CADAM

European Concerted Action on Dam-break Modelling

 

Jean-Charles GALLAND

 

EDF, Laboratoire National d'Hydraulique, 6 Quai Watier, 78400 Chatou, France

Tel.: +33 1 30 87 72 71, Fax: +33 1 30 87 80 86

Email: j-c.galland@edf.fr

 

Mark MORRIS

HR Wallingford, Howbery Park, Wallingford, Oxfordshire. OX10 8BA, UK

Tel: +44 1491 82 22 83, Fax: +44 1491 82 55 39

Email: m.morris@hrwallingford.co.uk

 

 

ABSTRACT

The nature of legislation in Europe addressing risk assessment for dam safety is varied, as are the numerical codes and techniques applied to predict the catastrophic failure of a dam and subsequent flood inundation. CADAM - a European Concerted Action on Dam-break Modelling - was established to gather and review knowledge and practice in order to optimise modelling technique and approach. The project runs for a two year period between February 1998 and January 2000. Topics covered include analysis and modelling of flood wave propagation, breaching of embankments and sediment effects. The programme of study is such that the performance of modelling codes are compared against progressively more complex conditions from simple flume tests through physical models of real valleys and finally to real dam-break cases. Conclusions from the study will be presented at a closing meeting to be held in Zaragoza, Spain in November 1999.

 

 

Keywords: Dam-break, numerical models, physical models, hydrodynamics, breach formation, sediment transport.

 

INTRODUCTION

The first legislation in Europe for dam-break risk analysis was presented in France in 1968, following the 1959 Malpasset dam-break that was responsible for more than 400 injuries. Since then, and especially more recently, many European countries have established legal requirements. However the techniques applied when undertaking the specified work can vary greatly. The perception of risks related to natural or industrial disasters has also evolved, leading to public demand for higher standards of safety and risk assessment studies. Considering the relatively high mean population density within Europe, a dam-break incident could result in considerable injury and damage ; efficient emergency planning is therefore essential to avoid or minimise potential impacts.

 

Dam-break analyses therefore play an essential role when considering reservoir safety, both for developing emergency plans for existing structures and in focussing planning issues for new ones. The rapid and continuing development of computing power and techniques during the last 15 years has allowed significant advances in the numerical modelling techniques that may be applied to dam-break analysis.

 

CADAM is funded by the European Commission and comprises a study programme running for a period of 2 years from February 1998. The project continues the work started by the IAHR Working Group (established by Alain Petitjean following the IAHR Congress in 1995) and has the following aims:

· The exchange of dam-break modelling information between participants, with a special emphasis on the links between Universities, Research Organisations and Industry.

· To promote the comparison of numerical dam-break models and modelling procedures with analytical, experimental and field data.

· To promote the comparison and validation of software packages developed or used by the participants.

· To define and promote co-operative research.

 

These aims are pursued through a number of objectives:

· To establish needs of industry, considering a means of identifying dam owners, operators, inspectors etc... throughout Europe.

· To link research with industry needs - encourage participation; distribute newsletters to dam owners and other interested parties.

· To create a database of test cases (analytical, experimental, real life) available for reference.

· To establish the state-of-the-art guidelines and current best practices for dam-break modelling within the technical scope of the Concerted Action. This leads towards establishing recommended European standard methods, procedures and practices for dam-break assessments.

· To determine future RTD requirements.

 

CONCERTED ACTION PROGRAMME

The project has participants from over 10 different countries across Europe. All member states are encouraged to participate, with attendance at the programme workshops open to all and to expert meetings by invitation. Also, links with other experts around the world are welcomed to ensure that state-of-the-art techniques and practices are considered. The programme of meetings planned for the presentation, discussion and dissemination of results and information is as follows:

 

Meeting 1 Wallingford, UK. 2/3^rd March 98 (Expert Meeting)

A review of test cases and modelling work undertaken by the group up to the start of CADAM, followed by a review of test cases considered during the last 6 months. Typical test cases include flood wave propagation around bends, over obstructions and spreading on a flat surface.

 

Meeting 2 Munich, Germany 8/9^th October 98 (Open Workshop)

Presentations and discussion on the current state of the art in breach formation modelling and sediment transport during dam-break events.

 

Meeting 3 Milan, Italy May6/7^th 99 (Expert Meeting)

Comparison and analysis of numerical model performance against physical models of real valleys. Update on breach modelling research.

 

Meeting 4 Zaragoza,Spain Nov 99 (Symposium)

A presentation of the results and conclusions drawn from the work of the Concerted Action over the two year study period (as well as the previous 6 months). The topics likely to be included are:

· Summary of the needs of industry

· Summary of benchmark test cases

· Review of state-of-the-art modelling software

· Review of state-of-the-art modelling practice

· Review of practical issues related to dam-break.

 

STUDY AREAS: PROGRESS SO FAR...

 

FLOOD WAVE MODELLING

The accurate prediction of flood wave propagation is fundamental to reliable dam-break simulation. In order to produce flood inundation maps and predict flood wave arrival times, both of which are essential outputs for use in emergency planning, it is necessary to predict both potential water levels and the progression of the flood wave downstream of the dam.

 

The performance of potential 'dam-break' numerical modelling schemes is being tested under progressively more difficult and complex conditions. The codes developed and applied by members of the CADAM group range from 1D schemes to more complex 1D/2D and 2D codes. Various numerical schemes, utilising different orders of accuracy, as well as various source term implementations, have been compared for a range of test cases (CADAM, March 98).

 

Test Cases

Initial test cases were relatively simple, with analytical solutions against which the numerical modelling results could be compared. These tests included:

· Flume with vertical sides, varying bed level and width. No flow - water at rest.

· Flume with (submerged) rectangular shaped bump. Steady flow conditions.

· Dam-break flow along horizontal, rectangular flume with a dry bed. No friction used.

· Dam-break flow along horizontal, rectangular flume with a wet bed. No friction used.

· Dam-break flow along horizontal, rectangular flume with a dry bed. Friction used.

 

These tests were designed to create and expose numerical 'difficulties' including shock waves, dry fronts, source terms, numerical diffusion and sonic points. Results were presented and discussed at the 2^nd IAHR Working Group meeting held in Lisbon, Nov. 96 (EDF, 1997).

 

Following these tests, a series of more complex tests were devised for which physical models provided data against which the numerical models could be compared. The aim was to check the ability of the codes to handle firstly, specific 2D features, and then important source terms. These tests were:

· Dam-break wave along a rectangular flume with 90° bend to the left.

· Dam-break wave along a rectangular flume with a symmetrical channel constriction.

· Dam-break wave along a rectangular flume expanding onto a wider channel (asymmetrical).

· Dam-break wave along a rectangular flume with 45° bend to the left.

· Dam-break wave along a rectangular flume with a triangular (weir type) obstruction to flow.

 

The first three test cases were presented and discussed at the 3^rd IAHR Working Group meeting in Brussels (UCL, June 97) and the remaining two at the 1^st CADAM meeting in Wallingford (CADAM, March 98).

 

Conclusions So Far...

From these initial tests it may be seen that many of the models offer results broadly comparable to the physical model or analytical solutions. Differences have been noted, however, between models using different numerical schemes to solve identical equations. Also noticeable is the variation in predicted arrival time of the flood wave, and reflected waves. This appears to depend upon the channel friction and the numerical method of implementing source terms.

 

All of the codes attempt to solve the Shallow Water Equations for dam-break situations. The question as to whether such equations are appropriate for use under all aspects of dam-break flow has been raised but with few alternatives offered at present. Furthermore, it can be concluded that the Shallow Water Equations allow for an accuracy which is 'acceptable' with respect to the general objectives of dam-break simulation (i.e. prediction of maximum water levels and wave arrival times), at least as far as pure hydrodynamics are concerned and providing efficient coding is used. However, it is also recognized that when other phenomena, such as sediment transport (see below) are included, a different set of equations may be more appropriate and this should be investigated further.

 

Whilst the differences noted for the flume tests appear to be relatively small, it is important to bear in mind that these may prove substantial when applied to a full dam-break study. This is supported by findings from a study undertaken in the US by Graham (CADAM, October 1998) comparing dam-break modelling results from two different modellers who were given identical original information. This study clearly demonstrates that large differences in predicted water levels and wave arrival times can arise from incorrect or differing assumptions with regards to channel or valley modelling.

 

As expected, 1D models have proved less accurate than 2D models under conditions where flow is obviously two dimensional, such as the spreading of a flood wave into a wider channel. The magnitude of error, and relevance to practical dam-break modelling, will be established during later model testing. Initial results suggest that the combined '1D model with 2D patches' may offer a proportionately large increase in accuracy for an acceptable increase in modelling time and effort.

 

Ongoing Test Cases

Two further stages of model testing are ongoing. The first is a comparison against dam-break physical models and the second against data from real dam-break events. By first undertaking fixed bed physical modelling of dam-break events it is planned that the performance of the numerical models may be assessed without the complications of sediment transport, debris, scour etc that feature prominently in any real failure event.

 

Two physical models being used. The first is of the Toce River in Italy and is at a scale of 1:100 covering a 5km length of the river valley. This model, constructed by ENEL, offers dam-break simulation along a real valley with a bunded lake in the centre. Flow into, out of and around the lake should prove challenging for the numerical models. The second model represented 4.5km of the Arade River valley in Portugal, and has been constructed at LNEC to a scale of 1:150. This model includes features such as a side valley, discharge onto a wide floodplain and areas of dead water storage. Discussion of modelling results may be found in the proceedings from the 3^rd CADAM meeting held in Milan (CADAM, May 1999).

 

EMBANKMENT BREACH FORMATION

The formation process for breaches is of particular importance when considering potential flood conditions relatively near to the dam site. The first day of the 2^nd CADAM meeting in Munich (CADAM, Oct 98), was devoted solely to this topic. Two test cases had also been specified for model comparison. The first based on physical modelling work undertaken at the Federal Armed Forces University in Munich and the second using data from the Finnish Environment Institute derived from a past collaborative research project undertaken with the Chinese. The modelling results submitted for comparison demonstrated the wide range of results predicted by current modelling approaches !

 

Whilst research continues in this area it is clear that there is still much work to be done. Conclusions drawn from group discussion, that summarise research priorities for the topic, are presented in the Munich meeting proceedings. Items identified under gaps in knowledge, understanding and practice include the need to link understanding in fluid and soil mechanics, the behaviour of cohesive materials, the erosion processes for real dams (rather than homogeneous models) and the piping / breach initiation phase.

 

To assist in promoting further research into breach formation CADAM has established a 'Breach Interest Group' via the Internet. The site may be accessed through the CADAM website and offers information and links about current research on breach formation and draws advice from modellers world-wide on key issues and problems in breach modelling.

 

DAM-BREAK SEDIMENT TRANSPORT

It is recognised that during a real dam-break event there can be significant sediment transport as well as debris transport. Large scale scour and deposition of material is likely to occur. This not only affects absolute flood water level and flood wave propagation but will lead to interaction with valley features and infrastructure so modifying behaviour of the flood wave. The build up and impact of debris can cause extreme local effects, attenuate flood waves by creating secondary dams and create new waves by causing secondary failures.

 

Given that such processes occur, it is logical that predicted water levels will vary significantly, as the bed level does. Current research using lightweight sediment to simulate transport under extreme conditions supports this theory (Capart et al, in CADAM, Oct. 98). The impact that such variations may have in comparison to the prediction accuracy of hydrodynamic models should be considered carefully to ensure a balanced approach for the development of tools contributing to an overall dam-break prediction.

 

Research is underway at a number of institutions across Europe looking at sediment, mud and debris flow. Details of this work may be found in the proceedings of the 2^nd CADAM meeting (CADAM, Oct 1998).

 

CONCLUSIONS

Numerical methods and computational hardware are now much more advanced than those available during the early eighties. This allows the models of today to simulate dam-break conditions more reliably and for more user friendly tools to be developed. CADAM offers a framework within which the performance of different numerical modelling schemes may be assessed, techniques and practice for modelling identified and areas where knowledge remains insufficient noted.

 

It is recognised that a dam-break model is created as a tool with which owners or authorities may assess the potential impacts of dam failure and take appropriate risk mitigation measures. To this end, the accuracy of flood level prediction and time of flood wave arrival is paramount. With this end user, and use, in mind, CADAM attempts not only to clarify and highlight numerical methods for modelling but also to identify related processes and how they interact with the estimation of flood conditions, including, for example, breach formation, valley roughness, debris, and sediment transport.

 

Full details of the work undertaken by CADAM may be obtained from the meeting proceedings, as published by the European Commission, or directly from the authors of this paper. In addition, information may be downloaded directly from the CADAM website, found at: www.hrwallingford.co.uk/projects/CADAM.

 

ACKNOWLEDGEMENTS

The members of CADAM are grateful for the financial support offered by the European Community to structure this concerted action programme. Without funding for research time or facilities, however, the project relies on contributions made by the group members, which is greatly appreciated. The teams who undertook modelling comprise :

 

Université Catholique de Louvain (Belgium), CEMAGREF (France), EDF/LNH (France), Université de Bordeaux (France), INSA Rouen (France), Federal Armed Forces University Munich (Germany), ENEL (Italy), Politechnika Gdanska (Poland), Universidade Tecnica de Lisboa (Portugal), Universidad de Zaragoza (Spain), Universidad Santiago de Compostela (Spain), Vattenfall Utveckling AB (Sweden), University of Leeds (UK), HR Wallingford (UK).

 

Special thanks should also go to the following organisations who have undertaken, or made available data from, physical modelling work:

 

Université Catholique de Louvain (Belgium), Université Libre de Bruxelles (Belgium), LNEC (Portugal), ENEL (Italy), Federal Armed Forces University Munich (Germany), The Finnish Environment Institute.

 

REFERENCES

EDF, 1997. Proceedings of the 2^nd AIRH Workshop on Dam-break Wave Simulation, Lisbon, Nov. 1996. Report LNH HE-43/97/016/B.

Université Catholique de Louvain, 1998. Proceedings of the 3^nd AIRH Workshop on Dam-break Wave Simulation, Brussels, June 1997.

CADAM, March 1998. Proceedings of 1^st CADAM workshop, Wallingford, March 1998.

CADAM, October 1998. Proceedings of 2^nd CADAM workshop, Munich, October 1998

CADAM, May 1999. Proceedings of 3^rd CADAM workshop, Milan, May 1999