Department of Chemical Engineering and Applied Chemistry, University of Toronto, Toronto, ON, Canada.
Ecotoxicol Environ Saf. 2010 Mar;73(3):231-9. doi: 10.1016/j.ecoenv.2009.05.008. Epub 2009 Jun 2.
A Monte Carlo analysis is used to quantify environmental parametric uncertainty in a multi-segment, multi-chemical model of the Venice Lagoon. Scientific knowledge, expert judgment and observational data are used to formulate prior probability distributions that characterize the uncertainty pertaining to 43 environmental system parameters. The propagation of this uncertainty through the model is then assessed by a comparative analysis of the moments (central tendency, dispersion) of the model output distributions. We also apply principal component analysis in combination with correlation analysis to identify the most influential parameters, thereby gaining mechanistic insights into the ecosystem functioning. We found that modeled concentrations of Cu, Pb, OCDD/F and PCB-180 varied by up to an order of magnitude, exhibiting both contaminant- and site-specific variability. These distributions generally overlapped with the measured concentration ranges. We also found that the uncertainty of the contaminant concentrations in the Venice Lagoon was characterized by two modes of spatial variability, mainly driven by the local hydrodynamic regime, which separate the northern and central parts of the lagoon and the more isolated southern basin. While spatial contaminant gradients in the lagoon were primarily shaped by hydrology, our analysis also shows that the interplay amongst the in-place historical pollution in the central lagoon, the local suspended sediment concentrations and the sediment burial rates exerts significant control on the variability of the contaminant concentrations. We conclude that the probabilistic analysis presented herein is valuable for quantifying uncertainty and probing its cause in over-parameterized models, while some of our results can be used to dictate where additional data collection efforts should focus on and the directions that future model refinement should follow.
采用蒙特卡罗分析方法对威尼斯泻湖多段多化学模型中的环境参数不确定性进行量化。利用科学知识、专家判断和观测数据，构建了描述 43 个环境系统参数不确定性的先验概率分布。通过对模型输出分布的矩（集中趋势、离散程度）进行比较分析，评估了这种不确定性在模型中的传播。我们还应用主成分分析（PCA）结合相关分析，以确定最具影响力的参数，从而深入了解生态系统的功能。我们发现，Cu、Pb、OCDD/F 和 PCB-180 的模拟浓度变化幅度可达一个数量级，表现出污染物和地点的特异性变化。这些分布通常与实测浓度范围重叠。我们还发现，威尼斯泻湖污染物浓度的不确定性具有两种空间变异性模式，主要由当地水动力条件驱动，将泻湖的北部和中部以及更孤立的南部盆地分开。尽管泻湖的污染物空间梯度主要由水动力条件塑造，但我们的分析还表明，在泻湖中部原地历史污染、当地悬浮泥沙浓度和泥沙埋藏速率之间的相互作用对污染物浓度的变化具有重要控制作用。我们得出结论，本文提出的概率分析方法对于量化不确定性和探究过度参数化模型中的不确定性原因非常有用，而我们的一些结果可用于指导应在哪里集中进行额外的数据收集工作以及未来模型改进应遵循的方向。
