The Numerical Simulation Of The Migration Of Heavy Metals In The Soil, And Uncertainty Analysis

With the rapid development of industry and agriculture, the contamination of heavy metals in soils is becoming increasingly serious, which poses a critical threat to the humen’s survival. This thesis studied heavy metals (Cu, Cd, Zn, Pb) transport in porous medium such as quartz sand, loamy sand and layered soils, and analysed the breakthrough curves (BTCs) with the solute transport models (two-site nonequilibrium model, TSM and one-site nonequilibrium model, OSM), based on HYDRUS-1D. Besides, it was studied that the effect of colloid (SiO2) on Zn, Cd transport in loamy sand and the colloid-facilitated solute tansport model (CFM) was used to simulate Zn, Cd transport. At last, parameters uncertainty was studied based on experimetal data of Cu transport in quartz sand. The main conclusions obtained are as follows:(1) The saturated water contentθs and dispersivityλwere identified from bromide breakthrough data and then were used in the remaining parameters identification. The two-site nonequilibrium model (TSM) coupled with the Freundlich adsorption equation gave good fits to the observed BTCs (R2>0.9, RMSE<0.8E-3), but failed to capture the peak concentrations. It was inferred from the estimated parameters that heavy metals adsorption was mainly controlled by time (the fraction of exchange sites assumed to be in equilibrium with the solution phase f less than 10% for Cu, Cd, Pb, and 13.2% for Zn) and the retardation factor Rd calculated from the transport simulation parameters was significantly less than that obtained from batch experiments, indicating that heavy metals sorption did not reach equilibrium in the columns. Sensitivity analysis results showed that the most sensitive parameter for the peak value of BTCs was the sorption parameterβand that for the peak time was the saturated hydraulic conductivity Ks. Both Ks andβcontrolled the fit between simulated and observed concentrations, as indicated by the RMSE.(2) The one-site nonequilibrium model (OSM) coupled with the Freundlich adsorption equation fitted the Zn, Cd concentrations in loamy sand soils better than the TSM. The small optimized parameters f indicated that heavy metals sorption was almostly rate-limited, controlled by time. Colloid transport was well described by the advection-dispersion equation coupled with the attachment-dettachment kinetics and the optimized attachment rate ka generally much greater than the detachment rate kd. Colloid facilitated the transport of Zn and restrained the transport of Cd. The CFM model described Zn and Cd transport behaviours, which were affected by colloid, reasonably well.(3) The retention of Cu, Cd and Zn in layered soils (loam and sandy loam) was studied in saturated condition. For single heavy metal, retention decreased in order:Cd>Zn>Cu, but for multi-heavy metals condition, the different order was observed:Cu>Cd>Zn. It revealed that adsorption of Cd, Zn was severely depressed by the presence of Cu and Cu has the highest affinity to the soil. The relative concentration of Zn was greater than one, which was called "snow plow effect", suggesting the overlapping of Zn desorbed and in solution soon after the background solution input. Theθs andλobtained seperatedly from tracer BTCs in loam and sandy loam could give good fits to the observed BTCs of Br" in layered soil. The BTCs of Cu, Cd and Zn in layered soils could all be fitted by OSM with high goodness of fit (R2>0.89, RMSE<0.05), but the optimized parameters were some uncertainty.(4) The results of numerical analysis revealed that the order in which the soil layers were stratified did not influence the effluent sulute concentration in a water-saturated condition, but there were striking differences in solute distribution within the clay-sand and sand-clay profiles.(5) The generalized likelihood uncertainty estimation (GLUE) approach was selected to evaluate the parameters uncertainty of the solute transport model with Latin Hypercube Sampling method. The results showed when onlyθs andλwere optimized the uncertainty was low and the regional sensitivity ofθs was greater thanλ. The narrowed ranges ofθs andλwere then used to analyse the uncertainty in the identification of parameters in TSM coupled with Freundlich adsorption equation for Cu transport. The analysis resulted in a large uncertainty inθs,λ, kp and small uncertainty inβ,ω, f. The uncertainty ofθs andλwere increased for the correlation of parameters. Regional sensitivities of these parameters in non-ascending order were f,ω,β,fp,λandθs as a result of that the sorption of Cu to quartz sand was rate-limited and f,ωwere both closely related to the kinetic adsorptive reaction.(6) There was a significantly large uncertainty inθs1,λ1,θs2,λ2 for modeling tracer transport in layered soils. Many parameter sets could perform equally well:the problem of equifinality. Parametersθs1 andθs2 were found to be high interrelated with the correlation coeffiecient-0.994.