LISFLOOD-FP 8.1: new GPU-accelerated solvers for faster fluvial/pluvial flood simulations

Abstract

Abstract. The local inertial two-dimensional (2D) flow model on LISFLOOD-FP, the so-called ACCeleration (ACC) uniform grid solver, has been widely used to support fast, computationally efficient fluvial/pluvial flood simulations. This paper describes new releases, on LISFLOOD-FP 8.1, for parallelised flood simulations on the graphical processing units (GPUs) to boost efficiency of the existing parallelised ACC solver on the central processing units (CPUs) and enhance it further by enabling a new non-uniform grid version. The non-uniform solver generates its grid using the multiresolution analysis (MRA) of the multiwavelets (MWs) to a Galerkin polynomial projection of the digital elevation model (DEM). This sensibly coarsens the resolutions where the local topographic details are below an error threshold ε and allows classes of land use to be properly adapted. Both the grid generator and the adapted ACC solver on the non-uniform grid are implemented in a GPU new codebase, using the indexing of Z-order curves alongside a parallel tree traversal approach. The efficiency performance of the GPU parallelised uniform and non-uniform grid solvers is assessed for five case studies, where the accuracy of the latter is explored for ε=10-4 and 10−3 in terms of how close it can reproduce the prediction of the former. On the GPU, the uniform ACC solver is found to be 2–28 times faster than the CPU predecessor with increased number of elements on the grid, and the non-uniform solver can further increase the speed up to 320 times with increased reduction in the grid's elements and decreased variability in the resolution. LISFLOOD-FP 8.1, therefore, allows faster flood inundation modelling to be performed at both urban and catchment scales. It is openly available under the GPL v3 license, with additional documentation at https://www.seamlesswave.com/LISFLOOD8.0 (last access: 12 March 2023).

Toward Street‐Level Nowcasting of Flash Floods Impacts Based on HPC Hydrodynamic Modeling at the Watershed Scale and High‐Resolution Weather Radar Data

“…In the future, it would be interesting to extend this modeling approach to bigger watersheds to test the overall run time, even considering more HPC‐enabled solutions to enhance the computational efficiency (Caviedes‐Voullième et al., 2023; Morales‐Hernández et al., 2021; Sharifian et al., 2023).…”

Sampling‐Based Methods for Uncertainty Propagation in Flood Modeling Under Multiple Uncertain Inputs: Finding Out the Most Efficient Choice

“…For probabilistic modeling of slowly propagating fluvial/pluvial floods over large catchments, a less complex (mathematically and numerically) physical solver may be more appropriate to expedite runtimes, such as the reduced acceleration solver on LISFLOOD‐FP (Beevers et al., 2020), which has a version for the GPU with grid‐resolution adaptivity to maximize runtime efficiency (Sharifian et al., 2023). Alternatively, physical solvers based on distributed hydrological modeling can be used for multi‐physics modeling such as to incorporate the feedbacks between groundwater and land surface processes (for example, Baroni et al., 2019).…”

“…Therefore, a uniform inflow discharge random variable, Q ( t , ξ Q ), follows: $$$Q\left(t,{\xi }_{Q}\right)=\overline{Q}(t)+{\xi }_{Q}\,{\sigma }_{Q}(t)$$$ where ξ Q is a random variable taking values in [−1,+1] and $${\sigma }_{Q}(t)=0.08\overline{Q}(t)$$ is the range of variation with respect to the mean $$\overline{Q}(t)$$. In flood modeling, a number of P mean inflow discharges, $$\overline{{Q}_{1}}(t)$$, …, $$\overline{{Q}_{P}}(t)$$ ( P > 1), can be given (Kesserwani & Sharifian, 2023; Sharifian et al., 2023). In this case, the variation in each of $${Q}_{1}\left(t,{\xi }_{{Q}_{1}}\right)$$, …, $${Q}_{P}\left(t,{\xi }_{{Q}_{P}}\right)$$ follows Equation , assuming that $${\xi }_{{Q}_{1}}$$, …, $${\xi }_{{Q}_{P}}$$ are not intercorrelated (Neal et al., 2013), leading to an uncertainty space dimension of D = P .…”

“…For the considered test cases (Section 2.5), a constant mean Manning coefficient value is used, thus further assuming $$\overline{n}(x,y)=\overline{n}$$ to reduce n ( x , y , ξ n ) to n ( ξ n ). Note that, Equation can be further used considering spatially varied mean Manning coefficient data, $$\overline{n}(x,y)$$, incorporating many land use types and channels (Sharifian et al., 2023), without loss of generality. However, using spatially varied roughness would lead to similar performance as with a uniform constant roughness (Alipour et al., 2022).…”

The Role of Different Total Water Level Definitions in Coastal Flood Modelling on a Low-Elevation Dune System

“…Numerical models can be used to provide crucial information to understand, predict, and, therefore, mitigate the impacts of coastal storms. Their development and availability have increased over the last decades, covering different areas: XBeach [3] and D-Morphology [4] for the analysis of morphodynamic changes, SWAN [5], WWIII [6], and FUNWAVE [7][8][9][10] for wave dynamics, Lisflood-FP [11][12][13], ROMS [14], SCHISM [15], TELEMAC [16], MIKE 21/3 [17], and D-Flow FM [4] for flood propagation. These numerical models can also be valuable tools if used in operational mode in support of early warning systems, providing relevant information for local authorities, such as flood extension or morphological impact, prior to the event.…”