近年来，在我国快速城市化的过程中，硬化的城市地表改变了原有的水文过程，由此导致的城市内涝问题一直制约着城市可持续发展。而在复杂多样的地表状况下管道出流的过程与影响因素尤为复杂，因此对城市降雨-管道出流过程的研究非常重要。由于城市降雨-管道出流过程涉及多过程相互作用的复杂水系统，实时监测数据较为稀缺，目前对于城市降雨径流过程的研究还不够充分。本研究基于降雨与雨水管网出流的实时在线监测数据，识别城市雨水管网出流的主要影响因素，进而探索主要影响因素、不同降雨模式和排水分区特征如何影响管网出流，之后通过构建暴雨洪水管理模型（SWMM），进一步探讨影响雨水管网出流的主要敏感参数，从而提高对雨水管网出流影响因素的理解。
本研究选择河南省鹤壁市国家海绵城市试点区内的部分排水分区为案例研究区，主要开展以下工作：辨识对管道出流有重要影响的降雨特征；探讨不同降雨模式和排水分区特征对管道出流的影响；构建排水分区SWMM模型，识别影响管道出流的敏感参数，探讨敏感参数随降雨条件的变化规律及参数的交互作用。本研究的主要结论如下:
（1）构建偏最小二乘回归模型PLSR，针对6个排水分区内三类管道出流特征的开展模拟，87.5%的模拟结果显示自变量方差贡献率为0.59-0.88，拟合优度为0.52-0.82，模拟效果较好。对预测变量的重要性投影值VIP值进行计算，结果显示：9个降雨特征中，对雨水管网出流影响最大的是总降雨量TP和最大60min降雨强度MAX60，其次为平均降雨强度RI和最大5min的降雨强度MAX5。主要影响因素的具体排序如下：对管道出流峰值PFR的影响：MAX60（VIP = 2.75）>TP（VIP = 2.35）>MAX5（VIP =1.72）>RI（VIP =1.48）；对单位面积出流量TIA的影响: MAX60（VIP =2.78）≈TP（VIP =2.73）>MAX50（VIP =1.74）>RI（VIP =1.09）；对径流控制率RC的影响:TP（VIP=2.00）≈MAX60（VIP = 1.93）>MAX5（VIP =1.87）>RI（VIP=1.44）。
（2）不同降雨模式下降雨特征和排水分区特征对管网出流的影响结果表明，强降雨模式条件(HR)下的管道出流峰值PFR比中等降雨条件(MR)下的PFR值高约1.5-10倍,强降雨模式下的单位面积出流量TIA值比中等降雨模式下TIA值高约2.7-17倍,相比于暴雨条件，非暴雨条件下前期累积雨量对管道出流的影响更为显著，尤其是前期5天累积雨量AP5甚至成为影响管道出流的主导因素；只有在强降雨条件下，排水分区特征才会对管道出流产生显著性影响，特别是排水分区面积和绿地比例对管道出流的影响较大。
(3)全局灵敏度性分析的灵敏参数识别结果表明，在不同降雨条件和不同目标函数的情况下，SWMM 模型参数的敏感性表现均不相同，表明SWMM 模型参数具有较大的不确定性；对纳什效率系数高度敏感的参数是不渗透率IMP,管道出流峰值误差PE的敏感参数排序为：不渗透率IMP>特征宽度WID=不透水区曼宁系数N-IMP;而峰现时间相对误差PT除了受到特征宽度WID,不透水区曼宁系数N-IMP的影响，不透水区贮水深度S-IMP和管道曼宁系数MANN在一定程度上对其也起着关键作用；管道出流总量误差TF主要受不渗透率IMP的影响，不渗透率对其起着决定性作用；不渗透率IMP对纳什效率系数NSE的敏感度和特征宽度WID对管道出流峰值PE的敏感度会随着最大降雨强度的增大呈现降低的趋势，与平均降雨强度或降雨等级关系不大。不渗透区贮水深度S-IMP对出流总量的敏感度会随着总降雨量或降雨强度的增加呈现降低的趋势；参数交互作用在SWMM模拟中是普遍存在的，它反映了实际降雨径流各个环节的相互影响，相比于其它降雨条件，暴雨条件下的参数间交互作用往往更为复杂更容易出现更高阶的参数相互作用（3阶4阶）。

Under the current background of rapid urban development, urban waterlogging caused by changes in urban hydrological processes has always restricted the sustainable development of cities, The process and influencing factors of storm drainoutflow under complex and diverse surface conditions are particularly complex. Therefore, to investigate the process of rainfall-drainoutflow is of significance. However, because the urban rainfall -drainoutflowprocess involves complex water system problems with multi-process interaction and feedback and real-time monitoring data is relatively scarce, the current understanding of urban rainfall drainoutflow is not sufficient. In this research, on the one hand, based on real-time online monitoring data, this paper carries out the identification of urban storm drain outflow influencing factors, and then explore how the main influencing factors including different rainfall patterns and drainage district characteristics specifically affect drainoutflow. On the other hand, based on the construction of Storm Water Management Models, this paper further discuss the sensitivity of water quantity simulation parameters to the drainoutflow, improving the understanding of the important factors affecting the drainoutflow.

This study selects part of the drainage area in the Sponge City pilot area of Hebi City, Henan Province as the case study area. The main tasks are as follows:

Identify the rainfall characteristics that have an important influence on the drainoutflow; Explore the influence of different rainfall patterns and drainage district characteristics on the drainoutflow; Construct the drainage district SWMM model to identify the sensitive parameters of the drainoutflow, explore the changes of sensitive parameters’sensitivity with the rainfall conditions and interactions between parameters. The main conclusions of this research are as follows:

The main conclusions of this research are as follows:

(1) This paper construct the Partial least square regression models to simulate the three tpyes characteristics of drainoutflow in six drainage district. 87.5% of the simulation results have a variance contribution rate R2 of 0.59-0.88 for independent variables, and the goodness of fit Q2 of 0.52-0.82, provig the simulation effect is better. The results of calculating the variables importance projection values VIP showed that:

among the 9 rainfall characteristics, 4 influencing factors have a major impact on runoff generation, of which the most significant impact on drainoutflow is TP and MAX60, followed by RI and MAX5. The specific sequence of the main influencing factors is as follows: Peak runoff PFR: MAX60 (VIP = 2.75)> TP (VIP = 2.35)> MAX5 (VIP = 1.72)> RI (VIP = 1.48); Outflow per unit area TIA: MAX60 (VIP = 2.78) ≈TP (VIP = 2.73)> MAX50 (VIP = 1.74)> RI (VIP = 1.09); runoff control rate RC: TP (VIP = 2.00) ≈ MAX60 (VIP = 1.93)> MAX5 (VIP = 1.87)> RI (VIP = 1.44).

(2) The effect of rainfall characteristics and drainage district characteristics on drainoutflow under different rainfall patterns showed that the PFR value under heavy rainfall pattern HR is about 1.5-10 times higher than that under moderate rainfall MR. pattern MR. The TIA values under the HR pattern is about 2.7-17 times higher than that under the MR pattern.; Compared to rainstorm events,the antecedent precipitation has a more significant effect on drainoutflow and is even be the dominant factor when rainstorm events with daily rainfall larger than 50mm are not considered.; Only under the HR pattern, the characteristics of the drainage district will have a significant impact on the drainoutflow, especially the area of drainage dstrict and the proportion of green landscapes have a greater impact on thedrainoutflow.

The sensitive parameter identification results of global sensitivity analysis showed that under different rainfall conditions and different objective functions, the sensitivity of parameters in the SWMM model is different, indicating that the SWMM model parameters have greater uncertainty; The highly sensitive parameter of Nash efficiency coefficient NSE is impermeability IMP, and the order of the sensitive parameters for peak flow rate error PE is: IMP>WID=N-IMP; The peak present time error PT is not only affected by WID and N-IMP, but also be affected by D-IMP and MANN to a certain extent.; The total drainoutflow error TF is mainly affected by the IMP which plays a decisive role; The sensitivity of IMP for NSE and the sensitivity of WID for PE show a decreasing trend with the increase of the maximum rainfall intensity and have little to do with the average rainfall intensity or rainfall level. The sensitivity of D-IMP for total runoff TF show a decreasing trend with the increase of total rainfall or rainfall intensity; Parameter interaction is common in SWMM simulation, which reflects the interactions between various links of actual rainfall -drainoutflow process;Compared with other rainfall conditions, the interaction between parameters under heavy rain condition is often more complicated and more likely to have higher-order parameter interactions (3rd order 4th order).