论文部分内容阅读
洪水预报(包括洪水实时预报和洪水风险预估在内)中存在着诸多的不确定性。在模拟研究过程中,为全局最优,研究人员偏向用不同的方法得到相同的结果,即“殊途同归”法,很多模型能产生合理的模拟结果,可以认为是多种假设共同作用的结果,如果有足够数据,这些假设有的可能会得到验证,有的则被拒绝。在“通用不确定性评估”(GLUE)中,参数是基于物理原因而随机设置的,在没有太多参数值信息的情况下,通常是均匀地取值。这些参数设置又被用于模型的输出,然后又使用一定的标准进行检测和评估,以提供每个参数的权重。这里的可能性涵义比统计范畴里的要宽范。在模型模拟以前,如果能够限定观测误差的范围,那么超出误差范围的那些模型预测结果就可舍弃。因此,任何这种类型的模型评估,都需要仔细考虑模型误差的不同来源。用实例展示了“通用不确定性评估”的过程,同时列出了实际问题的解决方法和未来发展。
There are many uncertainties in flood forecasting (including real-time flood forecasting and flood risk forecasting). In the process of simulation research, for the sake of global optimization, researchers tend to get the same result by different methods, that is, “the same way” method. Many models can produce reasonable simulation results and can be considered as the result of multiple hypotheses If there is enough data, some of these assumptions may be validated, others rejected. In Generic Uncertainty Evaluation (GLUE), parameters are set randomly based on physics, and usually take a uniform value without too much parameter value information. These parameter settings, in turn, are used to model the output and are then tested and evaluated using a standard that provides the weight of each parameter. The possibilities here are broader than those in the statistics area. Prior to model simulations, those models that exceeded the margin of error were discarded if they were able to limit the range of observed errors. Therefore, any such type of model assessment requires careful consideration of different sources of model error. An example is given to show the process of “universal uncertainty evaluation”, and the solutions and future development of practical problems are also listed.