An updated flow model called LISFLOOD-FP 8.1 enables more efficient flood inundation modeling at urban and catchment scales. It allows users to “coarsen” (lower) resolution on a non-uniform grid, which means it can operate up to 320 times faster than its predecessor to aid flood risk assessment.
The two-dimensional (2D) flow model LISFLOOD-FP is a hydrodynamic modeling framework that can simulate dynamic flood events and floodplain inundation over complex topographies. It has a broad range of applications and has been used to model morphodynamics, urban drainage, coastal flooding, population mapping and more.
Since its development by the University of Bristol in 2000, it has seen a number of developments and updates to refine its performance. This study presents v8.1 of LISFLOOD-FP.
The updated model: A new way to run the ACC uniform grid solver
LISFLOOD-FP breaks terrain into grids and predicts water depths in each at desired time steps, enabling it to track dynamic propagation of waves. LISFLOOD-FP v8.1 offers a new way to run the model’s ACCeleration (ACC) uniform grid solver, which can simulate gradual flooding over reasonably rough surfaces. Previously, the ACC solver had central processing unit (CPU)-specific optimizations and was accelerated using graphics processing units (GPUs).
LISFLOOD FP v8.1:
- Boosts the efficiency of the ACC solver’s existing architecture, by exploiting its ability to run parallelized flood simulations on the model’s GPUs rather than the CPU alone; and
- Presents a new non-uniform grid option for the ACC solver. This grid is generated via multi-resolution analysis of the digital elevation model (DEM). The solver sensibly coarsens (lowers) the resolutions of local topographic details that fall below a given error threshold, properly accounting for different classes of land use.
The results: Faster flood modeling at urban and catchment scales
The researchers tested v8.1 across five realistic case studies of fluvial and pluvial flooding over urban and catchment scales, a) assessing the efficiency of the GPU-parallelized ACC grid solvers (both uniform and non-uniform) and b) testing the accuracy of the non-uniform grid solver against its uniform predecessor.
On the GPU, the uniform ACC solver is 2–28 times faster than when run on the CPU and has an increased number of grid elements. If these elements are reduced and sensibly coarsened, the speed could be further increased by up to 320 times.
The predicted flood extents were deemed sufficiently accurate. For catchment scales, accuracy decreased with excessive coarsening on the non-uniform grid, while urban-scale simulations preserved accuracy with higher levels of coarsening.
The researchers concluded that LISFLOOD FP v8.1 allows faster flood inundation modeling at both urban and catchment scales.
Want to learn more about this research paper? Download it in full and dive in.
LISFLOOD-FP: The engine driving Fathom’s flood models
In September 2000, a research paper was published in the Journal of Hydrology, introducing the original LISFLOOD-FP,
Authored by Fathom’s Chairman Prof. Paul Bates and Ad P.J. de Roo, the paper had more than 1,000 citations and was the first time a grid-based 2D flood model had been described.
LISFLOOD-FP model is widely recognized as a critical tool for creating, testing and applying high-resolution, large-scale flood hazard simulations,
More than 20 years after it was created, newer, faster versions are being developed, and LISFLOOD-FP is still the powerful engine behind Fathom’s flood continental and global-scale models such as Fathom-US.