On the Use of Surrogate Model Ensembles for Bioretention Cell Design Optimisation under Current and Future Climate Conditions

Urbanisation and climate change are contributing to increased stormwater runoff, which exacerbates the risk of urban flooding and adversely affects our receiving waters. Bioretention cells (BRCs) can be used to mitigate these impacts, however finding their optimal design can be a time-consuming process, further exacerbated by using physical-based models. Surrogate models, an alternative to physical-based models, are investigated, to determine the impact architecture has on the performance of these stochastic models. Ultimately, the quality of the input set is the biggest determinant of model performance, and selecting the appropriate input set significantly improves model prediction accuracy. This surrogate is then used in a simulation-optimisation framework to streamline the optimisation of BRC design under current and future climate. BRC performance under future climate decreases in comparison with current climate, and existing design guidelines should be updated to reflect those findings.