Font Size: a A A

Study On Urban Runoff Contaminants Characteristics And A Prediction Model

Posted on:2019-10-15Degree:MasterType:Thesis
Country:ChinaCandidate:X WangFull Text:PDF
GTID:2381330545999712Subject:Municipal engineering
Abstract/Summary:PDF Full Text Request
With the large-scale city construction,the proportion of hardened pavement increased,and the runoff characteristics changed,resulting in a significant increase in urban surface runoff.At the same time,the severity of rainfall runoff pollution was not fully considered in the early stage of urban construction,which resulted in the inherent design defects of the urban drainage system.The result is that rainfall runoff pollution has not been effectively controlled,and has become the main influence factors of urban water pollution.To build the foundamental knowledge of the pollution characteristics of urban runoff is necessary for effective management and control measures.In this study,the surface runoff sampling and water quality test were firstly carried out on 5 types of urban surface including main road,branch road,square,roof and grass.The event mean concentration(EMC),first flush phenomenon and the influence factors of runoff pollution were studied.The results showed that:? The overall level of runoff pollution in the study area was low,and the EMC values of COD,SS,NH3-N,TP and TN of each sampling point were between 22.92?79.91,19.01?53.29,0.403?0.800,0.063?0.298 and 1.213?3.315 mg/L,respectively.Furthermore,the pollution level of COD,TN in our main road,branch road and grass sampling sites were inferior to the IV-class standard of Surface water environmental quality.And the SS concentration was inferior to the 2-class sewage treatment plant standard,which will aggravate the pollution of the receiving water.? The characteristics of rainfall in the study area have a significant impact on runoff pollution:the greater the rainfall intensity and the longer the dry period before the rain,the higher the EMC,the more significant the initial scour effect is.? There are significant differences in runoff pollution level between different types of area in the study area:the level of main road and branch road runoff pollution is relatively high;the level of SS pollution in roof runoff is higher,organic matter and nutrients and other pollution levels are low;the grass runoff is affected by the retainment and infiltration,so that the runoff is little,and the runoff pollution level of grass surface is high.? There is a certain correlation between the indexes of runoff pollution:at the level of P<0.01,COD/NH3-N,COD/TN and NH3-N/TN are highly linear correlated,and COD/SS,COD/TP,SS/,etc.are significantly linear correlated.The greater the rainfall intensity and the longer the dry period before the rain,the stronger the correlation between the pollution index(especially the COD/SS,COD/TN)in this study area.In addition,in this study,we used the previous literature and the measured data as sample,chose 10 influence factors:rain type,average rain intensity,peak rain intensity,rainfall duration,dry period before rain,atmospheric dust amount,PM10,vehicle flow,surface material and urban functional area,as model input,then we chose EMC and FF30 value of runoff pollution index COD as model output.Then a road runoff pollution assessment model was built based on support vector machine(SVM).The results show that:? The constructed EMC-SVM and FF30-SVM models have high prediction accuracy.The verification parameters RMSE,MBE of the EMC-S VM model are far less than the EMC average value,and the CE,CC parameters are 0.815 and 0.933;The verification parameters RMSE,MBE of the FF30-SVM model are also far less than the FF30 average value,and the CE,CC are 0.866 and 0.932 respectively.Therefore,it is feasible to assess roads' runoff pollution with these models,where the runoff sampling tests and monitoring are difficult to implement.? The radial basis function(RBF)and the k-folding cross validations were used as the kernel function,and validation method,which can result in a higher precision with faster convergence speed.The optimal parameters we found for EMC-S VM and FF30-SVM models(c,g)are(64,0.001953125)and(2,0.0625)respectively.? The correlation and principal component analysis of the sample data show that the roads' characteristics and rainfall characteristics are the main factors of the runoff pollution,among which the roads'characteristics,including the atmospheric dust reduction and traffic flow,are the most important factors.
Keywords/Search Tags:stormwater runoff pollution, EMC, first flush phenomenon, statistical learning, support vector machine
PDF Full Text Request
Related items