Pollution Distribution of Microplastics, Organophosphate Esters, and Per-/Poly-fluoroalkyl Substances in River Basins and Application of Machine Learning Models for Microplastic Ecological Risk Prediction
學生姓名:
曹宏菖
指導教授:
凌明沛
學期:
113上
摘 要:
OrganophosphateEsters(OPEs)andPer-/Poly-fluoroalkylSubstances(PFAS)arepersistentenvironmentalpollutantscommonlyusedinflameretardantsandplasticizers,andmicroplastics(MPs)canadsorbOPEsandPFAS,posingriskstoecosystemsandhumanhealth.ThisstudyexploresthepresenceofMPsandtheirrelationshipwithOPEsandPFAS,andevaluatestheuseofmachinelearningmodelstopredictMPsconcentrationsinsoilandassessrisks.Thestudyanalyzesdatafromthreeriverbasins:theconcentrationsofMPsandOPEsinsedimentsandbiotaintheLoireRiverBasin,France;theconcentrationsanddistributionofMPsandPFASinsurfacewaterinthePearlRiverBasin,China;andtheuseofvariousmachinelearningmodelstopredictMPsconcentrationsanddistributionintheTaihuLakeBasin,China,alongwithriskassessments.TheresearchresultsindicatethattheconcentrationsanddistributionofMPs,OPEs,andPFASareinfluencedbymultipleenvironmentalfactors.IntheLoireRiverBasin,MPsconcentrationsshowsignificantseasonalandregionalvariations,andfoundoutsmallerparticlesofsedimentshaveagreatercapacityforaccumulation.InthePearlRiverBasinfoundsignificantseasonalandregionaldifferencesinthedistributionofMPsandPFAS.InthePFASanalysis,long-chaincompoundsshowedhigherconcentrations,particularlyinindustrialareas.IntheTaihuLakeBasinrevealsthatMPsconcentrationsinurbanandagriculturalsoilsaresignificantlyhigherthaninforestsoils.Amongthemachinelearningmodels,thesupportvectormachinewithradialbasisfunctionkernelmodeldemonstratedthebestperformanceinpredictingthespatialdistributionofMPs.Insummary,MPsandOPEsintheLoireRiverBasinshownosignificantcorrelationinsedimentandbiotasamples,butsignificantlycorrelatedwithPFASinsurfacewaterinthePearlRiverBasin.ThedistributionofMPs,OPEs,andPFASintheriverbasinsisinfluencedbyvariousfactors.Machinelearningmodelshaveshownpotentialapplicationsinpollutionpredictionandriskassessment.
曹宏菖.pdf