From single-objective to multiple-objective multiple-rainfall events automatic calibration of urban storm water runoff models using genetic algorithms.

The calibration of storm water runoff models is a complex task. Early attempts focused on the choice of a performance criterion function that could capture all the facets of the problem into a single-objective framework. Subsequently, the awareness that a good calibration must necessarily take into account conflicting objectives led to the adoption of more sophisticated multi-objective approaches. Only recently, the focus has shifted towards effective ways of exploiting the mounting information provided by the availability of many sets of concurrent rainfall and flow measurements. This paper revisits through a case study the transition just elucidated: the calibration of a SWMM model applied to a catchment in Singapore is tackled through a single-objective, a multi-objective and a multi-objective multiple-event (MOME) paradigm respectively. A new approach to support the latter is presented herein. It consists in formulating the problem of model calibration as a multi-objective problem with m x r objective functions, where m and r are the number of performance criteria and rainfall events respectively, that must be optimized simultaneously. Results suggest that the new MOME framework performs significantly better than the others tested on the case study presented.