Predicting combined sewer overflow occurrences using rainfall depth and maximum intensity: a case study of Buffalo, USA

The occurrence of combined sewer overflow (CSO) is a pressing environmental issue in many cities. This study aims to predict CSO occurrence using rainfall event characteristics, including rainfall depth, maximum intensity, and duration, and to determine which characteristic is the best predictor. Buffalo, New York, was selected as a case study. The results indicate that the prediction accuracy ranges from 80% to 100% for rainfall depth and maximum intensity, while rainfall duration is not a good predictor. Furthermore, rainfall depth is more likely to be the best predictor for sewersheds with a larger area. Additionally, combining the three rainfall event characteristics using the decision tree can only improve the average prediction accuracy slightly, from 93% (using a single characteristic) to 95% (using three characteristics). Using rainfall event characteristics and this simple method can be an effective alternative to complex urban hydrological models and/or expensive monitoring for predicting CSO occurrence.