山洪灾害临界雨量确定方法对比研究--以雅安名山河流域为例

Comparative study on rainfall threshold methods of flash flood disaster:Take Mingshan River of Ya’an as an example

临界雨量是山洪灾害预警的重要指标,以雅安名山河流域为例,依据山洪灾情记录与逐小时降水数据,采用3种基于数据驱动的统计归纳法和基于灾变物理机制的FloodArea水文水力学方法,对比分析了名山河流域不同降雨历时(1、3、6、12、24 h)的山洪灾害临界雨量。通过研究发现,单站临界雨量法适用于雨量站分布均匀且站点密度高的区域,区域临界雨量法基于流域面雨量而非站点雨量,适用于雨量站布设稀疏的区域,其计算结果较实际情况偏低。灾害与降雨同频率法以县站长时间极值序列为样本,基于概率密度函数计算与灾害发生频率对应的临界雨量值,发现广义极值分布拟合效果最优、对数正态与耿贝尔分布次之。该方法数据需求简单,尤其适用于对灾情记录不全以及站点分布稀疏地区,但其存在山洪频率计算偏低情况下,临界雨量值计算结果偏高的问题,可依据实际情况采用合适的衰减系数,适当调整,减小误差。FloodArea水文模型更适用于精细化山区小流域山洪灾害过程的模拟,计算得到四级山洪灾害预警临界雨量值偏小,实际更接近三级风险雨量值。结果表明,统计方法倾向于高估,而水文模型指标则可能低估,山洪灾害临界雨量的推求可依据数据的实际情况综合采用多个方法进行复核订正。

Evidence suggests that the critical rainfall thresholds are among the most important factors for flash flood early warnings.In this article,Mingshan River of Ya’an is taken as an example for the contrastive analysis of four rainfall thresholds quantitative approaches,three data-driven statistical inductive approaches,and a hydrologic-dynamic approach are employed in this article to analyze the rainfall threshold of flash flood disaster of various rainfall duration(1,3,6,12,24 h).The results indicate that the single gauge thresholds approach applies to the scenarios with even-distributed and dense rain gauges.The regional thresholds approach employs areal precipitation instead of single gauge,is applicable to the scenarios with an inadequate amount of gauges,and the thresholds given by this approach are relatively low.The common frequency approach quantifies rainfall thresholds takes a long series of historic gauge data as the sample,and uses probability density functions to obtain the rainfall thresholds linked to the specific frequencies of local flash floods.The results given by the common frequency approach suggest that the fitting curve given by generalized extreme value distribution is optimal,while logarithmic normal distribution and Gumbel distribution are suboptimal.The simple data requirements are the distinctive advantage of this approach and could be applied in scenarios with inadequate flash flood records and rain gauges,but the shortcoming is that the quantified thresholds tend to be artificially high while the frequencies of flash floods are underestimated.Practically,this shortcoming could be reduced by bringing in appropriate attenuation coefficients.The hydrologic-dynamic approach,namely FloodArea hydro-model,applies to the refined simulation of the flash flood life cycle process of the small-sized montane basin.The shortcoming of this approach is that the calculated lowest early warning threshold tends to be overestimated and flash floods are triggered by higher precipitations.In general,thresholds given by data-driven statistical inductive approaches tend to be overestimated,while thresholds given by hydrologic-dynamic approach tend to be underestimated,and multi-methods examinations and modifications are imperative in practice.