干旱区绿洲耕层土壤重金属铬含量的高光谱估测

doi:10.16258/j.cnki.1674-5906.2021.10.014

生态环境学报 ›› 2021, Vol. 30 ›› Issue (10): 2076-2084.DOI: 10.16258/j.cnki.1674-5906.2021.10.014

干旱区绿洲耕层土壤重金属铬含量的高光谱估测

王雪梅¹^,²(), 玉米提∙买明¹, 毛东雷¹^,², 梁婷³

1.新疆师范大学地理科学与旅游学院,新疆乌鲁木齐 830054
2.新疆维吾尔自治区重点实验室（新疆干旱区湖泊环境与资源实验室）,新疆乌鲁木齐 830054
3.新疆师范大学科研处,新疆乌鲁木齐 830054

收稿日期:2021-05-03 出版日期:2021-10-18 发布日期:2021-12-21
作者简介:王雪梅（1976年生）,女,教授,博士,硕士研究生导师,研究方向为干旱区资源环境遥感技术应用研究。E-mail: wangxm_1225@sina.com
基金资助:
国家自然科学基金项目(41561051);新疆维吾尔自治区自然科学基金项目(2020D01A79)

Hyperspectral Estimation of Heavy Metal Chromium Content in Arable Soil of Arid Area Oasis

WANG Xuemei¹^,²(), YUMITI∙Maiming ¹, MAO Donglei¹^,², LIANG Ting³

1. College of Geographic Science and Tourism of Xinjiang Normal University, Urumqi 830054, China
2. Xinjiang Uygur Autonomous Region Key Laboratory (Xinjiang Arid Lake Environment and Resources Laboratory), Urumqi 830054, China
3. Scientific Research Department of Xinjiang Normal University, Urumqi 830054, China

Received:2021-05-03 Online:2021-10-18 Published:2021-12-21

摘要/Abstract

摘要：

土壤重金属铬含量的高光谱估测技术较传统测量方法具有无污染、方便快捷进行动态监测的优势。对新疆渭干河-库车河三角洲绿洲耕层土壤98个样品的原始光谱反射率R分别进行倒数1/R、对数lg(R)、倒数对数lg(1/R)以及微分变换。不同处理结果与实测土壤重金属铬含量进行相关分析从而筛选出具有极显著相关的特征波段（P<0.001）。以不同变换处理下的特征波段反射率作为自变量,土壤铬含量为因变量,采用多元线性逐步回归、偏最小二乘回归、BP神经网络和随机森林回归方法构建土壤重金属铬含量的高光谱估测模型,并对最优估测结果进行克里格空间插值。结果表明,（1）原始光谱反射率的微分变换处理可有效提升光谱与土壤重金属铬含量之间的敏感性,其中经微分变换后的土壤光谱反射率与铬含量的相关系数由0.487显著提高到0.669（P<0.001）。（2）综合比较各模型的训练集和验证集估测结果,经倒数对数一阶微分[lg(1/R)]′处理后的BP神经网络模型具有较高的估测精度和很强的稳定性,可作为研究区土壤重金属铬含量的最优估测模型,其决定系数（R_d²）在0.8以上,均方根误差（RMSE）小于6.5,相对分析误差（RPD）大于2。（3）研究区土壤重金属铬含量具有中等空间变异性,低含量区域主要分布在新和县和沙雅县的外缘地带,而位于库车市东北部区域的耕层土壤铬含量达到最高水平。受人类活动影响该绿洲的土壤污染问题日趋严重,耕层土壤重金属铬含量呈现出较高含量的空间分布。

关键词: 土壤铬, 高光谱估测, 特征波段, 多元线性逐步回归, 偏最小二乘回归, BP神经网络, 随机森林回归

Abstract:

Compared with the traditional surveying method, the hyperspectral method for estimating the content of heavy metal chromium in soil has the advantages of non-pollution, convenient and rapid dynamic monitoring. The inverse 1/R, logarithm lg(R), logarithm lg(1/R) and differential transformation of the original spectral reflectance R of 98 soil samples from the delta oasis of Weigan-Kuqa River in Xinjiang were carried out respectively. Correlation analysis was conducted between the results of different treatments and the measured soil heavy metal of chromium content, so as to screen out the characteristic bands with extremely significant correlation (P<0.001). With the characteristics of band reflectance under different transformation processing as the independent variable, the soil chromium content as the dependent variable, using multiple linear stepwise regression, partial least-squares regression and BP neural network and regression method to build the random forest soil heavy metal chromium content of high spectral estimation model, and conduct the optimal estimate result Kriging interpolation space. The results showed that: (1) Differential transformation of original spectral reflectance could effectively improve the sensitivity between spectral reflectance and soil heavy metal chromium content, and the correlation coefficient between spectral reflectance and soil heavy metal chromium content after differential transformation was significantly increased from 0.487 to 0.669 (P<0.001). (2) Comprehensively compare the estimation results of the training set and the validation set of each model, the inverse logarithms first-order differential [lg(1/R)]' after being processed of BP neural network model has high estimation precision and strong stability, can be used as the optimal estimate model of soil heavy metal chromium content in the study area, the decision coefficient (R_d²) is over 0.8, the root mean square error (RMSE) is less than 6.5 and residual predictive deviation (RPD) is greater than 2. (3) Soil heavy metal chromium content in the study area is showed moderate spatial variability. The low-content areas were mainly distributed in the outer edge of Xinhe County and Shaya County, while the topsoil chromium content in the northeastern area of Kuqa City reached the highest level. Due to the influence of human activities, the soil pollution of the oasis has become more and more serious, and the content of heavy metal chromium in surface soil of farmland shows a relatively high spatial distribution.

Key words: soil chromium, hyperspectral estimation, characteristic band, multiple linear stepwise regression, partial least squares regression, BP neural network, random forest regression

中图分类号:

X830
P237

王雪梅, 玉米提∙买明, 毛东雷, 梁婷. 干旱区绿洲耕层土壤重金属铬含量的高光谱估测[J]. 生态环境学报, 2021, 30(10): 2076-2084.

WANG Xuemei, YUMITI∙Maiming , MAO Donglei, LIANG Ting. Hyperspectral Estimation of Heavy Metal Chromium Content in Arable Soil of Arid Area Oasis[J]. Ecology and Environment, 2021, 30(10): 2076-2084.

图/表 9

参考文献 20

[1]	KHOSRAVI V, ARDEJANI F D, GHOLIZADEH A, et al., 2021. Satellite imagery for monitoring and mapping soil Chromium pollution in a mine waste dump[J]. Remote Sensing, 13(7):1277. DOI URL
[2]	LI Z Y, MA Z W, VAN DER KUIJP T J, et al., 2014. A review of pollution from mines in China: pollution and health risk assessment[J]. Science of the Total Environment, 468- 469C:843-853.
[3]	LIU K, ZHAO D, FANG J, et al., 2017. Estimation of heavy metal contamination in soil using remote sensing spectroscopy and a statistical approach[J]. Journal of the Indian Society of Remote Sensing, 45(5):805-813. DOI URL
[4]	RINKLEBE J, ANTONIADIS V, SHAHEEN S M, et al., 2019. Health risk assessment of potentially Toxic elements in soils along the central Elbe river, Germany[J]. Environment International, 126:76-88. DOI URL
[5]	TAN K, MA W B, WU F Y, et al., 2019. Random forest-based estimation of heavy metal concentration in agricultural soils with hyperspectral sensor data[J]. Environmental Monitoring and Assessment, 191(7):1-14.
[6]	TAN K, WANG H M, CHEN L H, et al., 2020. Estimation of the spatial distribution of heavy metal in agricultural soils using airborne hyperspectral imaging and random forest[J]. Journal of Hazardous Materials, DOI: 10.1016/j.jhazmat.2019.120987. DOI
[7]	WANG X, AN S, XU Y Q, et al., 2020. A back propagation neural network model optimized by mind evolutionary algorithm for estimating Cd, Cr, and Pb concentrations in soils using Vis-NIR diffuse reflectance spectroscopy[J]. Applied Sciences, 10(1):51. DOI URL
[8]	陈宇波, 薛云, 邹滨, 等, 2020. 有色金属矿区土壤铬污染遥感反演研究[J]. 中南大学学报(自然科学版), 51(10):2876-2884.
	CHEN Y B, XUE Y, ZOU B, et al., 2020. Research on remote sensing retrieval of soil Chromium pollution in nonferrous metal mining area[J]. Journal of Central South University (Science and Technology), 51(10):2876-2884.
[9]	陈元鹏, 张世文, 罗明, 等, 2019. 基于高光谱反演的复垦区土壤重金属含量经验模型优选[J]. 农业机械学报, 50(1):170-179.
	CHEN Y P, ZHANG S W, LUO M, et al., 2019. Empirical model optimization of hyperspectral inversion of heavy metal content in reclamation area[J]. Transactions of the Chinese Society for Agricultural Machinery, 50(1):170-179.
[10]	贺军亮, 崔军丽, 李仁杰, 2019. 土壤重金属铬的高光谱估算模型[J]. 实验室研究与探索, 38(7):8-11, 61.
	HE J L, CUI J L, LI R J, 2019. Hyperspectral estimation model of heavy metal Cr in soil[J]. Research and Exploration in Laboratory, 38(7):8-11, 61.
[11]	黄晔, 王雪梅, 2018. 塔里木盆地北缘绿洲-荒漠过渡带典型植物群落物种多样性分析[J]. 西南农业学报, 31(3):605-610.
	HUANG Y, WANG X M, 2018. Analysis of species diversity of typical plant communities in oasis-desert transition zone of Tarim basin northern margin[J]. Southwest China Journal of Agricultural Sciences, 31(3):605-610.
[12]	侯一峰, 汪洋, 张帅, 2020. 新疆渭干-库车河流域土地利用变化空间耦合及其生态效应[J]. 灌溉排水学报, 39(6):11-18.
	HOU Y F, WANG Y, ZHANG S, 2020. Change in land use and its consequence for ecological systems in Weigan-Kuqa river basin in Xinjiang[J]. Journal of Irrigation and Drainage, 39(6):11-18.
[13]	路杰晖, 李西灿, 王凤华, 2018. 土壤重金属铬含量高光谱灰色关联度估测[J]. 测绘科学技术学报, 35(5):508-512, 517.
	LU J H, LI X C, WANG F H, 2018. Hyperspectral grey relational estimation of heavy metal Chromium content in soil[J]. Journal of Geomatics Science and Technology, 35(5):508-512, 517.
[14]	刘彦平, 罗晴, 程和发, 2020. 高光谱遥感技术在土壤重金属含量测定领域的应用与发展[J]. 农业环境科学学报, 39(12):2699-2709.
	LIU Y P, LUO Q, CHENG H F, 2020. Application and development of hyperspectral remote sensing technology to determine the heavy metal content in soil[J]. Journal of Agro-Environment Science, 39(12):2699-2709.
[15]	王涛, 喻彩丽, 姚娜, 等, 2018. MLR和PLSR的沙壤土盐分含量光谱检测对比研究[J]. 干旱区地理, 41(6):1295-1302.
	WANG T, YU C L, YAO N, et al., 2018. A comparison of the salt content in sandy soil between the MLR model and PLSR model[J]. Arid Land Geography, 41(6):1295-1302.
[16]	中华人民共和国生态环境部, 2018. 土壤环境质量农用地土壤污染风险管控标准 (试行): GB15618—2018 [S]. 北京: 中国环境出版集团: 1-4.
	Environment of the People’s Republic of China, 2018. Soil environmental quality agricultural land soil pollution risk control standards (Trial): GB 15618—2018 [S]. Beijing: China Environment Publishing Group: 1-4.
[17]	张明月, 张奇栎, 王璐, 等, 2019. 东北黑土区土壤铬含量高光谱反演研究[J]. 遥感技术与应用, 34(2):313-322.
	ZHANG M Y, ZHANG Q L, WANG L, et al., 2019. Researh on Chromium retrieval of black soil with hyperspectral imagery in northeast of China[J]. Remote Sensing Technology and Application, 34(2):313-322.
[18]	章琼, 郭云开, 钱佳, 等, 2020. 像元二分模型的土壤重金属Cr的遥感反演[J]. 测绘科学, 45(4):58-63.
	ZHANG Q, GUO Y K, QIAN J, et al., 2020. Remote sensing inversion of soil heavy metal Cr based on dimidiate pixel model[J]. Science of Surveying and Mapping, 45(4):58-63.
[19]	张贤龙, 张飞, 张海威, 等, 2018. 基于光谱变换的高光谱指数土壤盐分反演模型优选[J]. 农业工程学报, 34(1):110-117.
	ZHANG X L, ZHANG F, ZHANG H W, et al., 2018. Optimization of soil salt inversion model based on spectral transformation from hyperspectral index[J]. Transactions of the Chinese Society of Agricultural Engineering, 34(1):110-117.
[20]	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.
	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.

样品类型 Type of samples	样品数 Number of samples	土壤重金属铬质量分数 w_{(Cr in soil)}/(mg∙kg^-1)					偏度 SKEW	峰度 KURT	变异系数 Coefficient of variation/%
样品类型 Type of samples	样品数 Number of samples	最小值 Min value	最大值 Max value	中位值 Medium value	平均值 Mean value	标准差 Standard deviation	偏度 SKEW	峰度 KURT	变异系数 Coefficient of variation/%
总体样品 All samples	98	15	86	51	52.2	13.31	0.34	0.33	25.48
训练集 Training set	70	15	82	51	51.9	12.65	0.30	0.68	24.37
验证集 Validation set	28	22	86	50.5	53.1	15.06	0.38	-0.17	28.38

样品类型 Type of samples	样品数 Number of samples	土壤重金属铬质量分数 w_{(Cr in soil)}/(mg∙kg^-1)					偏度 SKEW	峰度 KURT	变异系数 Coefficient of variation/%
样品类型 Type of samples	样品数 Number of samples	最小值 Min value	最大值 Max value	中位值 Medium value	平均值 Mean value	标准差 Standard deviation	偏度 SKEW	峰度 KURT	变异系数 Coefficient of variation/%
总体样品 All samples	98	15	86	51	52.2	13.31	0.34	0.33	25.48
训练集 Training set	70	15	82	51	51.9	12.65	0.30	0.68	24.37
验证集 Validation set	28	22	86	50.5	53.1	15.06	0.38	-0.17	28.38

光谱变换 Spectrum transform	特征波段数 Number of bands	特征波段分布范围 Band range	相关系数取值范围 Range of correlation coefficient	显著水平 Significant level
R	217	350-566 nm	[-0.464, -0.329]	P<0.001
1/R	216	350-565 nm	[0.329, 0.487]	P<0.001
lg(R)	216	350-565 nm	[-0.476, -0.333]	P<0.001
lg(1/R)	216	350-565 nm	[0.333, 0.476]	P<0.001
R′	213	351-762, 807-900, 1255-2349 nm	[-0.587, -0.328], [0.327, 0.562]	P<0.001
R′′	95	397-764, 772-889, 1235-2389 nm	[-0.564, -0.327], [0.328, 0.630]	P<0.001
(1/R)′	254	359-762, 807-990, 1254-2349 nm	[-0.627, -0.327], [0.330, 0.429]	P<0.001
(1/R)′′	98	382-762, 772-889, 1132-2389 nm	[-0.575, -0.327], [0.327, 0.669]	P<0.001
[lg(R)]′	224	356-762, 807-990, 1254-2349 nm	[-0.523, -0.328], [0.327, 0.620]	P<0.001
[lg(R)]′′	94	382-762, 772-889, 1235-2389 nm	[-0.647, -0.330], [0.327, 0.610]	P<0.001
[lg(1/R)]′	224	356-762, 807-990, 1254-2349 nm	[-0.620, -0.327], [0.328, 0.523]	P<0.001
[lg(1/R)]′′	94	382-762, 772-889, 1235-2389 nm	[-0.610, -0.327], [0.331, 0.647]	P<0.001

光谱变换 Spectrum transform	特征波段数 Number of bands	特征波段分布范围 Band range	相关系数取值范围 Range of correlation coefficient	显著水平 Significant level
R	217	350-566 nm	[-0.464, -0.329]	P<0.001
1/R	216	350-565 nm	[0.329, 0.487]	P<0.001
lg(R)	216	350-565 nm	[-0.476, -0.333]	P<0.001
lg(1/R)	216	350-565 nm	[0.333, 0.476]	P<0.001
R′	213	351-762, 807-900, 1255-2349 nm	[-0.587, -0.328], [0.327, 0.562]	P<0.001
R′′	95	397-764, 772-889, 1235-2389 nm	[-0.564, -0.327], [0.328, 0.630]	P<0.001
(1/R)′	254	359-762, 807-990, 1254-2349 nm	[-0.627, -0.327], [0.330, 0.429]	P<0.001
(1/R)′′	98	382-762, 772-889, 1132-2389 nm	[-0.575, -0.327], [0.327, 0.669]	P<0.001
[lg(R)]′	224	356-762, 807-990, 1254-2349 nm	[-0.523, -0.328], [0.327, 0.620]	P<0.001
[lg(R)]′′	94	382-762, 772-889, 1235-2389 nm	[-0.647, -0.330], [0.327, 0.610]	P<0.001
[lg(1/R)]′	224	356-762, 807-990, 1254-2349 nm	[-0.620, -0.327], [0.328, 0.523]	P<0.001
[lg(1/R)]′′	94	382-762, 772-889, 1235-2389 nm	[-0.610, -0.327], [0.331, 0.647]	P<0.001

模型 Model		训练集 Training set			验证集 Validation set
模型 Model		R_d²	RMSE	RPD	R_d²	RMSE	RPD
MLSR	R′	0.641	7.815	1.62	0.722	8.603	1.75
	R′′	0.632	7.968	1.59	0.700	9.140	1.65
	(1/R)′	0.857	5.173	2.45	0.807	8.331	1.81
	(1/R)′′	0.78	6.419	1.97	0.564	12.528	1.20
	[lg(R)]′	0.724	7.009	1.80	0.826	7.307	2.06
	[lg(R)]′′	0.633	7.564	1.67	0.548	10.970	1.37
	[lg(1/R)]′	0.724	7.012	1.80	0.886	7.309	2.06
	[lg(1/R)]′′	0.662	7.572	1.67	0.549	10.954	1.37
PLSR	R′	0.758	6.185	2.05	0.714	8.045	1.87
	R′′	0.653	7.396	1.71	0.614	9.326	1.61
	(1/R)′	0.773	5.987	2.11	0.652	8.936	1.69
	(1/R)′′	0.587	8.076	1.57	0.528	10.754	1.40
	[lg(R)]′	0.865	4.614	2.74	0.734	7.842	1.92
	[lg(R)]′′	0.643	7.501	1.69	0.505	10.769	1.40
	[lg(1/R)]′	0.787	5.793	2.18	0.760	7.552	1.99
	[lg(1/R)]′′	0.643	7.503	1.69	0.504	10.771	1.40
BPNN	R′	0.805	5.620	2.25	0.670	9.041	1.67
	R′′	0.722	6.694	1.89	0.595	9.487	1.59
	(1/R)′	0.872	4.948	2.56	0.678	9.061	1.66
	(1/R)′′	0.708	7.332	1.73	0.600	12.481	1.21
	[lg(R)]′	0.891	5.001	2.53	0.710	8.895	1.69
	[lg(R)]′′	0.624	8.056	1.57	0.648	10.999	1.37
	[lg(1/R)]′	0.822	5.652	2.24	0.852	6.399	2.35
	[lg(1/R)]′′	0.610	7.932	1.59	0.700	8.332	1.81
RFR	R′	0.862	5.522	2.29	0.556	10.232	1.47
	R′′	0.854	5.908	2.14	0.599	10.159	1.48
	(1/R)′	0.873	5.160	2.45	0.747	8.846	1.70
	(1/R)′′	0.851	5.697	2.22	0.537	10.424	1.44
	[lg(R)]′	0.867	5.396	2.34	0.657	9.480	1.59
	[lg(R)]′′	0.894	5.247	2.41	0.611	9.821	1.53
	[lg(1/R)]′	0.887	5.212	2.43	0.673	9.410	1.60
	[lg(1/R)]′′	0.873	5.432	2.33	0.579	10.108	1.49

干旱区绿洲耕层土壤重金属铬含量的高光谱估测

Hyperspectral Estimation of Heavy Metal Chromium Content in Arable Soil of Arid Area Oasis

RichHTML

PDF

可视化

摘要/Abstract

引用本文

使用本文

图/表 9

参考文献 20

相关文章 1

编辑推荐

Metrics

本文评价