Inversion of heavy metal content in urban water sediments by combining machine learning and hyperspectral remote sensing data
The urban water system is often called the blood vessel of the city,and sediments in the water system record important information related to changes in the urban environment.The rapid,efficient,and accurate acquisition of heavy metal content information in urban water system sediments is of great significance for urban environmental monitoring,ecological environment restoration,and sustainable development.In this paper,hyperspectral inversion of Cu,Zn,and Cd content in the sediments of Jihe River in Tianshui City,Gansu Province is conducted.After various mathematical transformations of the spectral data,characteristic bands with a strong correlation with the measured heavy metal content are selected as independent variables,and three inversion models based on artificial neural network(ANN),support vector machine,and stepwise multiple linear regression(SMLR)are constructed.The determination coefficient(R2)and root mean square error are selected to evaluate the accuracy of the model.The results show that:(1)the original spectral data can effectively highlight the spectral feature information after spectral transformation,and the feature band screening effects of different spectral transformation methods are different.The screening effects of first-order differential(FD),second-order differential(SD),and reciprocal logarithm first-order differential(AFD)are better than those of reciprocal logarithm(AT)and reciprocal logarithm second-order differential(ASD).(2)The R2 of the three inversion models is greater than 0.6,meaning all three models can effectively realize the inversion of heavy metal content in sediments.(3)The optimal inversion models of different elements are different.The best inversion model for Cu is the SD-ANN model,which has an R2 of 0.750;the best inversion model for Zn is the SD-SMLR model,which has an R2 of 0.962;and the best inversion model for Cd is the SD-SMLR model,which has an R2 of 0.761.The optimal inversion model for each element is related to the selection of the characteristic band(s),and the inversion of heavy metal content based on the characteristic bands of stream sediments is conducive to improving inversion accuracy.This study provides a reference for the rapid acquisition of heavy metal pollution information in stream sediments,and provides technical support for nondestructive environmental monitoring and the sustainable development of ecological environments.
stream sedimenthyperspectral inversionheavy metal elementscharacteristic bandcontent prediction
国家科技期刊平台
NSTL
万方数据
