Comparing HEC-HMS, PCSWMM, and Random Forest Models for Rainfall-Runoff Evaluation to Extreme Flooding Events

Abstract

Urban stormwater systems have faced flood management challenges from the extremity of catastrophic losses from flooding events in recent years. In this regard, various modeling approaches have been introduced to explore flood control options by evaluating flood risk and impacts, early forecasting flooding events, and planning emergency and recovery options (e.g., evacuation). However, decision-making can suggest different flood control options depending on the modeling strategies since the variety of flood models has historically led to intense debate on the “appropriate” approach for hydrologic modeling, with discussion focusing on the suitability of process parameterizations, data limitations and uncertainty, and computational limitations on hydrologic analysis. Thus, the selection of an appropriate modeling approach is a challenge. Therefore, this study compares rainfall-runoff simulation models to extreme flooding events in an urban watershed. This study selected hydrologic, hydraulic, and data-driven models, that is, HEC-HMS, PCSWMM, and random forest, applied to a real urban watershed. The models were developed using the meteorological, hydrologic, and spatial data obtained from the United States Geological Survey (USGS) website and extracted with the ARC-GIS tool. The model performance was evaluated using statistical error measures such as standard deviation ratio (RSR), and Nash-Sutcliffe efficiency coefficient (NSE). The results discussed the strengths and weaknesses of the three different models to evaluate rainfall-runoff relationships under extreme flooding events. The discussions in this study will be able to encourage flood managers to improve current rainfall-runoff simulation models in an integrated way to resiliently cope with extreme flooding events.