Liu Chengshuai, Li Wenzhong, Zhao Chenchen, Xie Tianning, Jian Shengqi, Wu Qiang, Xu Yingying, Hu Caihong
Yellow River Laboratory, Zhengzhou University, Zhengzhou, 450001, China.
Yellow River Laboratory, Zhengzhou University, Zhengzhou, 450001, China.
J Environ Manage. 2023 Oct 15;344:118482. doi: 10.1016/j.jenvman.2023.118482. Epub 2023 Jul 5.
In recent years, urban flood disasters caused by sudden heavy rains have become increasingly severe, posing a serious threat to urban public infrastructure and the life and property safety of residents. Rapid simulation and prediction of urban rain-flood events can provide timely decision-making reference for urban flood control and disaster reduction. The complex and arduous calibration process of urban rain-flood models has been identified as a major obstacle affecting the efficiency and accuracy of simulation and prediction. This study proposes a multi-scale urban rain-flood model rapid construction method framework, BK-SWMM, focusing on urban rain-flood model parameters and based on the basic architecture of Storm Water Management Model (SWMM). The framework comprises two main components: 1) constructing a SWMM uncertainty parameter sample crowdsourcing dataset and coupling Bayesian Information Criterion (BIC) and K-means clustering machine learning algorithm to discover clustering patterns of SWMM model uncertainty parameters in urban functional areas; 2) coupling BIC and K-means with SWMM model to form BK-SWMM flood simulation framework. The applicability of the proposed framework is validated by modeling three different spatial scales in the study regions based on observed rainfall-runoff data. The research findings indicate that the distribution pattern of uncertainty parameters, such as depression storage, surface Manning coefficient, infiltration rate, and attenuation coefficient. The distribution patterns of these seven parameters in urban functional zones indicate that the values are highest in the Industrial and Commercial Areas (ICA), followed by Residential Areas (RA), and lowest in Public Areas (PA). All three spatial scales' RE, NSE, and R indices were superior to the SWMM and less than 10%, greater than 0.80, and greater than 0.85, respectively. However, when the study area's geographical scale expands, the simulation's accuracy will decline. Further research is required on the scale dependency of urban storm flood models.
近年来，由突发暴雨引发的城市洪涝灾害日益严重，对城市公共基础设施以及居民的生命财产安全构成了严重威胁。城市雨洪事件的快速模拟与预测可为城市防洪减灾提供及时的决策参考。城市雨洪模型复杂且艰巨的校准过程被认为是影响模拟与预测效率和准确性的主要障碍。本研究提出了一种多尺度城市雨洪模型快速构建方法框架BK - SWMM，该框架聚焦于城市雨洪模型参数，并基于暴雨管理模型（SWMM）的基本架构。该框架包含两个主要部分：1）构建SWMM不确定性参数样本众包数据集，并将贝叶斯信息准则（BIC）与K均值聚类机器学习算法相结合，以发现城市功能区中SWMM模型不确定性参数的聚类模式；2）将BIC和K均值与SWMM模型相结合，形成BK - SWMM洪水模拟框架。基于观测到的降雨径流数据，通过对研究区域内三种不同空间尺度进行建模，验证了所提框架的适用性。研究结果表明了诸如洼地蓄水、地表曼宁系数、入渗率和衰减系数等不确定性参数的分布模式。这七个参数在城市功能区的分布模式表明，其值在工商业区（ICA）最高，其次是居民区（RA），在公共区（PA）最低。所有三个空间尺度的RE、NSE和R指标均优于SWMM，分别小于10%、大于0.80和大于0.85。然而，当研究区域的地理尺度扩大时，模拟的准确性将会下降。需要对城市暴雨洪水模型的尺度依赖性进行进一步研究。
