Author(s): V. Bellos; J.P. Carbajal; J.P. Leitao

Linked Author(s): Vasilis Bellos

Keywords: Flood prediction; Machine learning; Compound channel; FLOW-R2D; Emulator

Abstract: Flood prediction in a synthetic trapezoidal, compound channel is made, using an emulator of the computationally demanding 2D hydrodynamic model FLOW-R2D. Accurate flood prediction in reasonable computational time is of great importance for operational purposes, such as flood warning schemes. Accuracy in results requires fine scale modelling in the (usually 2D) hydrodynamic modelling. However, even if 2D hydrodynamic studies are feasible in practice, the required computational time is still prohibitive for real-time prediction. Therefore in real world applications, flood warning schemes are usually based on ad hoc, empirical approaches, although the results derived are characterised by significant uncertainties. One way for achieving results of higher accuracy with low computational cost, is the use of machine learning techniques. In this study, we present an example of using an emulator for a detailed 2D hydrodynamic simulator, named FLOW-R2D, in a hypothetical case study.

DOI: https://doi.org/10.3850/978-981-11-2731-1_281-cd

Year: 2018