靖江市某电镀集中区典型重金属迁移转化模型参数灵敏性分析

伴随重金属污染风险的不断提高,流域重金属迁移转化模型的构建和水体重金属的模拟预测受到广泛关注,关键参数的甄选是模型优化的重点.本文构建了某电镀集中区Ni、Cu重金属迁移转化数学模型,以Ni为例,采用标准秩回归分析方法(SRRC法)和Sobol多元自适应回归样条算法(Mars-Sobol法)对7个重金属模型参数进行敏感性分析,并针对确定的2个敏感性参数对Ni、Cu模型进行率定和验证.结果表明:1SRRC法的Ni河沙分配系数的敏感性占比为96.1%~99.7%,平均为99.2%,泥沙沉降速率为0.1%~3.3%,平均为0.5%.2Mars-Sobol法的Ni河沙分配系数的总敏感度占比为87.18%~93.44%,平均为90.28%;泥沙沉降速率为5.68%~10.68%,平均为8.21%;随水流运动方向,Ni河沙分配系数敏感性逐渐减低,泥沙沉降速率参数敏感性逐渐增强.3相较于SRRC法,Mars-Sobol考虑了参数间的交互作用,通过Ni、Cu迁移转化模型中敏感参数"河沙分配系数"和"泥沙沉降速率"的率定和验证,Ni、Cu模拟相对误差可控制在15.28%和14.46%,实现了重金属模型的高效和高精度预测.

Parametric sensitivity analysis for typical heavy metals migration and transformation model in electroplating industrial area of Jingjiang City

With the increasing risk of potential heavy metal pollution, the development of heavy metal migration and transformation models as well as the prediction of their fate in aquatic environments have attracted extensive attention. Optimizing the model parameters becomes particularly important to promote the accuracy of the model. In this study, the heavy metals (Ni and Cu) simulation model for an electroplating industrial area was constructed. Taking Ni as an example, the parametric sensitivity analyses for model were performed using the Standardized Rank Regression Coefficient (SRRC) and the Mars-Sobol methods. Two sensitive parameters were applied for model calibration and verification to improve the accuracy and efficiency in simulations. Results suggested that the sensitivity of the water-sediment Ni distribution coefficients obtained using SRRC method was between 96.1% and 99.7% (99.2% in average), and the sediment deposition rate was identified from 0.1% to 3.3% (0.5% in average). By using Mars-Sobol method, the total sensitivity of the water-sediment Ni distrubution coefficients was determined in a range of 87.18% and 93.44% (90.28% in average), and the sediment deposition rate was in a range of 5.68% to 10.68% (8.21% in average). Our results revealed that the sensitivity of water-sediment Ni distribution coefficients decreased along the flow direction, while the sediment deposition rate displayed an increasing trend. Compared with SRRC method, Mars-Sobol model had an higher accuracy in predictions with the consideration of interactions between multiple parameters. By carefully calibrating and verifying two sensitive parameters (i.e. water-sediment distribution coefficient and sediment deposition rate), the maximum relative errors of Ni and Cu models were reduced within 15.28% and 14.46%, respectively. Our work suggests the prediction accuracy and efficiency of heavy metal model could be achieved by optimizing sensitive parameters.