Research On Evaluation Of Environmental Risk Of Metal Elements In Atmospherically Deposited Particulate Matter Based On LIBS Technology Combined With Random Forest

The situation of urban air pollution is increasingly worse in China,as the persistently development of industrialization and urbanization,and haze phenomenon occurs frequently.Contaminated metals may destroy a balanced soil ecosystem（acid-base equilibrium）,and can eventually threaten human health due to their cumulative effect by food chain.The process of evaluation of metals environmental risk calculates the pollution index according to the content of metal,and finally accomplishes the evaluation of environmental risk.However,technologies of determination of metals based on atomic absorption spectrometry（AAS）can accquire excellent quantitative analysis results,but they require complex sample pre-processing and a long analysis cycle,which cannot meet the need of rapid analysis.Hence,it is urgent to establish an accurately and rapidly technology to evaluation of environmental risk of metals in atmospherically deposited particulate matter.Based on the actual demand for evaluation of environmental risk of metal elements,this paper takes atmospherically deposited particulate matter as object,and implements a study on evaluation of environmental risk of metals by laser-induced breakdown spectroscopy（LIBS）combined with random forest（RF）.With the help of variable selection strategy,an accurate and fast evaluation method of environmental risk was established to supply theoretical basis and technical support for evaluation of atmospheric environmental risk.The main research contents in this paper are as follows:（1）Aiming at the evaluation of single-factor of metal environmental risk of atmospherically deposited particulate matter,taking the enrichment factor（EF）and geo-accumulation index(I_geo)of metal elements analysis as the research objective,a method of evaluation of environmental risk for single metal elements（Cr,Zn,V,Cu and Pb）was established based on LIBS technology combined with RF,with the help of extracting important chemical information of samples by spectral pretreatment methods and variable importance measurement（VIM）.Under the optimal conditions of wavelet transform（WT）,variable importance threshold（0.04）and model parameters(n_tree=500;m_try=272),a RF calibration model was constructed for the EF analysis of Zn element.The results show that the R_p²and the MREP are 0.9834 and 0.0238,respectively.In addition,the RF calibration models for EF analysis and I_geo analysis of five metal elements also have excellent analytical ability.Therefore,LIBS technology combined with RF can be used as an effective tool for accurate evaluation of single-factor environmental risk of metals in atmospherically deposited particulate matter.（2）Aiming at the evaluation of environmental risk of multiple metals（Al,Cu,Zn,Mn,Pb,Cr,V and Ti）in atmospherically deposited particulate matter,taking comprehensive potential ecological risk index（RI）analysis as the research target,a method of evaluation of environmental risk of multiple elements was constructed by LIBS technique combined with RF,with the help of hybrid variable selection strategy.Successive projection algorithm（SPA）and genetic algorithm（GA）were applied to precisely select feature variables,and this section explored the impact on accuracy of RF calibration model as different hybrid variable selection methods（VIM-SPA and VIM-GA）.The result shows that the VIM-SPA-RF calibration model has excellent prediction performance（R_p²=0.9227;MREP=0.0490）.Therefore,compared with the single variable selection method（VIM）,the VIM-SPA method can reduce the operation time,and improve the accuracy of the model.（3）Aiming at the discriminant analysis of pollution grade of metal elements in atmospherically deposited particulate matter,with the help of pattern recognition method,the discriminant analysis method of metal elements pollution grade was established based on LIBS technology and RF,taking the pollution grade divided by EF as the research target.Under the optimal conditions of WT method and model parameters(n_tree=500,m_try=76;LV=5),the method of pollution grade discrimination based on principal component analysis（PCA）,RF discriminant model and partial least squares discriminant analysis（PLS-DA）discriminant model were set up.The result reveals that the predictive performance of RF discriminant model is excellent（the accuracy,sensitivity and specificity of discrimination of pollution grade of Zn are 0.9200,0.9000 and 0.9333,respectively;the accuracy,sensitivity and specificity of discrimination of pollution grade of Cr are 0.9200,1.0000 and 0.9000,respectively）.Hence,LIBS coupled with RF is employed for rapid discrimination of pollution grade of metal elements,supplying early warning for air pollution prevention and control.