Integrated modelling is a promising approach to simulate processes operating within complex environmental systems. It is possible, however, that this integration may lead to computationally expensive compositions. In order to retain the process fidelity without loss of accuracy, the use of Kriging metamodels is proposed to perform Monte Carlo simulation and sensitivity analysis, in lieu of compositions developed using the model linking standard OpenMI. Results from the Monte Carlo simulation showed that the metamodels were in a good agreement with the original responses. However, metamodels provided a less accurate approximation of the original output distribution for the composition which involved a stronger non-linear behaviour. The fast runtimes of the metamodels allowed for increased computational budgets leading to an accurate screening of the important parameters for an Elementary Effects Test. Overall, Kriging metamodels provided significant computational savings without compromising the quality of the outcomes, even using small training data sets.

集成建模是一种用于模拟在复杂环境系统中运行的过程的有前途的方法。但是，这种集成可能会导致计算上昂贵的合成。为了在不损失精度的情况下保持过程的保真度，建议使用Kriging元模型来执行蒙特卡罗模拟和灵敏度分析，以代替使用链接标准OpenMI的模型开发的成分。蒙特卡洛模拟的结果表明，元模型与原始响应非常吻合。但是，元模型为包含更强非线性行为的合成提供了不太精确的原始输出分布近似值。元模型的快速运行时间允许增加计算预算，从而可以准确筛选基本效应测试的重要参数。总体而言，即使使用较小的训练数据集，克里金元模型也可节省大量计算量，而不会影响结果的质量。