大气扩散模型统计性确认评估

使用计算模型预测物理过程之前,必须对该计算模型进行论证,确认评估通过后才能用于物理状态和过程的计算和预测。讨论了基于数据统计属性的确认评估对面向大气扩散物理现象的计算模型之适用性;探究了模型评估工作的基本内容、关键概念和通用原则;给出了确认评估参数的定义、选择原则及示例;分析了评估度量指标的内涵和适用条件;探讨了可接受模型需要满足的指标准则及统计性确认评估指标的选用;最后,指出了基于统计度量的确认评估方法的改进方向。扩散模型评估理论的探讨有助于提高基于模型仿真的大气扩散研究的准确性,也有利于高精度模型和基准试验的设计、开发与遴选。

Statistical validation assessment study on the atmospheric dispersion models

This paper attempts to make a discussion over the applicability of the statistics-based validation evaluation theory to the mathematical models for physical phenomena involving atmospheric dispersion.As is known,statistical validation assessment method can be adopted to assess the accuracy of the dispersion models,and,what is more,since it is also the only feasible approach to be adopted for the time being,though such model validation assessments can only be done indispensable for the argumentation of the application of computational models to predict physical processes.For,they are the only models that can be used for successful assessment of prediction of the state and process of physical phenomena,including the basic contents,key concepts and general principles of assessment for the dispersion behaviors,though the assessment principles include,but are not limited to,the inherent limitations of the accuracy of the model outputs,the relativity of the assessment results,the relevance,persistence and bi-directionality of the assessment activities,with the definitions,selection principles and examples of evaluation parameters being given.Furthermore,since the selection of the evaluation parameters depends on the objectives of the dispersion study and physical data of interest,it is also necessary to analyze the connotation and conditions of evaluation measures to be applied.In addition,it is also of great importance to study and define the statistical metrics and the expressions of the indicators needed based on the systematic errors and random errors on the ratio of the predicted values to the experimental values.Besides,it is also of great need to recommend the criteria and indicators for the statistical validation assessment to be met so as to make models acceptable.For example,the following three statistical indicators should be made most suitable for the use in making the performance validation evaluations of the atmospheric dispersion models,i.e.the geometric mean bias(MG),the geometric mean variance(VG) and the fraction of predictions within a factor of two of observations (FAC2).And,finally,we have pointed out the improvement and renovation orientations of the validation evaluation methods based on the statistical metrics as well.All in all,the theory for building up the dispersion evaluation model can be put in a very helpful position to improve the accuracy of the atmospheric dispersion research based on the model simulation.It can also be of great benefit to the design,development and selection of high precision models and benchmark experiments.