Author(s): James Ball

Linked Author(s):

Keywords: Floods; Urban; Model; Calibration

Abstract: Design flood estimation remains an ongoing problem. A common approach to obtaining the desired design flood quantiles is to use catchment models. When using this approach, it is desirable to calibrate the catchment model. Calibration of a catchment model requires the catchment model be used to predict the catchment response to a historical event; in other words, the catchment model is being used to predict the data that would have been recorded if monitoring were available for that event. The predicted flood characteristics are then compared with recorded flood characteristics using a metric of some form. Having calibrated the catchment model, the design flood quantiles are obtained for the desired climatic and catchment conditions using two generic approaches discussed in ARR - use of the model in a statistical sense (AEP Neutral) and in a deterministic sense (continuous) where the catchment model is used to generate data equivalent to what would have been recorded if monitoring were available for the desired climatic and catchment conditions. However, guidance in ARR is not provided on what value of calibration metric is required for robust prediction of design flood quantiles. This paper addresses that question using data from the Powells Creek catchment in Sydney, Australia. A SWMM based model of the Powells Creek catchment was calibrated against 25 events. The purpose of this calibration was finding the most generic parameter set; the most generic parameter set is the one that performs best over all events considered. Using the identified generic parameter sets, continuous time series of flow were predicted and analysed using FFA techniques. The design flood quantiles obtained from the catchment model were compared with those obtained from the monitored data over the same period for different values of calibration metric.

DOI: https://doi.org/10.3850/IAHR-39WC252171192022811

Year: 2022