河口海岸模型验证方法及其性能比较研究

Comparative Study on the Performance of Estuarine and Coastal Model Verification Methods

关键词:  河口海岸模型  验证  置信区间  统计模型  假定检验

Key words:  estuary coastal model  verification  confidence interval  statistical model  hypothesis test

摘要:

模型验证是河口海岸模型建立和应用的关键步骤，传统的方法包括主观评价、可视化技术、误差估计及统计检验等量化方法，但各种验证方法的性能优劣尚无明确的结论。为探讨各种方法的性能，本文对常用的几种量化方法的性能进行比较，并提出了一种基于统计检验的定量的模型验证方法。研究表明：即使在模型数据的误差明显时，Skill、相关系数和决定系数也可能保持较大的值，因此使用这些量化指标进行模型验证需谨慎。相对于绝对误差、均方根误差、Skill等常用量化指标，统计检验法既能有效地量化模型模拟的误差，亦能有效减小模拟结果中的异常值对整体模拟效果的影响。由于误差分布的原因，统计检验方法不适用于往复流的流向验证。

Abstract:

Model verification is a key step in the establishment and application of estuarine and coastal models. Traditional methods are quantitative methods including subjective evaluations, visualization technologies, error estimations and statistical tests. However, the performance of various verification methods has not been clearly defined. In order to explore the performance of various methods, this paper compares the performance of several commonly used quantitative methods and a quantitative model verification method based on statistical tests is proposed. It shows that even when the errors of model data are significant, the values of Skill, correlation coefficient and determination coefficient might maintain relatively large, so cautions should be exercised when using these quantitative indicators for model verification. Compared with commonly used quantitative indicators such as absolute error, root mean square error and Skill, the statistical test method can not only effectively quantify the error of model simulation, but also effectively reduce the influence of abnormal values in the simulation results on the overall simulation effect. Due to the distribution of error, the statistical test method might not be suitable for flow direction verification of alternating current.