Reducing Computational Runtime of Two-Dimensional Urban Inundation Model by Dynamic Domain Reshaping

Abstract

Coupling rainfall-runoff and one-dimensional (1D)/two-dimensional (2D) hydrologic models provides the current state-of-the-art approach for predicting the extent of urban inundation, although the computational cost for running 2D models is still a major concern. This paper presents a procedure to dynamically reshape 2D numeric domain according to flooding extent evolution. The numeric domain is updated at each selected time step for reshaping (TSR) and includes grid elements directly connected to the 1D stormwater network, flooded grid elements, or grid elements within a buffer distance threshold (BDT) identified in both previous criteria. Results for an urban catchment in South Brazil are presented considering two spatial resolutions and three return-period rainfall events. An optimal relation between TSR and BDT was found and resulted in runtime savings greater than 86% in comparison to adopting a fixed boundary as catchment limits, and greater than 69% in comparison to using a fixed boundary below the 70-m elevation (which represents the maximum elevation reached by the 100-year flood). The major concepts of the proposed algorithm are straightforward and could be tested in similar 2D models.