结合GBDT模型估测土壤中重金属浓度的独立验证结果

Figure 6 Independent verification results of the optimal sensitive band method combined with the GBDT model to estimate heavy metal concentration in soil

图7 Cu、Zn、Pb和Cr浓度的相关性分析

Figure 7 Correlation analysis of Cu, Zn, Pb, and Cr concentrations

图8 过滤法、包裹法和嵌入法所选择的土壤重金属敏感波段频率

Figure 8 Frequency of soil heavy metal sensitive bands selected by filtration, encapsulation, and embedding methods

参考文献 45

[1]	BESSADOK A, BEN A S, 2019. Feature selection for hyperspectral data using mutual information and MRMR methods[J]. Remote Sensing, 11(8): 1932-1945.
[2]	CHEN Z L, YIN W Y, LIU H T, et al., 2017. Review of monitoring petroleum-hydrocarbon contaminated soils with visible and near-Infrared spectroscopy[J]. Spectroscopy and Spectral Analysis, 37(6): 1723-1727.
[3]	GOMEZ C, LAGACHERIE P, COULOUM G, 2008. Continuum removal versus PLSR method for clay and calcium carbonate content estimation from laboratory and airborne hyperspectral measurements[J]. Geoderma, 148(2): 141-148.
[4]	GU X Y, GUO J C, XIAO L J, et al., 2022. Conditional mutual information-based feature selection algorithm for maximal relevance minimal redundancy[J]. Applied Intelligence, 52(2): 1436-1447.
[5]	HUANG Y, WANG L Y, WANG W J, et al., 2019. Current status of agricultural soil pollution by heavy metals in China: A meta-analysis[J]. Science of the Total Environment, 651(Part 2): 3034-3042.
[6]	JIN J M, 2021. A comparison of a gradient boosting decision tree, random forests, and artificial neural networks to model urban land use changes: The case of the seoul metropolitan area[J]. International Journal of Geographical Information Science, 35(4): 1-19.
[7]	LI B, XU X M, ZHANG L, et al., 2020. Above-ground biomass estimation and yield prediction in potato by using UAV-based RGB and hyperspectral imaging[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 162: 161-172.
[8]	LIU H, YU L, 2005. Toward integrating feature selection algorithms for classification and clustering[J]. IEEE Transactions on Knowledge and Data Engineering, 17(4): 491-502.
[9]	LIU Y, FENG H K, YUE J B, et al., 2022. Remote-sensing estimation of potato above-ground biomass based on spectral and spatial features extracted from high-definition digital camera images[J]. Computers and Electronics in Agriculture, 198: 107089.
[10]	LU Q, WANG S J, BAI X Y, et al., 2019. Rapid inversion of heavy metal concentration in karst grain producing areas based on hyperspectral bands associated with soil components[J]. Microchemical Journal, 148: 404-411.
[11]	MALIK A, JAMEI M, ALI M, et al., 2022. Multi-step daily forecasting of reference evapotranspiration for different climates of India: A modern multivariate complementary technique reinforced with ridge regression feature selection[J]. Agricultural Water Management, 272: 107812.
[12]	MEI Y S, JIN Y Z, XIAO H Y, et al., 2023. A spectral decomposition method for estimating the leaf nitrogen status of maize by UAV-based hyperspectral imaging[J]. Computers and Electronics in Agriculture, 212: 108100.
[13]	PARVEEN S, ALI K, VINOD K, 2019. Comparison of different approaches for modeling of heavy metal estimations[J]. SN Applied Sciences, 25(1): 780.
[14]	PENG Y P, WANG L, ZHAO L, et al., 2021. Estimation of soil nutrient content using hyperspectral data[J]. Agriculture, 11(11): 1129.
[15]	RIVERA J L, BONILLA C A, 2020. Predicting soil aggregate stability using readily available soil properties and machine learning techniques[J]. Catena, 187: 104408.
[16]	RONG G Z, ALU S, LI K W, et al., 2020. Rainfall lnduced landslide susceptibility mapping based on bayesian optimized random forest and gradient boosting decision tree models: A case study of Shuicheng county, China[J]. Water, 12(11): 3066.
[17]	ROSTAMI M, BERAHMAND K, FOROUZANDE S, 2021. A novel community detection based genetic algorithm for feature selection[J]. Journal of Big Data, 8: 2.
[18]	SHARMA P, GUPTA R, 2021. Soil heavy metal prediction using decision tree and its applications in environmental management[J]. Environmental Monitoring and Assessment, 193(2): 51.
[19]	SONG X Z, HUANG Y, TIAN K D, 2020. Near infrared spectral variable optimization by final complexity adapted models combined with uninformative variables elimination-a validation study[J]. Optik, 203: 164019.
[20]	SUN W C, LIU S, ZHANG X, et al., 2022. Estimation of soil organic matter content using selected spectral subset of hyperspectral data[J]. Geoderma, 49: 115653.
[21]	SUN W C, ZHANG X, 2017. Estimating soil zinc concentrations using reflectance spectroscopy[J]. International Journal of Applied Earth Observation and Geoinfirmation, 58: 126-133.
[22]	SWIERENGA H, WULFERT F, DE N O E, et al., 2000. Development of robust calibration models in near infra-red spectrometric applications[J]. Analytica Chimica Acta, 411(1-2): 121-131.
[23]	TIBSHIRANI R, 1996. Regression shrinkage and selection via the lasso[J]. Journal of the Royal Statistical Society. Series B: Statistical Methodology, 58(1): 267-288.
[24]	WANG Y, LI Y, ZHANG S, 2023. Soil high contribution band selection using random forest and ridge regression for improved prediction accuracy[J]. Environmental Monitoring and Assessment, 195(7): 456.
[25]	WANG Z, DING J L, ZHANG Z P, 2022. Estimation of soil organic matter in arid zones with coupled environmental variables and spectral features[J]. Sensors, 22(3): 1194.
[26]	WEI J, LI X, CHEN Y et al., 2021a. Estimation of soil heavy metal concentrations using laboratory-based hyperspectral data: A comparison of models and preprocessing methods[J]. Remote Sensing, 13(2): 234.
[27]	WEI L, YANG X, YANG Z, 2019. Application of decision tree and random forest models for soil pollution monitoring[J]. Environmental Pollution, 246: 329-338.
[28]	WEI Z, HAN Y, LL J X, et al., 2021b. Hyperspectral inversion of soil heavy metals in three-river source region based on random forest model[J]. Catena, 202: 105222.
[29]	WU L J, JIANG Q F, ZHANG Y, et al., 2022. Peroxidase activity in tomato leaf cells under salt stress based on micro-hyperspectral imaging technique[J]. Horticultrae, 8(9): 813.
[30]	XIE X, LIU H, CHEN W, 2021. Feature selection with genetic algorithm and decision tree for hyperspectral data classification[J]. IEEE Transactions on Geoscience and Remote Sensing, 59(9): 7864-7875.
[31]	YANG H B, LI F, WANG W, et al., 2021. Estimating above-ground biomass of potato using random forest and optimized hyperspectral indices[J]. Remote Sensing, 13(12): 2339.
[32]	YANG H B, YIN H, LI F, et al., 2023. Machine learning models fed with optimized spectral indices to advance crop nitrogen monitoring[J]. Field Crops Research, 293: 108844.
[33]	YANG H F, XU H, ZHONG H, 2022. Prediction of soil heavy metal concentrations in copper tailings area using hyperspectral reflectance[J]. Environmental Earth Sciences, 81(6): 183.
[34]	YANG H, ZHANG Q, 2020. Predicting soil properties and quality using GBDT based model and hyperspectral data[J]. Geoderma, 368: 114250.
[35]	YIN F, WU M M, LIU L, et al., 2021. Predicting the abundance of copper in soil using reflectance spectroscopy and GF5 hyperspectral imagery[J]. International Journal of Applied Earth Observation and Geoinformation, 102: 102420.
[36]	ZHANG C, PENG T, NAZIR M S, 2022. A novel hybrid approach based on variational heteroscedastic gaussian process regression for multi-step ahead wind speed forecasting[J]. International Journal of Electrical Power & Energy Systems, 136: 107717.
[37]	ZHANG S W, SHEN Q, NIE C J, et al., 2019. Hyperspectral inversion of heavy metal content in reclaimed soil from a mining wasteland based on different spectral transformation and modeling methods[J]. Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy, 211: 393-400.
[38]	ZHAO B, ZHANG Q, 2020. Improved stepwise regression and sequential projection algorithm for feature selection in hyperspectral image classification[J]. Journal of Applied Remote Sensing, 14(3): 335-348.
[39]	ZHAO M S, GAO Y F, LU Y Y, et al., 2022. Hyperspectral modeling of soil organic matter based on characteristic wavelength in east China[J]. Sustainability, 14(14): 8455.
[40]	ZHOU H F, WANG X Q, ZHU R R, 2022. Feature selection based on mutual information with correlation coefficient[J]. Applied Intelligence, 52: 5457-5474.
[41]	梅雪, 刘鸿雁, 吴龙华, 等, 2023. 基于HDXRF和ICP-MS的黔西北土壤重金属空间分布及影响因素研究[J]. 土壤, 55(2): 399-408.
	MEI X, LIU H Y, WU L H, et al., 2023. Spatial distribution of soil heavy metal contents and influencing factors in northwest of Guizhou based on HDXRF and ICP-MS[J]. Soils, 55(2): 399-408.
[42]	徐英杰, 张睿, 王丽, 2019. 中国土壤环境质量标准GB 1518—2018解析与应用[J]. 土壤学报, 56(2): 234-242.
	XU Y J, ZHANG R, WANG L, 2019. Analysis and application of the China soil environmental quality standard GB 1518—2018[J]. Acta Pedologica Sinica, 56(2): 234-242.
[43]	肖洁芸, 周伟, 石佩琪, 2023. 土壤重金属含量高光谱反演[J]. 生态环境学报, 32(1): 175-182. DOI
	XIAO J Y, ZHOU W, SHI P Q, 2023. Hyperspectral inversion of soil heavy metals[J]. Ecology and Environment, 32(1): 175-182.
[44]	赵玉玲, 杨楠楠, 张海霞, 等, 2020. 基于高光谱的邯郸市土壤重金属统计估算模型研究[J]. 生态环境学报, 29(4): 819-826. DOI
	ZHAO Y L, YANG N N, ZHANG H X, et al., 2020. Study on the statistical estimation model of soil heavy metals in Handan city based on hyper-spectral[J]. Ecology and Environmental Sciences, 29(4): 819-826.
[45]	魏复盛, 1990. 中国土壤元素背景值[M]. 北京: 中国环境科学出版社.
	WEI F S, 1990. Background values of soil elements in China[M]. Beijing: China Environmental Science Publishing.

高光谱敏感波段筛选与机器学习协同提升土壤重金属预测精度

Collaborative Enhancement of Soil Heavy Metal Prediction Accuracy Using Hyperspectral Sensitive Band Selection and Machine Learning

RichHTML

PDF

可视化

摘要/Abstract

引用本文

使用本文

图/表 10

参考文献 45

相关文章 10

编辑推荐

Metrics

本文评价

[1]	刘鸿林, 赵方凯, 杨磊, 沈琳钧, 杨恺丰, 李敏, 陈利顶. 城市公园土壤重金属污染及影响因素研究——以宁波市为例[J]. 生态环境学报, 2025, 34(5): 773-783.
[2]	刘东宜, 屈永华, 冯耀伟, 屈冉. 基于网格搜索优化CatBoost模型的GF-5卫星影像铬离子含量反演研究[J]. 生态环境学报, 2024, 33(9): 1460-1470.
[3]	吴文伟, 沈城, 沙晨燕, 林匡飞, 吴健, 谢雨晴, 周璇. 城市工业地块土壤重金属污染风险评价与源解析[J]. 生态环境学报, 2024, 33(5): 791-801.
[4]	唐舒娅, 王春辉, 宋靖, 李刚. 环象山港区域土壤重金属污染特征及风险评估[J]. 生态环境学报, 2024, 33(11): 1768-1781.
[5]	陈敏毅, 朱航海, 佘伟铎, 尹光彩, 黄祖照, 杨巧玲. 珠三角某遗留造船厂场地土壤重金属人体健康风险评估及源解析[J]. 生态环境学报, 2023, 32(4): 794-804.
[6]	肖洁芸, 周伟, 石佩琪. 土壤重金属含量高光谱反演[J]. 生态环境学报, 2023, 32(1): 175-182.
[7]	刘娣, 苏超, 张红, 秦冠宇. 典型煤炭产业聚集区土壤重金属污染特征与风险评价[J]. 生态环境学报, 2022, 31(2): 391-399.
[8]	邢树文, 许佳敏, 黄彬, 高锦婷, 韩丽. 钨尾矿重金属污染对茶园土壤动物群落结构及多样性的影响[J]. 生态环境学报, 2021, 30(9): 1903-1915.
[9]	他维媛, 康桢, 孟昭君, 金盛华, 杨幸, 郭龙飞, 赵东旭, 张馨. 秦岭典型停产关闭锌冶炼企业场地土壤重金属污染特征研究[J]. 生态环境学报, 2021, 30(7): 1513-1521.
[10]	茶丽娟, 周丹丹, 冯鸿娟, 赵淑媛, 冯凯萍. 两种野生食用菌对土壤重金属的富集特征研究[J]. 生态环境学报, 2021, 30(10): 2093-2099.