SEAMLESS-WAVE
SEAMLESS-WAVE is a developing “SoftwarE infrAstructure for Multi-purpose fLood modElling at variouS scaleS” based on "WAVElets" and their versatile properties. The vision behind SEAMLESS-WAVE is to produce an intelligent and holistic modelling framework, which can drastically reduce iterations in building and testing for an optimal model setting, and in controlling the propagation of model-error due to scaling effects and of uncertainty due statistical inputs.
Smart and flexible multiresolution solvers
Flow of water in a flood event frequently depends on small-scale topographic features, particularly for urban flood events. Flood simulations must represent many square kilometres with a resolution of a metre or smaller. Conventional flood simulations are performed on a static, uniform grid, but this becomes computationally prohibitive when the simulation area is very large and the finest resolution is very small since the resultant grid has many millions of cells.
Adaptive flood simulators (e.g. TUFLOW-HPC Quadtree) reduce the computational burden by using dynamic, or static, adaptivity on non-uniform grids to capture the important features of the topography and the flow. Where the topography or flow features are smooth, adaptive grid methods can coarsen the grid resolution to reduce the total number of cells.
Conventional adaptive grid methods can introduce simulation errors through heuristic interpolation of flow data that occurs across different grid-resolution patches on the non-uniform grid. These errors can result in inaccurate predictions of flood extent, water depth and flow velocity (e.g. see Fig. 14b in Kesserwani and Liang 2012). These methods can also be inflexible to practitioners, given the need to introduce extrinsic sensors to post-process slope data, to tune several parameters to find case-sensitive coarsening vs. refinement, and to alleviate error propagation.
Multiwavelets bases are naturally compatible with the piecewise-polynomial bases shaping the local solution structure within the Discontinuous Galerkin (DG) method, hence also the Finite Volume (FV) method featuring in industrial flood models (Ayog et al. (2021)). Multiwavelets’ multiresolution analysis enables to dynamically scale an adaptive solution that overcomes the majority of the above-mentioned shortcomings of conventional adaptive grid methods (Kesserwani et al. 2019; Kesserwani and Sharifian 2020). In other words, the multiwavelet-based adaptivity offers a solid mathematical foundation to design smart and flexible multiresolution models that:
-
deploy rigorous filter transforms for inherent encoding and decoding of the modelled data across cascades of resolution scales, and
-
only need one user-specified parameter to both directly controls adaptivity error and sensibly decide where to refine and where to coarsen.
We have developed and investigated (multi)wavelet-based adaptive approaches for flood modelling (Caviedes-Voulliéme and Kesserwani 2015; Gerhard et al. 2015). The first-order finite volume model (FV1) is combined with Haar wavelets to produce the Haar-FV1 (HFV1) model. Similarly, the second-order discontinuous Galerkin model (DG2) is combined with Alpert multiwavelets (Alpert 1993) to produce the multiwavelet-DG2 (MWDG2) model. A complete description of how one-dimensional (1D) hydrodynamic MWDG2/HFV1 solvers can be formulated is found in Kesserwani et al. (2019), with a diagnostic analysis on its potential for real-world problems and links to access the relevant codes and data.
To design two-dimensional (2D) hydrodynamic MWDG2/HFV1 solvers that are robust for real-world applications and gain even more efficiency for 2D hydraulic simulations (Caviedes-Voulliéme et al. 2020), a redesign has been proposed (Kesserwani and Sharifian 2020). This redesign adapts multiwavelets theory to fit the setting whereby 2D solvers are efficient and robust for real-world applications, and is a significant step forward to demonstrating how multiwavelet-based 2D solvers can be made operational for 2D flood inundation modelling.
Practicability and code accessibility
Details on how to access and run educative 1D and 2D codes are documented within Kesserwani et al. (2019) and Kesserwani and Sharifian (2020), for a range of synthetic and laboratory-scale test cases.
Moreover, fully GPU-resident versions of the (multi)wavelet-based adaptivity can be found in Kesserwani and Sharifian (2023), considering their abilities to pre-process high resolution DEMs and for real flood simulation case studies. The code has been integrated into the LISFLOOD-FP flood modelling suite (version 8.1 or higher), and its potential for tsunami generated coastal floods is presently under investigation.