Seminar - Development of computationally efficient flood inundation models for use with new high-resolution terrain data
The quantity, resolution and accuracy of remotely sensed terrain data have increased substantially over the last 5 years. Airborne LiDAR data down to 1m resolution are now available over whole regions, whilst new vehicle-mounted terrestrial LiDARs can provide DEMs with centrimetric horizontal resolution and millimetric relative accuracy over a few km2. Detailed modelling studies over limited areas have shown that small- scale terrain features captured in such data make a significant difference to the flooding patterns. However, this presents a considerable challenge for whole city scale flood risk analyses as the computational cost of performing hydraulic simulations at 1-2m spatial resolution for full dynamic events is currently prohibitive. Typical large scale flood risk analyses are run at ~50m spatial resolution even for urban areas where there are terrain feature length scales (buildings, embankments, walls, kerbs etc) considerably smaller than this. Such flooding analyses can be shown to significantly mis-estimate the risk because of their too coarse resolution, yet each halving of the grid resolution increases the computational cost by 1-2 orders of magnitude. Hence a ~1.5m grid model which could capture the necessary terrain features would be 105 to 1010 times more expensive to run than one at 50m.
This talk describes the development of a computationally efficient flood inundation model that overcomes this problem to allow whole city flood risk analyses (domain extents of 10-100 km2) at fine spatial resolution (1-5m). The new model is shown to be significantly faster than alternative methods (up to 1000x faster than diffusive wave storage cell models), yet yields accurate results for gradually varied sub-critical flows problems when compared to analytical solutions and remotely sensed flooding images. New high-resolution remotely sensed terrain data sets are presented and a number of example applications shown to demonstrate the ability of the model to be applied at or close to the native resolution of these data.
Paul Bates is Professor of Hydrology at the School of Geographical Sciences in the University of Bristol. He is Director of the Cabot Institute, a world-class multidisciplinary institute for research on all aspects of global environmental change, from basic science and social science to technological and policy solutions. It brings together some of Bristol’s most outstanding research – in natural hazards and risk, Bayesian statistics, uncertainty and decision-making, climate modelling, poverty, global insecurities and governance, and systems engineering.
Contact email@example.com for more information about this event.