Extremely Rapid, Lagrangian Modeling of 2D Flooding: A Rivulet-Based Approach
This figure compares flood depth predictions during Hurricane Harvey using traditional shallow water equations (left) and the new rivulet-based model (right). Both methods predict similar flood patterns, showcasing the rivulet approach's efficiency and accuracy.
Image courtesy of Daniel et al.
A novel, Lagrangian rivulet-based, 2D flood model, Torrent, has been developed. It is 10,000 to 10 million times more computationally efficient than the full solution of the shallow water equations, while achieving inundation area hit rates of between 0.8 and 0.9, relative absolute mean errors of 10%–20% across a wide range of flow depths, and comparable accuracy at forecasting empirical high-water mark levels following real-world events.
Estimates of potential flood inundation areas and depths are critical to informing the preparedness, response, and investment decisions of many government agencies and private sector organizations, especially under a changing climate. However, standard modeling approaches are often computationally intensive, limiting the scope of studies, the spatial resolution that can be achieved, and the ability to quantify forecast uncertainty. The combination of accuracy and efficiency provided by Torrent will significantly enhance real-time depth forecasts in the run-up to extreme events by supporting wider-area and finer-scale projections, enabling detailed sensitivity and uncertainty analyses related to future climate scenarios, and facilitating probabilistic depth and inundation area forecasts.
The rivulet-based algorithm is a novel approach that can support real-time risk assessments, studies of depth and inundation area sensitivity to climate change-driven variations in precipitation, the generation of large-scale, wide-area ensembles for uncertainty quantification, and probabilistic flood depth projections. A key benefit of the approach is that it allows researchers to adjust model parameters to balance computational complexity with projection accuracy, enabling more approximate projections to efficiently characterize a wide range of scenarios, with key events re-run at higher fidelity. In the present work, rivulet-based results were compared to shallow water equation-based (SWE) projections for the 2013 flooding in Boulder County, Colorado, and the 2017 landfall of Hurricane Harvey in Houston, Texas. The rivulet-based model demonstrated inundation area hit rates between 0.8 and 0.9 and relative mean absolute errors of 10%–20% across various flow depths when compared with the SWE solution. Additionally, it was found that the higher-resolution simulations enabled by the rivulet approach compensated for its simplified physics, performing comparably to SWE-based forecasts when validated against empirical high-water mark data from a real-world flood event.