基于局部回归的城市特征降雨研究

Determination of Characteristic Rainfall at Urban Scale Based on Local Regression Method

随着对海绵城市设计以及城市雨洪管理和面源污染管控等的精细化要求,科学表达特征降雨是其中关键的基础工作。特征降雨是指在某一降雨类型中能够代表该类降雨统计特征的单次连续降雨过程,传统的短历时暴雨雨型设计不适合长时间的降雨分析,也不宜用于表征不同降雨量下的典型降雨雨型。为此,以滇池流域20多年的逐小时降雨数据为基础,通过对降雨类型的划分,采用基于累积分布与Logistic回归处理的局部加权线性回归方法对降雨数据进行拟合,识别多降雨量类型下的降雨规律,形成特征降雨。结果显示,针对不同的降雨类型,参数优化后的局部加权线性回归模型可有效识别累积降雨量、降雨历时与降雨强度等特征,可为城市径流模拟提供典型降雨边界条件,对理解城市降雨规律及相关的雨污合流和面源污染治理等具有重要的支撑作用。

With the deep requirements of sponge city design,urban stormwater management and non-point source pollution control,scientific expression of characteristic rainfall is one of the key basic tasks.Characteristic rainfall refers to a single continuous rainfall process that can represent the statistical characteristics in a certain type of rainfall.The traditional short-duration rainstorm pattern design is not suitable for long-term rainfall analysis and cannot be used to identify typical rainfall patterns under different rainfall scenarios.Therefore,based on the hourly rainfall data of Dianchi Basin over the past 20 years,through the classification of rainfall types,the locally weighted linear regression method based on the cumulative distribution and Logistic regression processing was used to fit the rainfall process,identify the rainfall features under the multiple rainfall types,and eventually form the characteristic rainfall.The results showed that for different types of rainfall,the optimized locally weighted linear regression model could effectively identify the characteristics such as cumulative rainfall,rainfall duration and rainfall intensity,provide typical rainfall boundary conditions for runoff simulation,and play a supporting role in understanding the nature of urban rainfall and its related rainwater pollution and the strategies of non-point source control.