Liu Ruimin, Men Cong, Yu Wenwen, Xu Fei, Wang Qingrui, Shen Zhenyao
State Key Laboratory of Water Environment Simulation, School of Environment, Beijing Normal University, No. 19, Xinjiekouwai Street, Beijing, 100875, China.
State Key Laboratory of Water Environment Simulation, School of Environment, Beijing Normal University, No. 19, Xinjiekouwai Street, Beijing, 100875, China.
Chemosphere. 2018 Jan;191:922-936. doi: 10.1016/j.chemosphere.2017.10.070. Epub 2017 Oct 13.
To examine the variabilities of source contributions in the Yangtze River Estuary (YRE), the uncertainty based on the positive matrix factorization (PMF) was applied to the source apportionment of the 16 priority PAHs in 120 surface sediment samples from four seasons. Based on the signal-to-noise ratios, the PAHs categorized as "Bad" might drop out of the estimation of bootstrap. Next, the spatial variability of residuals was applied to determine which species with non-normal curves should be excluded. The median values from the bootstrapped solutions were chosen as the best estimate of the true factor contributions, and the intervals from 5th to 95th percentile represent the variability in each sample factor contribution. Based on the results, the median factor contributions of wood grass combustion and coke plant emissions were highly correlated with the variability (R = 0.6797-0.9937) in every season. Meanwhile, the factor of coal and gasoline combustion had large variability with lower R values in every season, especially in summer (0.4784) and winter (0.2785). The coefficient of variation (CV) values based on the Bootstrap (BS) simulations were applied to indicate the uncertainties of PAHs in every factor of each season. Acy, NaP and BgP always showed higher CV values, which suggested higher uncertainties in the BS simulations, and the PAH with the lowest concentration among all PAHs usually became the species with higher uncertainties.
为研究长江口（YRE）源贡献的变异性，基于正矩阵因子分解（PMF）的不确定性被应用于四季采集的120个表层沉积物样本中16种优先多环芳烃的源解析。基于信噪比，归类为“差”的多环芳烃可能会从自抽样估计中剔除。接下来，利用残差的空间变异性来确定哪些具有非正态曲线的物种应被排除。自抽样解的中值被选为真实因子贡献的最佳估计值，第5至95百分位数的区间代表每个样本因子贡献的变异性。结果表明，木草燃烧和炼焦厂排放的中值因子贡献与各季节的变异性高度相关（R = 0.6797 - 0.9937）。同时，煤炭和汽油燃烧因子在各季节的变异性较大，R值较低，尤其是在夏季（0.4784）和冬季（0.2785）。基于自抽样（BS）模拟的变异系数（CV）值被用于指示各季节每个因子中多环芳烃的不确定性。苊、萘并芘和苯并[g]荧蒽总是显示出较高的CV值，这表明在BS模拟中不确定性较高，并且所有多环芳烃中浓度最低的多环芳烃通常成为不确定性较高的物种。
