A Numerical Comparison of Three Sewer Infiltration Models

This study compares the RTK, AMM, GIM, NAM, and QvI models’ ability to accurately predict RDII for 10 selected sites across the U.S. and 975 identified rain events. Tests were conducted using automated optimization to measure the goodness of fit, tendency to overfit, extrapolation performance, and accuracy of calibrating to a short data period. The AMM models were most accurate overall, but not without important tradeoffs. Results emphasize thoughtful model selection and the careful use of modeler judgment to reduce overfitting. This also provides a common dataset and framework for benchmarking other models in future study.