Automatic Calibration Approach for Multiple Rain Events in SWMM Using Latin Hypercube Sampling

Abstract

The present study aimed for the development and application of an automated calibration approach based on Latin Hypercube Sampling (LHS). The proposed methodology was used for calibrating an urban drainage model of a small sewer network of Dresden, implemented in EPA Stormwater Management Model (SWMM). Based on predefined ranges, LHS was applied to generate 1000 calibration parameter sets, which were used to simulate 24 different rain events. Nash-Sutcliffe Efficiency (NSE) was used to assess the goodness-of-fit of the results. NSE values were stored in an m-by-n matrix, where m corresponds to the number of parameter sets and n is the number of rain events and converted into 0 if they were below a defined threshold and 1 otherwise. For each row the sum of columns was calculated. The resulting number, referred to as degree, represents the amount of rain events for which the specific configuration yields a good NSE. Then, higher degrees imply a more robust set of parameters. By selecting a threshold of NSE = 0.75 it was possible to determine 11 different configurations, for which 6 out of the 24 events fulfilled the threshold. Analysis of these positive solutions revealed that one of them, parameter set 18 (S18), consistently reported the best NSE values for all analyzed events, and therefore was selected as the best solution. Although, adequate validation results were obtained, additional improvements are needed. Moreover, this approach can also be used to perform sensitivity analyses and evaluate the influence of rain events on calibration efficiency.