Predicting which areas are at risk from flooding is fraught with difficulty. Often we are concerned about the risk from very large floods which have not yet occurred. Even when we do experience very large events, such as the 1953 east coast floods, the priority is rightly on saving lives and not collecting scientific data. We also know that flooding of urban areas involves complex flow patterns and is strongly influenced by micro-scale features such as road layout, buildings, walls and fences.
In this situation computer models that can simulate flood flow are an obvious answer, but although the mathematical and computing knowledge to build these has been around for nearly 50 years, two factors have prevented their widespread use. First, simulating the detail of flow through and between buildings for large city areas has, until five years ago, been beyond the scope of available computers. Second, even with the computer power to build a model at the right resolution, the data to describe all the building features which control flooding have, until recently, been lacking.
This situation began to change about 10 years ago and from the beginning a team of scientists from the Departments of Geographical Sciences and Civil Engineering has been at the forefront of this research. The starting point in the transformation of our ability to model floods was the purchase by the Environment Agency of an instrument called a LiDAR.
LiDAR stands for Light Detection and Ranging and the instrument works by firing a laser pulse at a target and timing how long it takes to be returned back to the sensor. The laser beam moves at the speed of light so we can therefore calculate very precisely how far away the target is. So far, so good; however, the neat trick is that it was designed to be mounted on a light aircraft and to fire many thousands of pulses per second whilst scanning from side to side. Using the information gained this way, we can calculate the height of the ground surface for each of the thousands of laser shots. Modern LiDAR systems fire 100,000 shots per second and can measure the land height every 25 cm to a vertical accuracy of less than 8 cm. Being a light-based system, LiDAR can also penetrate many vegetation canopies to measure the ground height beneath.
However, processing LiDAR data is not straightforward and the Bristol team, along with colleagues at the University of Reading, have spent considerable effort developing algorithms to separate out vegetation and buildings from ground surface laser ‘hits’. More recently, we have begun combining LiDAR and digital map data to deal with some of the more detailed features present in urban areas. This even goes as far as identifying and removing the parked cars and garden sheds from LiDAR data to create a true ‘bare earth’ terrain model.
How can we be sure our models are producing realistic results?
Being able to map urban areas in detail is, however, only part of the story and we also need to develop computer models capable of using the vast amounts of new data now available. With approximately 50 per cent of England and Wales mapped using LiDAR and risk estimates required for individual properties, such models need to be both efficient and accurate. Here, the Bristol team has been developing a suite of computer models of varying complexity to apply in different situations. One particular code, LISFLOOD-FP, is now used by researchers at over 20 universities worldwide, including some in developing countires such as Brazil, where it is used to address local problems. Collaborators have also worked in Egypt and Ghana. In addition, the ideas behind LISFLOOD-FP have been used in a number of other research and commercial models. This includes the model used by the Environment Agency to produce maps of extreme flooding for the whole of England and Wales. When the predictions from such models are combined with other geographical property databases, the potential to calculate likely damages, identify critical risks – such as care homes or hospitals in the floodplain – and plan safe evacuation routes is obvious.
But how can we be sure that our models are actually producing realistic results? Well, here again aircraft and satellites have a role to play. In particular, highly developed radar systems can be used to map flooding to a high level of detail whilst it happens. A particular recent success was the use of a military-specification airborne radar operated by QinetiQ Ltd, during the November 2000 floods. This was used four times over nine days to map flooding along the River Severn at Upton-on-Severn at one metre resolution and has given us unprecedented insights into flood dynamics that we can compare with our model predictions.
As in other areas, remote sensing technology for flood science is not standing still. Truck-mounted LiDAR systems which can map buildings in 3D and cheap, networked sensors which can be deployed in large numbers to transmit water depth information in real time are just two technologies which can potentially continue the revolution which began with airborne LiDARs. The develop-ment of new High Performance Computing approaches at Bristol will also allow us to build more detailed models of bigger areas than ever before. Whatever the future holds, we know that whilst we may never eliminate all flood risks, we may at least be better informed to mitigate and manage them.
Paul Bates/School of Geographical Sciences