Message Board

Respected readers, authors and reviewers, you can add comments to this page on any questions about the contribution, review,        editing and publication of this journal. We will give you an answer as soon as possible. Thank you for your support!

Name
E-mail
Phone
Title
Content
Verification Code
Volume 47 Issue 9
Sep.  2022
Turn off MathJax
Article Contents

ZHANG Xia, DING Songtao, CEN Yi, SUN Weichao, WANG Jinnian. Soil Heavy Metal Pb Content Estimation Method by Combining Field Spectra with Laboratory Spectra[J]. Geomatics and Information Science of Wuhan University, 2022, 47(9): 1479-1485. doi: 10.13203/j.whugis20200386
Citation: ZHANG Xia, DING Songtao, CEN Yi, SUN Weichao, WANG Jinnian. Soil Heavy Metal Pb Content Estimation Method by Combining Field Spectra with Laboratory Spectra[J]. Geomatics and Information Science of Wuhan University, 2022, 47(9): 1479-1485. doi: 10.13203/j.whugis20200386

Soil Heavy Metal Pb Content Estimation Method by Combining Field Spectra with Laboratory Spectra

doi: 10.13203/j.whugis20200386
Funds:

The National Key Research and Development Program of China 2017YFB0503800

the Special Project of China High-Resolution Earth Observation System 30-H30C01-9004-19/21

More Information
  • Author Bio:

    ZHANG Xia, PhD, professor, focuses on quantitative applications of hyperspectral remote sensing. E‐mail: zhangxia@radi.ac.cn

  • Corresponding author: DING Songtao, PhD candidate. E‐mail: dingst425@163.com; WANG Jinnian, PhD, professor. E-mail: wangjn@radi.ac.cn
  • Received Date: 2020-11-15
  • Publish Date: 2022-09-05
  •   Objectives  The pollution of heavy metal has become increasingly serious in recent years. The accumulation of heavy metals in the soil will be a threat to ecological balance and human health. Therefore, we need to obtain heavy metal content in soil quickly and accurately.  Methods  This paper proposes a method to combine field and laboratory spectra to construct a mechanism estimation model of soil lead (Pb). Firstly, direct standardization (DS) algorithm was employed to eliminate the influence of environmental factors on the field spectra. Secondly, in order to enhance the diversity of the samples, the laboratory spectra were introduced to joint modeling. Finally, the characteristic spectra of iron oxide were extracted for mod‍el‍ing to increase the model rationality.  Results  This method was validated by the spectra of 70 soil samples from Xiong'an farming area in Hebei province. The accuracy R2 of model established by full-band field spectra without DS correction was only 0.220 0. However, the accuracy R2 of model established by the proposed method in this paper reached 0.914 6.  Conclusions  It indicates that the model for estimating Pb content can be significantly improved by removing the influence of environmental factors on the field spectra, extracting the iron oxide characteristic spectra of the combining field spectra with laboratory spectra.
  • [1] Wong S C, Li X D, Zhang G, et al. Heavy Metals in Agricultural Soils of the Pearl River Delta, South China[J]. Environmental Pollution, 2002, 119(1): 33-44 doi:  10.1016/S0269-7491(01)00325-6
    [2] Gmochowska W, Pietranik A, Tyszka R, et al. Sources of Pollution and Distribution of Pb, Cd and Hg in Wrocław Soils: Insight from Chemical and Pb Isotope Composition[J]. Geochemistry, 2019, 79(3): 434-445 doi:  10.1016/j.chemer.2019.07.002
    [3] Khosravi V, Doulati Ardejani F, Yousefi S, et al. Monitoring Soil Lead and Zinc Contents via Combination of Spectroscopy with Extreme Learning Machine and Other Data Mining Methods[J]. Geoderma, 2018, 318: 29-41 doi:  10.1016/j.geoderma.2017.12.025
    [4] 彭小婷, 高文秀, 王俊杰. 基于包络线去除和偏最小二乘的土壤参数光谱反演[J]. 武汉大学学报·信息科学版, 2014, 39(7): 862-866 http://ch.whu.edu.cn/article/id/3034

    Peng Xiaoting, Gao Wenxiu, Wang Junjie. Inversion of Soil Parameters from Hyperspectra Based on Continuum Removal and Partial Least Squares Regression[J]. Geomatics and Information Science of Wuhan University, 2014, 39(7): 862-866 http://ch.whu.edu.cn/article/id/3034
    [5] Guerrero C, Zornoza R, Gómez I, et al. Spiking of NIR Regional Models Using Samples from Target Sites: Effect of Model Size on Prediction Accuracy[J]. Geoderma, 2010, 158(1/2): 66-77
    [6] Ge Y F, Morgan C L S, Ackerson J P. VisNIR Spectra of Dried Ground Soils Predict Properties of Soils Scanned Moist and Intact[J]. Geoderma, 2014, 221/222: 61-69 doi:  10.1016/j.geoderma.2014.01.011
    [7] Zou B, Jiang X L, Feng H H, et al. Multisource Spectral-Integrated Estimation of Cadmium Concentrations in Soil Using a Direct Standardization and Spiking Algorithm[J]. Science of the Total Environment, 2020, 701, DOI:  10.1016/j.scitotenv.2019.134890
    [8] Zou Bin, Jiang Xiaolu, Feng Huihui, et al. Multisource Spectral-Integrated Estimation of Cadmium Concentrations Insoil Using a Direct Standardization and Spiking Algorithm[J]. Science of the Total Environment, 2020, 701: 134890 doi:  10.1016/j.scitotenv.2019.134890
    [9] 陈元鹏, 张世文, 罗明, 等. 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, 2019, 50(1): 170-179 https://www.cnki.com.cn/Article/CJFDTOTAL-NYJX201901018.htm

    Chen Yuanpeng, Zhang Shiwen, Luo Ming, et al. Empirical Model Optimization of Hyperspectral Inversion of Heavy Metal Content in Reclamation Area[J]. Transactions of the Chinese Society for Agricultural Machinery, 2019, 50(1): 170-179 https://www.cnki.com.cn/Article/CJFDTOTAL-NYJX201901018.htm
    [10] Rathod P H, Rossiter D G, Noomen M F, et al. Proximal Spectral Sensing to Monitor Phytoremediation of Metal-Contaminated Soils[J]. International Journal of Phytoremediation, 2013, 15(5): 405-426 doi:  10.1080/15226514.2012.702805
    [11] Wu Y Z, Chen J, Ji J F, et al. A Mechanism Study of Reflectance Spectroscopy for Investigating Heavy Metals in Soils[J]. Soil Science Society of America Journal, 2007, 71(3): 918-926 doi:  10.2136/sssaj2006.0285
    [12] Malley D F, Williams P C. Use of Near-Infrared Reflectance Spectroscopy in Prediction of Heavy Metals in Freshwater Sediment by Their Association with Organic Matter[J]. Environmental Science & Technology, 1997, 31(12): 3461-3467
    [13] Sun W C, Zhang X, Sun X J, et al. Predicting Nickel Concentration in Soil Using Reflectance Spectroscopy Associated with Organic Matter and Clay Miner‍als[J]. Geoderma, 2018, 327: 25-35 doi:  10.1016/j.geoderma.2018.04.019
    [14] Zhang X, Sun W C, Cen Y, et al. Predicting Cadmium Concentration in Soils Using Laboratory and Field Reflectance Spectroscopy[J]. Science of the Total Environment, 2019, 650: 321-334 doi:  10.1016/j.scitotenv.2018.08.442
    [15] Kooistra L, Salas E A L, Clevers J G P W, et al. Exploring Field Vegetation Reflectance as an Indicator of Soil Contamination in River Floodplains[J]. Environmental Pollution, 2004, 127(2): 281-290 doi:  10.1016/S0269-7491(03)00266-5
    [16] 葛云思. 土壤中铁氧化物对铅吸附特性的影响研究[D]. 上海: 华东师范大学, 2016

    Ge Yunsi. Study on the Influence of Iron Oxides on the Adsorption Characteristics of Pb in Soils[D]. Shanghai: East China Normal University, 2016
    [17] Galvão L S, Vitorello Í. Variability of Laboratory Measured Soil Lines of Soils from Southeastern Brazil[J]. Remote Sensing of Environment, 1998, 63(2): 166-181 doi:  10.1016/S0034-4257(97)00135-1
    [18] 徐彬彬, 季耿善, 朱永豪. 中国陆地背景和土壤光谱反射特性的地理分区的初步研究[J]. 环境遥感, 1991(2): 142-151 https://www.cnki.com.cn/Article/CJFDTOTAL-YGXB199102008.htm

    Xu Binbin, Ji Gengshan, Zhu Yonghao. A Preliminary Research of Geographic Regionalization of China Land Background and Spectral Reflectance Characteristicsof Soil[J]. National Remote Sensing Bulletin, 1991(2): 142-151 https://www.cnki.com.cn/Article/CJFDTOTAL-YGXB199102008.htm
    [19] Ji W, Viscarra R R A, Shi Z. Accounting for the Effects of Water and the Environment on Proximally Sensed Vis-NIR Soil Spectra and Their Calibrations[J]. European Journal of Soil Science, 2015, 66(3): 555-565 doi:  10.1111/ejss.12239
    [20] 邹滨, 涂宇龙, 姜晓璐, 等. 土壤Cd含量实验室与野外DS光谱联合反演[J]. 光谱学与光谱分析, 2019, 39(10): 3223-3231 https://www.cnki.com.cn/Article/CJFDTOTAL-GUAN201910046.htm

    Zou Bin, Tu Yulong, Jiang Xiaolu, et al. Estimation of Cd Content in Soil Using Combined Laboratory and Field DS Spectroscopy[J]. Spectroscopy and Spectral Analysis, 2019, 39(10): 3223-3231 https://www.cnki.com.cn/Article/CJFDTOTAL-GUAN201910046.htm
    [21] Leardi R, Lupiáñez G A. Genetic Algorithms Applied to Feature Selection in PLS Regression: How and When to Use Them[J]. Chemometrics and Intelligent Laboratory Systems, 1998, 41(2): 195-207 doi:  10.1016/S0169-7439(98)00051-3
    [22] Vohland M, Besold J, Hill J, et al. Comparing Different Multivariate Calibration Methods for the Determination of Soil Organic Carbon Pools with Visible to near Infrared Spectroscopy[J]. Geoderma, 2011, 166(1): 198-205 doi:  10.1016/j.geoderma.2011.08.001
    [23] Wang J J, Cui L J, Gao W X, et al. Prediction of Low Heavy Metal Concentrations in Agricultural Soils Using Visible and Near-Infrared Reflectance Spectroscopy[J]. Geoderma, 2014, 216, DOI:  10.1016/j.geoderma.2013.10.024
    [24] Saeys W, Mouazen A M, Ramon H. Potential for Onsite and Online Analysis of Pig Manure Using Visible and near Infrared Reflectance Spectroscopy[J]. Biosystems Engineering, 2005, 91(4): 393-402 doi:  10.1016/j.biosystemseng.2005.05.001
  • 加载中
通讯作者: 陈斌, bchen63@163.com
  • 1. 

    沈阳化工大学材料科学与工程学院 沈阳 110142

  1. 本站搜索
  2. 百度学术搜索
  3. 万方数据库搜索
  4. CNKI搜索

Figures(4)  / Tables(2)

Article Metrics

Article views(285) PDF downloads(25) Cited by()

Related
Proportional views

Soil Heavy Metal Pb Content Estimation Method by Combining Field Spectra with Laboratory Spectra

doi: 10.13203/j.whugis20200386
Funds:

The National Key Research and Development Program of China 2017YFB0503800

the Special Project of China High-Resolution Earth Observation System 30-H30C01-9004-19/21

Abstract:   Objectives  The pollution of heavy metal has become increasingly serious in recent years. The accumulation of heavy metals in the soil will be a threat to ecological balance and human health. Therefore, we need to obtain heavy metal content in soil quickly and accurately.  Methods  This paper proposes a method to combine field and laboratory spectra to construct a mechanism estimation model of soil lead (Pb). Firstly, direct standardization (DS) algorithm was employed to eliminate the influence of environmental factors on the field spectra. Secondly, in order to enhance the diversity of the samples, the laboratory spectra were introduced to joint modeling. Finally, the characteristic spectra of iron oxide were extracted for mod‍el‍ing to increase the model rationality.  Results  This method was validated by the spectra of 70 soil samples from Xiong'an farming area in Hebei province. The accuracy R2 of model established by full-band field spectra without DS correction was only 0.220 0. However, the accuracy R2 of model established by the proposed method in this paper reached 0.914 6.  Conclusions  It indicates that the model for estimating Pb content can be significantly improved by removing the influence of environmental factors on the field spectra, extracting the iron oxide characteristic spectra of the combining field spectra with laboratory spectra.

ZHANG Xia, DING Songtao, CEN Yi, SUN Weichao, WANG Jinnian. Soil Heavy Metal Pb Content Estimation Method by Combining Field Spectra with Laboratory Spectra[J]. Geomatics and Information Science of Wuhan University, 2022, 47(9): 1479-1485. doi: 10.13203/j.whugis20200386
Citation: ZHANG Xia, DING Songtao, CEN Yi, SUN Weichao, WANG Jinnian. Soil Heavy Metal Pb Content Estimation Method by Combining Field Spectra with Laboratory Spectra[J]. Geomatics and Information Science of Wuhan University, 2022, 47(9): 1479-1485. doi: 10.13203/j.whugis20200386
  • 近年来,土壤重金属污染已经成为一个严重的环境问题,尤其是在城市快速扩张、工业迅速发展的国家和地区[1],研究表明,中国耕地土壤重金属污染的概率约为16.67%,土壤重金属污染的面积大约占据耕地总面积的1/6。Pb是土壤中常见的一种重金属污染物,渗透进土壤中后移动性较差,残留时间长,不仅影响农作物质量,人类的健康[2],而且会影响整个生态系统的安全。因此,开展土壤重金属Pb含量调查和监测方法研究具有重要意义。高光谱遥感通过获取连续且精细的土壤反射光谱,能够体现出土壤成分的细微变化,具有低成本、大范围快速监测土壤重金属状况的潜力[3]

    迄今为止,大多数研究聚焦于实验室光谱建模,很少用到野外光谱进行建模[4]。野外光谱会受到环境因素(土壤粒径和含水量等)的影响,且由于土壤样本难以获取,样本数一般偏少,导致数据差异性不足。目前去除环境因素的光谱转换算法主要包括Spiking[5]、外部参数正交化[6](external parameter orthogonalization,EPO)算法以及直接矫正(direct standardization,DS)算法,这些环境因素去除算法在土壤有机质(碳)和粘土含量等属性参量高光谱反演方面有较好的应用,但在土壤重金属反演方面的应用还鲜有触及[7]

    现有研究大多以全谱段建模为主[8-9],未能充分利用土壤组分对土壤重金属的吸附作用机理。土壤光谱曲线上很难直接探测到土壤重金属的自身光谱特征,直接通过土壤重金属的光谱响应特征反演重金属含量不可行[10-11],然而通过借助重金属元素与土壤光谱活性物质(有机质、黏土矿物、铁氧化物)之间的吸附或赋存关系,可以间接反演土壤重金属元素含量[12],此为土壤重金属的高光谱间接反演机理。Sun等[13]提出基于土壤光谱活性物质特征谱段的重金属反演方法,提取对Ni起主要吸附作用的有机质和黏土矿物的特征谱段反演Ni含量,将反演精度R2由全谱段建模的0.67提升到0.98;Zhang等[14]提取有机质的特征谱段对Cd含量进行反演,显著提升了野外光谱的估算精度,但这种方法对土壤其他重金属反演的有效性还有待验证。

    本文以中国河北雄安一般农作区为例,利用野外光谱开展重金属Pb含量反演的研究,在去除环境因素对野外光谱影响的基础上,通过结合野外光谱与实验室光谱建模的方法,提高样本的差异性,以期提高反演精度。同时研究中提取了对重金属Pb起主要吸附作用的土壤光谱活性物质的特征谱段作为模型的输入,探讨了特征谱段建模相对于全谱段建模的有效性。

  • 土壤样本采集的时间为2018-09,采集地点主要在河北雄安新区的雄县和安新县的一般农作区农田。野外光谱采集使用的是SVC HR1024i地物光谱仪,测量波长范围为350~2 500 nm,共获取70个样点的土壤样本及配套野外土壤光谱。将土壤样本在实验室风干研磨后分别过20目和100目筛制成标准样,一部分用于电感耦合等离子体质谱仪(inductively coupled plasma-mass spectrometer,ICP-MS)测定Pb含量,另一部分用于测定实验室光谱,测量仪器同野外光谱采集。

  • 野外光谱测量时,大气中的水汽在1 400 nm和1 900 nm处存在强吸收,且在1 900~2 500 nm的光谱存在较严重噪声[15]。为去除大气水汽等噪声影响,并保留尽可能多的光谱区间,将1 800 nm之后的光谱波段从土壤野外光谱中剔除,对350~1 800 nm光谱使用分段SG(Savitzky Golay)滤波进行去噪处理,其中890~1 020 nm和1 330~1 520 nm为中等噪声,采用窗口大小为15的二次多项式,其余区间为轻度噪声,采用窗口大小为7的二次多项式。利用去除噪声后的350~1 800 nm的野外光谱进行土壤铁氧化物特征谱段提取及重金属含量估算。

    实验室光谱在350~1 800 nm存在不同程度的光谱噪声,同样采取分段SG滤波处理,350~500 nm和1 000~1 800 nm为中度噪声区间,采用窗口大小为15的二次多项式进行噪声去除;剩余的500~900 nm为轻度噪声区间,采用窗口大小为7的二次多项式进行噪声去除。

  • 土壤中对Pb起主要吸附作用的物质为铁氧化物[16],Galvão等[17]和徐彬彬等[18]的研究表明,铁氧化物的吸收特征在500 nm和950 nm附近的吸收峰。因此,本文从雄安70条有效土壤光谱中提取500 nm和950 nm为中心的铁氧化物吸收峰,提取波段区间为[Bm-W/4,Bm+W/4],其中Bm是最大的吸收波段,W是吸收区域的宽度,最终铁氧化物特征谱段提取出的光谱波段范围为450.7~523.2 nm及914.2~1 027.9 nm。

  • DS算法是一种常用的环境因素去除算法,其目的是为了通过转换数据集,根据对应的野外光谱和实验室光谱之间的关系,计算出转换矩阵并对野外光谱进行转换,去除野外光谱上环境因素(如水分、颗粒大小和温度)的影响[19]

    DS算法计算步骤如下:野外光谱(Xfield)和实验室光谱(Xlab)的矩阵大小都是m×pm是转换光谱的数量,p是波段的数量。DS算法的模型为:

    式中,B是由p×p个未知参数构成的转换矩阵,由实验室光谱和野外光谱矩阵共同决定;E为残差矩阵,表示为:

    式中,dsT是由基线差异产生的p×1的矩阵;λ是所有列向量值都为1的p×1的矩阵。将式(2)代入式(1)有:

    为了计算未知的转换矩阵B,首先需要定义一个大小为m×m的中心化矩阵Cm,可表示为:

    式中,Im是一个m×m的单位矩阵。将式(3)两边同时乘以Cm,因CmλdsT=0,并且CmXfieldCmXlab分别为野外光谱矩阵和实验室光谱矩阵的中心化矩阵,分别用X¯fieldX¯lab表示,则式(3)可以表示为:

    经过最小二乘后变为:

    式中,+表示X¯field的广义逆矩阵。

    基于转换矩阵B可以计算出残差矩阵E的值,根据式(2),将式(3)两边同时乘以(1/mλT有:

    式中,X¯fieldX¯lab分别是由野外光谱(Xfield)和实验室光谱(Xlab)对每列求均值得到的1×p的矩阵。野外光谱可以根据下式进行转换,得到去除环境因素影响的野外光谱:

    在DS转换的过程中,需要选择有代表性和差异性的土壤样本作为转换集,用于计算转换矩阵,因此本文用Kennard-Stone算法来筛选样本,具体步骤如下[20]:(1)计算两两样本之间距离,选择距离最大的两个样本;(2)分别计算剩余样本与已选两样本之间的距离;(3)对于每个剩余样本而言,计算其与已选各样本之间的最短距离,选择这些最短距离中相对最大的距离所对应的样本作为新入选的样本;(4)重复步骤(3),直至所选样本的个数等于事先设定的数目为止。

  • 在高光谱建模中,波段选择有利于降低模型复杂度,同时提高模型估算精度。遗传算法(genetic algorithm,GA)是一种随机的全局寻优算法,在偏最小二乘回归(partial least squares regression,PLSR)建模中被认为是一种有效的波段选择算法[21]。GA-PLSR已经被用于土壤反射光谱估算有机碳和重金属含量研究[22],能够降低模型复杂度,提高模型估算精度。本文采用GA-PLSR构建土壤重金属含量反演模型。

    参考已有GA-PLSR建模研究[23],将GA参数设置为:染色体20个,迭代次数1 000次,代际间隙90%,基因变异概率10%。本文中用均方根误差(root mean square error,RMSE)作为GA算法中适应度的评判标准,即目标函数,用RMSE值小的个体替换父一代中RMSE值大的个体,不断反馈产生新解进行下一轮迭代,直到达到最大的迭代次数,将RMSE最小的个体作为最优解输出,认为此时建立的PLSR为最优的模型。

  • 精度评定采用预测均方根误差(root mean square error of prediction,RMSEP)、相对分析误差(ratio of prediction to deviation,RPD)和决定系数(R2)3个评价指标,RMSEP值越小,RPD值越大,R2值越接近1,说明反演模型的精度越高;反之,RMSEP值越大,RPD值越小,R2值越小,说明反演模型的精度越低。本文模型优劣参考现有的土壤属性含量高光谱估算的评价标准[24]:出色模型,R2≥0.9;良好模型,0.9>R2≥0.8;近似模型,0.8>R2≥0.65;具有一定估算能力,0.65>R2‍≥0.50;不具备估算能力,0.50>R2

    本文构建的GA+PLSR模型的技术路线如图 1所示,分别对野外光谱和实验室光谱去噪后,以实验室光谱为参考对野外光谱进行DS校正,联合两者的铁氧化物特征波段用于模型的建立。

    Figure 1.  Flowchart of the Proposed Method

  • 本文将经过数据预处理后的70个样本按照2∶1的比例划分为训练集和测试集,训练集包括47个样本,测试集包括23个样本,训练集用于进行PLSR模型建立时训练模型,测试集用于代入训练后的模型中估算重金属Pb的含量,并计算评价指标,评估模型的估算能力。样本集的数据统计如表 1所示,从表 1中可看出划分的训练集和测试集与总样本集的统计特征基本一致。

    样本集 最小值 最大值 平均值 标准差
    总样本集 3.57 28.59 15.59 5.80
    训练集 3.57 28.59 15.46 5.71
    测试集 3.96 26.06 15.85 6.10

    Table 1.  Statistics of Pb Content in Sample Set/(mg·kg-1)

  • DS转换算法涉及到转换集的大小设定,而转换集选择的样本个数影响反演模型的精度。不同转换集大小对应的反演精度实验结果如图 2所示(由于3种精度指标同步变化,此处仅展示R2)。从图 2中可以看出,当转换集的样本个数为30时,反演模型的精度最高,因此本文选取30个样本的转换集进行转换矩阵计算。

    Figure 2.  Accuracy Change with Conversion Dataset Size

    将转换集中对应的野外光谱和实验室光谱按照DS转换算法进行计算得到转换矩阵,用转换矩阵对70个样本的野外光谱进行处理,获得DS校正后的野外光谱。校正效果如图 3所示。

    Figure 3.  Reflectance Spectrum Curves

    图 3可看出,经过DS转换后的野外光谱,波形与实验室反射光谱相似,光谱也变得更为平滑,表明通过DS转换,能够较好地去除环境因素(土壤粒径、含水量等)对野外光谱的影响。

  • 本节分别采用原始野外光谱、DS校正后的野外光谱、联合DS校正的野外光谱与实验室光谱进行建模,比较3种光谱的重金属Pb估算精度,以及全谱段建模与铁氧化物特征谱段建模的反演精度。

    为了避免实验的偶然性,每组建模实验均运算5次,取精度最优值作为模型的反演精度。

  • 图 4(a)为未经DS校正的野外光谱单独建模的结果,R2仅为0.513 7,属于具有一定估算能力的模型,可见直接利用野外光谱进行重金属Pb含量反演时精度较低。

    Figure 4.  Estimation Results of Pb Content

    图 4(b)为DS校正后的野外光谱独立建模结果,R2为0.627 6,精度有了明显提升,说明DS算法在去除野外光谱的环境因素影响方面是有效的。所建模型属于具有一定估算能力的模型,从散点图可看出,部分样本的预测值和测量值之间还存在着较大的差距,精度尚有待进一步提升。

    结合DS校正的野外光谱和实验室光谱的建模结果见图 4(c),样本点紧密分布于拟合线附近,精度R2达到0.914 6,属于出色模型,由此可见,通过联合建模增强了样本的差异性,从而使得反演精度得到大幅提高。

  • 3类光谱数据分别以全谱段和铁氧化物特征谱段作为GA-PLSR输入时的反演精度,结果如表 2所示。

    光谱波段 光谱类型 RMSEP RPD R2
    全谱段 野外光谱 5.271 1 1.157 7 0.220 0
    DS野外光谱 3.493 4 1.746 8 0.657 4
    DS野外与实验室联合光谱 1.932 4 3.158 0 0.895 2
    铁氧化物特征谱段 野外光谱 4.161 9 1.466 3 0.513 7
    DS野外光谱 3.641 9 1.675 6 0.627 6
    DS野外与实验室联合光谱 1.743 9 3.499 2 0.914 6

    Table 2.  Comparison of Modeling Accuracy

    表 2可知,除了DS校正野外光谱的全谱段与铁氧化物特征谱段反演精度相当外,其他两类光谱数据的铁氧化物特征谱段反演精度均高于全谱段反演精度。根据本文模型评价标准,对于未进行DS校正的野外光谱而言,通过铁氧化物特征谱段提取,建模精度R2从0.220 0提升到0.513 ‍7,精度得到了大幅的提升,反演模型从不具备估算能力提升为具有一定估算能力;联合建模的精度R2从0.895 2提升到0.914 6,反演模型从良好模型提升为出色模型,证明了提取500 nm和950 nm的铁氧化物特征谱段对于提高土壤重金属Pb反演精度的有效性。

  • 本文针对环境因素对土壤野外光谱乃至土壤重金属含量高光谱遥感反演的影响、土壤样本不足及全谱段建模机理性不足问题,以土壤重金属Pb为例,开展了基于野外光谱的土壤重金属含量反演方法研究。采用DS算法去除环境因素对野外光谱的影响,联合野外与实验室光谱,提取对Pb起主导吸附作用的铁氧化物的特征谱段用于反演建模,在增强样本的差异性的同时,增强反演的机理性,以雄安一般农作区土壤数据集为例,分析了该方法对提高Pb含量反演精度的有效性。

    研究表明,DS转换算法能够去除环境因素对野外光谱的影响,与原始野外光谱建模相比,反演精度R2从0.513 7提升到了0.627 6;联合野外光谱与实验室光谱建模的方法有利于提高样本的差异性,引入噪声极少的先验知识后,反演精度得到大幅提升,R2达到0.914 6,使得反演模型达到了出色模型的标准。

    除了DS校正的野外光谱,全谱段反演精度与铁氧化物特征谱段反演精度大致相当外,其他光谱类型的结果皆为铁氧化物特征谱段反演精度显著优于全谱段反演精度,表明通过提取对重金属Pb起主要吸附作用的铁氧化物的特征谱段,降低了数据的冗余度,提高了反演的机理性,对于提高模型反演精度是有效的,且本文提取的铁氧化物波段在500 nm和950 nm的吸收特征对于重金属Pb来说确实是敏感有效的特征谱段。

    由于野外光谱与同步或准同步获取的航空/卫星高光谱图像在土壤自然状态上最为接近,基于野外光谱的反演方法研究将为高光谱遥感图像反演土壤重金属含量奠定基础。本文方法通过环境因素去除以及基于铁氧化物特征谱段建模,联合野外与实验室光谱显著提高了土壤Pb含量反演精度,未来可尝试将该方法应用于同一研究区或同类的多个研究区的多源光谱数据(实验室、野外、图像光谱),探讨提高基于高光谱图像反演土壤重金属含量精度的可行性。

Reference (24)

Catalog

    /

    DownLoad:  Full-Size Img  PowerPoint
    Return
    Return