Workshop Extreme Discharges

Day 1 - 18 April 2005

Registration with lunch - from 11.00 a.m.

13.00 - 13.10
Welcome + introduction and goals of the workshop by Prof. Dr. Manfred Spreafico, President of CHR

Theme block I - How extreme is extreme?

Chair: Prof. Dr. Dieter Gutknecht
Rapporteur: Prof. Dr. Rolf Weingartner

13.10 - 13.30
Extreme floods in the Elbe catchment - Prof. Dr. Uwe Grünewald, Brandenburg Technical University Cottbus

13.30 - 13.50
Extreme discharges in the Meuse basin - Dr. Marcel de Wit, RIZA Arnhem

13.50 - 14.10
Does the perception of extremity change? An ongoing case study in the Sure river basin - Hugo Hellebrand, Centre de Recherche Public Gabriel Lippmann Luxembourg

14.10 - 14.30
Extreme scenarios and flood risk management - Dr. Bruno Merz and Dr. Annegret Thieken, GeoForschungszentrum Potsdam

14.30 - 15.00
Coffee and Thee + Poster Session

15.00 - 15.20
Extreme precipitation in Germany - Dr. Gabriele Malitz, German Weather Sevice Berlin

15.20 - 15.40
The August flood 2002 in the eastern Erz mountains and its statistical evaluation - Prof. Dr. Andreas Schumann, Ruhr University Bochum

15.40 - 16.00
Climate change and hydrological extremes - Prof. Dr. Christoph Schär, Atmospheric and Climate Science, ETH Zürich

16.00 - 16.20
Regional flood process types - Dr. Ralf Merz and Prof. Dr. Günter Blöschl, Technical University Vienna

16.20 - 17.00
Discussion on theme I

17.00 - 17.30
Snacks and refreshments + Poster Session

Theme block II - Determination of extreme discharges

Chair: Prof. Dr. Andreas Schumann
Rapporteur: Dr. Marcel de Wit

17.30 - 17.50
Derived frequency methods for estimating flood from rainfall frequencies - Prof. Dr. Günter Blöschl and Dr. Ralf Merz, Technical University Vienna

17.50 - 18.10
Estimating design river discharges using generated series and Bayesian statistics - Houcine Chbab, RIZA Lelystad

18.10 - 18.30
Improving the estimation of the magnitude of extreme floods by considering dominant runoff processes and historical floods - Dr. Felix Naef, Institute for Hydromechanics and Water Management, ETH Zürich

18.30 - 18.50
Modeling series of extreme flood events for (re)insurance purposes - Dr. Jens Mehlhorn, Swiss Reinsurance Company Zürich

19.30
Participants dinner

Day 2 - 19 April 2005

Theme block II (continuation)

08.30 - 08.50
Uncertainty in flood quantiles from basin and river models - Dr. Ferdinand Diermanse and ir. Henk Ogink, WL | Delft Hydraulics

08.50 - 09.10
Fuzzy logic approach for reducing uncertainty in flood forecasting - Dr. Shreedar Maskey, UNESCO-IHE Delft

09.10 - 09.50
Discussion on theme II

09.50 - 10.20
Coffee and Thee + Poster Session

Theme block III - Can we assess the effects of human interventions on extreme events?

Chair: Prof. Armin Petrascheck
Rapporteur: Dr. Günter Blöschl

10.20 - 10.40
Extreme value statistics for instationary conditions - Methods for consideration of climate change in the determination of design discharges - Prof. Dr. András Bárdossy, Institute for Hydraulics, University of Stuttgart

10.40 - 11.00
Effects of widely spread retention measures on the flood development in the Rhine - Dipl.-Ing. Heinz Engel, Federal Institute for Hydrology, Coblence

11.00 - 11.20
Impact assessment of flood mitigation measures - Dr. Matthijs Kok, HKV Lelystad

11.20 - 11.40
Risk management in the community - Reinhard Vogt, Flood Protection Centre Cologne

11.40 - 12.00
Increasing flood losses indicate increasing flood risk. What are the responsible parameters? - Dr. Wolfgang Kron, Munich Reinsurance Company

12.00 - 13.30
Lunch

13.30 - 13.50
Transboundary effects of flooding and flood reducing measures along the Rhine in Northrhine-Westfalia (Germany) and Gelderland (Netherlands) - Dr. Rita Lammersen, RIZA Arnhem and Dr. Bernd Mehlig, Landesumweltamt Northrhine-Westfalia, Düsseldorf

13.50 - 14.10
Large scale simulation of land use change effects on floods in the Rhine basin (results from the LaHoR project) - Prof. Dr. Axel Bronstert, University of Potsdam

14.10 - 14.30
Hydra-models, a way to assess the influence of climate change and river programs on future dike heights and the probability of dike failure - Robert Slomp, RIZA Lelystad

14.30 - 15.00
Discussion on theme III

15.00 - 15.30
Coffee and Thee + Poster Session

15.30 - 16.00
Final discussion, summary and conclusions of the workshop

16.00
Closing of the workshop 

Dieter Gutknecht (chair)
Institute of Hydraulics and Water Resources Engineering
Technical University Vienna
Karlsplatz 13
A-1030 Vienna
Tel. +43-1-58801 22300
gutknecht@hydro.tuwien.ac.at

Henk Ogink
WL | Delft Hydraulics
Postbus 177
NL-2600 MH Delft
Tel. +31-15-285 8507
henk.ogink@wldelft.nl

Eric Sprokkereef
Secretariat CHR
Postbus 17
NL-8200 AA Lelystad
Tel. +31-320-298603
eric.sprokkereef@rws.nl

Rolf Weingartner
Department of Geography  University Bern
Hallerstrasse 12
CH-3012 Bern
Tel. +41-31-631 8874
wein@giub.unibe.ch

András Bárdossy
Institute of Hydraulic Engineering
Faculty of Hydrology and Geo-hydrology
University Stuttgart
University area Vaihingen
Pfaffenwaldring 61
D-70550 Stuttgart
Tel. +49-711-6854 663
bardossy@iws.uni-stuttgart.de  

Brief report on the first presentation block:
"How extreme is extreme?"

Reporter: Rolf Weingartner

The question behind the first block of presentations, "How extreme is extreme?" may - as the various presentations clearly showed - be answered in many very different ways. One speaker [5] said appropriately: "The meaning of ‘relatively rare' is defined by many authors themselves, or they refer to DIN 4049". A team of authors [4] sought to define extreme events verbally based on Sarewitz & Pielke (2000):

• extreme events are inherently contextual,
• extreme means ‘something rare, big, different',
• ‘extremeness' implies an event's behaviour to cause change.

And a poster author [9] even succeeded in capturing the subject of the meeting with gentle irony as follows: Extrem = rare • rare • rare.

The difficulties with the definition of "extreme" are connected to a large extent with the relativity of what is experienced as an extreme event. This is evident, among other things, from the fact that the perception of "extreme" may change over time. A good example is provided by the studies in the Sure catchment basin in Luxembourg: "Floods of the same magnitude become more important today" [3]. An essential factor here is the increase, observed in many places, of the vulnerable areas and the potential for damage. The Elbe high water of 2002 was seen above all as a "man made extreme" [1] as "there is much evidence that there have been higher waters in the past in the Elbe area" [1]. Extreme high water has always been defined primarily by the extent of the damage. Only since natural events have been quantitatively recorded have any more objective evaluation criteria been available.

As extreme events can rarely be measured directly, in general they have to be defined indirectly, usually by means of an analysis and regionalisation of extreme precipitation. There were reports of experiences in Germany, where various tools for estimating rare to very rare precipitation events have been developed [5]. There is still a great need for research in order to derive maximised and extreme localised precipitation levels from selective heavy rain values determined on the basis of extreme statistics. An approach developed in Austria seeks to define (extreme) high water levels by means of regional process types, in order to facilitate a high water estimate which is ultimately more closely oriented towards the process on the regional scale [8]. Greater proximity to the process is a very justified demand; in the case of extreme situations, account must be taken of processes and process combinations, which are often not observed in the case of ordinary events [4].

Various talks and discussions at the KHR Workshop clearly showed that a great deal of information, extremely revealing and relevant to practice, can be derived from a comprehensive analysis of past extreme events [1, 2]. An analysis of high water statistics benefits substantially from the inclusion of historical events [6]. In earlier times, extreme events were often identified as mavericks and excluded from the observations! The actual scientific challenge consists in the interdisciplinary analysis of (historical) extreme situations, with cooperation between historians, hydrologists and hydraulicists [1]. In large catchment basins, international co-operation is very important [2]; here, KHR could have an important role as "facilitator". High water chronologies therefore need to be examined objectively [1]. Such chronologies can contribute to the development of more plausible high water scenarios [4], enabling an improvement in high water risk management.

Overall, the presentations in the first block can be viewed as an initial move towards tackling the still-problematic research area of extreme high water. One speaker [4] from the first presentation block summed it up: "We tend to overestimate our knowledge!" While retrospective analysis has already been developed to a substantial extent - not least on the basis of the major events of recent years - the aspects of looking ahead, estimating or forecasting are still in the early stages. A further complicating factor is that climate change further increases uncertainty in the context of extreme situations, so that protective measures to today's safety standards will become ever more expensive in the future [7]. Which emphasises the truth in the statement: "Climate is what you affect, extreme events is what you get" [7].

[1] Uwe Grünewald: Extreme floods in the Elbe catchment
[2] Marcel de Wit: Extreme discharges in the Meuse basin
[3] Hugo Hellebrand: Does the perception of extremity change? An ongoing case study in the Sure river basin
[4] Bruno Merz, Annegret Thieken: Extreme scenarios and flood risk management
[5] Gabriele Malitz: Extreme precipitation in Germany
[6] Andreas Schumann: The August flood 2002 in the eastern Erz mountains and its statistical evaluation
[7] Christoph Schär: Climate Change and Hydrological Extremes
[8] Ralf Merz, Günter Blöschl: Regional flood process types
[9] Peter Schmocker: Classification of extreme floods 

Brief report on the second presentation block:
"Determination of extreme discharges"

Reporter Marcel de Wit

Introduction

The determination of extreme discharges is needed to assess flood risks. Such assessment is essential for policy, design of buildings and infrastructure and insurance purposes. An underestimation of flood risk may result in large damages, whereas an overestimation of flood risk increases the design costs. Depending on the potential damage, recurrence intervals used for design are most often in the range of once every 102 years to once every 103 years. The problem is that for most rivers the observation record is too short to accurately estimate design discharges. Aim of this session was to present methods to define and determine extreme flow rates. This summary is mainly based on the studies presented in session 2, but also includes results from relevant aspects presented in the other two sessions. The outline of this summary is as follows: i) limitations and capabilities of flood frequency analysis, ii) alternative approaches for the determination of extreme discharges, iii) other topics related to the determinations of extreme discharges, and iv) a synthesis of the workshop session.

Limitations and capabilities of flood frequency analysis

Traditionally flood frequency analyses are used to estimate the recurrence interval and magnitude of extreme floods. Their applicability was illustrated by the fact that during the workshop flood frequency curves were presented for many rivers. Merz (R.) even combined information from a large number of Austrian catchments to derive compound flood frequency distributions for specific flood types. The limited capability of flood frequency analysis to determine extreme discharges was illustrated by several examples with observed extremes that did not fit the distribution functions; Schumann (Saxonia-2002), Naef (Saltina-1993), and Bloschl (Zwetl-2002). The limitations of flood frequencies are not only caused by the limited length of the observation records. Naef showed that discharge measurements during extreme floods are often not reliable. Merz (B.) and Bardossy questioned whether long discharge records are stationary given the upward trend in observed flood peak records. Also the non-homogeneity and the extrapolation of the discharge records was discussed as a limitation of flood frequency analysis.

Alternative approaches for the determination of extreme discharges

Blöschl presented a quasi-analytical derived flood frequency model that is able to account for both flood type and seasonality. Data from 50,000 runoff events in Austria were used to develop and validate this method. Main advantage of the method is that it combines information from different catchments and it allows for an analysis of flood generating processes. Naef performed similar experiments for Swiss catchments. De Wit/Torfs (Meuse) and Diermanse/Chbab (Rhine) presented results from a stochastic rainfall generator coupled with hydrological and hydraulic models. This instrument was used to generate long (103 years) discharge records. Such an instrument proved to be a valuable tool to explore flooding conditions that are more extreme than covered by the observed records (Mehlig).

Other topics related to the determination of extreme discharges

Several presenters stressed that the extremeness of an event should not only be quantified in terms of discharges/water levels but also in terms of damage. This issue was addressed by Mehlhorn who presented flood risk assessment tools for (re)insurance purposes. Such tools not only consist of a flood hazard element, but also quantify vulnerability and value distribution. Naef/Thieken/Grünewald showed examples of the use of historical information for the analysis of extremes. They suggest that there may be a lot of unexplored information in written archives. The use of paleoflood analysis was called for, but not covered during the workshop. Maskey, who presented a fuzzy logic approach for reducing uncertainty in flood forecasting, addressed the issue of flood forecasting during extreme events.

Synthesis

Flood frequency analysis is still widely used for the determination of extreme discharges. The limitations of this method are well known and this has motivated the development of additional methods that range from statistical to process based. There was an overall agreement that these methods are complimentary and their applicability largely depends on the aims of the determination and the availability of reliable data. Further improvements can be expected in the coming years through the access to new data (ongoing monitoring, exploration of archives, spatial data etc.) and the implementation of successful methods and experiences from all over Europe. 

Brief report on the third presentation block:
"Effects of human interventions"

Reporter: Günter Blöschl

The purpose of block 3 was to discuss how best to assess effects of human interventions on floods, and what these effects may be in different contexts. The contributions dealt with these issues in four sub-themes. The first sub-theme addressed statistical methods of how to detect the effects of interventions in flood data. András Bárdossy proposed statistical methods for dealing with non-stationarity. The methods are data based or index methods and allow to perform extreme value statistics in the presence of trends in the data. While this comes at the cost of increasing the number of parameters and some assumptions on the type of the non-stationarity, the methods do relax the limitations of many extreme value statistical methods of assuming stationary populations.

The second sub-theme addressed the issue of how to best assess the effects of interventions in catchments. Axel Bronstert performed large scale hydrological simulations to examine the effect of land use on the water balance and hydrological extremes at various scales. Not surprisingly, the net effect of urbanisation, for example, can be large in small catchments where the increase in sealed surface area can be large relative to the total catchment area. In large catchments the analysis showed the effect of land use change to be very small. Heinz Engel, on the same sub-theme, examined the relative role of interventions in catchments and along the channels. There is a delicate interplay of the retention measures on the Rhine in terms of their total effect on flood mitigation.

This issue has been examined in more detail in sub-theme three in which the discussion focused on how best to examine the effects of hydraulic structures on the flood regime of rivers. Heinz Engel provided a detailed analysis of the polder retention at the Rhine river. Polder size and shape are the main controls and there is the issue of upstream/downstream stakeholder perspectives which can be at stark variance. Flood forecasting plays an important role in flood management. Similar flood retention issues have been expanded upon by Bernd Mehlig from an international perspective.

Sub-theme four focused on how to assess flood risk and mitigate it including residual risk. This involved probability methods and management issues. Matthijs Kok examined the impact of mitigation measures and how to deal with the total probability of failure. Methods of probability based design are now becoming more widely available. A delicate issue of flood management are decisions on when to flood polders. Robert Slomp used hydraulic modelling as a tool of assessing dike heights and dike safety. For the example of Cologne, Reinhard Vogt illustrated flood management and community based approaches of raising flood awareness. Wolfgang Kron presented flood damage issues from the re-insurance perspective.

The ensuing discussion identified a number of research issues that need more attention in the future. The participants felt that the analysis of spatial hydrological patterns in river basins need much more attention when assessing human interventions on extremes, and more generally, assessing extremes at large. More specifically, research should focus on how to best quantify and combine the spatial patterns of flood controls, be it rainfall or antecedent soil moisture, as well as the coincidence of flooding of tributaries. Assessing the probabilities or risk of such coincidence in floods and their controls is a key issue in better understanding and managing floods in large basins. In a statistical context, the participants felt there is a need for better conceptual underpinnings of statistical methods. A wealth of information is available on process controls that is currently used in deterministic methods and this should also be used in statistical methods. How best to include this information statistically should be high in the research agenda on extremes and assessing the effects of interventions. Finally, it was clear from the discussion that flood forecasting plays a key role in flood management. To better constrain forecasting uncertainties, research needs to focus on probability-based methods of flood forecasting.