Uncertainty analysis of environmental simulation models

我们通过环境模拟模型对复杂环境系统的基本理解和数学表示，以及对环境变量或量的测量，只是对自然真实状态的近似。近似程度因所研究的环境系统的位置和类型而异。基于环境模拟模型预测的健全的环境管理决策应明确考虑这些近似值传播到整体模型预测的不确定性。我计划在未来五年及以后调查的研究问题，最终集中在开发强大的和有效的技术，用于评估环境模拟模型预测的不确定性建模研究，测量校准数据可用。 将制定模型校准和不确定性分析的新方法，并与各种环境建模案例研究的现有方法进行比较。这项研究将特别侧重于水文或流域模拟模型，预测流域内的流量，养分和其他污染物的传输。所有的真实的案例研究将主要位于安大略的农业流域。这项研究的结果将是感兴趣的所有工程师，研究人员和决策者谁是使用环境模拟模型。然而，这项研究最重要的好处将是加强基于环境建模的决策。这项研究将开发工具，通过这些工具，可以对更多的建模研究进行模型预测的不确定性分析。 通过对未来环境系统行为的不确定性评估，而不是确定性分析，决策者可以更有效地权衡风险和成本，从而改进环境系统的管理。

Our basic understanding and mathematical representation of complex environmental systems via environmental simulation models, as well as the measurement of environmental variables or quantities, are only approximations of the true state of nature. The degree of approximation varies by location and type of environmental system being studied. Sound environmental management decisions based on environmental simulation model predictions should explicitly consider the propagation of these approximations into overall model prediction uncertainty. The research questions I plan to investigate over the next five years and beyond are ultimately focused on developing robust and efficient techniques for assessing environmental simulation model prediction uncertainty for modelling studies where measured calibration data are available.   New methods for model calibration and uncertainty analysis will be developed and compared with existing approaches for a variety of environmental modelling case studies. This research will focus specifically on hydrologic or watershed simulation models that predict flow, nutrient and other pollutant transport within a watershed. All real case studies will be agricultural watersheds located mainly in Ontario. The findings of this research will be of interest for all engineers, researchers and decision-makers who are using environmental simulation models. However, the most important benefit of this research will be enhanced decision-making based on environmental modelling. This research will develop tools through which uncertainty analyses of model predictions can be conducted for more modelling studies.  With an uncertainty assessment of future environmental system behaviour, rather than a deterministic analysis, decision-makers can more effectively trade-off risks and costs and therefore improve the management of environmental systems.