CNES JC2 Poster

Anticipating flood-induced morphological damage from high-resolution topography

Anticipating where floods erode and deposit sediment is essential for assessing hazards and understanding landscape evolution.

Conventional flood models hit a computational wall: a simulation covering just 3 km² at metre resolution can take tens of hours.

Our new method computes 400 km of the Waiau Toa / Clarence river valley in less than 5 hours, across 12.52 billion terrain cells, by combining out of core (OOC) terrain processing with GPU simulations of simplified flood dynamics at peak discharge.

Method references

Terrain processing: Barnes (2016), Barnes, Lehman & Mulla (2014) and Barnes (2017) underpin tiled depression filling, flow directions and global flow accumulation for terrain too large to fit in memory.
Inertial flow: Bates, Horritt & Fewtrell (2010), de Almeida et al. (2012) and Sridharan et al. (2020) provide the local inertial formulation and numerical refinements used here for efficient GPU simulations of peak-flow depth and shear stress.

Code availability. The code will be released as open source once stable. The current implementation builds on pyfastflow, pytopotoolbox and LSDTopoTools, combining GPU flow processing inspired by FastFlow (Jain et al., 2024) with GraphFlood methods (Gailleton et al., 2024) to accelerate convergence to a stationary solution in a hybrid local inertial solver.

Next steps

I will use DEM time series derived from Pléiades, Pléiades Neo and CO3D to simulate floods for each observation date and track changes in channel width and shear stress. Integrating sediment fluxes will then allow me to test whether these patterns can explain the observed erosion and deposition—and whether the initial topography alone can anticipate their location and magnitude.

Base data

The base DEM uses open LiDAR elevation data made available by Toitū Te Whenua — Land Information New Zealand (LINZ); thanks to LINZ and the contributing data providers for making these datasets accessible.

Browse and download the data through the LINZ Data Service, or access the elevation catalogue and cloud data via the LINZ elevation repository.