Flood Simulation

Generally, floods are one of the most common natural risks to human beings because most of the populated areas in the world are vulnerable to flood disasters, whereas other natural hazards such as earthquakes, volcanic activities, landslides and avalanche are particular to certain regions only.  In long-term analysis, flood disasters account for about one third of all natural catastrophes (evaluated by frequency of occurrence and economic losses induced), and cause more than half of the life losses due to huge flood-induced human cost in the developing and heavily populated countries such as China and Bangladesh.  Fortunately, the number of deaths caused by floods has decreased greatly due to the progress in flood warning methods in recent years.  An example is the Yangtze River in China.  The 1931 and 1954 floods killed 145,492 and 33,000 people respectively.  Whilst the 1998 flood was the largest during the previous 44 years in China, the loss of life was reduced to 3,656.  However, world-wide, floods are likely to become increasingly severe and more frequent due to climate change, population growth, change of land-use, irrigation, deforestation, and urban development of flood plains.  Therefore flood disasters should attract more attention, and more work is needed on flood risk assessment and mitigation.

Flood risk assessment normally involves in running a hydraulic model many times within a Monte Carlo platform to predict the probability of flooding.  Therefore accurate simulation of flooding waves is critical in order to provide a reliable flood risk map.  Because of the restriction of available data for parameterisation and validation and the concern of high computational cost, simplified two-dimensional numerical models have been popular for predicting large-scale fluvial and coastal flooding.  A simplified model normally ignores most of the dynamical aspects induced by flooding waves in order to simplify the governing equations and thus to reduce computational cost.  However, flooding waves generally involve very complicated hydraulic process, e.g. subcritical and supercritical flow, shock-type discontinuity, and wet-dry interface.  Accurate prediction of the evolution of flooding waves, including routing and arriving time, and its interaction with structures would become impossible without accurate representation of the complex hydrodynamic effects, which is beyond the scope of a simplified model.  In recent years, rich sources of high-resolution data provided by remote sensors and airborne scanning laser altimetry (LiDAR) become available for model parameterisation and validation.  Together with the modern development of computational power and simulation techniques, it is possible to predict large-scale flooding using the full two-dimensional solution of shallow water equations.

An adaptive quadtree grid based shallow water equation solver is reported to save computational cost more than six times compared with its counterpart on a uniform grid.  With a quadtree grid, the geometry of a flood plain can be accurately represented with local fine grids applying to those areas with structures, dramatic change of elevation and vegetation height, etc.  A quadtree grid can also easily adapt to the local flow features and moving wet-dry interface.  Therefore the dynamically adaptive quadtree-grid based Godunov-type shallow water equation solver can provide accurate prediction of complex hydraulic process with relatively low computational cost, and is a ideal tool for simulating a large-scale flood event.

Back to Research