| In order to alleviate a series of prominent water environment problems such as frequent waterlogging and deterioration of water ecology in the process of urbanization and improve the efficiency of stormwater management measures,urban storm water model has been widely used in the planning,design and management of urban rainwater system.While SWMM is widely used at home and abroad,its parameters are numerous.There are complex correlations among them,and a large number of parameters can not be directly measured,resulting in high uncertainty of simulation results.In order to improve the accuracy of parameter calibration and the simulation accuracy of SWMM model,this paper explores the correlation between key parameters and the possible influencing factors in parameter calibration process.Based on Bayesian statistical theory and advanced DREAM（zs）（DiffeRential Evolution Adaptive Metropolis）sampling method,an algorithm for parameter calibration of SWMM model is constructed.Then this paper explored a series of factors that may affect the calibration effect,such as rainfall intensity,rainfall pattern and hydrological characteristics of catchment area.Then the influence of these factors on parameter calibration was analyzed from the following aspects:the sample trajectory and of posterior distribution of parameters,correlation between parameters,and fitting of simulated and measured values.The purpose of this study is to find out the best conditions for parameter calibration of SWMM model,so as to further improve the simulation accuracy of SWMM model.The main research methods and conclusions are as follows:（1）Establishment of SWMM hydrological parameter calibration algorithm based on Bayesian statisticsBased on Bayesian statistical theory and advanced DREAM（zs）sampling algorithm,an appropriate likelihood function was selected and the Bayesian theory-DREAM（zs）calibration algorithm was programmed in Matlab.A numerical model was used to demonstrate the reliability and efficiency of the calibration algorithm.By integrating SWMM with the Bayesian-DREAM（zs）calibration algorithm,the data interaction between the algorithm and SWMM was realized in the iteration process,which can be directly used in the hydrological parameter calibration of SWMM.（2）Study on the influence of rainfall intensity on the calibration effect of SWMM hydrological parametersNine types of rainfall with different rainfall intensities were designed.The established calibration algorithm was used to explore the influence of rainfall intensity on the calibration effect of eight hydrological parameters in SWMM.The parameters were analyzed from the aspects of parameter posterior distribution,correlation between parameters,fitting between simulated and measured values.The results show that Width has a large linear correlation with N-imperv and N-perv,Slope with S-imperv and S-perv.The interaction among parameters has a great influence on the identification effect of Width,Slope and Zero-imperv.The sensitivity of the three parameters is obviously reduced.Imperv,S-imperv and N-imperv have higher sensitivity and can be better recognized,less affected by rainfall intensity.The other parameters can only be better recognized under some rainfall intensity.According to the parameter calibration results under different rainfall intensities,the comprehensive ranking for the recognition effect of parameters in this study is:Imperv>S-imperv>N-imperv>Zero-imperv>N-perv>S-perv>Width>Slope.（3）Study on the influence of surface hydrological characteristics in catchment area on the calibration effect of SWMM hydrological parametersTypical impervious rates of 20%,50%and 80%were designed to represent the regions with different development level,and two typical slopes of 0.4%and 4%were designed to represent the catchment areas of plain and hilly cities respectively.Five kinds of rainfall intensity were selected to further explore the influence of surface hydrological characteristics on parameter sensitivity and recognition effect.Finally,the suitable rainfall intensity tables for parameter calibration under different hydrological characteristics are obtained for reference.The results show that the topography of catchment area has certain regularity on the calibration effect of some parameters.S-imperv has better sensitivity and recognition effect when the catchment slope is larger（except when the rain is heavier）.With 20%and 50%impermeability,catchment slope has some influence on the calibration effect of N-imperv,N-perv,S-perv and Zero-perv,and it is more obvious when the rainfall intensity is higher.The influence of catchment development level on sensitivity and identification effect of parameters is quite different,and there is no significant regularity.The development level and topography of catchment area have little influence on the posterior distribution of SWMM hydrological parameters,and will not change the relative sensitivity and recognition effect among parameters.（4）Study on the influence of other external factors（the rainfall peak type and the ampunt of calibrated data）on the calibration effect of SWMM model hydrological parametersIn addition to rainfall intensity,the study further explored the effects of rainfall patterns（single,double and three peak）on parameter calibration effect.The results show that the rain pattern has a great influence on the sensitivity and recognition effect of parameters.Except for Slope and S-Imperv,the calibration effect is slightly better in single-peak rain pattern,the sensitivity and recognition effect of other parameters are better in double-peak rain pattern.Increasing the amount of calibration data,i.e.using two sets of rainfall and corresponding flow data to calibrate at the same time,only improves the sensitivity and recognition effect of the two parameters,while the complexity of the algorithm and the calculation cost of the calibration process increases significantly.The SWMM model of a watershed in Yuelai,Chongqing is constructed.Combined with the established calibration algorithm and the corresponding monitoring data,three rainfall events were used for calibration.During the validation period,two rainfall events were randomly selected to validate the calibrated parameters.The results show that the parameter posterior distribution law of three regular rainfall events is consistent with the previous research conclusion.The fitting of the simulated and observed values is good in both the periods of validation and validation. |
PDF Full Text Request