| Home Contact us Gallery Site Map Resource Centre Search | ||
| About us People News Publications Education & Training Events | ||
Download report(3.17MB PDF)
Foreword
Author's Note
Executive Summary
Introduction
• Atmospheric
Hazards
• Geological Hazards
• Hydrological
Hazards
• Climate Change
Sources & Further reading
| Hazard & Risk Science Review
2006 5. Hydrological Hazards | |
|
Along with windstorms and earthquakes, floods make up one of the ‘big three’ natural hazards, responsible for most deaths, damage, and economic and insured losses. While climate change is forecast to contribute to more flooding through, for example more extreme precipitation events, and ‘wetter’ tropical cyclones, other factors are also contributing to an increasing flood threat, in particular anthropogenic factors such as urbanisation and changes in land use. For major river basins, a substantial rise in the frequency of serious floods has already been demonstrated for the second half of the 20th century (figure 12). Flooding, therefore, will undoubtedly be a major issue in coming decades, and improving flood forecasting will take on ever-increasing importance.  Figure 12. Flooding of the Elbe at Dresden in August 2002. Major flood events are certain to become more common as the world continues to warm. Courtesy: Stefan Malsch Flood forecasting issues It is timely, therefore, that Gabriel Arduino 3 of the World Meteorological Organisation in Geneva, and colleagues, have edited a comprehensive special issue of Hydrology and Earth System Science, dedicated to advances in flood forecasting. In a concise, but excellent overview of the subject, Arduino and co-authors address the most critical issues, including precipitation forecasting, the different types of hydrological models - most notably data-driven and process-driven, the coupling of such models with precipitation forecasts, assessing flood risk, operational flood forecasting, and assessing predictive uncertainty. In this latter context, they stress that - to be of value - a flood warning must be associated with an objective measure of uncertainty, which generally increases with the lead-time required to implement flood protection measures. The authors point out that quantifying uncertainty is becoming increasingly relevant, as political pressures mount to provide medium- to long-term flood forecasts from numerical weather predictions. One aspect of the uncertainty theme is tackled by Matt Horritt 29 of the University of Bristol (UK), who presents a methodology for validating uncertain flood inundation models, in a paper published in the Journal of Hydrology. Horritt points out that without a means of comparing uncertain model predictions with the small number of observed flood events, it is impossible to assess the performance of uncertain models, particularly with respect to large magnitude events. To address this issue, the author introduces a means for comparing uncertain flood model predictions with single observations of flood extent. In a case study, this is applied to a stretch of the UK’s River Severn, for which satellite imagery of a number of recent flood events is available. Horritt’s methodology provides insights into the differences between model calibration and validation processes, and into the effectiveness of satellite sensor data in constraining uncertain flood model parameters, permitting uncertainty and precision in model predictions to be quantified in a meaningful and useful manner. Returning to the Arduino et al. special issue of Hydrology and Earth System Science, Jens Bartholmes of the EC’s Joint Research Centre at Ispra (Italy), and Ezio Todini 6, of the University of Bologna, look at the problem of coupling meteorological and hydrological models to improve flood forecasting. They focus on a recent issue that has resulted in a disparity between the two types of model. Increased computer capabilities now permit the utilisation, in flood forecasting, of distributed hydrological models with very fine mesh sizes of between decimetres and a few kilometres. In contrast, today’s meteorological models, designed to provide quantitative precipitation forecasts, tend to produce average values based upon much coarser mesh sizes of between around 10 and 200 km. In order to examine the effects of such a mismatch, the authors built a hydrological model for part of the River Po, using a 1 x 1 km mesh, and coupled this with a range of European meteorological models with different spatial resolutions. In respect of improving flood forecasting, Bartholmes and Todini feel that the results were in part encouraging, and in part disappointing. The main problem hindering better deterministic quantitative flood forecasts for an extended time horizon (several days ahead), the authors note, lies in the poor performance of the lower resolution meteorological models. Until the precipitation prediction issue is tackled, on the meteorological side, and the uncertainty issue on the flood modelling side, Bartholmes and Todini conclude that only warnings, as opposed to deterministic, quantitative flood forecasts are feasible for an extended time horizon. Continuing the theme of flood forecasting through utilizing weather prediction data, Ben Gouweleeuw 28 of NASA’s Goddard Space Flight Centre, and colleagues, take a look - in the same volume of Hydrology and Earth System Science - at how medium-range probabilistic weather prediction may be used for operational flood forecasting. They note that following similar developments in short- to medium- range weather forecasting over the last decade, operational flood forecasting is now beginning to adopt a probabilistic ‘ensemble’ approach, rather than the more traditional ‘single solution’ or ‘best guess’ method. This is managed through combining ensemble weather prediction systems with rainfall-runoff models. Gouweleeuw and colleagues, developed a system along these lines, which combines - for floods in the last decade on the Meuse (Germany) and Odra (Poland) rivers - the large-scale hydrological model LISFLOOD, with deterministic and Ensemble Prediction Systems (EPS) hindcasts of the global Numerical Weather Prediction (NWP) model of the European Centre for Medium-Range Weather Forecasts (ECMWF). The authors note that, in respect of improving aspects of forecasting, such as the timing of flood peaks, the system proved far better at replicating the 1995 Meuse flood, than the 1997 Odra event, leading them to suggest that, although encouraging, its performances may differ from event to event and/or from region to region. Predicting extreme precipitation While many factors are involved in the development of major floods, extreme precipitation is clearly a major one, if not the major one. Predicting incidences of extreme precipitation is, therefore, an important element of flood forecasting itself. Such prediction is, however, notoriously difficult, due - amongst other things - to the inherent uncertainties in short-range quantitative precipitation forecasts. In another paper in the special issue of Hydrology and Earth System Science, edited by Arduino and others, Kai Sattler and Henrik Feddersen 58, of the Danish Meteorological Institute, focus on the prediction of extreme rainfall events over European river basins. Using a limited-area ensemble approach, they address three periods of c. 10 days duration, around heavy rain events that caused or contributed to disastrous river flooding in Europe: the November 1994 Piemonte (Italy) flood, the January 1995 Rhine and Meuse floods, and the July 1997 Odra flood. The ensemble predictions were measured against actual precipitation at the time of the three floods and their performance assessed, in particular over the affected river catchments. The two ensemble designs tested proved to be superior to that of a single control forecast over the river basins, and each ensemble had particular strengths that suggested a combination of the two might prove even more effective. Assessing flood risk and hazard Traditionally, flood defences are designed by specifying an exceedance probability and by demonstrating that the flood defences are capable of coping with floods that correspond to this exceedance probability. Writing in Natural Hazards, however, Heiko Apel 2 of GeoForschungsZentrum, and colleagues, suggest that a more comprehensive assessment of flood risk can be provided through application of a simple, probabilistic model, rather than the more sophisticated and complex deterministic models used by others. This, they indicate, should include both the estimate of the flood hazard (e.g. runoff and associated probability), and the consequences of the flooding (e.g. property damage and loss of life). Apel and his group present such an assessment, developed for Germany and involving a large number of simulations run within a Monte Carlo ‘framework’, and describe its application to a stretch of the Rhine river between the cities of Cologne and Rees. Here, they show how it can be used to quantify many aspects of flood risk, including levee failure, and damage costs. They highlight the versatility of the approach, and propose its general application to the integrated assessment of flood risk in flood-prone areas, for cost-benefit assessment and risk-based design of flood protection measures, and as a decision-support tool for flood management. In an interesting paper in Quaternary International, Philipp Schmidt-Thomé 60 of the Geophysical Survey of Finland, and colleagues, present colourful, broad-brush, first impressions of economic risk due to flooding (and earthquakes) across Europe. The paper incorporates a series of maps developed as part of the thematic project on Natural & Technological Hazards, of the European Spatial Planning Observation Network (ESPON). The maps express the spatial significance of floods (and earthquakes) as an economic risk, by combining hazard potential and vulnerability data. These types of plot are designed to highlight those areas that are economically most likely to suffer as a result of a major flood or earthquake, thereby facilitating targeted mitigation policies and responses. With respect to flooding, the end product of the study is the first aggregated flood map of Europe, based on events over the period 1987 to 2002, a flood intensity may be determined from the above, and a flood risk map. Large flood events are seen to concentrate in north-western Romania, southeastern France, central and southern Germany and eastern England. Once vulnerability is factored in, however, it is apparent - perhaps not surprisingly - that the areas of highest economic risk are located in central Europe. Flood modelling data for insurance purposes In respect of river flooding, insurers face two problems; firstly, assessing their exposure, and secondly, determining how best to price the flood element of their insurance products. Both issues are addressed by Richard Sanders 57 of Willis Ltd., and co-authors, in Hydrology and Earth System Science, who look at the insurance implications of recent flood events in Europe, and the issues surrounding the insurance of future events. Sanders and his colleagues focus, in particular, on the flood risk requirements of insurers, and how these can best be met. In this regard, the authors provide a very useful summary of the current situation, drawing attention to the data requirements of national and regional flood models in the context of the accuracy of available data on property location. Noting that terrain data are often the weakest component of sophisticated flood models, Sanders and his co-researchers discuss various sources of such data and looks at their advantages and disadvantages. They conclude that the NEXTMap DTM (Digital Terrain Model) series, from Intermap Technologies, is well suited to flood risk identification and mapping. This is based on IFSAR technology, which uses airborne radar to generate high resolution DTMs within a limited time-scale. IFSAR data for the UK were collected from altitudes of between 20 and 28,000 ft, providing a vertical accuracy for elevation data of ± 1 m. «back to top« |
Hydrological Hazards Flood forecasting issues Predicting extreme precipitation Assessing flood risk and hazard Flood modelling data for insurance purposes |