烟台地区土壤重金属镍高光谱估测模型
|
摘要
快速监测土壤重金属污染程度,对发展精细农业、保障食品安全和社会经济可持续发展具有重要意义。本文基于山东省烟台市的70个土壤样本数据,首先分析了土壤重金属镍的分组光谱特性;对土壤光谱反射率进行一阶微分、倒数的一阶微分、对数的一阶微分等六种变换并计算出光谱反射率变换值与土壤镍含量的相关系数,根据极大相关性原则选取光谱特征;然后建立基于BP神经网络的土壤重金属镍含量光谱估侧模型;并利用其它2种建模方法对镍含量进行建模,验证BP神经网络模型的有效性。结果表明,土壤光谱反射率随镍含量的升高而降低,呈现负相关性;以(1/R1015)′、(1/R2286)′、(1/ln(R925))′和(1/ln(R1911))′为估测因子,所建镍含量估侧模型的决定系数为R2=0.912,平均相对误差为14.279%。研究表明,利用高光谱技术定量估测土壤镍含量是可行的。
Rapid monitoring of heavy metal pollution in soil is of great significance for developing precision agriculture,ensuring food safety and sustainable development of society and economy. Based on the data of 70 soil samples in Yantai city of Shandong Province, this paper first analyzes the spectral characteristics of the heavy metal nickel group spectral characteristics. Then six transformations of soil spectral reflectance such as first order differential, reciprocal first order differential and logarithmic first order differential are adopted, the correlation coefficients between transformed spectral value and soil nickel content were calculated respectively. The spectral characteristics are selected according to the principle of maximum correlation. Finally, BP neural network was used to estimate nickel content of soil heavy metals based on spectral features, and use two other modeling methods to model the nickel content to verify the effectiveness of the BP neural network model. The results showed that the spectral reflectance of soil decreased with the increase of nickel content and showed negative correlation. Take(1/R1015)′、(1/R2286)′、(1/ln(R925))′ and(1/ln(R1911))′ as the estimation factors, Using BP neural network model, the determination coefficient of nickel content was 0.912, the average relative error is 14.279%. The research shows that it is feasible to estimate the nickel content directly by hyperspectral technology.
Rapid monitoring of heavy metal pollution in soil is of great significance for developing precision agriculture,ensuring food safety and sustainable development of society and economy. Based on the data of 70 soil samples in Yantai city of Shandong Province, this paper first analyzes the spectral characteristics of the heavy metal nickel group spectral characteristics. Then six transformations of soil spectral reflectance such as first order differential, reciprocal first order differential and logarithmic first order differential are adopted, the correlation coefficients between transformed spectral value and soil nickel content were calculated respectively. The spectral characteristics are selected according to the principle of maximum correlation. Finally, BP neural network was used to estimate nickel content of soil heavy metals based on spectral features, and use two other modeling methods to model the nickel content to verify the effectiveness of the BP neural network model. The results showed that the spectral reflectance of soil decreased with the increase of nickel content and showed negative correlation. Take(1/R1015)′、(1/R2286)′、(1/ln(R925))′ and(1/ln(R1911))′ as the estimation factors, Using BP neural network model, the determination coefficient of nickel content was 0.912, the average relative error is 14.279%. The research shows that it is feasible to estimate the nickel content directly by hyperspectral technology.
引文
[1]吴昀昭.南京城郊农业土壤重金属污染的遥感地球化学基础研究[D].南京:南京大学,2005
[2]Pichtel J,Sawyerr HT,Czarnowska K.Spatial and temporal distribution of metals in soils in Warsaw,Poland[J].Environmental Pollution,1997,98(2):169-174
[3]Wong YS,Tam NFY,Liu WK,et al.Heavy metal pollution in roadside urban parks and gardens in Hong Kong[J].Sci Total Environ.,1987,59:325-328
[4]李巨宝,田庆久,吴昀昭.滏阳河两岸农田土壤Fe、Zn、Se元素光谱响应研究[J].遥感信息,2005(3):10-13
[5]Wu YZ,Chen JJ,Ji JF,et al.Feasibility of reflectance spectroscopy for the assesment of soil mercury contamination[J].Environmental Science&Technology,2005,39(3):873-878
[6]柴世伟,温琰茂,张云霓.广州郊区农业土壤重金属含量与土壤性质的关系[J].农村生态环境,2004,20(2):55-58
[7]Kemper T,Sommer S.Estimate of heavy metal contamination in soils after a mining accident using reflectance spectroscopy[J].Environmental Science&Technology,2002,36(12):2742-2747
[8]解宪丽,孙波,郝红涛.土壤可见光-近红外反射光谱与重金属含量之间的相关性[J].土壤学报,2007,44(6):982-993
[9]徐明星,吴绍华,周生路,等.重金属含量的高光谱建模反演:考古土壤中的应用[J].红外与毫米波学报,2011,30(2):109-114
[10]夏芳,彭杰,王乾龙,等.基于省域尺度的农田土壤重金属高光谱估测[J].红外与毫米波学报,2015,34(5):593-598,605
[11]葛哲学,孙志强.神经网络理论与MATLAB R2007实现[M].北京:电子工业出版社,2007
[2]Pichtel J,Sawyerr HT,Czarnowska K.Spatial and temporal distribution of metals in soils in Warsaw,Poland[J].Environmental Pollution,1997,98(2):169-174
[3]Wong YS,Tam NFY,Liu WK,et al.Heavy metal pollution in roadside urban parks and gardens in Hong Kong[J].Sci Total Environ.,1987,59:325-328
[4]李巨宝,田庆久,吴昀昭.滏阳河两岸农田土壤Fe、Zn、Se元素光谱响应研究[J].遥感信息,2005(3):10-13
[5]Wu YZ,Chen JJ,Ji JF,et al.Feasibility of reflectance spectroscopy for the assesment of soil mercury contamination[J].Environmental Science&Technology,2005,39(3):873-878
[6]柴世伟,温琰茂,张云霓.广州郊区农业土壤重金属含量与土壤性质的关系[J].农村生态环境,2004,20(2):55-58
[7]Kemper T,Sommer S.Estimate of heavy metal contamination in soils after a mining accident using reflectance spectroscopy[J].Environmental Science&Technology,2002,36(12):2742-2747
[8]解宪丽,孙波,郝红涛.土壤可见光-近红外反射光谱与重金属含量之间的相关性[J].土壤学报,2007,44(6):982-993
[9]徐明星,吴绍华,周生路,等.重金属含量的高光谱建模反演:考古土壤中的应用[J].红外与毫米波学报,2011,30(2):109-114
[10]夏芳,彭杰,王乾龙,等.基于省域尺度的农田土壤重金属高光谱估测[J].红外与毫米波学报,2015,34(5):593-598,605
[11]葛哲学,孙志强.神经网络理论与MATLAB R2007实现[M].北京:电子工业出版社,2007