NCRdays€2004 · NCRdays€2004 Research€for€managing€rivers:...

NCRdays 2004

Research for managing rivers:present and future issues

November 4 – 6

B. Makaske &A.G. van Os (eds.)

October 2005

Publication of the Netherlands Centre for River StudiesNCR publication 262005

ISSN 1568234X

Proceedings NCR days 2004 ii

PrefaceThese proceedings are the product of the NCRdays 2004, held 46 November 2004 in Wageningen.The NCRdays are a yearly conference at which mainly young scientists present their ongoingresearch on a wide variety of fluvial subjects. Since 2000, the NCRdays have been organised inrotation by different institutes represented in the Netherlands Centre for River studies (NCR). With theNCRdays 2004, organised by Alterra, the first lustrum of this event was reached.

The conference centre ‘De Wageningse Berg’, hosting the NCRdays 2004, is an excellentlocation for a conference on rivers: it sits on top of a steep hill offering a magnificent view on the riverNederRijn and its floodplain (see Photo). In this inspiring setting we welcomed 141 participants, thegreatest number since the start of the conference series. Among them were 22 participants giving anoral presentation. Additionally, 24 posters were presented. For the first time, four workshops wereorganised, with the objective to initiate discussion on how to match ‘supply and demand’ of researchon different issues of present interest. The idea to include workshops in the conference programmewas an outcome of the evaluation of the NCRdays 2003, endorsed by the 5 year evaluation of NCR.After two days of presentations and workshops, a field trip to various recent projects on improvingenvironment and safety on the river NederRijn (Fig. 1) was a proper conclusion of the conference.

The 46 contributions (oral presentations and posters) to the conference resulted in the 43papers in this proceedings volume. The papers have been arranged into sections that basicallyrepresent the various sessions of the conference. In the review process, we were helped by membersof the NCR Programme Committee: Eelco van Beek, Gerard Blom, Ipo Ritsema, Rob Leuven, JanRibberink, Erik Mosselman (replacing Kees Sloff) and Remko Uijlenhoet. They are thanked for theircareful work.

We also wish to thank Tine Verheij of the conference agency Routine, Jolien Mans of NCR,the chairmen of the sessions Freek Huthoff, Nadine Slootjes, Saskia van Vuren, Ivo Thonon andMenno Straatsma, the workshop speakers Peter Glas, Almar Otten, Jan Al, Udo Boot and HenkVerkerk, and our keynote speakers Pavel Kabat and Joost de Ruigh. Funding by the NetherlandsOrganization for Scientific Research (NWO) is gratefully acknowledged.

During the first five years of its existence, the NCRdays have proven to be an attractiveplatform for exchange of ideas and discussion serving the community of developers and users ofexpertise on rivers. The challenge for the next lustrum will be to maintain the intimate and informalatmosphere with possibly still increasing numbers of participants. Given the positive experience of2004, we foresee a bright future for the NCRdays and confidently look forward to the 2005 edition thatwill be organised by NITGTNO.

Bart Makaske, Henk Wolfert & Ad van Os

iii Proceedings NCRdays 2004

At various righthand pages you will find photographs giving an impression of the NCRdays 2004

Figure 1. An oblique aerial view on ‘De Wageningse Berg’, the NederRijn and its floodplain between Arnhem andWageningen (view looking northeast). The area shown (12.5 x 20 km) also covers the destinations of the field trip after theconference. (based on the Geomorphological Map of the Netherlands, scale 1:50.000 by Alterra).

Introduction NCR days 2004

Proceedings NCR days 2004 2

ContentsPreface ................................................................................................................................................. iiAbstract ............................................................................................................................................... 4Samenvatting .................................................................................................................................... 5

NCRdays 2004; Introduction ..................................................................................................... 6

Flood Management and DefenceR.H.G. Jongman, B. Makaske, C.R. Padovani, M. van Eupen & S.A.M. van RooijRelating river change, biodiversity and landuse consequences: the Taquari River, Pantanal, Brazil ... 8K.S. MeijerAssessing the relationship between river flows and human wellbeing; a case study in Bangladesh .. 11P.R. van Oel & M.S. KrolClimate change and adaptive water demand........................................................................................ 14K.M. de BruijnResilience of the lowland part of the Mekong River .............................................................................. 16D.F. Kroekenstoel & R. LammersenTransboundary effects of extreme floods on the Lower Rhine.............................................................. 18E. Kater & A.J.M. SmitsPutting the cyclic rejuvenation strategy into practice: symbiosis between safety and nature............... 22M. van Ledden, M.E. Groot Zwaaftink & G.J. AkkermanRijnstrangen/Lingewaarden: a tap for the Rhine branches ................................................................... 26A.J. Nienhuis & B. StalenbergOptimal design of multifunctional flood defences in urbanized areas ................................................... 28M.H. WinnubstCommunication strategies in river management; research plan for comparison of two cases in theNetherlands ........................................................................................................................................... 30S.V. MeijerinkDesigning institutions for water management ....................................................................................... 32M.J. Kolkman, P.A.T.M. Geurts, A. van der VeenWhat’s on a decision maker’s mind? Identifying barriers in information flows between actors inintegrated water management using mental model mapping ............................................................... 34H.P. Wolfert, L.C.P.M. Stuyt, A.G.M. Hermans, J. Kruit, R.J.W. Olde Loohuis & F. KlijnPlanning a green river as a solution to increasing discharge in an urbanizing area on the River Rhine............................................................................................................................................................... 37J.A.E.B. Janssen, J.L. de Kok, M.S. Krol, S.J.M.H. Hulscher & R.M.J. SchielenRapid assessment methodology for river management with application to the Lower Meuse proposedresearch................................................................................................................................................. 40G.T. Raadgever, M.J. Booij, J.A.P.H. Vermulst & S.J.M.H. HulscherDamage due to low flows on the Meuse ............................................................................................... 42Y. Huang, J.L. de Kok & A.E. MynettIntegrated flooddamage and risk assessment ..................................................................................... 44S.P.J.M. van de Pas & N.G.M. van den BrinkBaseline MIXER: a GIS application for managing geographic information and hydraulic models forrivers ...................................................................................................................................................... 46T.P.F. Koopmans & J.H.J. EbbingRiver widening: from understanding the subsurface to digging for ‘gold’.............................................. 48

HydrologyH.A. Peeters, M.J.M. de Wit & R. UijlenhoetFloods in the Meuse basin: contribution of tributaries........................................................................... 50R. Leander, H. Buiteveld, M.J.M. de Wit & T.A. BuishandApplication of a weather generator to simulate extreme river discharges in the Rhine and Meusebasins .................................................................................................................................................... 54P. Aalders, M.J.M. de Wit, L. Bolwidt, P. Warmerdam, P.J.J.F. Torfs, R. Leander & T.A. BuishandA 3,000 year discharge simulation in the Meuse basin with a stochastic weather generator and theHBV model ............................................................................................................................................ 56

Introduction NCRdays 2004

3 Proceedings NCRdays 2004

H.C. Winsemius, H.H.G. Savenije, W.M.J. Luxemburg, H. Havinga, F. Diermanse, S. Tijm, E. Sprokkereef & H.van de LangemheenA quick scan forecasting tool for prescreening probabilistic weather forecasts on their seriousness . 58M. ten Heggeler, A. Berne, R. Uijlenhoet, L. Delobbe, Ph. Dierickx & M. de WitHydrological application of areal rainfall estimates from the Wideumont weather radar over the Ourthecatchment: preliminary results............................................................................................................... 60F. Fenicia, P. Matgen, L. Pfister & H.H.G. SavenijeLearning from the data: a stepped calibration approach....................................................................... 63H. Hellebrand, J. Juilleret, R. van den Bos & L. PfisterTowards a gridbased regionalisation of storm flow coefficients........................................................... 66N.J. de Vos, T.H.M. Rientjes & L. PfisterGroundwater levels as state indicator in rainfallrunoff modelling using Artificial Neural Networks...... 70G.P. Zhang, H.H.G. Savenije, F. Fenicia, T.H.M. Rientjes & P. ReggianiImplications of hydrological modelling and observations in the Alzette river basin............................... 74Y. van der Velde, R. Brunt, R. van Montfoort & R. StuurmanSurface water management during droughts in peat areas .................................................................. 78

EcologyD.S.J. Mourad, M. van der Perk, K. Piirimäe, E. Loigu & J. DeelstraDevelopment of nutrient loads from headwaters to lowland rivers........................................................ 80F.P. Sival, B. Makaske, G.J. Maas & J. RunhaarFloodplain sedimentation regulating vegetation productivity on small rivers? ...................................... 82M.W.A. de Haan, A.J.M. Jansen, J. Grijpstra & C.J.S. AggenbachEcological impact of changes in groundwater withdrawal in river forelands ......................................... 84R.S.E.W. Leuven, G.W. Geerling, S. Gerrits, H.J.R. Lenders & R.J.W. de NooijCumulative effect assessment of physical reconstruction and landuse changes on riverine biodiversity............................................................................................................................................................... 87

Sediments, Hydraulics and MorphologyD. Maljers, S.H.L.L. Gruijters & J.G. VeldkampInaccuracies in estimated grain size parameters and their implication on geological models.............. 90R.M. Frings & M.G. KleinhansSupplylimited transport of bedload sediment at the IJsselkop............................................................ 94L.J. Bolwidt, P. Jesse & R.M. FringsMorphological behaviour around bifurcation points; preliminary results of recent measurements ....... 96F. Huthoff & D. AugustijnChannel roughness in 1D steady uniform flow: Manning or Chézy? .................................................... 98M.W. Straatsma3D float tracking: insitu floodplain roughness estimation ................................................................... 102D. Noordam, A. Blom, H. van der Klis & S.J.M.H. HulscherVariations in roughness predictions (flume experiments) ................................................................... 104A.J. Paarlberg, C.M. DohmenJanssen, S.J.M.H. Hulscher & A.P.P. TermesEffect of main channel roughness on water levels .............................................................................. 106L. Haitel, C.M. DohmenJanssen, S.J.M.H. Hulscher & A. BlomEffect of climate change on bedforms in the Rhine and consequences for navigation ...................... 108S. van Vuren & H.J. BarneveldNavigability of the Niederrhein and Waal in the Netherlands; a stochastic approach......................... 110H. van der Klis & H.R.A. JagersStochastic modelling of twodimensional river morphology ................................................................ 114D.S. van MarenSediment density stratification and river channel patterns in the lower Yellow River, China.............. 116G. Erkens, W.Z. Hoek & E.A. KosterCausal relationships between climate change and natural river behaviour in the Rhine delta during thelast 15,000 years ................................................................................................................................. 118

NCR Supervisory Board....................................................................................... 121NCR Programme Committee ............................................................................... 121NCR Publications series ...................................................................................... 122Colophon............................................................................................................... 125



Abstract

NCR is the abbreviation for the Netherlands Centre for River studies. It is a collaboration of nine majorscientific research institutes in The Netherlands, which was established on October 8, 1998.

NCR’s goal is to enhance the cooperation between the most important scientific institutes in the fieldof river related research in The Netherlands by:• Building a joint indepth knowledge base on rivers in The Netherlands in order to adequately

anticipate on societal needs, both on national as well as international level;• Strengthening the national and international position of Dutch scientific research and education;• Establishment of a common research programme.

NCR strives to achieve this goal by:• Committed cooperation, in which the actual commitment of the participating parties is expressed;• Offering a platform, which is expressed by the organisation of meetings where knowledge and

experiences are exchanged and where parties outside NCR are warmly welcomed.

The committed cooperation and collaboration is based on a programme. This programme was firstpublished in October 2000 and was actualised in August 2001.The platform function is expressed amongst others by the organisation of the socalled annual NCRdays. The publication at hand contains the proceedings of the NCRdays, organised on November45, 2004.

The proceedings of the NCRdays 2004 are subtitled “Research for managing rivers: present andfuture issues”.This is subdivided in the themes (i) Flood Management and Defence, (ii) Hydrology, (iii) Ecology and(iv) Sediments, Hydraulics and Morphology. They cover to a large extend the research which isperformed in The Netherlands nowadays.



Samenvatting

NCR staat voor Nederlands Centrum voor Rivierkunde. Het is een samenwerkingsverband dat op 8oktober 1998 is opgericht door negen wetenschappelijke onderzoekinstituten in Nederland.

Het doel van NCR is het bevorderen van samenwerking tussen de belangrijkste wetenschappelijkeinstituten op het gebied van rivieronderzoek in Nederland door:• het opbouwen van een kennisbasis van voldoende breedte en diepte in Nederland omtrent rivieren

waardoor adequaat kan worden tegemoet gekomen aan de maatschappelijke behoefte, zowelnationaal als internationaal;

• het versterken van het wetenschappelijke onderwijs en onderzoek aan de Nederlandseuniversiteiten;

• het vaststellen van een gezamenlijk onderzoekprogramma.

NCR wil dit doel op twee manieren bereiken:• via gecommitteerde samenwerking; hierin komt het daadwerkelijke commitment van deelnemende

partners tot uiting;• via het bieden van een platform; deze functie uit zich in het organiseren van bijeenkomsten, waarop

kennis en ervaringen worden uitgewisseld; andere partijen zijn daarbij van harte welkom.

De gecommitteerde samenwerking geschiedt op basis van een programma. Dit programma is inoktober 2000 voor het eerst in het Nederlands gepubliceerd en geactualiseerd in Augustus 2001.De platformfunctie komt onder andere tot uiting in het jaarlijks organiseren van de zogenaamde NCRdagen. Voorliggende publicatie bevat de “proceedings” van de NCRdagen die gehouden werden op 4en 5 november 2004.

De proceedings van de NCRdagen 2004 dragen de subtitel ““Research for managing rivers: presentand future issues”, vrij vertaald “Onderzoek ten behoeve van rivierbeheer: heden en toekomst”.

De verschillende thema’s van de NCRdagen 2004, (i) Hoogwaterbescherming en beheer, (ii)Hydrologie, (iii) Ecologie en (iv) Sedimenten, Hydraulica en Morfologie, dekken een groot gedeeltevan het hedendaagse onderzoek dat in Nederland op rivierkundig gebied wordt uitgevoerd.



NCRdays 2004; IntroductionA.G. van OsProgramming secretary NCR, Netherlands Centre for River studies, P.O. Box 177, 2600 MH Delft, The Netherlands;[email protected]

The Netherlands Centre for River studies(NCR) is a collaboration of the majordevelopers and users of expertise in theNetherlands in the area of rivers, viz. theuniversities of Delft, Utrecht, Nijmegen, Twenteand Wageningen, UNESCOIHE, ALTERRA,TNONITG, RIZA and WL | Delft Hydraulics.NCR’s goal is to build a joint knowledge baseon rivers in the Netherlands and to promotecooperation between the most importantscientific institutes in the field of river studies inthe Netherlands.

NCR has two key functions:• Network or platform function: this function is

reflected in the organisation of meetings atwhich expertise and experience areexchanged; other parties are very welcometo attend.

• Researchorientated and educational cooperation: in which a real commitment of thepartners is reflected.

To perform its first key function NCR aims toprovide an open platform for all peopleinterested in scientific research andcommunication on river issues.To that end NCR organises once a year thesocalled NCRdays, where on two ongoingconsecutive days scientists present their riverstudies, in order to maximise the exchange ofideas and experiences between theparticipants and to provide the researchers asounding board for their study approach andpreliminary results. Based on these contacts

they can improve their approach and possiblyestablish additional cooperation.

NCR organised these NCRdays for the fifthtime on November 4th and 5th, 2004 in DeWageningse Berg in Wageningen, theNetherlands.In the publication at hand the presentationsand posters presented are summarized.

The statistics of the 2004 days are verysatisfying: we reached an all time high as faras number of participants is concerned: some140 participants distributed evenly over theNCR partners and other institutes andconsultancy agencies (Fig. 1).The development of the participation over theyears is given in Fig. 2 (next page).Also the presentations and posters werereasonably distributed over NCR partners andnonNCR participantsIn total 27 oral presentations were given and25 posters could be seen and discussed. Infact much more presentations were proposed,but the organisers had to limit the amount to20 (plus 7 key note presentations) to give theparticipants opportunity for the poster sessionsand discussions.

The NCR Programme Committee decided in2003 to establish the NCRdays Presentationand Poster Awards. They both consist of aCertificate and the refunding of theparticipation costs for the NCRdays.

Number of participants per institute

0

5

10

15

20

RIZA

ALTERRA

IHE RU UU

TUDW

L UTNIT

GW

U

V&W/R

WS/K

NMI

CONS

Others

Figure 1. Number of participants per institute



Figure 4. NCRdays Poster Award 2004

Paul Aalders

Figure 3. NCRdays Presentation Award 2004

Roy Frings

The participants determined the winners. Tothat end each participant received fourevaluation forms (two for a specificpresentation and two for a specific poster) atthe registration desk. They were selected atrandom.The participants took their ‘evaluation job’ veryseriously. This added considerably to theliveliness of the discussions during theintermissions and poster sessions.The poster sessions are a very important partof the NCRdays. We use the ‘Hyde ParkCorner approach’ where the primary posterauthors are given the opportunity in ‘twominutetalks’ to give the participants anappetite to come and see the posters and

discuss the content with the authors. Thisworked again very well.

The winners of the NCRdays Awards wereannounced at the end of the NCRdays.

The NCRdays Presentation Award 2004 (Fig.3) was won by Roy Frings for her presentation‘Supplylimited transport of bedload sedimentat the IJsselkop’ (see page 94).The NCRdays Poster Award 2004 (Fig. 4)was won by Paul Aalders for his poster ‘A3,000 year discharge simulation in the Meusebasin with a stochastic weather generator andthe HBV model’ (see page 56).

Participation over the years

0102030405060708090

100110120130140150

posters others

posters NCR

presentations others

presentations NCR

participants others

participants NCR

Figure 2. Development of participation over the years

2000 2001 2002 2003 2004

Flood Management and Defence


Relating river change, biodiversity and landuseconsequences: the Taquari River, Pantanal, BrazilR.H.G. Jongman 1, B. Makaske 1, C.R. Padovani 2, M. van Eupen 1 & S.A.M. van Rooij 11 Alterra, Wageningen University and Research Centre, P.O. Box 47, 6700 AA Wageningen, The Netherlands;[email protected] EMBRAPAPantanal, Caixa Postal 109, 79320900 Corumbá, Mato Grosso do Sul, Brazil

IntroductionThe Taquari is a tributary of the ParaguayRiver in southwestern Brazil (Fig. 1). One ofthe problems that has developed in the lastthirty years is the permanent flooding of thesavannah over an area of 11.000 km2 on theTaquari alluvial fan in the Pantanal. In thisarea, an interdisciplinary research project wascarried out (within the framework of thePartners for Water research program), whichfocused on: (1) the analysis and modelling ofthe ongoing process; (2) the consequences forbiodiversity and land use of the lower Taquarifloodplain; (3) capacity building for theorganisation of integrated river management atthe basin level including relevant stakeholders. The upper parts of the Taquari catchmentrepresent one of the major erosive areas of thehighlands around the Pantanal, consisting ofsandy soils. According to the localstakeholders this erosive character hasresulted in the inundations of the floodplain ofthe lower Taquari because of silting up of theriver channel.Erosion in the upper catchment is believed tohave strongly increased as a result of clearingof the natural vegetation (Fig. 1).

Project approachThe project approach was to carry out jointBrazilianDutch research on river managementfocusing at understanding the Taquari system.Modelling of the river system and its land coverand land use involved construction of a DigitalElevation Model (DEM) and a river dischargemodel. Geomorphological analysis of remotesensing data and collection of new field data(sampling for 14C dating and grainsizeanalysis) yielded an impression of riverdynamics at various time scales. The DEMwas constructed by the Institute for GeoInformation Science and Earth Observation(ITC) and was used by WL | Delft Hydraulicsas a basis for a river flow model for theanalysis of the river changes. The water inputfrom the Planalto was considered as a givenparameter (‘black box’): the Taquari at Coxim(Fig. 1) is the only input of surface water intothe plains. The DEM and the hydrologicalmodel, with important ecological knowledge ofEMBRAPAPantanal made it possible todevelop scenarios on the consequences of theongoing processes for ecotopes, land use andspecies. To provide river managers andstakeholders with insight in the consequencesof planning and management options for theriver system, it was necessary to analyse boththe socioeconomic and the ecohydrologicalconsequences of the changes in the riversystem. For the analysis of these impacts,socioeconomic and ecological scenarios weredeveloped for different hydrological andclimatic changes affecting the river system. The last, and for all institutes involvedmost difficult, aspect of the project was theintegration of socioeconomic consequenceswith the natural processes. Use was made ofinterviews with stakeholders and analysis ofexisting economic data. It is supposed thatdata and knowledge present in the institutions,with farmers and civil society are sufficient tomake a first start with the construction of theDecision Support System. In a specialworkshop the principles of decisionmakingand the need for a decision unit werediscussed as well as the principles of multicriteria evaluation. Three scenarios (a dry,

Figure 1. The Taquari River and the Pantanal basin insouthwestern Brazil. Forestclearing following ‘nationalcolonisation’ of the upper Taquari catchment washypothesized to have environmental impact on the lowerTaquari (increased flooding). Hydrological modelling in thepresent project, however, demonstrated only limited impactof landuse changes on Taquari discharge.



average, and wet scenario) were evaluated todemonstrate the principles of spatial multicriteria evaluation.

ResultsThe results of the project are various. Afundamental product is the DEM of the studyarea, with an altitude accuracy of 0.10 m (Fig.2; Maathuis, 2005).

Other basic research products are thereconstructed geomorphological history, andthe geomorphological map of the Taquarialluvial fan. The geomorphological analysis(Makaske, 2005) shows that the floodingproblems in the area are associated with twomajor avulsions: the Caronal avulsion on themiddle fan and the Zé da Costa avulsion onthe lower fan (avulsion is defined as adiversion of river flow from an existing channelonto the floodplain, eventually resulting in anew river main channel). In addition to thesetwo avulsions, many crevasses exist in thelevees of the Taquari on the middle and lowerfan (Fig. 3). Our understanding of the hydrology of thestudy area has considerably increased. Agroundwater map and a flooding map of theriver basin were produced. Longitudinal andtransverse hydraulic measurements werecarried out and a discharge model was set up.It was demonstrated that increased dischargeof the Taquari River, leading to avulsions andflooding, mostly results from increasedprecipitation and to a much lesser extent fromchanges in land use in the catchment (Querneret al., 2005). As a result of research efforts in variousfields, there is now an uptodate ecotope mapfor the study area. Much existing ecologicalknowledge was organised in such a format thatit could be included. These data were used forscenario development on the recognition ofchanges with impact analysis for biodiversityusing the OSIRISLEDESS model and LARCHspecies models. Decision support scenarioswere worked out in a special workshop inAugust 2005 and the results were presentedand discussed in November 2005 withstakeholders and researchers.

Research can only have an impact onsociety when it presents a coherent vision onthe future of the river basin and if there is astructure for decision making, and amanagement organisation. The objective of theproject was to develop better understanding ofthe impact of human influences on thePantanal basin and to be able to understandthe functioning of the Upper Paraguay RiverBasin (UPRB) as a whole. This means thatthere had to be a strong link between researchof ecological and landuse aspects,technology, management and policy. Theproject helped to identify opportunities foreconomically feasible use of the system, andfor its management (Jongman, 2005). Threeimportant lessons can be learned from thisproject.• The erosion and sedimentation processes

in the basin are so intense that technicalsolutions without a river basinmanagement organisation attackingerosion and sedimentation processes areuseless.

• Flood pulses are essential ecologicalprocesses in rivers for productivity andbiodiversity. The comparison betweendisturbed and undisturbed rivers deliversimportant knowledge also for rivermanagement in Europe.

• Making water management work andsustainable depends on regional coordination and political commitment atsupraregional level. Cooperationbetween sectors and stakeholders appearssometimes difficult as each group isengraved into its own issues, priorities andviews. This is not only true for policymakers and research groups, but also forcivil society organisations.

ReferencesJongman, R.H.G. (ed.), 2005. PantanalTaquari; tools for

decision making in integrated water management.Alterra Special Publication 2005/02, Alterra,Wageningen, 40 p. and CDROM.

Maathuis, B., 2005. The Digital Elevation Model. In: R.H.G.Jongman (ed.), PantanalTaquari; tools for decisionmaking in integrated water management. AlterraSpecial Publication 2005/02, Alterra, Wageningen,pp. 4862 (in final report on CD).

Makaske, B., 2005. Avulsions, flooding and sedimentationin a geomorphological perspective. In: R.H.G.Jongman (ed.), PantanalTaquari; tools for decisionmaking in integrated water management. AlterraSpecial Publication 2005/02, Alterra, Wageningen,pp. 3447 (in final report on CD).

Querner, E., R. Jonker, C. Padovani, B. Soriano & S.Galdino, 2005. Impact of climate change andagricultural developments in the Taquari River basin,Brazil. In: T. Wagener, S. Franks, H.V. Gupta, E.Bøgh, L. Bastidas, C. Nobre & C. de Oliveira Galvão(eds.), Regional hydrological impacts of climaticchange – impact assessment and decision making.Proceedings of Symposium 6, IAHS ScientificAssembly at Foz do Iguaçu, Brazil, April 2005. IAHSPublication 295, pp. 1925.

Figure 2. The final DEM of the Taquari River floodplainin the central Pantanal (Maathuis, 2005). Area shownmeasures ~300 km in width; north is up.



Figure 3. A small crevasse in the natural levee of the Taquari River routing water from the main channel (left) to the floodplain(right). Note the remains of sandbags on the foreground, which were used by the local inhabitants to close the entrance ofthis small channel.



Assessing the relationship between river flows and humanwellbeing; a case study in BangladeshK.S. MeijerSection of Water Resources, Faculty of Civil Engineering and Geosciences, Delft University of Technology, P.O.Box 5048, 2600GA Delft, The Netherlands; [email protected]

AbstractEnvironmental Flow Assessment methods,developed to assess what part of the flowregime should be maintained in a river toprotect the river ecosystem, do not include theimportance of this river ecosystem for humanwellbeing. This paper discusses a conceptualmodel to take human wellbeing into account inEnvironmental Flow Assessments, andpresents the results of applying the model in acase study in Bangladesh.

IntroductionWater resources development may change theflow regime of a river. Changes to the flowregime affect the river ecosystem andsubsequently the lifes of the people dependingon it. The negative effects of flow regimechanges were recognised in the 1950s. Sincethen, methods to assess Environmental FlowRequirements are being developed. Environmental Flow Assessments wererecently recognised by many internationalorganisations as a tool to ward off socialconflict and environmental degradation due tothe overexploitation of water in river basins ofthe world (IUCN, 2004). Environmental FlowAssessments first focussed on specific speciesand developed towards considering the entirenatural ecosystem. People who depend on thegoods and services provided by the riverecosystem did not receive much attention.How to take the needs of the people intoaccount in assessing Environmental Flows isthe subject of the research presented in thispaper. First, a conceptual model was developedwhich describes the links between human wellbeing and river flows (Fig.1). The second stepwas to test the model in a case study. Theresults of this case study, carried out inBangladesh, are the main topic of this paper.

Conceptual modelThe conceptual model starts with the total wellbeing of the stakeholders, which may be partlyrelated to water. The waterrelated aspectsrely on certain river ecosystem goods orservices which require a certain flow regime.The required internal flow regime can consistof discharge, water depth and flow velocity at

the location where the goods and services areavailable. The external flow regime is the flowregime at a location where this can becontrolled, for example at a dam or a weir. Atall levels the context should be considered tounderstand the importance of a certain flowrequirement for people’s wellbeing. Theblocks on the right side represent the variouspeople in different roles. A river managershould take the wellbeing of the stakeholdersinto account in a river basin plan and directactors to maintain a certain flow regime.

Case study area and methodsThe case study was carried out in theNortheast of Bangladesh along the Surma river(Fig. 2). The Barak river which originates inIndia bifurcates at the border with Bangladeshinto the Surma and the Kushiyara. The areabetween the two rivers is lowlying and isflooded every year during the monsoon season(Fig. 3). The recession cultivation of rice andthe fisheries, which are important income andfood provision sectors in the area, are adaptedto the rise and fall of the water level. In theselected floodplain area of 400 km2 liveapproximately 285 000 people. In the case study three methods wereused: (1) interviews in four villages along theSurma River (Fig. 2); (2) study of reports aboutthe SurmaKushiyara basin; (3) interpretationof 1D simulation.

Figure 1. Conceptual model



Case study resultsAnalysis of the interviews identified recessionagriculture, fisheries and domestic use of riverwater as the main goods and servicesprovided by the ecosystem. Also important arethe negative effects of flooding: damage tocrops and settlements. Income generationbased on ecosystem goods and services isimportant for 4080% of the population, 2570% of the population depends entirely on thenatural ecosystem for income. Of this lastgroup of people, 80% is considered to be poor,according to the standards of the villagers. Intotal, the villagers consider 50% of thepopulation to be poor. Upscaling to the floodplain area asindicated in Fig. 2 resulted in minimum andmaximum water depth requirements for theboth the floodplain and the river itself for everymonth. It was assumed that the objective fordefining flow requirements was to maintaincurrent use of ecosystem goods and services.The 1D simulation results were used tocalculate to what extent the floodplain and riverarea met these requirements. Table 1 shows the result of the 1Dsimulations for the floodplain area. For theyear 20002001 the requirements on thefloodplain are met for all purposes exceptAman paddy cultivation. For Aman paddycultivation only 20% of the required land wasavailable. The main restriction which causesthis low availability is the minimum requirementof 5 cm water depth in August and September,while a large area remained dry in the year20002001.

Conclusions• The main ecosystem goods and services

of the SurmaKushiyara floodplain are theenabling of cultivation and the provision offish, which generate food and income forthe population. Flooding is the maincharacteristic to sustain the current use ofthe river ecosystem, direct use of the riveris less important. Although flooding isimportant for agriculture and fisheries,most of the people prefer to have no flood,because depth and timing of flooding areunpredictable.

• Concerning the methods used in the casestudy, it can be concluded that interviewshelp to understand what people use andhow important this is for their wellbeing.To understand the relationship betweenthe (internal) river flow regime and theavailability of ecosystem goods andservices, expert knowledge is, however,required.

• The conceptual model proved useful forunderstanding the relationships betweenpeople, the river ecosystem and the flowregime. For a thorough understanding ofthe importance of the river flow regime forhuman wellbeing it is essential to considerthe different levels of the context asindicated in the conceptual model.

Figure 2. Location of the case study area andselected villages

Figure 2. Location of the case study area and selectedvillages



ReferenceIUCN, 2004. International organisations accept

‘environmental flows’ as solution to social confict overwater. Electronic news release IUNC Water & NatureInitiative, 19 Aug. 2004. IUCN, Gland, Switzerland.

Table 1. Comparison of floodplain requirements and simulation results.

Function Period Available area(ha)

Required area(ha)

Percentage ofrequirement met

Aus paddy April June 28,389 8,996 316Aman paddy JulyNovember 3,174 14,171 22Boro paddy December April 4,358 4,316 101Vegetables All year 21,039 1,258 1,672Overwintering offish

October May 4,030 18 22,389

Spawning of fish JuneSeptember 13,968 1,463 955

Figure 3. Crosssection of SurmaKushiyara rivers andfloodplain



Climate change and adaptive water demandP.R. van Oel & M.S. KrolDepartment of Water Engineering and Management, Faculty of Engineering Technology, University of Twente, P.O. Box 217,7500 AE Enschede, The Netherlands; [email protected]

IntroductionImpacts of global climate change imposepressure on humanenvironment systems. Theagricultural sector is particularly vulnerable. Inresponse to changes in water availability,farmers might change their activities. Theresearch focus is on agricultural landusechanges that influence water demands.System dynamics are represented in asimulation model that helps to explore effectsof water management strategies in the nearfuture. Subject of study is the public irrigationarea of IcóLima Campos in the JaguaribeRiver Basin in Ceará, Brazil (Fig. 1). In thisresearch, complexity theory (Axelrod, 1997)and the concept of CommonPool Resources(Ostrom et al., 1994) are guiding principles forthe description of the complex humanenvironment system.

MethodsThe research methodology is divided into thefive steps described below.1. Collection of quantitative data. Both

remote sensing techniques and statisticalmethods are applied to analyse systemdynamics in recent history.

2. Collection of qualitative data in the casestudy area. Farmerheuristics with respect

to decisionmaking on landuse changeare formulated.

3. Building a MultiAgent Simulation (MAS)model. In this research, AgentBasedModeling (ABM) for landuse and coverchange (Axelrod, 1997; PahlWostl, 2002;Parker et al., 2002; Hare & Deadman,2004) adds value to the assessment ofagricultural vulnerability to droughts byconfronting individual farmer heuristics toan environment that is evolving throughsystem dynamics (Fig. 2).

4. Model calibration and validation. A timeseries of remote sensing images andinterviews with experts are used. Acomparison of the casestudy area to theMorada Nova area (upper right in Fig. 1)will also be done.

5. Exploring future developments. Theresulting MAS tool will be used to explorepossible future developments in land useand evaluate water managementstrategies.

ResultsTrends in crop cultivation that suggestadaptive responses to droughts were found.These changes might be directly or indirectlytriggered by drought. Between 1990 and 2002some interesting changes in aggregated localwater demand emerged. Figure 3 shows localeffective precipitation (a), water volumes in aregionally strategic reservoir (b), changes inlocal crop cultivation (c) and the correspondingchanges in water demand (d).

Figure 1. Lower right: Ceará in the world. Lower left:IcóLima Campos area and Orós reservoir from space.

Figure 2. Farmer decisionmaking on landuse changeunder influence of system dynamics.



ConclusionsA drastic decrease in water demand emergedrecently. The underlying system dynamics arerelated to individual and collective farmerdecisionmaking. This justifies taking step twoand model system dynamics through MultiAgent Simulation.

ReferencesAxelrod, R., 1997. The complexity of cooperation: agent

based models of cooperation and collaboration.Princeton University Press, Princeton, 248 p.

Hare, M. & P. Deadman, 2004. Further towards ataxonomy of agentbased simulation models inenvironmental management. Mathematics andComputers in Simulation 64, pp. 2540.

Ostrom, E., R. Gardner & J. Walker, 1994. Rules, gamesand commonpool resources, University of MichiganPress, Ann Arbor, 392 p.

PahlWostl, C., 2002. Agent based simulation in integratedassessment and resources management. In: A.Rizzoli & T. Jakeman (eds), Integrated assessmentand decision support. Proceedings of the 1st biennialmeeting of the International Environmental Modellingand Software Society, Vol 2, pp. 239250.

Parker, D.C., T. Berger, & S.M. Manson, 2002. Agentbased models of landuse/landcover change; reportand review of an international workshop, October 47,2001. LUCC Report Series 6, LUCC Focus 1 Office,University of Indiana, Bloomington, 124 p.

Figure 3. Recent changes in Municipio Icó(from www.sidra.ibge.gov.br).



Resilience of the lowland part of the Mekong RiverK.M. de BruijnFaculty of Civil Engineering and Geosciences, Delft University of Technology, P.O. Box 5048, 2600 GA Delft, The Netherlandsand WL | Delft Hydraulics, P.O. Box 177, 2600 MH Delft, The Netherlands; [email protected]

AbstractThe paper studies the usefulness of theresilience concept for flood risk managementof the lowland part of the Mekong River. Thepaper argues that applying resilience is usefulbecause it results in more insight into therelationships between floods and the socioeconomy and in more and better solutions forflood risk management.

IntroductionThe application of the resilience concept toflood risk management is expected to result innew visions and improved strategies.Resilience, and its opposite, resistance, areboth system characteristics. Applying theseconcepts in flood risk management thusrequires a systems approach. Flood riskmanagement systems are defined as acombination of the lowland river and theadjacent floodprone area with both its physicaland socioeconomic characteristics (De Bruijn,2004). The resistance of this systemdetermines which discharges can still bedischarged through the river without causingfloods, while the resilience determines theease of the system to recover from floods. Inorder to evaluate the usefulness of theresilience concept, it has been applied to thelowland part of the Mekong River in Cambodia.

The main question of the case studyreads: what is the current resilience of thelowland part of the Mekong River Basin andhow do different strategies affect the systemand its reactions to the discharge regime?

In the Mekong flood risk managementsystem, the socioeconomy and floods arestrongly related. Normal annual floodings inthe monsoon season do not cause damage,because the system is so much adapted to theannual flood pulse, that floods can be regardedessential for the wellfunctioning of the socioeconomy. However, extreme floods (e.g. 1996,2000, 2001 and 2002) caused lots of damageand casualties. In the future these impacts areexpected to rise, because the area isdeveloping fast, population is expected todouble in the next 50 years and also the floodfrequencies are expected to increase.

Quantifying resilienceResilience cannot be measured, but it can bequantified by indicators (De Bruijn, in press).

De Bruijn (2004a) explains that resilience andresistance reflect the reaction of a system topeak discharges. The indicators thereforecover the three aspects that describe areaction of the system to peak discharges:amplitude which is the severity of the reactionto peak discharges, the graduality of theincrease of reaction to increasingly severepeak discharges and the recovery rate fromfloods. As an indicator for the amplitude of thereactions to the whole regime of peakdischarges the expected annual damage(EAD) can be used. The graduality is assessedby a comparison of the percentual increase ofdamage and discharge. To assess therecovery rate, the recovery capacity of thesystem is analysed (De Bruijn, in press).

To determine the indicator values for theMekong River, first the peak dischargeprobabilities and the peak discharge volumesand durations were analysed. Secondly, anumber of discharge events were simulatedwith a quasi2D Mike11 model (Fujii et al.,2003). Thirdly, a damage module wasdeveloped to quantify the damagecorresponding with these discharge events.Finally, a recovery capacity analysis wasperformed.

ResultsThe resilience of the Mekong system is not ashigh as expected. The EAD of the lowland partof the Mekong River Basin is high: 12 M€/yr or1 M€/inhabitant. (The EAD of the Netherlands’Rhine is 5 M€/yr or 0.28 M€/yr per inhabitant).The graduality of the increase of damage withincreasing peak discharges is 0.79, however,which is comparatively high. The Lower RhineRiver, for example, has a graduality of 0.25.The recovery capacity of the Mekong is nothigh, because the economic and socialcharacteristics of the system limit recovery.The recovery capacity scores a 6 only on ascale from 1 to 10, whiles the Rhine incontrast, scores a 9.

FutureIn the future economic growth, populationincrease and an increase in peak dischargesare expected. Because flood risk managementstrategies and economic developments aredifficult to separate, three storylines of three



alternative futures were developed, whichcombine socioeconomic developmentscenarios and flood risk managementstrategies: (1) continuation of current flood riskmanagement strategy; (2) agriculturaldevelopment combined with a resilient floodrisk management strategy; (3) rapid economicdevelopment combined with a more resistantflood risk management strategy.

The first future results in a decreasedresilience because the EAD increasessignificantly. The second future involvesagricultural development (Fig. 1), an improvedflood early warning system and a stepwiseimprovement of water management includingirrigation, drainage and flood management.This results in an increased recovery rate,while the EAD and graduality do not changesignificantly. In the third future, the world’slarge donors such as the JICA, ADB and theWorld Bank finance a Flood Action Planconsisting of embankments. Agriculturechanges to a more export focused agricultureand industrialisation and urbanisation occur.

This future results in an increased EAD,decreased graduality, and a reduction of therecovery rate. Evaluation of the three futuresshowed that the resilience strategy is verypromising: this strategy enhances socioeconomic development whilst not harmingnature, land scenery or increasing thesensitivity of the system to uncertainties. Theresistant future results in increased economicopportunities but also in an increased floodrisk, and it negatively affects nature and landscape. Besides, it makes the system moresensitive to uncertainties.

Conclusions• The current resilience of the Mekong is not

as high as expected. The graduality is veryhigh indeed, but the amplitude is also highand the recovery rate is low. Becausefrequent floods cause significant damage,the amplitude of the reaction to theseasonable discharge patterns is high. Thelow recovery rate is mainly caused bypoverty.

• The assessment of the resilience of thissystem showed that even in systems thatseem wholly adapted to annual floods,resilience is not necessarily high.

• The resilience future looks promising, incomparison to the resistance strategy.Since the resilience strategy can beimplemented step by step, is cheaper thanthe resistance strategy and hasadvantages for agriculture and fishery, it iscertainly a strategy worth considering forthe future.

• Adopting a systems approach andconsidering the whole range of peakdischarges resulted in a better insight inthe flood risk management of the MekongRiver. Because it takes into account moreaspects of a system’s reactions to a wholerange of peak discharges, it helps toidentify more and other solutions.

ReferencesDe Bruijn, K.M., 2004. Resilience and flood risk

management. Water Policy 6, pp. 5366.De Bruijn, K.M., in press. Resilience indicators for flood

risk management systems of lowland rivers.International Journal of River Basin Management 2(3).

Fujii, H., H.Garsdal, P. Ward, M. Ishii, K. Morishita & T.Boivin, 2003. Hydrological roles of the Cambodianfloodplain of the Mekong River. International Journalof River Basin Management 1 (3), pp. 114.

Parker, D.J. (ed), 2000. Floods, Volume I. Routledge,London.

Figure 1. About 80% of the population is farmer. Riceis the most important crop in Cambodia.



Transboundary effects of extreme floods on the LowerRhineD.F. Kroekenstoel & R. LammersenInstitute for Inland Water Management and Waste Water Treatment (RIZA), PO Box 9072, 6800 ED Arnhem, The Netherlands;[email protected]

AbstractUnder extreme conditions extreme peakdischarges can develop in the Rhinecatchment area. At discharges between 11000m3/s and 16000 m3/s largescale floodingoccur along the Lower Rhine in Germany.When dikes overflow or break flows parallel tothe Rhine will develop, also resulting in theflooding of areas with a higher protection level.In case of largescale flooding along the LowerRhine the peak discharge at Lobith is reduced.Under extreme discharge conditions plannedfloodreduction measures in Germany havelittle effect on the discharge (and water level)at Lobith.

IntroductionAfter the floods of the Rhine in 1993 and 1995three parties signed a Joint Declaration on Cooperation in the field of flood protection in1997. These parties were: (1) the Province ofGelderland (The Netherlands); (2) the Ministryof Transport, Public works and WaterManagement (The Netherlands); (3) theMinistry of Environment and NatureConservation, Agriculture, and ConsumerProtection of North Rhine – Westphalia(NordrheinWestfalen) (Germany). To investigate the effects of extreme floodsof the Rhine in NordrheinWestfalen the project‘Transboundary effects of extreme floods atthe Lower Rhine’ was commissioned at theend of 2001. Transboundary refers to theboundary between Germany and theNetherlands. The project was executed by theRijksinstituut voor Integraal Zoetwaterbeheeren Afvalwaterbehandeling (RIZA) in Arnhem,the Provincie Gelderland, theLandesumweltamt (LUA) in Düsseldorf, andthe Bundesanstalt für Gewässerkunde (BfG) inKoblenz. Under the project title ‘Extreme floodand flood protection along the Rhine (FAR)’this project was cofunded by the EUInterregIIIB North West Europe programme(Lammersen, 2004).

ObjectivesGoal of this project was to answer thequestions below.• How much water can be expected in the

study reach (Fig. 1) from the Rhinecatchment under extreme conditions (atAndernach and Lobith)?

• What is the discharge capacity of theLower Rhine?

• What happens on the Lower Rhine whenthe discharge capacity is exceeded?

• What are the effects of flood reductionmeasures?

Figure 1. Research area and study area (circle).



Research methodA stochastic weather generator, developed byKNMI, was used to produce an artificial timeseries of 1000 year of precipitation andtemperature. The input consisted of 30 yearsof measured meteorological data of 34different weather stations in the Rhinecatchment area. The generated time series,with the same statistics as the historical data,was then put into a rainfallrunoff modelcovering the whole Rhine basin (HBV) andwas transformed into discharge. A selectionwas made of the 16 most extreme events,based on the HBV results at Andernach andLobith (Fig. 1). These 16 extreme events werethen put into a 1dimensional flood routingmodel to compute the 16 highest discharge

waves at Andernach in a more accurate way.This part of the project was carried out by theBfG. With the two most extreme dischargewaves at Andernach flood simulations havebeen performed using the 2dimensionalmodel DelftFLS. A DelftFLS model was madeof the Rhine downstream of Andernach(Rhinekm 642) using a 100 m x 100 m grid ontop of a digital terrain model. In this modeldikes and flood walls were modelled as gridcells. When the water level reaches the dikelevel a dike collapse occurs. In case of a floodwall, or a natural levee, the floodwall or leveesimply overflows and no collapse is simulated.Two situations have been considered: the year2002 and 2020, with the dike levels of 2002and 2020 respectively. The input of the 2Dmodel consisted of the discharge at Andernachand the tributaries of the Rhine. The outputconsisted of information about locations of adike collapse or an overflow, flow into theprotected area, flood patterns inside theprotected area, effect on the discharge waveand finally the discharge capacity of the Rhine.This part of the project was carried out by theProvince of Gelderland. The results of the 2D flood simulationswere then transferred to a 1dimensionalSOBEKmodel. Each dike collapse or overflowwas modelled as a retention basin. Parameterslike surface area, capacity, inflow and outflowwere based on information from the DelftFLSmodel. Using the SOBEKmodel the effect offlood reduction measures in NordrheinWestfalen was studied. Computations weremade with the eight most extreme dischargewaves, with and without dike collapse oroverflow. Two situations have beenconsidered: the year 2002 and 2020, with theflood reduction measures finished in 2002 and2020 respectively. This part of the project wascarried out by RIZA and LUA.

Figure 2. Discharge at Andernach and Lobith, withand without dike overflow, for the eight most extremedischarge waves.

Figure 3. Flooding along the Lower Rhine (situation 2002)



ResultsUnder extreme conditions extreme dischargescan develop in the Rhine catchment area, upto 17800 m3/s at Andernach and 18700 m3/s atLobith, when dike overflows are notconsidered. When dike overflows on the UpperRhine and Lower Rhine are taken into accountthese numbers reduce to 15300 m3/s atAndernach and 15500 m3/s at Lobith (Fig. 2). At discharges between 11000 m3/s and16000 m3/s largescale flooding occurs alongthe Lower Rhine (Fig. 3). First areas along thesouthern part of the Lower Rhine will beflooded (from Bonn/Köln toDüsseldorf/Dormagen), and at higherdischarges also areas along the middle part ofthe Lower Rhine (from Düsseldorf/Dormagento the mouth of the Ruhr River). Furtherdownstream no flooding occurs in thatsituation, except for near Emmerich. In thepresent situation the flood wall at Emmerich istoo low and transboundary floods can occur atdischarges exceeding 14000 m3/s. In 2020,when the flood wall will have been raised, no

transboundary flooding can occur anymore.When dikes collapse or overflow, flows parallelto the Rhine will develop, also resulting in theflooding of areas with a higher protection level.In case of largescale flooding on the LowerRhine the peak discharge at Lobith is reduced(Fig. 4). At some locations the water flowsback into the river (bypasses; Fig. 5). The planned flood reduction measures inNordrheinWestfalen are most effective in caseof floods equal to the flood of 1995 (~12000m3/s at Lobith), but have little effect underextreme conditions (Fig. 5). Measures inGermany affect the water levels in theNetherlands and the other way around.Measures along the Lower Rhine in Germanyreduce the maximum water levels in theNetherlands between 0 and 0.06 meter (Fig.5). The measures in the Netherlands have adecreasing effect in upstream direction (as faras Ruhrort), with a maximum water levelreduction of 0.30 meter at the GermanDutchborder.

Figure 4. Maximum discharge with and without dike overflow (situation 2002, situation with dike overflow on Upper Rhine).



Final remarksCooperation between upstream anddownstream areas is crucial in the future, sincethe river Rhine has no boundaries. As a resultof this study there is a discussion about thepossibilities for optimizing some of the plannedretention basins, i.e. to deploy these retentionbasins at more extreme conditions (higherflood levels than in 1995). Presently, additionallargescale dike improvements along theLower Rhine are not considered.

AcknowledgementsAcknowledgements go the German – Dutchflood study group for their supervision and trustand all colleagues at the Bundesanstalt fürGewässerkunde, Landesumweltamt NRW,Provincie Gelderland and RIZA whocontributed to the success of this project.

ReferenceLammersen, R., 2004. Grensoverschrijdende effecten van

extreem hoogwater op de Niederrhein, Eindrapport.Provincie Gelderland / Rijkswaterstaat Directie OostNederland, Arnhem, 159 p.

Figure 5. Effect of planned flood reduction measures in NordrheinWestfalen



Putting the cyclic rejuvenation strategy into practice:symbiosis between safety and nature

E. Kater & A.J.M. SmitsCentre for Water and Society, Faculty of Science, Radboud University Nijmegen, P.O. Box 9010, 6500 GL Nijmegen, TheNetherlands; [email protected]

AbstractUnbridled growth of shrubs and forests in thenature restoration projects of the regulatedMeuse River and Rhine River floodplains,reduce the water discharge capacity beyondacceptable levels. To meet both hydraulic andecological conditions a new floodplainmanagement strategy will be furtherelaborated and applied to the Beuningenfloodplain (300 ha; Waal River). Thismanagement strategy is referred to as CyclicFloodplain Rejuvenation (CFR) and comprisesnew institutional arrangements, periodic(cyclic) interventions in the morphology andvegetation of the floodplains and theapplication of innovative managementtechniques. The research project is funded bythe national research program ‘Living withWater’.

IntroductionSince the publication of Plan Ooievaar (DeBruin et al., 1987) the land use of manyfloodplains along the Meuse and Rhinebranches in the Netherlands has beentransformed from agriculture to naturemanagement. Although the implementation ofthis policy can be considered as a success, thevegetation development in some floodplainscauses a dangerous decrease of the waterdischarge capacity (Jesse, 2004). Within the BSIKresearch program ‘Livingwith Water’ this project is defined to developand apply a management strategy thatcombines both nature and safety objectives.The research project is also a building stone ofan international and more comprehensiveInterregIIIbproject named ‘Freude am Fluss’.The ‘Freude am Fluss’ project focuses onchanges in land use via local initiatives, newmarket mechanisms and technical innovations.These changes in land use will generate morespace for the river and the riparian vegetationbut also requires a new view on management.The ‘Freude am Fluss’ project is carried out byFrench, German and Dutch governmentalorganisations, academic institutions andconsultants.

Casestudy Beuningen floodplain,Waal RiverSince 1991 the shift from agriculture to naturerestoration has been carried out successfully inthe Beuningen floodplain (300 ha; Waal River).However, since this transformation the growthof shrubs and trees gradually exceeded thestandard value of hydraulic roughness. Figure1 shows the increase in hydraulic roughnessdue to vegetation development between 1985and 2003. As a consequence, the maximumwater level linked to the standard maximumwater discharge in this river section increasedby 5.5 cm (Mannaerts, 2004). This causes anunacceptable situation. In cooperation with theriver and nature managers, it was decided tofurther elaborate and apply a newmanagement strategy to the Beuningenfloodplain to compensate for the increasedhydraulic roughness without affecting theecological values. This new managementstrategy is named Cyclic FloodplainRejuvenation.

The new management strategy:Cyclic Floodplain RejuvenationIn this research ‘Cyclic FloodplainRejuvenation’ (CFR) is considered as thesystem of natural processes in nonregulatedrivers that is responsible for building up andbreaking down morphology and vegetation.The main processes are erosion,sedimentation and vegetation succession.

Figure 1. Increase in hydraulic roughness in theBeuningen floodplain, 19852003



In natural rivers the combined action of theseprocesses results in cyclic rejuvenation ofmorphology and vegetation, and therefore in anatural regulation of the discharge capacity(Smits et al., 2000). In regulated rivers like the Rhine branchesthese natural processes cannot act freelybecause of the presence of for instance dams,weirs, dikes and groynes. As a consequencethe natural rejuvenation cycle is broken, andvegetation tends to develop to the climaxstage (forest). The result of this development isan increase of hydraulic roughness, andsubsequently an increase of the risk of flooding(Duel et al., 2001; Baptist et al., 2004). The basic idea of the CFRstrategy is to‘repair’ the broken rejuvenation cycle byhuman interventions. Therefore we have tounderstand the natural rejuvenation processesand if possible, imitate them. In practice, thismeans setting back succession stages to apioneer situation (e.g. removal vegetation,lowering floodplains or digging side channels). The CFRstrategy can contribute to reestablish the discharge capacity, becausenormally pioneer stages have a lower hydraulicroughness (Van Velzen et al., 2003). TheCFRstrategy will also result in more variationof succession stages, and therefore in a higherbiodiversity. These are the two main reasonsthat the CFRstrategy is a promising solutionfor the realisation of a symbiosis betweensafety and nature. Besides, the CFRstrategyprovides also opportunities for sand or gravelexcavations which can reduce themanagement costs. In summary, applying the CFR strategy tonature restoration projects in floodplains mayrealise a symbiosis between safety and nature.However, many knowledge gaps still need tobe filled before this new management strategycan be applied and scaled up to entire riversections. The most important knowledge gapshave been identified and incorporated into thisresearch project named ‘Symbiosis betweenSafety and Nature’. The planned research activities (20042008) address the following issues: (1)institutional arrangements of floodplain andriver management; (2) spatial and temporalapplication of the CFR strategy; (3) costeffective, innovative management techniques.

Institutional arrangementsDuring the last two centuries theresponsibilities and management tasks of thefloodplains in The Netherlands has hardlybeen changed. However, because of largescale transformation of agricultural use of

floodplains to nature management and theupcoming measures within the context of thenational flood defence project ‘Room forRivers’, a new and dynamic situation hasevolved. This requires a thorough analysis andpossible adjustments of the existinginstitutional arrangements between involvedstakeholders.

Spatial and temporal application ofthe CFRstrategyBecause the strategy implies periodic (cyclic)interventions at different locations, the spatialand temporal application of CFRmeasures isimportant. The preliminary study activitiesfocused on the Beuningen floodplain anddemonstrated that various scale levels need tobe addressed:• studying the ecological aspects should

involve the river section on both sides ofthe river between Nijmegen and Tiel (ca.40 km);

• considering the hydraulic effects of CFRinterventions, solutions for solving theproblems in the Beuningen floodplain canbe searched up to ca. 15 km downstream;

• an analysis of the institutionalarrangements has to be carried out at thelocal, regional and national scale.

Costeffective, innovativetechniquesAs stated, within the CFRstrategyinterventions are necessary on a regular basis.Therefore the development of costeffectivetechniques is crucial for a successfulimplementation of this strategy. The casestudy of the Floodplains Beuningen focuses onCFRinterventions realised by applying costeffective, innovative techniques. For example acombination between subsurface sandexcavation techniques and removal offloodplain vegetation will be investigated.Another promising technique is the use of atransformed agricultural machine to removeyoung trees effectively.

Deliverables of the projectThe ‘Living with Water’ research project‘Symbiosis between Safety and Nature’ willgenerate the following deliverables in 2008:• handbook CFRstrategy for river and

floodplain managers, focusing on concreteinterventions measures, planning in spaceand time and organisational and logisticrecommendations;

• design and realisation of CFRinterventions in the Beuningen floodplain



which will result in a river managementpermit delivered by the river manager;

• a series of scientific reports and articles inpeer reviewed scientific journals dealing withnature management, institutionalarrangements and technical innovations infloodplain management.

ReferencesBaptist, M.J., W.E. Penning, H. Duel, A.J.M. Smits, G.W.

Geerling, G.E.M. van der Lee & J.S.L. van Alphen,2004. Assessment of cyclic floodplain rejuvenation onflood levels and biodiversity in the Rhine River. RiverResearch and Applications 20(3), pp. 285297.

Duel, H., M.J. Baptist & W.E. Penning (eds.), 2001. Cyclicfloodplain rejuvenation; a new strategy based onfloodplain measures for both flood risk management

and enhancement of the biodiversity of the riverRhine. NCRPublications 142001, 72 p.

Jesse, P., 2004. Hydraulische weerstand in (natuur)ontwikkeling. De verandering van de hydraulischeruwheid van acht natuurontwikkelingsprojecten, april2004. RIZAwerkdocument 2003.124X, RIZA,Arnhem.

Mannaerts, J.J.H.M., 2004. Beuningse Uiterwaard Rivierkundige toets 2003. Rijkswaterstaat, DirectieOost Nederland, Arnhem, 10 p.

Smits, A.J.M., P. Nienhuis & R. Leuven (eds.), 2000. Newapproaches to river management. Backhuys, Leiden.

Van Velzen, E.H., P. Jesse, P. Cornelissen & H. Coops,2003. Stromingsweerstand vegetatie in uiterwaarden.Deel 1 Handboek versie 12003. RIZArapport2003.028, RIZA, Arnhem.





Rijnstrangen/Lingewaarden: a tap for the Rhine branchesM. van Ledden, M.E. Groot Zwaaftink & G.J. AkkermanRoyal Haskoning, P.O. Box 151, 6500 AD Nijmegen, The Netherlands; [email protected]

IntroductionThe high waters of 1993 and 1995 gave rise toa new policy of the Dutch Rhine branches‘Ruimte voor de Rivier’ (‘Room for the river’),consisting of measures creating moredischarge capacity in the river system.Staatsbosbeheer has developed anotheroption for this, ‘Lonkend Rivierenland’, withRijnstrangen/Lingewaarden as a central part ofthe plan. This plan consists of a new ‘river’ (i.e.an embanked floodway) along a formerchannel of the Rhine between the German Dutch border and the Pannerdensch Kanaal(Rijnstrangen). The second section of the newriver is along the river Linge, starting atDoornenburg on the Pannerdensch Kanaal andending at Druten on the River Waal. Thepresent research aims at investigating theeffects of such a new river branch on waterlevels and the discharge distribution in theDutch Rhine branches.

1DModelWe applied a onedimensional model of theDutch Rhine branches, the socalled‘Rijntakkenmodel’. The new riversRijnstrangen and Lingewaarden were includedwith several crosssections. The bed level ofthe crosssections follows the bed level in thearea. The shapes of the various crosssectionsin Rijnstrangen and Lingewaarden werederived from sketches made by BureauStroming. At the beginning of each section aweir is present with a crest level equal to thefloodplain level. As boundary conditions, thecurrent design discharge at the upstream end(16000 m3/s) and Qh relationships at thedownstream ends have been imposed.

ResultsA key result is that the discharge distribution atPannerdensche Kop can be regulated within alarge range. It is possible to divert more watertowards the south (i.e. the River Waal), butalso to the north (i.e. NederRijn/IJssel),depending on the weir levels in theRijnstrangen and Lingewaarden (Fig. 1).Besides, the new river decreases the waterlevel in various parts of the Rhine branches.Figure 2 shows that effect on the water levels

can be up to 60 cm. This drop contributes tothe aim of the project ‘Ruimte voor de Rivier’.

Conclusions and outlookIncluding the new rivers Rijnstrangen andLingewaarden results in:• a regulation possibility to divert more water

towards the northern (i.e. the IJssel andNederRijn) or the southern branches (i.e.the Waal);

• a significant lowering of the water levels.Effects on the morphology will be subject offurther research

Figure 1. Effect on the discharge distribution in theRhine branches. The numbers in the boxes refer to thenew discharge in each section.



Figure 2. Water level and effect on the water level at a discharge at Lobith of 16.000 m3/s for the present situation and thesituation with Rijnstrangen/Lingewaarden. A negative waterlevel effect means a decrease in water level causedRijnstrangen/Lingewaarden



Optimal design of multifunctional flood defences inurbanized areasA.J. Nienhuis 1 & B. Stalenberg 21 Section of Urbanism, Faculty of Architecture, Delft University of Technology, P.O. Box 5043, 2600 GA Delft, The Netherlands;[email protected] Section of Hydraulic Engineering, Faculty of Civil Engineering and Geosciences, Delft University of Technology, P.O. Box5048, 2600 GA Delft, The Netherlands

AbstractPartly due to urbanization in the RhineMeusedelta, it is often impossible to adjust theexisting dikes and thus alternative solutionsare needed. Improving the flood defence is nolonger a water management and hydraulicissue only, but it has also become an issue ofspatial planning and design. Goal of theplanned research is to find a method to designmultifunctional flood defences in urban areas.This research will focus on the design of newflood defences in new water managementsystems that have a surplus value for thespatial structure of city and landscape.

IntroductionTraditionally, multiple use of space is appliedin the Dutch wetlands. A dike, for example,was often both a flood defence and a road. Adike can therefore be considered as an axis ofurbanism. Most settlements have beenfounded near rivers because of the transportpossibilities and the occurrence of suitablesoils for agriculture. Gradually, these earlysettlements have developed into urban areas.The development of smart water managementsystems is a characteristic element in theprocess of urbanization of the Dutch lowlands.These systems play multiple roles in the urbanpatterns with respect to: flood protection,drainage, military defence, transportation,drinking water supply and recreation. Partly in response to urbanization, naturalriver floodplains have been reclaimed andquay walls have been raised. Anticipatedincreases in flood water levels, due to climatechange and sealevel rise (Ministerie vanVerkeer en Waterstaat, 2000, p.12) havecaused a new challenge in flood defence.In present urban landscapes it is oftencomplicated to adjust the traditional dikes andthere is a need for alternative solutions.Improving flood defence is no longer a watermanagement and hydraulic issue only, but ithas also become an issue of spatial planningand design. Various interests have to be takeninto account.

Flood protection in urbanized areasIn general, the solution for increased floodwater levels is sought in two directions: (1)‘room for the river’ [e.g. dike relocation or‘green’ rivers (embanked floodways parallel topresent river courses)] and (2) strengtheningthe current flood defence structures. Whereasthe latter approach is the ‘classic’ solution,‘room for the river’ is an alternative in theNetherlands that is presently investigated. Inresponse to the floods of 1993 and 1995 (Fig.1) the government has encourageddevelopment of this new approach towardsflooding, because it was doubted whetherstrengthening of the current flood defencewould be sufficient on the long term. Within anurban context, strengthening (raising) theexisting flood defence is often consideredproblematic and harmful for the city view. Onthe other hand, how to apply the ‘room for theriver’ approach to urban areas is still a matterof debate. We feel that both directions shouldbe combined in order to obtain an acceptablesolution.

Figure 1. Flooding in the city of Roermond, 1995.



It should also be recognized that watermanagement systems in urban areas, aremultifunctional and flood control is not the onlyaspect that has to be taken into account. Manyother difficulties occur at the urban riverfrontsuch as conflicts between traffic demands, thedesire for higher building densities and betterlivability. It should be stressed also that thecurrent historical city view (e.g. Fig. 2) is aresult of a constantly changing relationshipbetween the city and her waterfront. This isdue to the changing economic relationships,changing views on the relationship betweencity and water, and changing opinions aboutflood defence. We feel that placing the currentdemand for a new urban flood defencestrategy into this perspective, may lead to auseful extrapolation into the future. Thechallenge is to develop one integralmultifunctional solution for the currentproblems of the urban waterfront.

Planned researchThe goal of our research is to find a method todesign multifunctional flood defences in urbanareas. The research focuses on the design ofnew flood defences in new water managementsystems, which have a surplus value for thespatial structure of city and landscape. In thedesign process the two directions in flooddefence, ‘room for the river’ and strengtheningof the current flood defence structures will beexplored. Solutions may involve new types ofurban settlements and building constructionsin wetlands. Questions about weighing of theinterests, decisionmaking and organisationalaspects will also be addressed in our research.

They play in a complex field with variousactors, in a government policy context as wellas in a private context. An integrated approachwill be used, which covers relevant aspects inthe fields of flood defence, traffic circulation,aesthetics of the city view, landscapearchitecture and livability. Futhermore,probabilistic design and risk analysis will betaken into account.

Current activitiesPresently, rivers in the Netherlands andseveral cities on the Rhine branches and theMeuse are investigated, with the mainresearch questions given below.• What are the characterizations of the rivers

in the Netherlands?• How were the cities founded and in what

way did they develop?• How did and do they cope with floods?A broad reference work will be the outcome ofthe first phase of the research, which will bethe starting point of finding a method to designmultifunctional flood defences in urban areas.Updates of the results of our research can befound at:www.waterbouw.tudelft.nl/public/stalenberg.

ReferenceMinisterie van Verkeer en Waterstaat, 2000. Anders

omgaan met water; waterbeleid voor de 21e eeuw.Ministerie van Verkeer en Waterstaat, Den Haag, 70p.

Figure 2. Dordrecht: flood defence within the context of a historic city centre.



Communication strategies in river management; researchplan for comparison of two cases in the Netherlands

M.H. WinnubstCentre for Water and Society, Faculty of Science, Radboud University Nijmegen, P.O. Box 9010, 6500 GL Nijmegen, TheNetherlands; [email protected]

IntroductionThe floods that occurred in the Netherlands in1993 and 1995 confronted the Dutchgovernment with the consequences of achanging climate and spatial developments inthe Rhine and Meuse river basins. Peakdischarges of these rivers are expected toincrease in the future. The traditional technicalprotection measure of dike reenforcement, isno longer considered a sustainable solution.Alternative ways of reducing the dischargepeaks are spatial measures like dike relocationor lowering the floodplain or a combination oftechnical and spatial measures. Since thegovernment started the project ‘Room for theriver’ in 2000, possible measures for the Rhineand the lower Meuse have been studied.Examples are the Overdiepse polder retentionbasin on the Meuse and the dike relocation atLent on the Rhine.

In the framework of the European ‘Freudeam Fluss’ project, a research plan has beenmade for a comparison of the casesOverdiepse polder and Lent concerning thecommunication strategies of government andinhabitants, with the objective to determine thefactors that cause local communities to be proor against ‘Room for the river’ plans. In thispaper the two cases are briefly described andthe outline of the planned research ispresented.

Overdiepse polderThe Overdiepse polder (Fig. 1) measures 550ha in area and has 94 inhabitants and 19enterprises, mostly farms. Since thegovernment considered the Overdiepse Polderas one of the options for a flood retentionbasin, the inhabitants organized themselvesand took the initiative in preparing a plan. Thereason for their active participation was thenotion that a proactive and cooperativeattitude would help to clear away theuncertainty concerning their future livingcircumstances.

Various options had been explored of whichthree plans were elaborated. Finally, in June2004 government decided that the Overdiepsepolder could start with the realization of the socalled ‘Terp plan’ (Fig. 2). This plan includes acombination of technical solutions and spatialmeasures, such as dike heightening on thesouthside of the polder and the relocation ofhouses and farms in the polder to elevatedlocations on the dike. The plan would reducefloodwater levels in the river by approximately30 cm. The inhabitants are now discussinghow to execute the plan and to decide who canstay and who might be relocated outside thepolder.

Dike relocation LentThe Waalkade of Nijmegen and the dike ofLent on the opposite river bank (Fig. 1) form abottleneck in the Waal River during peak flows,affecting the safety of local communities. Since2000, the municipality of Nijmegen, theprovince of Gelderland, the waterboard

Figure 1. The locations of the Overdiepse polder andLent within the central Netherlands.

Figure 2. The Overdiepse polder with the ‘Terp plan’measures. Area shown 7 km wide; north is up



Rivierenland and Rijkswaterstaat have cooperated to find a solution for this problem.After an inventory Rijkswaterstaat (part of theMinistry of Transport, Public Works and WaterManagement) proposed to relocate the dikesnear Lent inland (Fig. 3). The inhabitantsprotested against this plans. Since they hadnot been involved in the preparation phase andthe design of the plan, the inhabitants decidedto develop their own plan. With support ofexperts they prepared the plan ‘LentseWarande’ (Fig. 4). In this plan the current dikewill be maintained and the floodplain will beexcavated. The area between the dike and theWaalsprong, an newly built urban quarter ofNijmegen some 350 m inland, could bedeveloped as a park, partially with temporarybuildings. If necessary, this area could provideroom for the Waal in the future.

After the presentation of this alternativeplan, the municipality of Nijmegen, theprovince of Gelderland, waterboard Rivierlandand Rijkswaterstaat signed an institutionalarrangement. They agree with the execution ofan Environmental Impact Assessment (EIA) ofthe two plans and the establishment of aproject organisation including: (1) a steeringcommittee, (2) a project group, and (3) a groupof representatives of local citizens and theirorganisations. After the EIA, the steeringcommittee will advice the State Secretary whowill decide which plan has to be realized. Afterthis decision the EIA will be published andinhabitants of Lent have a say on the results. Ifthese reactions are judged as relevant, theywill be included in the elaboration of thedefinitive plan.

Research outlineThe following steps will be taken in theinvestigation of the cases described above.The research question is: what are the factorsthat cause local communities to be pro oragainst ‘room for the river’ plans? The first stepis to get insight into the perception of eachstakeholder of the project, especiallyconcerning the communication between theactors in the project and between the actorsand the wider audience. Before conclusionscan be drawn about the strategies used, it isnecessary to understand the startingpoint ofeach stakeholder and the relationshipsbetween the different actors in the project. Thesecond step is the analysis of the projectstructure and project process itself. The thirdstep is the formulation of theoreticalrecommendations based on a review of therelevant literature in order to create aframework for the subsequent analysis andcomparison of the cases.

The results of these case studies mayhelp to get insight into the key factors ofparticipation and support of local communitiesin ‘Room for the river’ plans. The results areexpected to reveal underlying motives forstakeholder proposition and opposition. Theresults will be used for developing a jointplanning approach for Freude am Fluss riverprojects.

Figure 3. Overview of the Lent dike relocation plan. Areashown aproximately 2 km wide; view looking north

Figure 4. Overview of the ‘Lentse Warande’ plan.Areashown aproximately 2 km wide; view looking north.



Designing institutions for water managementS.V. MeijerinkNijmegen School of Management, Radboud University Nijmegen, P.O. Box 9108, 6500 HK Nijmegen, The Netherlands;[email protected]

AbstractMost water management problems are multilevel and multiactor problems characterizedby a high degree of substantive and strategicuncertainty. For social scientists it is aninteresting question which institutions areneeded for solving these problems. In thispaper it is argued that new practices ofnetwork governance (interactive, participatoryor open planning processes) are a promisingalternative to state power, but that moreresearch is needed on sources of networkgovernance failure. Based on observationsmade of the room for the river policy process, itis concluded that we are in need of anintelligent combination of strategies of networkgovernance and state power for solving ourwater management problems.

Introduction‘Institutions matter’ is a famous statement bythe Nobel prize winner for economics North.Institutions are, in the widest sense, rules.They can either be formal, such as Acts andthe Dutch ‘House of Thorbecke’, or informal,such as the Dutch consensus decision makingculture. Anyone who has ever participated inan international research or policy project mayhave experienced why institutions matter(Meijerink, 1999).

Institutions often show considerableinertia. Nevertheless, some institutions can bepurposefully designed (De Bruijn et al., 2002).We may for example decide on theintroduction of a market for water services or a‘watertoets’ for decision making on landusepolicies (Wiering & de Rooij, 2004). In thispaper we address the issue of institutionaldesign for solving wicked water managementproblems.

Wicked water managementproblemsMost water management problems are wickedproblems. Characteristic to these problems arethat multiple governmental and nongovernmental parties at multiple levels ofgovernment are involved in problem solving.These parties generally have different problemperceptions and policy preferences. Moreover,resources needed for problem solving, such as

legal, financial and political resources, aredistributed amongst them. Finally, wickedpolicy problems are characterized byuncertainty (Koppenjan & Klijn, 2004). The‘Room for the River’ issue, for example, ischaracterized by both substantive (riverdischarges expected) and strategic uncertainty(e.g. about strategic behavior of regional andlocal parties).

Markets, hierarchies and networksfor water managementIt is useful to think about the institutionsneeded for solving such wicked watermanagement problems. Basically we can drawon three types of institutions: markets,hierarchies and networks (Thompson et al.,1991). Markets are very good at providingprivate goods. We might think aboutpossibilities for organizing a market for drinkingwater supply, sewerage and/or waste watertreatment, though we should be extremelycareful with that, and it is necessary to protectpublic values, such as water quality or equalaccess to water services. For the provision ofpublic goods, such as dikes, or commongoods, such as clean water resources, there isa serious risk of market failure. In these casesmarkets do either produce negativeexternalities, such as water use or pollution tothe detriment of others, or free riders, i.e.parties that do not pay for a good or service,but nevertheless enjoy its benefits.

Because of these market failuresgovernment plays an active role in waterresources management in most countries. Inthe past decades, however, the water sectorhas experienced that state power or hierarchyis not very successful in solving wickedproblems either. Stakeholders that feel theyare worse off with newly developed policies ascompared to the status quo often try tofrustrate policy implementation successfully.Hierarchy invokes strategic behavior, and inpolicy controversies scientific research is oftenused strategically. Rather than a disinterestedsearch for truth, the policy process, then, ischaracterized by partisan use of researchresults and reports. The river dikestrengthening controversy of the eighties is aclear example of such a ‘dialogue of the deaf’



(Sabatier & JenkinsSmith, 1993), andtherefore of government failure.

For these wicked policy problems networkgovernance (or interactive decision making) isa promising alternative, mainly as it aims atconditioning a joint learning process. We maydistinguish between processes of substantive,strategic and institutional learning (Koppenjan& Klijn, 2004). Substantive learning is learningabout causeeffect relationships, policyalternatives and impacts of these alternatives.Strategic learning refers to learning about theperceptions and preferences of others, and theneed to take into account these other parties’perspectives by developing more cooperativestrategies. Finally, institutional learning refersto the development of shared norms andexpectations, and the development of a cultureof trust. The processes of deliberation andnegotiation aimed at the preparation of aregional advice in the Dutch Room for theRivers project are an interesting example ofsuch learning. The parties involved learnedabout the many policy alternatives, the manypossible combinations and their impacts. They,however, also learned about possibilities tocombine different policy objectives in multipurpose plans, and by that to address differentproblems at the same time. In spite of thesesubstantive and strategic learning processes,the relationship between some partiesremained rather tense, and a culture of trusthas hardly developed. Among other things, thismay be explained by the rather coercivestrategies the Dutch national government hasused in the controversy over the designation ofemergency flooding areas (Meijerink, 2004).

Network governance failureAs more experiences have been gained withthe new practices of network managementnow, policy scientists have begun to addressthe sources of network management failure. Inspite of the rather positive observations madeof the Room for the Rivers policy process sofar, it should also be noted that there has beena permanent risk that problems and costs arepassed on to other parties or levels ofgovernment. This particularly concerns parties’willingness to take policy measures for thebenefit of areas and parties situated moredownstream. Moreover, not in all cases it willbe possible to reach a consensus or

negotiated agreement. In the end, we may wellneed state power to solve these dilemmas ofnetwork governance.

Hierarchy or state power should neitherbe used to simply impose policies nor shouldinteractive policy making be used to createpublic support for policies that already havebeen decided upon. Hierarchy, however, maybe used fruitfully to create a sense of urgency,which implies that deliberations andnegotiations take place within the ‘shadow ofhierarchy’: if parties will not be able to reach anagreement, central government will have totake a decision in the end. Finally, state powermay be used to impose conditions thatsafeguard coordination at higher levels ofscale. The safety objectives for the Dutchrivers imposed by the Dutch nationalgovernment are a good example of that.

ConclusionsWhilst policy scientists have given ampleattention to sources of market failure andgovernment failure since long, they have onlyjust begun to address sources of networkgovernance failure. From recent experienceswith network governance in Dutch rivermanagement we may learn that strategies ofnetwork governance have been rathersuccessful so far, but that there are somedilemmas of network governance as well, andthat we may well need state power to solvethese dilemmas.

ReferencesDe Bruijn, H., E. ten Heuvelhof & R. in ’t Veld, 2002.

Procesmanagement, over procesontwerp enbesluitvorming, Academic Service, Schoonhoven.

Koppenjan, J. & E.H. Klijn, 2004. Managing uncertaintiesin networks. Routledge, London.

Meijerink, S.V., 1999. Conflict and cooperation on theScheldt river basin. Kluwer, Dordrecht.

Meijerink, S.V., 2004. Rivierbeheer als leerproces, eentussentijdse evaluatie van de PKB Ruimte voor deRivier. Bestuurswetenschappen 5, pp. 406427.

Sabatier, P.A. & H.C. JenkinsSmith (eds.), 1993. Policychange and learning, an advocacy coalitionapproach. Westview, Boulder.

Thompson, G., J. Frances, R. Levacic & J. Mitchell (eds.),1991. Markets, hierarchies & networks, thecoordination of social life. Sage, London.

Wiering, M.A. & P.A.E. de Rooij, 2004. De watertoets:nieuwe spelregels voor water en ruimte.Bestuurswetenschappen 5, pp. 391405.



What’s on a decision maker’s mind? Identifying barriersin information flows between actors in integrated watermanagement using mental model mappingM.J. Kolkman 1, P.A.T.M. Geurts 2, A. van der Veen 11 Department of Water Engineering and Management, Faculty of Engineering Technology, University of Twente, P.O. Box 217,7500 AE Enschede, The Netherlands; [email protected] Department of Research Methods and Statistics, Faculty of Business, Public Administration and Technology, University ofTwente, P.O. Box 217, 7500 AE Enschede, The Netherlands

AbstractThis research studies the relation betweenmental models and the decision processoutcome, in the specific case of the Zwollestorm surge barrier. Differences in mentalmodels between stakeholders will result indifferent lines of argumentation leading todifferent solution alternatives. The finaloutcome is considered suboptimal from apurely technical scientific point of view.

IntroductionA mental model (Doyle & Ford, 1998) containsthe elements and relations a stakeholderconsiders relevant for his position in thedecision making process. A mental modelrestricts information flows to only thoseaspects that affect him. Restrictions may be onthe scale (geographical boundaries, timehorizon, level of detail) and on the processesand relations considered relevant (includingphysical, biological, legal, financial, social).The mental model represents a causal chain ofargumentation that starts from the originalproblem and contains selected data andinterpretation thereof, to present convincingevidence for a favoured solution. The mentalmodel can be ‘run’ to simulate the effects ofintended actions. Different stakeholders mayuse the same starting point and the same data,

but with different interpretations, to arrive atdifferent effects. These effects are subsequently evaluatedin the frame against five major perspectives(Courtney, 2001). Perspectives determinewhat stakeholders see as their interests.Perspectives differ between stakeholders,influence every step of the decision makingcycle, and will result in the creation or supportof different alternative solutions. Theperspectives are related to a stakeholder’s(professional, institutional and personal)position in the decision making process, andare indicated with the letters T, O, P, E, A..

T: Technical A functional and rationalorientation with regard to systembehaviour.

O: Organizational A manager’s interpretiveorientation with regard to institutional andlegal responsibilities and consequences.

P: Personal A political and individualorientation with regard to position andpower.

E: Ethical A moral orientation with regard tocodes of conduct and values (e.g.environment).

A: Aesthetic An orientation on the beautyand harmony of a design.

Meaning

Frame of perception

Perspective types (T, O, P, E, A)

Mentalmodel

Real world data flow

Information flow II

Effect modelpredictions

Alternativeselection

Solutionspace

generation

Alternativeanalysis

Weightingbenefits& costs

Choice

Problemrecognition

Implementation

Evaluation

Problemdefinition I

Influences

Figure 1. Framework of analysis (Kolkmanet al., 2004).

Offer protection

Dike failurein Salland

NWstorm +high discharge Vecht +

low dischargeWeteringen

Threateningscenario's

WB21 &"Waternood"

Dikes complywith

WWK norm

Responsibilityof

Water Board

Dikes are notunsafe now

Dike improvementín accordancewith WWK96

Do nothing Zwollebarrier

Not functionalat high

discharge

Change of legalstatus of

Weteringen

Contradictswith

Additionalmeasures

Barrier controlby autonomous

computer

Backflow

Largequantities

Water frominside dike ring

Water fromoutside

dike ring

Duration

Too small

Inundation ofpolders

Bufferingwater in canals +

Weteringen

by means of

Is disputedforconsists of

differentalternatives

safety assessment

are partof

Prevents

has valuehas value

dependson

hascharacterist

is causedby

Requires Protectsagainst

Causes

Protectsagainst

Results in

Isclosed

by

are

is reported in EIAreport tooriginate from

Figure 2. Selected part of the mental map of a Ttypeactor. Ovals represent elements disputed by one or moreother stakeholders.



Feedback from the frame influences themental model (loop II in Fig.1). This is aniterative learning process that occurscontinuously during the decision makingprocess. Decision disputes are resistant toresolution by appeal to facts or reasonedargumentation present within the mental modelbecause stakeholders’ conflicting framesdetermine what counts as a fact and whatarguments are taken to be relevant andcompelling. A mental model’s general structurecan be explained by a stakeholder’s dominantperspective. We expect to find categories ofmental models that differ between types ofstakeholders. Each category has its preferredtypical alternative solution. The actual outcomeof the decision making process, however,cannot be predicted from the mental models.

Materials and methodThe theoretical framework is applied to theZwolle storm surge barrier case. The caseconcerns the improvement of sections 1 and 2of dike ring 53, as required by the act ‘Wet opde Waterkering 1996’ (WWK’96). Theresearcher was bootstrapped using the EIA(Environmental Impact Assessment) projectand related documents, and preliminary talkswith selected stakeholders. Mental models andframe elements were elicited from 14stakeholders using semistructured interviews(Kolkman & Van der Veen, 2004). Interviewswere processed into mental maps (Fig. 2). Atotal of 67 disputed map elements wereanalysed in an overview grid (Fig. 3). The mainmap elements were processed into a causaldecision explanation model (Fig. 4).

Results and discussionWhile all actors generally start from the sameunderstanding of the system, some crucialdetails differ between (groups of) actors.These details concern assumptions anduncertainties present in model calculations[MHWs (Design high water levels), frequenciesand inundation severity], and the exact natureof historical data (on flooding of polders).Interpretations of data exhibit much morevariation, e.g. on the questions below.• Does the WWK’96 inevitably prescribe a

closed dike ring?• Can innovating concepts like a risk

approach and norm differentiation beapplied to dike ring 53?

• How is the distribution of institutionalresponsibilities and accompanyingexpertise between Water Board, Provinceand Ministry to be interpreted?

• Can the city centre of Zwolle be flooded bywater from the river Vecht?

• Is the city centre of Zwolle safe for floodingwhen the storm surge barrier is closed?

Mental models of some actors show elementsof careful construction toward a desiredoutcome, or are reconstructed at a later time tolegitimise the outcome and to accommodatenew events. Remarkably none of the actorssupports the original dike improvement plan.

Disputedelement

Observations from interviews Scores 1 2 3 4 5 6 7 8 9 10 11 12 13 14

2 Dike ringprinciple

Must be physically closed …vs… canhave an open discharge canal oncondition that the safety normremains guaranteed.

Closed(G),Open(O)

G O G G G G O G G,O

G G (G)

16 ProvinceWWKapproval

Province must dissociate and limitto assessing the reasonability of theEIA report contents …vs… can fullyparticipate on contents aspects alsofrom the start of the project

Distance(R),Involvement(M)

R M R (M) M R R*

R M R

33 Watersystemdischargepeaks

Discharge peaks from the riverVecht and the Sallandse Weteringendo not coincide …vs… in the pastthe discharge peaks were alwaysobserved to coincide

Differ(N),Coincide(S)

S S N S N S * S

Figure 3. Example rows of the analysis overview grid, showing relevant generalized disputed concepts with stakeholderscores. The scores can be related to a stakeholders’ dominant frame perspective. The stakeholders interviewed are listedacross the header of the table with a code number (114).



ConclusionThe Tperspective apparently has ‘lost’ theargumentation ‘battle’ against the Operspective. The P and Eperspectivesexplain the project delay. The resultingapproach to separate water protection andwater control does not solve the floodingproblems in Salland, upstream of the Zwollebarrier. The case results confirm thetheoretical framework. Different stakeholderperspectives can be related to different(groups of) mental model elements. Anoverview grid is suitable to analyse disputedconcepts. The mental models revealassumptions and interpretations implicitlypresent in the various alternative solutions,identify barriers in communication andinformation flow, and can be used to explainthe decision process outcome.

ReferencesCourtney, J. F., 2001. Decision making and knowledge

management in inquiring organizations: toward a newdecisionmaking paradigm for DSS. Journal ofDecision Support Systems 31, pp. 17–38.

Doyle, J. K. & D. N. Ford, 1998. Mental models conceptsfor system dynamics research. System DynamicsReview 14(1), pp. 330.

Kolkman, M. J., M. Kok, A. van der Veen, 2004. Conceptmapping as a new tool to visualise the use ofinformation in decisionmaking. Physics andChemistry of the Earth 30 (45), pp. 317332.

Kolkman, M. J. & A. van der Veen, 2004. Mental models the root of precautionary approach? Civil Engineeringand Management research reports, 2004R008 /WEM011 (Int. r. no. 15684652). Civiele techniek(Faculteit CTW), Universiteit Twente, Enschede, 29 p.(abstract available athttp://upem.er.dtu.dk/programme.htm)

Figure 4. Top part of causal decision explanation model



Planning a green river as a solution to increasingdischarge in an urbanizing area on the River RhineH.P. Wolfert 1, L.C.P.M. Stuyt 1, A.G.M. Hermans 1, J. Kruit 1, R.J.W. Olde Loohuis 1 & F. Klijn 21Alterra, Wageningen University and Research Centre, P.O. Box 47, 6700 AA Wageningen, The Netherlands;[email protected] WL | Delft Hydraulics, P.O. Box 177, 2600 MH Delft, The Netherlands

IntroductionFuture land management in the upper part ofthe Rhine delta, The Netherlands, will face twoproblems in the near future. The urban fringesof the two mediumsized cities of Arnhem andNijmegen will expand strongly, and urbansprawl may be a threat to the quality of theenvironment. At the same time, more space isneeded for the safe discharge of river floods,which are expected to increase in the nearfuture due to climatic change. Since the RiverRhine floods of 1993 and 1995, this is apolitical issue of high urgency. We investigated whether a new, large riverbypass, called floodway in the USA or greenriver in The Netherlands, will provide a solutionto both problems mentioned (Wolfert et al.,2004). Our example was the Mississippi delta,where the Morganza and the Bonnet Carrefloodways were constructed following theGreat Flood of 1927 to pass floodwaters fromthe Mississippi River to the Gulf of Mexico,thus safeguarding the city of New Orleans fromflooding.

The planning of such a green river in acultural landscape will involve major land usechanges and many people will be involved inthe decision making. Therefore, aims of thestudy were: (1) to indicate possibilities for newtypes of land use and to visualize the future

landscape and (2) to demonstrate the impacton the water levels in the river system duringflood events. The results were compared withthe effects of retention polders in the samearea, which is another option of whichexamples exist along the Upper Rhine inGermany.

Plan designThe Green River Lingewaarden comprises tworeaches (Fig. 1). The upstream Rijnstrangenreach is surrounded by old dikes as these3200 ha of land were regularly flooded until the1960s. The downstream, new Lingewaardenreach is planned in the former floodbasin inbetween two embanked Rhine distributaries,the Rivers Waal and NederRijn. In the lowestpart of this floodbasin, 2900 ha of green river isdesigned with a minimum width of 500 m nearimportant highway and railway crossings inorder to reduce construction costs, but muchwider where there are no builtup areas atpresent or envisaged in the very near future, inorder to allow some backswamp restoration inthis area.

For each of the two options water levelsand dikes heights required were calculated.Water depth was calculated using data onpresent water levels – assuming the greenriver will prevent maximum water levels to rise– and data on altitude. In case a green river ispreferred, new dikes have to be up to 8 m

Figure 1. The green river, planned along the present Rhine distributaries (present embanked floodplain in blue). Area showsmeasures ~45 km in width; north is up.



high, while a choice for the option of retentionbasin requires dikes of more than 10 m high.As a green river, the area will discharge wateras soon as the present embanked floodplainsare flooded and thus will functionapproximately once in one or two years as afloodway. As a retention basin part of thedischarge of extreme floods events is stored inthe area, and released as soon as the waterlevels in the river system get lower, an eventwhich is estimated to occur only once inapproximately 600 years.

Future landscapesFuture land use was explored based on thequalities of the present landscape, spatialdevelopments in land use, and future floodingfrequencies and water levels. It was assumedthat the frequent presence of water in a greenriver will induce land use changes, but that therare inundation of a retention basin will notlead to changes. Thus, the option of a greenriver will enhance functions such as nature andrecreation. The frequent presence of watermay attract building highquality residencesalong the new dikes. The new dikes may beused for new types of recreation infrastructuresuch as long distance footh pats and cycletracks, that enable citizens to enjoy theirsurroundings more than before. Based upon the local qualities of variousparts of the area, new land use combinationswere described to occur in the various parts ofthe green river (Fig. 2). In the Rijnstrangenreach emphasis will be on agriculture withnature, in the eastern part of the Lingewaardenreach development plans with urban parksmay be anticipated (Fig. 3, see page 19), whilein the western part of the Lingewaarden reachcontinuation of agricultural use will conservethe highly esteemed openness of the presentlandscape.

River managementWhen the socalled design discharge willincrease from its present 16.000 m3/s to afuture 18.000 m3/s, the new green river willdischarge 20003000 m3/s, which will lead to adrop in water levels of approximately 60100cm along the present dikes. In the case of agreen river, there is no impact on water levels

in the downstream river reaches in the delta,as would be the case when retention polderswere constructed. However, that option wouldonly lead to a 3540 cm drop in water levels inthe study area, which is not sufficient in thelong run. The new design discharge wouldrequire a retention basin of 810 times the sizeof the area investigated here. The construction of a green river does notchange the discharge distribution over thevarious Rhine distributaries profoundly, butmore research is needed on this. Anadvantage of the option of a green river is thatit does not require any operationalmanagement during the rise of the waterlevels. In contrast, a retention basin must beopened and closed with precision on the rightmoment, otherwise it will not effectively reducethe flood peak water level in the river system.The decision on whether to open the basin ornot are seen as a great risk of failure.

ConclusionsCompared to the option of retention basin,construction of a green river will be the best inorder to stop the present process of urbansprawl, because of the frequent inundations.These inundations will also induce new typesof land use contributing to the environmentalquality of the urban areas, and will enhancedevelopment plans with highqualityresidences. Realization of the latter maycontribute to the financing of the constructionof new dikes. Besides we argued that a greenriver is not only a sufficient but also a robustsolution to the increasing river discharges, as itrelies on natural functioning withoutmanagement procedures, resulting in a muchlower risk of failure during floods.We concluded that a green river is a rigoroussolution, but deserves the attention of decisionmakers, because it provides longlasting

opportunities for an environment which is safeand pleasant to live in.

ReferenceWolfert, H.P., L.C.P.M. Stuyt, A.G.M. Hermans, J. Kruit,

R.J.W. Olde Loohuis & F. Klijn, 2004. Bergendestroming KAN. Alterrarapport 973, Alterra,Wageningen, 23 p.

Figure 2. New landuse combinations in the various parts of the green river



Figure 3. The future landscape in the urban part of the Lingewaarden reach



Rapid assessment methodology for river management withapplication to the Lower Meuse proposed researchJ.A.E.B. Janssen 1, J.L. de Kok 1, M.S. Krol 1, S.J.M.H. Hulscher 1 & R.M.J. Schielen 21 Department of Water Engineering and Management, Faculty of Engineering Technology, University of Twente, P.O. Box 217,7500 AE Enschede, The Netherlands; [email protected] Institute for inland Water Management and Waste Water Treatment (RIZA, P.O.Box 9072, 6800 ED Arnhem, The Netherlands

IntroductionRecent peak discharges in several largeEuropean rivers, such as the rivers Meuse (Fig.1), Rhine, and the German Elbe River led to theawareness that flood safety is still a key aspectin strategic river management. However, thisaspect is sometimes conflicting with otherfunctions, such as spatial planning, ecology andagriculture. Tradeoffs have to be madebetween the various objectives. This projectaims at improving strategic river planning byproviding a flexible and integrated frameworkfor river management.

Problem identificationVarious tools apply to make the required tradeoff in river management (Nieuwkamer, 1995;Schielen et al., 2001; Matthies et al., 2003;Ministerie van Verkeer en Waterstaat,Rijkswaterstaat Directie Limburg, 2003).Decision support systems, different models andqualitative studies can be used to assess riverstrategies in an integrated way. Althoughconsiderable research effort has been spent onthe development of integrated tools thepractical applicability often remains limited dueto a number of reasons: (1) the occurrence ofblank spots where no measures were or couldbe assessed; (2) a lack of flexibility to cope withchanging future conditions or enduserrequirements; (3) difficulties with integration ofqualitative and quantitative model concepts.

Integrated water management requires amethod that allows for quick iteration in case ofchanging conditions either in the field ofclimatologic change or policy or enduserrequirements. Furthermore, to cope with blankspots, the method should be based on minimaldata requirements and relatively simplecalculations. Finally, qualitative and quantitativemodel aspects should be integrated. Theongoing Integrated Exploration Meuse (Fig. 2),one of the largest river planning projects in theNetherlands, serves as a case study for thisresearch.

Research methodologyThe research comprises the following steps.• Qualitative systems analysis: consulting

endusers, stakeholders and selectedexperts to make an inventory of relevantindicators, measures, and scenarios forfuture conditions. Modeling effects:

Figure 1. Peak discharge in the Meuse, Geulle 2003.

Figure 2. Lower Meuse; study area for IEM (Ministerievan Verkeer en Waterstaat, Rijkswaterstaat DirectieLimburg, 2003).



collecting data and theoretical concepts tomodel to describe the relations in thequalitative systems network. The availabilityof the Planning Kit Meuse, developed byWL Delft and involving the hydraulic effectsof measures, allows for calibration ofresults.

• Evaluation of river strategies: testingrobustness of river strategies formulated forthe program Integrated Exploration Meuse(IEM) under the formulated scenarios andranking the strategies.

• Verification of the methodology: testing thegeneric applicability of the method bymeans of a second case study.

Integration of qualitative and quantitativeaspects takes place by using fuzzy set theory.This approach originates from social scienceand will allow for integration of qualitativeknowledge in the model, resulting in agenerically applicable approach (Fig. 3).

AcknowledgementsThis project is partially embedded in theInstitute for Governance Studies of theUniversity of Twente. Results of the IVM Studyand the Planning Kit Meuse, developed by WL |Delft Hydraulics, will be used.

ReferencesMatthies, M., J. Berlekamp, S. Lautenbach, N. Graf, S.

Reimer, B. Hahn, G. Engelen, M. van der Meulen, J.L. de Kok, K.U. van der Wal, H. Holzhauer, Y. Huang,M. Nijeboer & S. Boer, 2003. Pilotphase für denAufbau eines Enstcheidungsunterstützungssystem(DSS) zum Flusseinzugsgebietsmanagement amBeispiel der Elbe. Bundesanstalt für Gewässerkunde,Projektgruppe Elbe Ökologie, BMBFForschungsvorhaben FKZ 339542A, KoblenzBerlin,102 p.

Ministerie van Verkeer en Waterstaat, RijkswaterstaatDirectie Limburg, 2003. Integrale verkenning Maas;advies, hoofdrapport en achtergronddocumenten.Ministerie van Verkeer en Waterstaat, RijkswaterstaatDirectie Limburg, Maastricht.

Nieuwkamer, R.L.J., 1995. Decision support for rivermanagement, Ph.D. thesis, University of Twente,Enschede.

Schielen, R.M.J., C.A. Bons, P.J.A. Gijsbers, W.C. Knol,2001. DSS Large Rivers, Interactive FloodManagement and Landscape planning in RiverSystems: development of a Decision Support Systemand analysis of retention options along the LowerRhine river, Final Report IRMASponge04WP. NCRpublication 132001, Netherlands Centre for RiverStudies, Delft, 61 p.

Figure 3. Application of the RAM approach.



Damage due to low flows on the MeuseG.T. Raadgever 1, M.J. Booij 1, J.A.P.H. Vermulst 2 & S.J.M.H. Hulscher 11 Department of Water Engineering and Management, Faculty of Engineering Technology, University of Twente, P.O. Box 217,7500 AE Enschede, The Netherlands; [email protected] Royal Haskoning, P.O. Box 1754, 6201 BT Maastricht, The Netherlands

AbstractIn this research the damage due to lowdischarges on the Meuse has been analysedto get a better view on the scope of the lowflow problem. The research area consists ofthe Dutch and the Flemish part of the Meuseupstream of Roermond and the canals fed byit. A model has been developed to assess thetotal damage and the distribution over differentregions and economic sectors in a number ofsituations. Total damage varies from a fewmillion Euros in a medium dry year to ten timesthat much in an extreme dry year. Most of thedamage occurs in the navigation sector andthe power generation sector. Climatologicaland economical development will increasefuture damage substantially. On the otherhand, much can be gained by applyingappropriate management strategies.

IntroductionThe main aim of this research is to createinsight into the damage due to low dischargesof the Meuse (Fig. 1). This insight can helpwater managers (e.g. Rijkswaterstaat) toevaluate international, national and regionalagreements concerning the distribution of lowflows. It can also give direction in thedevelopment of strategies to alleviate the lowflow problem now and in the future. Theresearch area consists of the Dutch and theFlemish part of the Meuse upstream ofRoermond and the canals fed by it (like theAlbert Canal, Juliana Canal and ZuidWillemsvaart).

Damage modelA model has been developed to assess thetotal damage and the distribution of thedamage over several regions and economicsectors in a number of situations. The modelcalculates the damage that would occur in thecurrent system, if it would be confronted withcertain characteristic discharge series, in threesteps. First, the distribution of the dischargeover the main branches of the water system isdetermined. Then water shortages arequantified for each economic sector, bycomparison of water supply and demand. Inthe last step, the financial damage is assessedfor the relevant economic sectors: navigation,agriculture and power generation. Damage tonavigation and agriculture is caused by watershortages, but damage to power generation iscaused by high temperatures of the riverwater, which is used for cooling purposes.

ResultsThe damage that can be expected in thecurrent situation has been computed for anumber of characteristic years, based on theyearly cumulative discharge deficit. Thedamage varies from about 6 million Euros in a50%dry year to over 30 million Euros in a 1%dry year (Fig. 2). In a 50%dry year almost90% of the damage occurs in the powersector, but in a 1%dry year that fraction is onlyabout 30%. In a 1%dry year most of thedamage occurs in the navigation sector. Thedamage is caused by the increasing delay ofships at locks, when more economical lockprocedures are applied. The damage tonavigation is particularly high on the AlbertCanal in Flanders, where – in contrast to in theNetherlands no pumps are installed to pumpback the locking water. In the agriculturalsector substantial water shortages do occur,but the damage caused by these shortages isnegligible. To develop insight into the possible futureincrease of damage, the model has beenapplied to a number of scenarios forclimatological and economical development. Inthe most extreme of the two applied climatechange scenarios, the damage in a 10%dryyear nearly doubles in 100 years time. Asubstantial part of the extra damage is causedFigure 1. Low flows on the Meuse in 2003



by an increase in the water temperature. Thescenario for economical development predictsan increase in damage of about 20% in 10years time. The increase in damage is mainlycaused by an increased intensity of ships onthe Juliana Canal and the Lateral Canal. If theeconomical growth will continue with the samerate, the increase in water demand will addmore to the low flow problem then thedecrease of the water availability due toclimate change. Finally, the model has been used toidentify beneficial solutions to the low flowproblem, by assessing the damage that occursunder several management strategies. Thedamage to power generation is not taken intoconsideration in the evaluation of managementstrategies, because the strategies do notinfluence the water temperature. Since thedamage to agriculture is already very limited,the strategies are mainly aimed at decreasingthe damage to navigation. The appropriate management strategieslead to a decrease of the total damage ofnearly 20% to over 50% in a 10%dry year.First of all, much can be gained by adjustmentof the water distribution over the economicsectors and the regions in the research area.To decrease total damage, more water shouldbe made available for navigation at the JulianaCanal and the Albert Canal. The distribution ofwater over the river system can however notbe adjusted without (political) effort. Besideoperational measures, a couple of appropriatestrategical measures can be identified, amongwhich the installation of pumps to pump backthe locking water.

DiscussionThe possible fault in the calculated damage iscaused by various sources. First of all, a feweconomical sectors and processes areexcluded from the model. Secondly, the modelschematisation is a little inaccurate due to thefixed discharge distribution and the simplifieddamage functions. Finally, the uncertainty inthe input and the model parameters – mainlythe parameters for the power generation andnavigation sector cause quite someuncertainty in the output. Therefore, the totaluncertainty in the calculated damage issubstantial.

Conclusions and recommendationsLow flows on the Meuse are mainly a problemto navigation and power generation. Part of thedamage is unavoidable and has to beaccepted. Nonetheless, the total yearlydamage can be decreased substantially byapplying appropriate management strategies.Especially with regard to the expectedincrease of the low flow problem due to climatechange and economical development, it seemsto be wise to implement a combination ofoperational and strategical measurses. Toestablish a widely accepted package ofeffective and efficient measures, it isrecommended to consider all financial andnonfinancial effects to the interested parties.

Figure 2. Damage to the economical sectors in the Netherlands (NL) andFlanders (FL) in characteristic years.



Integrated flooddamage and risk assessmentY. Huang 1, J.L. de Kok 1 & A.E. Mynett 21 Department of Water Engineering and Management, Faculty of Engineering Technology, University of Twente, P.O. Box 217,7500 AE Enschede, The Netherlands; [email protected] WL | Deflt Hydraulics, P.O. Box 177, 2600 MH Delft, The Netherlands

IntroductionFlood losses used to be assessed statistically,based on the risk analysis approach (CUR,1990; Vrijling, 2001). With the development ofremote sense technology and driven by therapid climate change, the physicalbasedapproach becomes more frequently applied(Van de Sande et al., 2003). Both approachesappear important in decision making: theformer in longterm flood defense planning andthe latter in shortterm flood mitigationmanagement. The study of natural systems calls for anintegrated approach in risk assessment. InEurope, shifted from structural measures tononstructural measures, modern flood riskmanagement has moved towards improvingflood mitigation through the improvement offlood warning and modelling systems(PenningRowsell et al., 1994). However, indensely populated river basins in countriessuch as China, structureal measures remainimportant (Yin et al., 2001). Thus, riskassessment is required to be able to analyzethe possible outcomes of any plan, strategy orproject, at different temporal and spatialscales. In this study, attention shall be paid tothe issues mentioned below.

• Most of the damage functions were depthbased. Improvement is needed due to theneglect of quantified inclusion of importantvariables such as velocity (Kelman, 2004;Fig. 1).

• Previous uncertainty analysis of flood riskassessment focuses on internalparameters involved in the risk model (e.g.NRC, 2000). External uncertainty such asuncertainty propagation through hydraulicmodels is rarely reported.

Methodology and case studyKey components and processes of theintegrated flood damage and risk assessmentsystem are shown in Figure 2. A case studyhas been set up on the river reach nearSandau in the River Elbe in Germany. MonteCarlo simulation propagates uncertaintythrough the system. The dike effect is studiedwith an artificial dike break simulated withSOBEK2D.

ResultsThe system can assess flood risk with: (1)expected annual damage; (2) damageassociated with a certain flood event; (3) theimpact of flood mitigation measures. Here, twotypical results are presented.1. A risklevel map (Fig. 3) – This map has

been obtained by combining the maps ofpercentage damage and the velocitydistribution. The idea is to distinguish fourrisk levels indicated by the indexes of R1to R4, which are corresponding to activitiesto be taken for decision making. The levelsof risk can be different nationwide.

2. The dikebreakeffect (Fig. 4) – This effectwas assessed using Monte Carlosimulation of the eventbased damage

Figure 2. Framework of the integrated flood riskassessment system.

Figure 1. Flood damage caused by high flow velocity(from BBC 2004:http://news.bbc.co.uk/1/hi/in_pictures/3571748.stm).



model. The result shows that damage isalways higher when dike break occurs.However, this result only shows the localimpact. Nonlocal impacts can be obtained byenlarging the modelling area downstreamtowards the area of interest. This work is still inprogress.

Conclusions• An integrated flood risk assessment

system has been developed. It can beused to provide multidimensional aids toflood management decision making inlongterm planning and shorttermoperations of the flood defence system.

• Inclusion of velocity improves the damagedriven forces.

• Integrated uncertainty analysis assistsflood risk presentation with morecomprehensive information.

AcknowledgementsThis work has been carried out partly for theproject Elbe_DSS funded by the GermanFederal Institution of Hydrology (BfG).

ReferencesCUR/TAW, 1990. Probabilistic design of flood defences,

Central for Civil Engineering Research and Codes,report 141, Gouda.

Kelman, I. & R. Spence, 2004. An overview of floodactions on buildings. Engineering Geology 73, pp. 297309.

NRC, 2000. Risk analysis and uncertainty in flood damagereduction studies. National Academy Press, WashingtonD.C. (http://www.nap.edu)

PenningRowsell, E., & M. Fordham, 1994. Floods acrossEurope flood hazard assessment, modelling andmanagement. Middlesex University Press, London.

Van der Sande, C.J., S.M. De Jong, & A.P.J. De Roo,2003. A segmentation and classification approach ofIKONOS2 imagery for land cover mapping to assistflood risk and flood damage assessment. Internationaljournal of Applied Earth Observation andGeoinformation 4, pp. 217229.

Vrijling J.K., 2001. Probabilistic design of water defensesystems in The Netherlands. Reliability Engineering andSystem Safety 74, pp. 337344.

Yin, H. & C. Li, 2001. Human impact on floods and flooddisaster on the Yangtze River. Geomorphology 41, pp.105109.

Figure 3. Risklevel map

Figure 4. The dikebreakeffect



Baseline MIXER: a GIS application for managinggeographic information and hydraulic models for riversS.P.J.M. van de Pas & N.G.M. van den BrinkInstitute for Inland Water Management and Waste Water Treatment (RIZA), Postbus 9072, 6800 ED Arnhem, The Netherlands;[email protected]

AbstractRiver engineering in the Netherlands nowrelies heavily on detailed geographicinformation. Information ofn dikes, elevationand vegetation is gathered and maintained inan instrument called ‘Baseline’. From thisdatabase the required hydraulic, ecologic ormorphologic models are drawn. This is donelargely automatically. Recently the Baselinedata controlling environment was extendedwith the facility to incorporate changes to thedatabase. This extension is called the MIXER.

What is Baseline?Baseline is a GISoriented database formanaging and processing spatial data for 1D,2D or 3D hydraulic models. It is an automatedswitch between ground truth data and modelinput for models such as SOBEK and SIMONA(WAQUA and DELFT3D) models. It providesstructures for data managing and applicationsfor editing. Baseline was developed in ArcInfo, mainlyusing the Arc Macro Language (AML). Someparts have been programmed in Fortran andC++. The Baseline application requires eitherArcInfo 7 or ArcGIS 8, and runs on UNIX, NTor XP platforms. At the moment RIZA isworking on a conversion to the ArcGISprogramming environment.

Baseline and dataThe workflow of Baseline involves three datalevels.1. Primary data The Baseline input

originates from three types of datasources: depth soundings points en lineswith height information, topography andfiles that describe ecotopes. If necessary,data can be preprocessed before beingsaved in the Baseline database.

2. Derived data Baseline softwareapplications process the primary data intoderived data like a Digital Elevation Model,Roughnessdata, Weirs and Sobeksections.

3. Applications Primary and derived dataserve as input for hydraulic andmorphological models. Baseline containsapplications to generate input files forSOBEK, WAQUA and DELFT3D.

Relations with primary andsecondary processesCalibrated WAQUA, DELFT3D and SOBEKmodels for the river Rhine and the river Meusehave been made with Baseline. The WAQUAmodels are typically used to assess the designwater levels. SOBEK models are regularlyused for flood prediction in the lower parts ofthe Rhine and Meuse. Because all data aremanaged in Baseline, it is easy to updatethese models on a yearly basis.

Baseline MIXERIn 2003 the Baseline data controllingenvironment was extended with the facility toincorporate changes to the database. Thisextension is called the MIXER. It is used toalter the database conform proposed humaninterventions or autonomous changes withinthe prototype. The extension enables the userto include a large number of changes to thedata model of the prototype automatically inone simple action. This addition enables us to make strictdistinction between the actual geometry andriver management measures. Measures aredefined as specific changes of the geometry.Strict distinction between actual situation andmeasures not only promotes the integrity of thedatabases itself, it also makes the informationof proposed measures transparent andsuitable for exchange by means of email.Because the information is entirely geographicit is accessible to anyone that has access toArcGIS software. It does not require any priorknowledge of hydraulic models.

Baseline and large studies for RivermanagementBaseline is used in a number of large studiesfor river management, such as Room for Riverand the river Meuse project. In these projectsBaseline proved to be a suitable tool formodelling any change in river geometry.Examples are widening or lowering the mainchannel, lowering of the floodplains (e.g. Fig.1), removal of bottlenecks in the river and the



planning of scenic areas. The use of Baselinein such projects is very beneficial becauseresults and hydraulic models are reproducibleand objective. Nowadays, by development of the MIXER,Baseline is an essential tool for hydraulic datamanagement in projects such as ‘Maaswerken’and ‘Room for the River’. Measures thatincrease the discharge capacity of the riversare processed in Baseline, and then convertedwith the MIXER automatic into the hydraulicmodels. With these models, the effects of thedifferent measures are assessed.For instance, when studying the IJsseldeltaarea in the Netherlands, two plan alternativeswere made for the same area. In one of themodels the proposed construction of a siltdepot in the lake was incorporated, in the othermodel it was not. In order to be able to makecalculations for the current situation and thesituation after completion, both modelswere made and managed by Baseline.

ConclusionWe think the approach of database andMIXER is such a success that it should beextended further. Until now soil compositionand features have not been included in thedatabase. These are probably the next twocharacteristics to be gathered within thedatabase. More challenging are other featureslike economic value, ecologic value, fauna andnavigation intensity. Because the system ofdatabase and MIXER is so well fitted for alkinds of geographic network analyses, wethink further extension of the databases isprobably the best way to fulfill the promises oftrue integral river management.

Figure 1. Example of the use of Baseline MIXER.(a) Elevation Model in actual situation. (b) Elevation Model after a possible measure



River widening: from understanding the subsurface todigging for ‘gold’T.P.F. Koopmans & J.H.J. EbbingSyncera GeoData, Dr. Stolteweg 54a, 8025 AX Zwolle, The Netherlands; TKO@syncerageodata.nl

AbstractThe subsurface plays an important role in thenecessary subsequent steps when dealingwith the planning of river widening. Itdetermines, together with design and location,the quality and quantity of the materials thatwill be removed. Geological,geomorphological, socioeconomic andenvironmental information therefore is,together with riskassessment, of great interestfor planners and decision makers in riverwidening projects.

This paper presents the differentnecessary subsequent steps when dealingwith the planning of river widening: startingwith a phase of choosing where a preliminarydesigned project should or may take place,followed by the (sub)surface effects phase(including an EIA Environmental ImpactAssessment) plus an insight in the rawmaterial flows, and a phase of collection ofnecessary additional new information,eventually leading to the implementation of theriverwidening project itself. Knowledge of thesubsurface could and probably should play acrucial role, especially in combination withother more superficial data and socioeconomic considerations.

IntroductionThe subsurface plays an important role in thenecessary subsequent steps when dealingwith the planning of river widening. Fourphases are distinguished within the riverwidening process: (1) the investigation ofgeological information, together with severalother important themes, to understand thesurface and the subsurface; (2) the creation ofan EIA (‘MER’) describing the (interaction of)surface and subsurface effects of the riverwidening processes; (3) the collection of extrainformation on various subjects; (4) a finalphase that leads to implementation of the riverwidening project and restoration of the area,often after many years of discussion anddeliberation. Below, special emphasis is givento the important role of the subsurface withineach phase mentioned.

Phase 1: geological information...understanding the earthIn the first phase (choosing where apreliminary designed project should or maytake place) describing and understanding the(sub)surface provides necessary input for thedecision making process. A world ofinformation is often available about thesubsurface and the surface, mainly fromborehole data and various thematic maps.Combined they provide insight in geology,morphology, geotechnology and hydrology.Using 3dimensional geological modelling,geological interpretation techniques,geostatistics and Geographical InformationSystems (GIS) the subsurface can bedescribed and understood (e.g. Fig. 1). In thisway the subsurface can provide the necessaryinput for the decision making process. It canwork as a solid underlay for other relevantthematic maps.

Phase 2: combining information effects at the surfaceIn the next phase subsurface data arecombined with other demands regarding riverwidening. Such as, the preliminarytechnological project design, floodcontrol andmore superficial data (environmental data,pollution data, socioeconomic data, laws andpolicy, etc). In this way a wellbalancedassessment can be made where to undertakeriver widening.

Figure 1. Depth of the top of the coarse sand deposits belongingto the Kreftenheye Formation in the Tiel area along the riverWaal.



Many of these items are dealt with in anEnvironmental Impact Assessment (EIA) or‘MER’ in Dutch.Important items are a careful planning ofmaterial flows, combination and clustering ofdifferent projects, and dealing with landuse,nature and mineral planning in a smart way.Together with insight in the release ofeconomically interesting raw materials, andsustainable use of reusable materials theseare all important aspects of the assessmentand may help making a wellbalanced choice.

Phase 3: getting extra informationIn this phase a decision about the exactlocation for the project and the final design willbe narrowed down. Getting closer to the phaseof implementation it is often considerednecessary to collect extra information (e.g. Fig.2). Information that may influence the choicefor the exact location of the new bypassesand dikes. Information that may lead topossible benefits when the removal of pollutedsoil can be avoided, or that createspossibilities for nature restoration incombination with mineral extraction (e.g. Fig.3). Welldocumented information is especiallyimportant as environmental and planning lawsare strict. As well as to address nature andenvironmental groups in a proper way to avoidslowing down unnecessarily the decisionmaking process.

Phase 4: digging for ‘goldenopportunities’Finally, when the riverwidening project will beimplemented, often only after many years ofdeliberation and discussion, welldocumentedand readily available information will enhancethe implementation process. Leading toefficient digging and building, wellbalancedmaterial flows and sustainable raw materialmanagement. Maybe, together with socioeconomical benefits, some financial benefitsfrom economically interesting buildingmaterials can be obtained. In this way riverwidening can be considered ‘digging for goldenopportunities’.

ReferenceRijkswaterstaat, 2004. Beton en metselzand uit de

Noordzee? Eindrapport van de PIA SubwerkgroepZeezand; resultaten van de haalbaarheidsstudienaar beton en metselzandwinning voor deHollandse en Zeeuwse kust. PublicatiereeksGrondstoffen 2004/1. ExpertisecentrumBouwstoffen, Rijkswaterstaat, Dienst Weg enWaterbouwkunde, Delft, 108 p.

Figure 2. Example of a simple lithostratigraphic profilebased on extra borehole data (North Sea).

Figure 3. Example of the calculation of the quantity ofextractable coarse sand beneath a layer of fine sand (afterRijkswaterstaat, 2004).

Hydrology


Floods in the Meuse basin: contribution of tributariesH.A. Peeters 1, M.J.M. de Wit 2 & R. Uijlenhoet 11 Chair of Hydrology and Quantitative Water Management, Department of Environmental Sciences, Wageningen University, Nieuwe Kanaal11, 6709 PA Wageningen, The Netherlands; [email protected] Institute for Inland Water Management and Waste Water Treatment (RIZA), P.O. Box 9072, 6800 ED Arnhem, The Netherlands

AbstractIf a flood wave of a river coincides with floodwaves of tributaries, extreme floods can occur.The floods of 1993, 1995, 2002 and 2003 inthe Meuse basin are analysed to see ifpatterns can be derived with regard to peakconvergence. The data show that a flood wavein the Meuse basin is generated at differentlocations and at different times; this is due todifferences in precipitation patterns andgeographical features of the area. The Meusedischarge at Borgharen often peaks before thedischarge at Ampsin and Chooz, which areboth located upstream of Borgharen. Due tooccurrence of different flood waves during aflood event, measures taken to reduce peaksat one place along the Meuse can havenegative effects on the peak discharge at otherplaces along the Meuse. Hence, floodmanagement should be done at basin level.

IntroductionIn the last decade the frequency andmagnitude of floods in the Meuse basin havebeen relatively large. The volume of thesefloods is mainly influenced by initial storagevolume and precipitation volume. The shape islargely determined by hydraulic properties ofthe river network and the convergence of floodwaves of the main river and its tributaries. Ifthe flood wave of the river coincides with floodwaves of tributaries extreme floods can occur.This motivates the need to tune floodmanagement of the river and the tributaries.Aim of the study is to analyse how flood wavesof the Meuse and tributaries coincide duringfloods. The main question is: Which interactionpatterns between the Meuse and thetributaries can be derived from the flood dataof 1993, 1995, 2002 and 2003 and how canthis be implemented within the objectives ofthe “Flood Action Plan” for the river Meuse?The information obtained allow for aquantitative analysis of hydrological processesunder extreme conditions for the entire Meusebasin (Fig. 1). A similar study was performedMeuse basin (i.e. Limburg only).

Material and methodsThe flood data used in this study consist ofprecipitation and hourly discharge data fromgauging stations located in 14 tributaries andat 10 locations on the Meuse, of which three inFrance, eleven in Belgium, two in Germanyand eight in the Netherlands. The data for thisstudy are provided by DIREN, METSETHY,Staatlicher Umweltamt, AWZ and RWS.

For the different stations, the time ofoccurrence of peak discharge was determined;the peak time. As some stations are notlocated at confluences of the Meuse andtributaries, the peak time was corrected withthe propagation time of a flood wave fromstation to confluence. In case data waslacking, this time was estimated. Secondly thetime difference between the Meuse flood wavepassage at the confluence and the dischargepeak time of the tributary was determined (Fig.2). Results provide information whether Meuseand tributary peaks coincide. Finally the actual(discharge at the confluence during passage ofthe floodwave in the Meuse) and potential

Figure 1. The Meuse basin with locations of monitoringstations and tributaries mentioned in Figure 2. Themonitoring stations used in this study are only aselection out of the total number of monitoring stationsthat are operated at the different hydrological andmeteorological institutes

Hydrology


(observed peak discharge) contribution of thetributaries were determined. This allows aquantitative analysis of the contribution of thetributaries to the peak discharges at Chooz,Ampsin and Borgharen.

ResultsGeographically, the Meuse basin can bedivided in 3 zones. In zone 1 and 3 the riverflows through a wide river valley; in zone 2 theriver valley is narrow and steep. As a result,the propagation time of the flood wave is muchlonger in zones 1 and 3 than in zone 2. Figure2 shows average peak times at the differentlocations with Borgharen as reference (t=0),plotted against the distance. The dotted linedisplays the assumed peak time of a location.This line is estimated, assuming propagationtimes in the Meuse (zone 2) that are based ongeneral observations (Berger, 1992). It wouldbe better to perform this analysis with hydraulicmodels, but that is beyond the scope of thisstudy.

Analysis of the peak times shows that thetributaries in zone 2 on average peak beforethe tributaries in zone 1 and zone 3, mostlybefore the Meuse peak passage. However inzone 2 it may occur that the flood wave on theMeuse is generated only by the discharge ofthe tributaries. This explains the generation ofdifferent flood waves. The peak discharges ofthe Lesse, Sambre and Ourthe often coincidewith the peak on the Meuse. The peakdischarge at Chooz is caused mainly by the

Semois. The Lesse and Sambre cause thepeak discharge at Ampsin. For the dischargeat Borgharen the contribution of the Ourthe,Sambre, Lesse and Amblève are significant.

The discharge of zone 1 becomes moresignificant for the discharge at Borgharen if theflood period extents due to a sequence ofprecipitation events. Also some patterns withregard to peak order can be derived. For theevents considered it appears that thedischarge of the Ourthe system peaks insequence: Vesdre, Amblève and Ourthe.

Discussion of uncertaintiesThe propagation time differs for every flood.However by assuming a constant propagationtime for some tributaries, this is not taken intoaccount and peak times at confluences candiffer a few hours. For example, in reality thepeak lines of Ampsin and Borgharen could beconnected, but due to limiting data no univocalpicture can be derived. In this study the time ofmaximum discharge at a station is taken as thepeak time. In reality the considered dischargediffers only little from the discharge in thehours before or after the peak. Thus, if the lagbetween Meuse peak and tributary peak issmall (e.g. 3 hours), than the peaks virtuallycoincide.

The discharge data used are measured atthe tributary stations; in reality the dischargeconfluencing with the Meuse is larger. The

Figure 2. Average propagation of discharge peaks in the Meuse basin

Hydrology


information needed to correct the discharge isnot available. Furthermore, due to differencesin discharge (hourly) and precipitation (daily)records, it is not possible to determine anexact time of concentration of subcatchments.In a next step of this study precipitation datawill be included to support the dischargeanalysis.

ConclusionThe analysis presented here shows thegeneral pattern of the confluence of floodwaves in the Meuse basin. Due to theoccurrence of different flood waves during aflood event, measures taken to reduce peaks

at one place along the Meuse can havenegative effects on the peak discharge atanother place along the Meuse. This study canbe used as a first inventory needed to tunemeasures that aim at a reduction of flood riskat the scale of the entire Meuse basin.

ReferencesDe Wit, M., R. van der Veen & L. van Hal, 2004.

Samenvallen hoogwaterpieken Maas en zijrivieren,RIZAwerkdocument 2004.126x, RIZA, Arnhem.

Berger, H.E.J., 1992. Flow forecasting for the RiverMeuse, Ph.D. thesis, Delft University of Technology,Delft.

Hydrology


Hydrology


Application of a weather generator to simulate extremeriver discharges in the Rhine and Meuse basinsR. Leander 1, H. Buiteveld 2, M.J.M. de Wit 2 & T.A. Buishand 11 Royal Netherlands Meteorological Institute (KNMI), P.O. Box 201, 3730 AE De Bilt, The Netherlands; [email protected] Institute for Inland Water Management and Waste Water Treatment (RIZA), P.O. Box 9072, 6800 ED Arnhem, TheNetherlands

AbstractA weather generator has been developed tosimulate longduration sequences (1000 yearsor more) of daily discharges using ahydrological/hydraulic model. Resultingextreme value distributions of 10day rainfallamounts and daily river discharges arediscussed.

IntroductionThe traditional method for the estimation of thedesign discharge is based on the extrapolationof the distribution of recorded annual dischargemaxima to a mean return period of 1250 years.Disadvantages of this method are that strongextrapolation is required, and that dischargerecords are potentially inhomogeneous.Furthermore, considering annual dischargemaxima gives no insight in the shape andduration of the flood peaks. Since the mid1990s a new methodology is underdevelopment, which aims at the simulation oflongduration discharge sequences. Besides ahydrological model (HBV) and a hydraulicmodel (SOBEK), also a stochastic weathergenerator is involved to synthetically generaterealistic long daily sequences of rainfall andtemperature for the river basin.

The weather generatorRainfall and temperature at different locationsin the drainage area are simultaneouslysimulated by ‘nearestneighbour’ resampling. Amajor advantage of this resampling method isthat both the spatial association of daily rainfallover the drainage basin and the dependencebetween daily rainfall and temperature arepreserved, without making assumptions aboutthe underlying joint distributions. Toincorporate autocorrelation, one first searchesthe days in the historical record (here 10),whose characteristics are most similar to thoseof the previously simulated day, referred to as‘nearest neighbours’. One of these nearestneighbours is then randomly selected using adecreasing kernel and the observed values forthe day subsequent to that nearest neighbourare adopted as the simulated values for thenext day. The search for nearest neighbours of

the previously simulated day is based onquantities like the basinaverage rainfall andtemperature for that day, the rainfall total of thepreceding four days and the fraction of thebasin for which the daily rainfall exceeds thewetday threshold of 0.3 mm. The effect ofseasonal variation is reduced by restricting thesearch for nearest neighbours to days within amoving window of 61 days, centred on thecalendar day of interest. The variables beingsimulated do not necessarily play a role in theselection of nearest neighbours (e.g. arealrainfall for a subcatchment) but are supposedto be closely related to those variablesinfluencing the selection. More details aboutnearestneighbour resampling can be found inBuishand & Brandsma (2001).

ResultsFigure 1 compares the distribution of the 10day winter maxima of basinaverage rainfall forfour 3000year simulations for the Meuse basinwith the corresponding distribution of thehistorical winter maxima (October throughMarch) for the period 19611998. Theseextremes are considered because large riverdischarges are often caused by large multidayrainfall amounts in winter. The figure showsthat the weather generator is capable ofreproducing the distribution of the 10dayrainfall extremes well. Furthermore, thehistorical maximum 10day rainfall amounts islargely exceeded in the simulations. It is

Figure 1. Gumbel plot of winter maxima (historical andsimulated) 10day basinaverage rainfall for the Meuse.

Hydrology


expected that this leads to higher flows thanthose observed. Figure 2 displays thedistribution of annual maxima of dailydischarge for the Rhine at Lobith from three50year simulations, to which Gumbeldistributions are fitted. Large differencesbetween the simulations are found if thesedistributions are extrapolated to return periodsin the order of 1000 years. Figure 3 shows thesame for three simulations of 1000 years,except that threeparameter GEV distributionsare fitted to the data. These show aconsiderably smaller spread for long returnperiods.

ConclusionsThe weather generator reproduces observedproperties of extreme rainfall well and is alsocapable of simulating more extreme eventsthan have been observed in the past. Largeextrapolation of distributions fitted to observeddischarge maxima are very uncertain. The useof a weather generator with ahydraulic/hydrological model can help reducethe uncertainty in the estimated extreme flowsfor long return periods.

ReferenceBuishand, T.A. & T. Brandsma, 2001. Multisite simulation

of daily precipitation and temperature in the Rhinebasin by nearestneighbor resampling. WaterResources Research 37, pp. 27612776.

Figure 2. Gumbel plot of annual maxima of dailydischarge of the Rhine at Lobith from three 50yearsimulations.

Figure 3. Gumbel plot of annual maxima of dailydischarge of the Rhine at Lobith from three 1000yearsimulations.

Hydrology


A 3,000 year discharge simulation in the Meuse basin witha stochastic weather generator and the HBV modelP. Aalders 1,4,5, M.J.M. de Wit 2, L. Bolwidt 2, P. Warmerdam 1, P.J.J.F. Torfs 1, R. Leander 3 & T.A.Buishand 31 Chair of Hydrology and Quantitative Water Management, Department of Environmental Sciences, Wageningen University,

Nieuwe Kanaal 11, 6709 PA Wageningen, The Netherlands2 Institute for inland Water Management and Waste Water Treatment (RIZA), P.O. Box 9072, 6800 ED Arnhem, The

Netherlands3 Royal Netherlands Meteorological Institute (KNMI), P.O. Box 201, 3730 AE De Bilt, The Netherlands4 present address: De Dommel Water Board, P.O. Box 10.001, 5280 DA Boxtel, The Netherlands5 [email protected]

AbstractA stochastic weather generator whichgenerates longterm rainfall and temperaturerecords has been linked to the hydrologicalmodel HBV to provide more insight in theestimation of design discharges. This studysummarizes the results of the application ofthe rainfall generator in the Meuse basin.Therefore, a discharge simulation of 3,000years has been performed.

IntroductionFlood protection along the main Dutch rivers isbased on design water levels with a givenprobability of exceeding. The estimation of thedesign discharges is currently based on theextrapolation of the measured discharges atBorgharen (Meuse) and Lobith (Rhine).However, the determination of designdischarges from statistical analyses of themeasured peak discharges faces variousproblems. First, it is unknown howrepresentative the relatively short measureddischarge records are. Secondly, thedischarge record is potentially non

homogeneous because of changes in theupstream basin, the river geometry andclimate. Third, the choice of frequencydistributions is also a point of uncertainty.Therefore, a new methodology has beenproposed to provide a better physical basis forthe estimation of the design discharge of theDutch rivers. This new methodology is knownas rainfall generator.

Rainfall generator for the MeusebasinThe application of the rainfall generator for theMeuse basin consists of a KNMI stochasticweather generator linked to the hydrologicalHBVMeuse model (Fig. 1). KNMI generated a3,000 year record of precipitation andtemperature data for the entire Meuse basin(Leander & Buishand, 2004). This record hasbeen routed through the rainfallrunoff moduleof the HBVMeuse model resulting in adischarge record containing 3,000 years ofdaily discharge data.

Figure 1. Schematized image of rainfall generator methodology

Hydrology


ResultsFigure 2 contains 3,000 annual dischargemaxima (daily values) derived from the rainfallrunoff simulation using input of the weathergenerator. The dots show a random temporaldistribution and no trend was found in thesimulated data record. The generated datasetshows an underestimation in the highest rangeof the annual discharge maxima if fitted to aGumbel distribution (Fig. 3). However, thefigure also shows that the rainfall generatorresulted in peak discharges larger thanhistorically observed.

ConclusionsThe simulation of 3,000 years shows that therainfall generator is capable of generatingextreme discharge events that are larger thanobserved. A Gumbel plot of annual dailydischarge maxima derived from measured andgenerated records reveals that extremes arehard to fit. This counts both for the observedextremes (1993 and 1995) and the generatedextremes.

ReferenceLeander, R. & T.A. Buishand, 2004. Rainfall generator for

the Meuse basin; development of a multisiteextension for the entire drainage area. KNMIpublication 196III, De Bilt.

Figure 3. Gumbel plot of annual daily discharge maxima (m3/s) calculated byHBV for measured (dots) and generated (stars) meteorological input data

Figure 2. Simulated annual discharge maxima (daily values).

Hydrology


A quick scan forecasting tool for prescreeningprobabilistic weather forecasts on their seriousnessH.C. Winsemius1, H.H.G. Savenije1, W.M.J. Luxemburg1, H. Havinga1,4, F. Diermanse2, S. Tijm3,E. Sprokkereef 4 & H. van de Langemheen41 Faculty of Civil Engineering and Geosciences, Delft University of Technology, P.O. Box 5048, 2600 GA Delft, TheNetherlands; [email protected] WL | Delft Hydraulics, P.O. Box 177, 2600 MH Delft, The Netherlands3 Royal Netherlands Meteorological Institute (KNMI), P.O. Box 201, 3730 AE De Bilt, The Netherlands; [email protected] Institute for Inland Water Management and Waste Water Treatment (RIZA), P.O. Box 9072, 6800 ED Arnhem, TheNetherlands

AbstractThe flood forecasting model FEWSRhine iscapable of computing discharges with leadtimes of 10 days, including computation ofweather uncertainties induced by EPS weatherforecasts. However, the computation timeneeded is too large. A quick scan tool wasdeveloped in this study to give a first indicationof discharges induced by EPS weatherforecasts. Three simplifications with regard toFEWSRhine were applied: (1) rainfall runoffmodelling was lumped per subbasin, using theHYMOD rainfall runoff model structure; (2)river routing was based on a multiple linearregression equation containing discharges insubbasins upstream from Lobith in thepreceding days; (3) only three subbasins wereincluded in the regression equation (Lippe,Mosel and Neckar) since discharges betweenneighbouring subbasins are strongly autocorrelated. Calibration and validation scoresprove that quite accurate daily averagedresults can be obtained using these modellingsimplifications. The computation of dischargesfrom EPS weather forecasts shows aconsiderable amount of bias, partly due toanomalies in the model but mostly due toinaccuracies in the predictions of theprecipitation intensity by EPS.

Introduction and problemdescriptionSince the floods in 1993 and 1995 in the Rhineand Meuse basins, efforts have been put intothe improvement of flood forecasting systemsand extension of lead time. For the Rhinebasin, the last forecasting system produced isFEWSRhine, developed by WL | DelftHydraulics and RIZA. It enables theforecasting of floods at Lobith 10 days aheadusing weather forecasts and waterlevelrecordings as input. The large lead times result in a largerinfluence of weather uncertainties ondischarge forecasts. Therefore, thecomputation of discharges from EPS

(Ensemble Prediction System) weatherforecasts was enabled in FEWSRhine. EPScomputes different possible weather statesfrom perturbed initial conditions. The EPS ofthe European Centre for Medium rangeWeather Forecasts produces 50 possibleweather states, called ‘ensemble members’. Adisadvantage of the ensemble computations isthat it takes too much computation time to runall EPS members for purposes of realtimeflood forecasting. Therefore, the objective ofthis study is to define a quick scan tool toobtain a first estimate of daily averageddischarges resulting from full EPS forecasts.

Description of simplified modelThree simplifications were applied with regardto the FEWSRhine model.• The distribution of the modelling of sub

basins of the Rhine over several rainfallrunoff models, was aggregated into onelumped model per subbasin. Instead ofthe HBV structure, the HYMOD structure(Vrugt et al., 2002) was used formodelling. It contains routines for snow,subsurface flow and groundwater flow(Fig. 1).

• The SOBEK model between Maxau andLobith was replaced by a multiple linearregression equation containing dischargesfrom subbasins upstream from Lobith inthe preceding days and the discharge atLobith at time = t – 1. The lag times thatare taken into account in this model arebased on travel times between Lobith andthe confluence of the Rhine and its majortributaries.

• The discharges between neighbouringsubbasins are assumed to be autocorrelated due to similarities betweenhydrology and meteorology, which isespecially the case in winter periods. Thismakes it possible to exclude autocorrelated subbasins from computation inthe regression equation. Only discharges

Hydrology


from the Lippe, Mosel and Neckar areincluded.

CalibrationCalibration of the rainfall runoff models wascarried out by trial and error using theNash/Sutcliff criterion as objective function. Acalibration data set of October 1997 untilDecember 1998 was used. The models werevalidated on a data set of September 1994until February 1995.

The regression equation was derived bysetting up a matrix containing discharges fromthe three sampled subbasins at time = t –travel time to Lobith and a vector, containingdischarges at Lobith at time = t. The regressioncoefficients that described the best fit werederived using the least squares method.Results from the regression equation arepresented in Figure 2.

EPS resultsComputation of the EPS forecast of January21, 1995, which should describe the flood of1995, shows a considerable amount of bias,which is partly caused by the regressionequation and bias in the rainfall runoff modelsbut for the largest part by underestimations ofthe precipitation intensity by the EPS system(Fig. 3). Due to inaccuracies in the EPSsystem, the spread in EPS in the first two daysof forecasting is unreliable. Therefore it couldbe considered to use only a highresolutiondeterministic weather forecast for the first twodays.

Conclusions• A model concept based on linear

regression technique containing onlysamples of subbasin discharges asvariables can produce quite accurate dailyaveraged discharges at Lobith.

• This model concept is able to produce aquite accurate indication of probability ofoccurrence of floods at Lobith, accordingto EPS forecasts.

• Compared to the physical model FEWSRhine, this model is able to compute anEPS forecast with a lead time of 10 dayswithin a much shorter computation time.

• EPS computations show some bias inpeaks, which is partly caused byanomalies in the hydrological model, andpartly by average underestimation inprecipitation intensity by EPS.

ReferenceVrugt, J.A., W. Bouten, H.V. Gupta & S. Sorooshian, 2002.

Toward improved identifiability of hydrologic modelparameters: the information content of experimentaldata. Water Resources Research 38 (12), 1312,doi:10.1029/2001WR001118.

Figure 1. HYMOD rainfall runoff model including a snowroutine. The states and parameters in the HYMODstructure: T = Temperature (deg. C); SP = Snowpack (mm);SM = Soil moisture (mm); P = Precipitation (mm/day); AET= Actual Evaporation (mm/day); EP1 = Excess rainfall 1(mm/day); EP2 = excess rainfall 2 (mm/day); S(t) =Average soil moisture (mm); F(c) = Distribution of fieldcapacity over the watershed (); CMAX = maximum fieldcapacity (mm); CMIN = soil moisture at location where F =CMAX (mm); alpha = separation coefficient of excessrainfall over quick flow and slow flow zones (); Kq = Quickflow recession constant (day1); Ks = Slow flow recessionconstant (day1); Qs = Slow flow (m3/s); Qq = Quick flow(m3/s); Q = Total flow (m3/s).

Figure 2. Calibration time series of the model for Lobith,September 1994 February 1995.

Figure 3. EPSdriven hindcast of the 1995 flood event atLobith using the EPS quick scan tool

Hydrology


Hydrological application of areal rainfall estimates from theWideumont weather radar over the Ourthe catchment:preliminary resultsM. ten Heggeler 1, A. Berne 1, R. Uijlenhoet 1, L. Delobbe 2, Ph. Dierickx 3 & M. de Wit 41 Chair of Hydrology and Quantitative Water Management, Department of Environmental Sciences, Wageningen University,Nieuwe Kanaal 11, 6709 PA Wageningen, The Netherlands; [email protected] Royal Meteorological Institute of Belgium, Ringlaan 3 / Avenue Circulaire, B1180 Brussels, Belgium3 Ministère de l'Équipement et des Transports, Service d'Études Hydrologiques, Boulevard du Nord 8, B5000 Namur, Belgium4 Institute for Inland Water Management and Waste Water Treatment (RIZA), PO Box 9072, 6800 ED Arnhem, The Netherlands

AbstractThis paper presents a first assessment of thehydrometeorological potential of a Cbanddoppler weather radar recently installed by theRoyal Meteorological Institute of Belgium nearthe town of Wideumont in the southernArdennes region.

IntroductionWageningen University (WU), the RoyalMeteorological Institute of Belgium (RMI) andthe Hydrological Service of the WalloonRegion of Belgium (METSETHY) haverecently established a research collaborationto investigate whether an improvedassessment of the spacetime structure ofprecipitation, as can be obtained with thenewly installed weather radar in Wideumont,will lead to an improved understanding of thehydrometeorology of Ardennes catchments, inparticular the Ourthe (Berne et al., 2005).

Methods and materialsThe Royal Meteorological Institute of Belgium(RMI) recently installed a new Gematronik Cband doppler weather radar near the town ofWideumont (535 m.a.s.l.), in the southernArdennes area (Province of Luxemburg), closeto the border with the Grand Duchy ofLuxembourg. The radar, which is installed on a

50 m high tower, performs every 15 min. a 10elevation volume scan of the 3D structure ofthe rainfall field out to a distance of 240 kmand similarly a doppler scan to a range of 120km. The range resolution of the radar data istypically 0.5 km. An operational precipitationproduct is generated every 5 min. To perform a first assessment of thehydrological potential of the Wideumontweather radar, the Ourthe catchment upstreamof Tabreux was selected as study area. Figure1 shows the 1597 km2 catchment with thelocation and elevation of the 10 METSETHYrain gauges covering the area. Also included isa cartesian 1x1 km2 grid resampled from thepolar volume scan reflectivity data of theWideumont weather radar, which is locatednear the southern tip of the catchment. Tworainfall events were considered, namely therainfallrunoff events of 410 May 2002 and 17March 2003. METSETHY kindly providedhourly raingauge and discharge data for theOurthe at Tabreux for these events. RIZA kindly provided the opportunity to runa fully calibrated HBVmodel (Lindström et al.,1997) of the Ourthe catchment. A 30yearcalibrated HBV model parameter data set fromRIZA (19681998) was employed to choose“optimal” initial conditions for both selectedevents by searching in the database foranalogous hydrographs during the monthpreceeding the discharge events, employingthe NashSutcliffe parameter as an errorcriterion (Ten Heggeler, 2004).

Results and discussionThe mean areal rainfall estimated from thesecond elevation of the Wideumont weatherradar volume scan reflectivity data using thestandard MarshallPalmer reflectivityrain raterelation (without adjustment to rain gauges)showed a 42% underestimation with respect tothe gauge average rainfall for 45 May 2002and a 12% underestimation for 1 March 2003(Table 1). The second rather than the firstelevation was employed to minimize thepossible adverse effects of ground clutter.

Figure 1. Ourthe catchment upstream of Tabreux(1597 km2) with the location and elevation (metersabove sea level) of the 10 METSETHY rain gaugesand a cartesian 1km x 1km grid resampled from thepolar data of the Wideumont weather radar (locatednear the southern border of the catchment)

Hydrology


The observed underestimation may beattributed to an erroneous reflectivityrain raterelation and/or to raininduced attenuation.Obviously, application of the radarestimatedmean areal rainfall to the gaugecalibratedHBVmodel for the Ourthe upstream ofTabreux produced an underestimation of thepredicted with respect to the measureddischarge for the event of 410 May 2002. Asimilar analysis for 17 March 2003 wasimpossible because the available radar datacovered only one day. Recall that HBV is a lumped rainfallrunoffmodel, in principle not ideal to assess theimpact of spatial rainfall variability ofprecipitation. Nevertheless, in this limitedsetting there is still an interesting application ofthe power of weather radar, namely its spatialcoverage. Approximately 1600 1x1 km2 radarpixels cover the Ourthe basin upstream ofTabreux. In a Monte Carlo simulationframework one can assume the radar data torepresent the actual rainfall field (1600 “raingauges”) from which a “true” areal averagerainfall can be calculated. Also, one canrandomly pick (without replication) 10 “gauges”from the 1600 pixels and compute thearithmetic mean of those 10 numbers. Thisrandom drawing can be repeated say 1000times to assess (in a very simple fashion sincegauge locations are chosen independentlywithout imposing for instance a minimum intergauge distance) the uncertainty in estimatingthe areal average rainfall over a 1600 km2

catchment from only 10 hourly gaugeobservations. The resulting sampling distribution of themean areal rainfall appears to be nearlysymmetrical. In addition, the ratios of thequantiles with respect to the “true” mean arealrainfall appear to remain remarkably constantover time during the events considered. It turnsout that the hydrological uncertainty associated

with this rainfall sampling uncertainty (±25% onan hourly basis, see Fig. 2) is of the sameorder of magnitude as the uncertaintyassociated with the initial conditions asestimated from the 30year database (Fig. 3).

Conclusions and recommendationsThe mean areal rainfall over the ~1600 km2

Ourthe catchment upstream of Tabreuxestimated from the Wideumont weather radarusing the standard MarshallPalmer reflectivityrain rate relation (without adjustment to raingauges) shows biases between +128% and 42% with respect to the corresponding gaugeestimates for six selected precipitation events.For two rainfall events the radarestimatedmean areal rainfall is applied to the gaugecalibrated (lumped) HBVmodel for the Ourtheupstream of Tabreux, resulting in a significantunderestimation with respect to the observeddischarge for one event and a closer match foranother. The uncertainty in the hourly dischargefrom the ~1600 km2 Ourthe catchmentupstream of Tabreux associated with thesampling uncertainty of the mean areal rainfallestimated from 10 rain gauges evenly spreadover the catchment amounts to ±25% for thetwo events analyzed. This uncertainty is of thesame order of magnitude as that associatedwith the initial conditions. The development of accurate and robustprocedures for correcting for raininducedattenuation and the vertical profile of reflectivityis the topic of ongoing investigations. Themajor floods which occurred during the 20022003 winter season will be studied as part ofthe ongoing collaborative research project. Wealso foresee a comparison of HBV with runoffestimates from the Hydromax river flowforecasting model, which is currently usedoperationally at METSETHY.

Figure 2. Uncertainty in the discharge at Tabreux forthe period 17 March 2003 calculated using the(lumped) HBVmodel due to uncertainty in the meanareal rainfall. The black line indicates the dischargecalculated using the mean areal rainfall from the 10rain gauges, the gray lines correspond to thedischarges calculated using the 20% and 80%uncertainty limits on the mean areal gaugederivedrainfall over the 1597 km2 catchment area (Fig. 1).

Figure 3. Sensitivity of the simulated discharge atTabreux for the event of 17 March 2003 due to anuncertainty of ±5 mm in the initial content of the fastrunoff reservoir of the HBV model. The black lineindicates the discharge calculated using the meanareal rainfall from the 10 rain gauges (identical to theblack line in Fig. 2), the gray lines correspond to thedischarges calculated using a 5 mm increase and a 5mm decrease in the initial content of the fast runoffreservoir (UZ).

Hydrology


AcknowledgementsThe second author acknowledges financialsupport from the European Commissionthrough a Marie Curie Postdoctoral Fellowship(Contract EVK1CT200250016). The thirdauthor acknowledges financial support fromthe Netherlands Organization for ScientificResearch (NWO) through aVernieuwingsimpuls/VIDI grant (Project016.021.003). The collaboration between WU,RMI and METSETHY is supported by theEuropean Commission as part of IntegratedProject FLOODsite (Contract GOCECT2004505420, see also http://www.floodsite.net/).

ReferencesLindström, G., B. Johansson, M. Persson, M. Gardelin &S. Bergström, 1997. Development and test of thedistributed HBV96 hydrological model. J. Hydrol. 201, pp.

272288.Ten Heggeler, M., 2004. Hydrological modelling and areal

average rainfall estimates of the Ourthe catchmentderived from the wideumont weather radar andprecipitation gauges. M.Sc. thesis, WageningenUniversity.

Table 1. Characteristic precipitation events selected by the RMI, the corresponding total mean arealrainfall accumulations over the Ourthe catchment upstream of Tabreux as derived from radar andraingauges, and the resulting bias of the uncorrected radar rainfall estimates with respect to the gaugeestimates

Date Type of event Radar (mm) Gauges (mm) Bias (%)4 May 2002 convective 19.1 32.7 420 July 2002 convective 10.2 9.8 +419 August 2002 convective 7.3 3.2 +12824 November 2002 stratiform 9.1 7.6 +2030 January 2003 stratiform 4.2 2.4 +751 March 2003 stratiform 9.1 10.3 12

Hydrology


Learning from the data: a stepped calibration approachF. Fenicia 1,2, P. Matgen 1, L. Pfister 1 & H.H.G. Savenije 21 Public Research Center – Gabriel Lippmann, 162a Avenue de la Faïencerie, L1511 Luxembourg, G.D. Luxembourg2 Section of Water Resources, Faculty of Civil Engineering and Geosciences, Delft University of Technology, P.O. Box 5048,2600 GA Delft, The Netherlands; [email protected]

AbstractFollowing a ‘topdown’ modelling approach amodel structure has been developed based onan analysis of the observed discharge at astreamgauge station in Luxembourg.Individual model components have beenidentified starting with a simple structure, andincreasing model complexity when the modelshowed some difficulties in matchingobservations.

IntroductionThe ‘topdown’ or ‘downward’ approach(Klemes, 1983; Sivapalan, 2003) in a causeeffect relationship can be considered as theprocess of analyzing the effects and trace backto the causes that might have generated them.The ‘bottomup’ or ‘upward’ approach wouldperform the opposite operation, starting fromthe causes and combining them trying toachieve the desired effect. The ‘topdown’approach can be considered as a strategy oflearning from data. In the present case, we adopted the ‘topdown’ approach in the development of aconceptual model structure, starting from asimple concept and adding or modifyingprogressively in an iterative way processesand components. Modifications have beenconsidered based on physical reasoning, andhave been applied when model performancesin reproducing observations could beimproved.

A model structure developed in this wayinvolves components and parameters that canbe directly associated to specific hydrographcharacteristics. In the calibration phase thoseparameters should be adjusted to match thehydrograph characteristics that theypresumably influence. To perform this operation we think that,instead of an ‘allatonce’ calibration approach,where all parameters are calibrated together ina single optimization run, a stepped approachwould be more appropriate. The stepped approach proposed hereconsists of: (1) associating some modelparameters with specific physical processesthat can be identified analysing the recordedtime series; (2) defining some objectivefunctions that represent performancemeasures for specific hydrographcharacteristics; (3) calibrating separately in aniterative manner each group of parametersassociated with each objective function. In this way we aim to overcome someproblems that may arise when ‘allatonce’calibration approaches are used. When thoseapproaches are applied, in fact, someparameters might be calibrated versusobjectives on which they have little influence.This would cause some effects that we try toavoid, like the tendency of fitting certainaspects of the simulations at the expense ofothers, the generation large uncertainties forthe representation of certain processes, andcompensation of internal structure errors byparameter adjustment.

MethodsThe conceptual model used consists of aninterception component, a soil moisturereservoir, a fast reacting reservoir, and a slowreacting reservoir (Fig. 1). Two lagfunctionsare used to offset the fluxes that enter theslow and the fast reservoir. A runoff coefficientdependent on the soil moisture reservoirseparates infiltration from surface runoff Waterreaches the fast reservoir by surface runoff,and the slow reservoir by percolation andpreferential recharge.

Figure 1. Model structure

Hydrology


The stepped approach is applied hereto calibrate three key aspects of a hydrographsimulation: the low flow recession, the highflow events, and the lag time of the system.Groups of parameters and functionaldependencies associated with each of thoseaspects are assessed independently in thefollowing steps: (1) estimation of a StorageDischarge relation for the slow reservoir bycalculating a Master Recession Curve (MRC);(2) calibration of high flows by maximising theNash Sutcliffe (NS) coefficient, which issensitive to high flows; (3) calibration of timelags by maximising the correlation coefficientof simulated and observed discharge. After thisa new estimate of the storagedischargerelation is made, by correcting the modelledpercolation flux, and the procedure is repeateduntil convergence is reached. The modelparameter defining the interception process iscalibrated iteratively within a fixed feasiblerange within the second calibration step. The MRC (Lamb & Beven, 1997)calculation consists of sorting hydrographrecessions into ascending order, based on thelowest tailend discharge (Fig. 2). Allrecessions are then concatenated to form asingle recession curve. This procedureexcludes storm flow effects from the MRC. TheMRC therefore represents the longtermrecession of a catchment. The first objective of the stepped approachis to have a model that has a response that isclose to the MRC for low flows. For this reasonwe use the calculated MRC to determine a firstestimate of the storagedischarge relation thatwe apply to the slow reservoir of the model.After that we calibrate other model parametersto maximise the NashSutcliffe coefficient,which provides a good performance measurefor high flow simulations, due to the square ofthe difference between observed andsimulated discharge.

Subsequently, the lag time is calibrated tomaximise the correlation coefficient. At thispoint a new estimate of the storagedischargerelation is calculated by taking into account themodelled percolation that enters the slowreservoir.

ApplicationThe procedure has been applied to theHesperange catchment in Luxembourg (292km2), which is located upstream ofLuxembourg City. Land use of the study areais dominated by pastures and forests, while thegeology is characterised by Marls, Limestoneand Sandstone. For calibration of parametersin the second and third calibration steps theAdaptive Cluster Covering algorithm(Solomatine, 1999) has been used. Figure 3 shows the calculation of thestoragedischarge relation for the slowreservoir based on the MRC. After threeiterations the curves are already quite close,we therefore stop iterating. It is interesting tonotice that while the first estimate of thestoragedischarge relation resembles anexponential function, the second and thirdestimates are close to a linear function,suggesting that a linear reservoir applies.Hence, the nonlinear reservoir of Lamb &Beven (1997), in our case, appears to be theresult of percolation. In Fig. 4 a simulation run is shown.Observed and simulated hydrograph andoutflow from the slow reservoir arerepresented. Also the calculated MRC isrepresented. In recession periods the modelfollows the MRC quite closely (red line).

Figure 2. Master Recession Curve calculation.

Figure 3. Storagedischarge relation for the slowreservoir calculated for three iterations.

Hydrology


ConclusionsIn this particular case the stepped approachwas applied with good results, and, beingintegrated with a ‘topdown’ modellingapproach, it has been useful to identifycomponents and parameters that neededimprovement. Therefore it helped thedevelopment of the model structure. Webelieve that the application of such anapproach gives best results when modelparameters can be associated with specificprocesses that can be separately calibrated.

ReferencesKlemes, V., 1983. Conceptualization and scale in

hydrology. Journal of Hydrology 65, pp. 1–23.Lamb, R. & K.J. Beven, 1997. Using interactive recession

curve analysis to specify a general catchment storagemodel. Hydrology and Earth System Science 1, pp.101113.

Sivapalan, M., G. Blöschl, L. Zhang & R.Vertessy, 2003. Downward approach to hydrological

prediction. Hydrological Processes 17, pp. 21012111.

Solomatine, D.P., 1999. Two strategies of adaptive clustercovering with descent and their comparison to otheralgorithms. Journal of Global Optimization 14(1), pp.55–78.

Figure 4. Hydrograph simulation at Hesperange

Hydrology


Towards a gridbased regionalisation of storm flowcoefficientsH. Hellebrand, J. Juilleret, R. van den Bos & L. PfisterPublic Research Center – Gabriel Lippmann, 162a Avenue de la Faïencerie, L1511 Luxembourg, G.D. Luxembourg;[email protected]

IntroductionFlood awareness in the Grand Duchy ofLuxembourg became apparent after the majorfloods of 1993 and 1995. In 1996 a start wasmade by the public authorities to develop adense measuring network throughout theGrand Duchy of Luxembourg to gain insight ingroundwater and surface water behaviour. Inour study we improved a method, originallydeveloped by Pfister et al. (2002), to identifybasins that are capable of high runoffproduction. For this purpose basin specificstorm flow coefficients were calculated with aview to regionalization.

Study areaThe study area lies within the Grand Duchy ofLuxembourg and is part of the Mosel basin(Fig. 1). It contains 29 gauged basins varyingin size from about 4 km2 up to 1000 km2.Rainfall data is collected at 14 meteorologicalstations throughout the study area. Bothrainfall and discharge are recorded on a 15

minute time step. The amount of yearly rainfallis about 800 mm and the rainfall patterns arecharacterised by a strong negative West Eastgradient. Furthermore, rainfall totals are higherin the northern part than in the southern partsof the country. The basins are located ondifferent lithological substrata. The northernpart of the country, which belongs to theArdenne massif, is called Oesling and here thedominant rocks are schists. Marls, sandstoneand limestone represent the dominant lithologyin the middle and south of the country, calledthe Gutland.

MethodologyIn our study we improved a method todetermine runoffproducing areas using stormflow coefficients. Pfister et al. (2002),developed this method for mesoscale basinswith short hydrological data sets and it is ableto classify basins on the basis of theirresponse to rainfall. The method consists of

Figure 1. The Grand Duchy of Luxembourg, the location of the Mamer River basin and the lithology of the Mamer River basin.

Hydrology


five steps to determine the behaviour of mesoscale basins with a view to regionalise theirspecific storm flow coefficients: it comprisesareal rainfall input (step 1) and storm flowseparation (step 2) by a digital recursive filtertechnique (Nathan & McMahon, 1990). Thestorm flow coefficient C for the winter period iscalculated by plotting storm flow and rainfall ina double mass plot for each basin (step 3).Then, a regression analysis is performed todetermine the influence of lithology, land useand morphological properties of the basin onits specific storm flow coefficient (step 4).These parameter categories are commonlyused in literature (Mazvimavi, 2003). With theequation derived from the regressions analysisit becomes possible to regionalise the stormflow coefficient C (step 5). A final step, yet to be taken, is to apply theregression formula to a raster grid for the studyarea. It then should become possible todetermine high runoff producing areas,independently from the physiographiccharacteristics of basins.

ApplicationBy plotting total storm flow and total rainfall forrelevant events in a double mass curve astrong seasonal variation occurs. Duringsummer the curve remains almost horizontal,implying low to very low storm flowcoefficients. During the winter period a fairlyconstant slope is reached (Fig. 2). The stormflow coefficient C for a basin during this periodis obtained by analysing the trend in thedouble mass curve for several years for winterperiods only (Fig. 3). The steeper the slope ofthe double mass curve, the higher the stormflow coefficient. Furthermore, this storm flowcoefficient is basin specific. The Cvalue for two discharge stations inthe Mamer River, one at Mamer and one atSchoenfels near the outlet of the basin, are

respectively 0.72 and 0.55 (Figs 4 and 5).Upstream of Mamer, lithology consists ofMarls; downstream the lithology is acombination of Marls and Sandstone (Fig. 1).Marls can be considered as an impermeablesubstratum, whereas sandstone as asubstratum is fractured and the sandyweathered zone allows deep percolation.

This difference in lithology might be anexplanation for the differences in Cvalues. Toanalyse the influence of other factors on the Cvalue a regression analysis was performed. For the regression analysis 3 threeparameter categories were selected: (1)lithology, (2) land use and (3) geomorphology.Each category has several parameters that willbe used to explain the Cvalue. The firstpreliminary results from the regressionanalysis show that basin size is not stronglylinked with the Cvalue. Relationships betweenall other parameters show a nonlinearrelationship with the Cvalue. The parameterfor impermeability seems to be inverselycorrelated with the Cvalue.

ConclusionsThe regression analysis as applied so farshows promising results. Also an analysis onthe date of the inflection points in the doublemass curve (i.e. the change from summer towinter conditions in the basins) indicates thatbasins behave differently in time. This, incombination with the Cvalue, can give extrainformation about the importance of the runoffproducing areas. Once the regression formulais found it will be applied to a raster grid for thestudy area. We hope it then becomes possibleto determine high runoff producing areas,independently from basins. The method’stransferability will be tested on a larger scaleby applying it in the region of the RhinelandPalatinate (Germany).

Figure 2. Cumulative storm flowrainfall plot for theEisch at Hagen; 6 years of measurements. The nearhorizontal ‘steps’ represent summer periods, whereasthe steep sections represent winter periods

Figure 3. Cumulative storm flowrainfall plot for the Eischat Hagen (6 years of measurements; winter periods only;mean storm flow coefficient of all winter events =0.6631).

Hydrology


ReferencesMasvimavi, D., 2003. Estimation of flow characteristics of

ungauged catchments. Ph.D. thesis, WageningenUniversity, Wageningen.

Nathan R. J. & T.A. McMahon, 1990. Evaluation ofautomated techniques for base flow and recessionanalysis. Water Resources Research 26 (7), pp.14651473.

Pfister, L., JF. Iffly & L. Hoffmann, 2002. Use ofregionalized stormflow coefficients with a view tohydroclimatological hazard mapping. HydrologicalSciences 47, pp. 479491.

Figure 4. Cumulative storm flowrainfall plot for theMamer at Mamer (6 years of measurements; meanstorm flow coefficient of all winter events = 0.7218)

Figure 5. Cumulative storm flowrainfall plot for theMamer at Schoenfels (6 years of measurements;mean storm flow coefficient of all winter events =0.5464)

Hydrology


Hydrology


Groundwater levels as state indicator in rainfallrunoffmodelling using Artificial Neural NetworksN.J. de Vos 1, T.H.M. Rientjes 1,2 & L. Pfister 31 Section of Water Resources, Faculty of Civil Engineering and Geosciences, Delft University of Technology, P.O. Box 5048,2600 GA Delft, The Netherlands; [email protected] Institute for GeoInformation Science and Earth Observation (ITC), P.O. Box 6, 7500 AA Enschede, The Netherlands3 Cellule de Recherche en Environnement et Biotechnologies (CREBS), Centre de Recherche Public – Gabriel Lippmann, 162aAvenue de la Faïencerie, L1511, Luxembourg, G.D. Luxembourg

AbstractArtificial Neural Networks (ANNs) have beensuccessfully used for the simulation of rainfallrunoff behaviour in the Hesperange catchment(Luxembourg). Groundwater level informationwas used with the ANN models as an indicatorof the hydrological state of the catchment andwas found to be valuable as additional modelinput.

IntroductionThe various interacting processes that involvethe transformation of precipitation intodischarge are complex and spatially as well astemporally variable, which makes rainfallrunoffmodelling using knowledgedriven approachesfar from a trivial task (Beven, 2001). Theseapproaches often suffer from excessive datarequirements, large computational demands,and calibration problems. We thereforeinvestigated the potential of a datadrivenapproach.

Datadriven rainfallrunoff modellingapproaches are based on extracting and reusing information that is implicit in hydrologicaldata, without directly taking into account thephysical laws that underlie the hydrologicalprocesses. Our approach involved Artificial

Neural Networks (ANNs), which have provento be a successful modelling tool in varioushydrological applications (ASCE TaskCommittee on Application of Artificial NeuralNetworks in Hydrology, 2000). We developedand tested multilayer feedforward ANNmodels using a data set from the Hesperangecatchment in Luxembourg (Fig. 1). Severaldesign aspects of ANN modelling were

investigated. Moreover, the influence of addinggroundwater level information as additionalmodel input was tested.

Artificial Neural NetworksANNs (Fig. 2) use dense interconnection ofsimple computational elements, known asneurons. Optimisation algorithms attempt tooptimise the ANN’s internal structure (i.e. theweights on the internal connections) in acalibration procedure. The goal of mostalgorithms is to match the response of theANN to sample input data with accompanyingsample output data.

Advantages of ANNs are: (1) ability tosimulate nonlinearity; (2) no assumption of ana priori solution structure is needed (nonparametric technique);

Figure 1. Catchment and data.

Hydrology


(3) ability to discern relevant from irrelevantinformation;(4) compactness, flexibility, and lowcomputational demands.

Disadvantages of ANNs are: (1) someANN design aspects have to be determinedthrough trial and error; (2) training is complex;(3) good performance during training does notimply good operational performance.

RainfallRunoff Modelling UsingANNsANNs can be used as empirical rainfallrunoffmodels by letting them simulate the relationsbetween the input and output of a hydrologicalsystem (e.g. Hsu et al., 1995; Shamseldin,1997; Rajurkar et al., 2004). The input signalsto an ANN rainfallrunoff model should containas much information as possible on the futuredischarge. Low or overlapping informationcontent of input signals, however, may result indeteriorated performance. Some variables thatare relevant to discharge prediction are listedin Table 1.

Table 1. Relevant ANN inputs for rainfallrunoffmodelling.

Input variables Information contentRainfall Driving force for runoff

productionEvaporation,temperature

Losses in waterbalance, indicators ofseason

Previousdischarge,groundwaterlevels, soilmoisture, rainfallindex

Indicator ofhydrological state

ANN DesignIn our study, we represented the dimension oftime in the ANN model approach using tappeddelay lines: multiple time series values over acertain window in time are presented asseparate network input signals. ANNs with onehidden layer of neurons between the inputunits and output neuron were found to beeffective and parsimonious model structures.The optimal number of hidden neurons in allANN models was found to be around thesquare root of the number of input units. Wechose the LevenbergMarquardt algorithm forANN training. This robust algorithm convergedfastest and produced the most accuratesimulation results

ResultsThe low time resolution of the data (days),considering the size of the catchment and thetime scales of the dominant runoff processes,constrained the forecasting capability of ourmodels.Figure 3 shows the results of a simulation overthe entire length of the time series, in whichthe current discharge is simulated usingcurrent and previous rainfall and evaporationvalues, and previous values of the discharge.This is the ideal case in which the forecastlead time is zero.Table 2 shows the mean performance (over 10runs) of various ANN model types for a onedayahead forecast of Q. Figs 4 and 5 showthe best results of model structures 2 and 3.Model structure 2 performs much better thanmodel structure 1: the groundwater levelinformation represents the hydrological statewell.

Figure 2. Multilayer feedforward ANN for rainfallrunoffmodelling

Figure 3. Simulation results over complete data period(19962001); forecast lead time equal to zero

Hydrology


Table 2. Mean performance of various ANNmodel structures for onedayahead forecast ofQ

ANN Inputs Performance Variables Window of

timeRMSE R2

1 PE

t6 to t0t4 to t0

4450 0.277

2 PEGW

t6 to t0t4 to t0t6 to t0

3350 0.582

3 PEQ

t6 to t0t4 to t0t2 to t0

3120 0.631

4 PEQGW

t6 to t0t4 to t0t2 to t0t6 to t0

3070 0.655

P = precipitation;E = evaporation;GW = groundwater level;Q = discharge;RMSE = Root Mean Squared Error;

R2 = Nash Sutcliffe coefficient.

The performance of model structure 3 showsthat using previous discharge values is aneven better way of representing hydrologicalstate. Combining both state indicators (modelstructure 4) results in a minor performancegain: there is considerable information overlap.

ConclusionANNs are capable of modelling thetransformation between rainfall and runoff.However, the time resolution of the availabledata was too low (in proportion to the averagelag time between peak rainfall and peakdischarge) for the forecast performance to bevery good. Groundwater level information maybe successfully used as additional ANN modelinput for rainfallrunoff simulation. It is a goodindicator of the hydrological state of thecatchment.

ReferencesASCE Task Committee on Application of Artificial Neural

Networks in Hydrology, 2000. Artificial neuralnetworks in hydrology, II: hydrologic applications.Journal of Hydrologic Engineering 5(2), pp.124137.

Beven, K.J., 2001. How far can we go in distributedhydrological modelling? Hydrology and EarthSystem Sciences 5(1), pp. 112.

Hsu, K.L., H.V. Gupta, & S. Sorooshian, 1995. Artificialneural network modeling of the rainfallrunoffprocess. Water Resources Research 31(10), pp.25172530.

Rajurkar, M.P., U.C. Kothyari, & U.C. Chaube, 2004.Modeling of the daily rainfallrunoff relationship withartificial neural network. Journal of Hydrology 285,pp. 96113.

Shamseldin, A.Y., 1997. Application of a neural networktechnique to rainfallrunoff modelling. Journal ofHydrology 199, pp. 272294.

Figure 4. Model structure 2 forecasting results over validation period (groundwater levels as state indicator)

Hydrology


Figure 5. Model structure 3 forecasting results over validation period (previous discharge as state indicator)

Hydrology


Implications of hydrological modelling and observations inthe Alzette river basinG.P. Zhang 1, H.H.G. Savenije 1, F. Fenicia 1, T.H.M. Rientjes 1,2 & P. Reggiani 31 Section of Water Resources, Faculty of Civil Engineering and Geosciences, Delft University of Technology, P.O. Box 5048,2600 GA Delft, The Netherlands; [email protected] International Institute for GeoInformation Science and Earth Observation (ITC), P.O. Box 6, 7500 AA Enschede, TheNetherlands3 WL | Delft Hydraulics, P.O. Box 177, 2600 MH Delft, The Netherlands

IntroductionPhysically based rainfallrunoff models havebeen evolved remarkably over the pastdecades along with recognition anddescriptions of different runoff processes.Infiltration excess overland flow and saturationoverland flow are two major runoff generationmechanisms that are widely adopted in modelapproaches, e.g. REW approach (Reggiani etal., 1999, 2000, Reggiani & Rientjes, 2005;Zhang et al., 2003, in press). However,subsurface storm flow processes, such asmacropore flow, preferential flow and pipeflow, generally are not well addressed in thosemodel approaches, even though they havebeen extensively investigated (e.g. Uhlenbrooket al., 2002). In this study, we applied the REW approach toa subcatchment of the Alzette river basin inLuxembourg. Modelling results showed thatstream hydrographs are generally wellsimulated.

However, the simulated saturationoverland flow area fractions for most of theREWs (subwatersheds) are relatively largecompared to the values reported in theresearch for other catchments. Throughanalyzing the modelling results, reexaminingthe model structure and evaluating the fieldobservations, we conclude that quicksubsurface flow processes are dominant in theAlzette river basin and should be furtherinvestigated. Therefore, a quick subsurfaceflow component should be taken into accountin the model code to improve therepresentation of not only the streamhydrograph, but also the other state variables,such as groundwater level and saturated areafractions.

Figure 1. The Alzette river basin and the model area

Hydrology


Modelling watershed response inthe Alzette basinThe Representative Elementary Watershed(REW) approach is applied to set up the modelto simulate the watershed response to rainfallevents. Using physically based ordinarydifferential equations, the model describes themost dominant hydrological processes, i.e.saturation overland flow, unsaturatedsubsurface flow, saturated groundwater flowand river channel flow. In addition, interceptionand groundwater outflow are taken intoaccount. The Alzette river basin (Fig.1) is mainlylocated in the Luxembourg part of the Parisgeological basin. The Hesperange subbasin(Fig. 1), for which the rainfallrunoff model hasbeen built, is selected for this study. Since thebeginning of the 1990’s, a dense hydroclimatological observation network has beenestablished in the Alzette basin (Pfister &Hoffmann, 2002).

Numerical simulationsThe Hesperange subbasin is divided into 15REWs (Fig. 1) using the 3rd order threshold ofthe Strahler order system. Runoff simulationsare carried out using daily rainfall and potentialevaporation time series as external drivingforce. The simulation period is from01/01/1998 to 12/31/1998. Initial conditions(e.g. soil moisture content) and soil parametersare set to uniformly distributed over eachREW. Geometric properties of each REW arederived from the DEM data. For brevity, weonly present here the modelling results of the

stream hydrograph at the outlet of theHesperange subbasin (Fig. 2) and thesaturation overland flow area fraction for eachREW (Fig. 3).

Discussion and conclusions• The general dynamics of the watershed

responses, represented by the dischargeat the outlet, is well simulated (Fig. 2).

• Topographic control on the saturated areaformation is well represented by themodel: the steeper the hillslope, thesmaller the saturated area (Fig. 3), andvice versa.

• Recession limbs for some of the peakdischarge events are less well modelledsince the current model is lacking amechanism to describe this hydrologicalresponse.

• The simulated saturated area fractions formany of the REWs are relatively high.

•••••••••••••

This is due to the fact that the saturationoverland flow is the only mechanism inthe current model to represent all quickflow components. Thus, a relatively largesaturated area is needed to support thestorm runoff generation in order toreproduce the stream hydrograph.

• Field observations in the study areaindicate that saturated areas are ratherlocalized and concentrated in the valleys.It also has been noticed that macroporesin soils and fractures in weathered rocksare existing. Therefore, quick subsurface

Figure 2. Simulated and measured discharges

Hydrology


flow processes are likely to be dominantin this basin.

• The REW approach and the modelapplied here are capable of describing amesoscale watershed rainfall runoffrelation. However, the model structureneeds to be improved when applied to acatchment like the Alzette river basinwhere subsurface storm flow processesprevail.

AcknowledgementThis research is funded by Delft Cluster project‘Oppervlakte water hydrologie’.

ReferencesPfister, L. & L. Hoffmann, 2002. Experimental hydro

climatological atlas of the Alzette river basin Grandduchy of Luxembourg. Centre de Recherche Rublic Gabriel Lippmann, Cellule de Recherche enEnvironnement et Biotechnologies, Luxembourg.

Reggiani, P., S.M. Hassanizadeh, M. Sivapalan & W.G.Gray, 1999. A unifying framework of watershedthermodynamics: constitutive relationships. Advancesin Water Resources, 23(1), pp. 1539.

Reggiani, P., M. Sivapalan & S.M. Hassanizadeh, 2000.Conservation equations governing hillsloperesponses: physical basis of water balance. WaterResources Research, 36(7), pp. 18451863.

Reggiani P. & T.H.M. Rientjes, 2005. Fluxparameterization in the representative elementarywatershed approach: application to a natural basin.Water Resources Research 41 (4), W04014.

Uhlenbrook, S., M. Frey, C. Leibundgut, & P. Maloszwski,2002. Hydrograph separations in a mesoscalemountainous basin at event and seasonal timescales.Water Resources Research, 38(6), pp. 114.

Zhang G., P. Reggiani, T.H.M. Rientjes & S.M.Hassanizadeh, 2003. Modeling rainfallrunoff relationby Representative Elementary Watershed approach.In: R.S.E.W. Leuven, A.G. van Os, & P.H. Nienhuis(eds.), Proceedings of NCRdays 2002; currentthemes in Dutch river research, NCRpublication 202003.

Netherlands Centre for River Studies, Delft, pp. 2022.Zhang G.P., F. Fenicia, T.H.M. Rientjes, P. Reggiani &

H.H.G. Savenije, in press. Modeling runoff generationin the Geer river catchment with improved modelparameterization to the REW approach. Physics andChemistry of the Earth.

Figure 3. Simulated saturation overland flow area fraction

Hydrology


Hydrology


Surface water management during droughts in peat areasY. van der Velde, R. Brunt, R. van Montfoort & R. StuurmanNetherlands Institute of Applied Geoscience (NITG) TNO, P.O. Box 80015, 3508 TA Utrecht, The Netherlands;[email protected]

IntroductionDuring the dry summer of 2003, surface waterwas added to peat areas to prevent decreasingwater table levels and oxidation of the peat.Great efforts have to be made to ensure thesehigh water levels, however, effectiveness ofhigh surface water levels during dry summersis doubtful. To understand water flow patternsand flow directions within peat soils, wecombined field measurements with modelresults. As a result, key parameters thatinfluence the interaction between surfacewater level and groundwater level were found.

MethodsWe selected three field sites to monitor surfacewater levels, groundwater levels andgroundwater quality. (De Meije, Vlietpolder andIlperveld). Each site has its own localhydrology (e.g. Fig. 1), which determines waterflow directions.

Important differences between sites are theamount of infiltration, compaction of the peatand vegetation. De Meije and Ilperveld arenature conservation areas (with very loosepeat structures and therefore high hydraulicconductivities), whereas the Vlietpolder is anagricultural site. In this paper, the fieldmeasurements and model results of De Meijeare presented and discussed.

MeasurementsTime series of groundwater levels in De Meijeshow that the effect of surface water levelchange on the groundwater levels stronglydepends on the ditch density. Making use of along ‘prikstok’, crosssections of winter andsummer electrical conductivity (EC) andtemperature (Fig. 2) have been measuredbetween two ditches. The ditches contain high(120 mg/l) chloride concentrations duringsummer and low concentrations (15 mg/l)during winter. The temperature crosssections

for summer and winter are fairly symmetricalas expected. A high temperature duringsummer in the center of the field might becaused by preferential flow. The Ec crosssections correspond well to the temperatureprofiles. In summer it shows the same fastdownward flow in the center of the profile. From these measured profiles we canconclude that in De Meije preferential flow isan important process. Mechanisms that causethis preferential flow, like compaction from thetop, change of hydraulic conductivity related toinward flux from ditches, or influence ofvegetation, are still unknown.

ModelsA modflow model study has been conducted ofthe impact of lowering the surface water levelunder natural evaporation of 40 cm in 2months. Figure 3 shows a sample result for DeMeije. In the summer months (between day160 and day 255, xaxis) the surface waterlevel decreased by 40 cm. The figure showsthe drawdown (expressed with a color scale) inthe meadow (Yaxis represents distance fromditch) resulting from a decreasing surfacewater level in summer. Hydrus 2D has been used tounderstand the flow patterns within thefield crosssection of De Meije. The ditchescontain high (120 mg/l) chloride concentrationsduring summer and low concentrations (15mg/l) during winter. The water level of theditches was kept constant during simulations.The main overall flow direction is downward,but close to the ditches the main flow direction

Figure 1. Schematic overview of subsurfacelithology and groundwater levels in De Meije.

Figure 2. Winter and summer profiles of soil electricalconductivity (EC; upper graphs) and winter andsummer profiles of soil temperature (TEMP; lowergraphs).

Hydrology


is horizontal (more so in summer than in winterdue to evapotranspiration). This process iscreating ‘bananas’, zones withrelatively high chloride concentrations (Fig. 4).

Model results do not clearly resemble themeasurements of EC profiles in De Meije, butcould explain the large variety in areas withlow and high concentrations observed.

ConclusionsKey parameters influencing interactionbetween surface water level and groundwaterlevel are: (1) ditch density and (2) the localhydrology. The measured ECprofiles do notmatch the Hydrus model results. Thus,unknown factors influence water flow patterns,like heterogeneity of the subsurface and soilproperties.

Future work• Continuing fieldwork on groundwater

quality to understand flow patterns.• More research on heterogeneity and

physics of peat soil properties and surfacewater sediment resistivity.

• Production of a sensitivity map (i.e. a mapwith peat areas sensitive to surface waterlevel changes during dry summers).

Figure 4. Hydrus 2D calculation of concentration patterns in winter within a crosssection(40 m)

Figure 3. Modflow calculations of the effect oflowering the surface water level by 40 cm during adry summer.

Ecology


Development of nutrient loads from headwaters to lowlandriversD.S.J. Mourad 1, M. van der Perk 1, K. Piirimäe 2, E. Loigu 2 & J. Deelstra 31 Department of Physical Geography, Faculty of Geosciences, Utrecht University, P.O. Box 80.115, 3508 TC Utrecht, TheNetherlands; [email protected] Department of Environmental Engineering, Tallinn Technical University, Ehitajate tee 5, 19086 Tallinn, Estonia3 Jordforsk (Norwegian Centre for Soil and Environmental Research), Frederik A. Dahls vei 20, N1432 Ås, Norway

AbstractEutrophication problems in shallow waterbodies triggered the development of large(drainage basin) scale models to simulate theorigin, transfer and retention of nutrients.Although these models are well capable ofsimulating longterm average nutrient loads,the spatial and temporal distribution of transferand retention processes is still unknown.Therefore, we took water quality samples, andmeasured discharges at 80100 locations in a900 km2 catchment in Estonia during six fieldcampaigns. Samples were taken from tiledrains, sources, and lower and higher orderstreams. First results for nitrate show theimportance of seasonality, land cover,hydrological connectivity, and instreamretention for the development of nutrient loads.We will use this information to improve processdescriptions in largescale models.

IntroductionEutrophication of shallow water bodies byelevated nutrient concentrations is one of themost common environmental problems. Theinput of nutrients is a complex function ofnutrient emissions, their transfer throughvarious hydrological pathways, temporarystorage and decay in soil and groundwater,

and retention processes in the drainage basin.De Wit (2001) and Mourad et al. (2005) haveproved that it is well possible to model longterm average nutrient fluxes within a drainagebasin, using large, easily to obtain GISdatasets. Fluxes are simulated on basis ofrelatively easy to obtain statistical data aboutpoint and diffuse emissions, and maps withland cover, elevation, drainage network,hydrological fluxes and residence times.However, the spatial and temporal distributionof processes that dominate the transfer ofnutrients from the land surface to the riveroutlet are still unknown, because the modelsare usually calibrated only on average nutrientloads from a limited amount of river locations.Only when the spatial and temporal distributionof nutrient concentrations and loads at themedium (catchment) scale (1001000 km2) isknown, we can improve process descriptionsin drainage basin scale models and makethem suitable for the evaluation of mitigationmeasures to prevent excessive nutrient loads.

MethodsWe carried out six field campaigns in thelowland catchment of the Ahja jõgi in Estonia(area: 900 km2), which is characterized byalternating agricultural areas and forests andpeat bogs. Higher order rivers are typicallysituated in wide seminatural valleys filled with

Figure 1. Nitrate concentrations in summer 2003

Figure 2. Nitrate concentrations in spring 2003

Ecology


riparian wetlands and surrounded by forestedslopes. The lower order streams often have amore direct connection to the agriculturalfields. There is one town in the catchment,Põlva (6500 inhabitants). During six fieldcampaigns (Summer 2002; Autumn 2002,Spring 2003, Summer 2003, Autumn 2004,Spring 2004), water samples were taken atapproximately 80100 different locationsspread over the catchment. These locationsare covering streams of all sizes and types,including tile drains, sources and first orderstreams as well as higher order rivers.Additionally, we estimated stream dischargewhere possible. The water samples wereanalyzed for the major water quality ions in theLaboratory of Physical Geography, UtrechtUniversity. Concentrations (mg l1) of the mainnutrient species NH4

+, NO3, and PO4

3+ wereplot as maps and river profiles for all fieldcampaigns. Discharge measurements wereused to calculate average daily loads (kg d1)for the sampling periods. Here we restrictourselves to nitrate (NO3

).

ResultsFigure 1 shows the spatial distribution ofnitrate concentrations during summer baseflowconditions. High nitrate concentrations arefound downstream of wastewater outlets, andin agricultural areas where fields are directlyconnected to streams. Further downstream,concentrations tend to decrease. However,during spring flood, high nitrate concentrationsare present almost everywhere (Fig. 2), in bothlower and higher order streams. The nitrateconcentration profile for the Ahja jõgi (Fig. 3)shows that concentrations in spring 2003 arethe highest (around 78 mg l1), followed byconcentrations in spring 2004 and autumn2003 (46.5 mg l1).

Concentrations in summer 2002, summer 2003and autumn 2002 (all baseflow periods) aretypically 15 mg 11. The daily load profile ofthe Ahja jõgi (figure 4) shows even moredifference between the seasons. Note thatloads quickly increase in downstream directionduring spring 2003, and to a lesser extentduring spring 2004 and autumn 2003.Retention inhibits a downstream increase ofloads during baseflow periods. Loads, mainlydeveloping upstream, do not increase furtherdownstream.

Conclusions and outlookThe first results reveal that there is a strongseasonal variation in nitrate concentrationsand loads, caused by both higherconcentrations and higher discharges in thewinter half year. The spatial variation in nitrateconcentrations and loads in the catchmentroughly follows land cover: In agricultural areasand near point emissions, concentrations andloads are higher than in forested areas.Moreover, the connectivity of the hydrologicalpathways from field to stream is important. In(summer) baseflow periods, nitrate export islimited by retention in bed sediments, whichseems absent during spring flood. For theimprovement of drainage basin scale models,we will focus on seasonality and hydrologicalconnectivity.

ReferencesDe Wit, M.J.M., 2001. Nutrient fluxes at the river basin

scale. I: the PolFlow model. Hydrological Processes15, pp. 743759.

Mourad, D.S.J., M. van der Perk, G.D. Gooch, E. Loigu, K.Piirimäe, & P. Stålnacke, 2005. GISbasedquantification of future nutrient loads into LakePeipsi / Chudskoe using qualitative regionaldevelopment scenarios. Water Science andTechnology 51(34), pp. 355363

Figure 3. Nitrate concentrations profile of the Ahja jõgiriver for the six seasons

Figure 4. Nitrate loads profile of the Ahja jõgi river for thesix seasons

Ecology


Floodplain sedimentation regulating vegetationproductivity on small rivers?

F.P. Sival, B. Makaske, G.J. Maas & J. RunhaarAlterra, Wageningen University and Research, P.O. Box 47, 6700 AA Wageningen, The Netherlands; [email protected]

AbstractSediment input and associated nutrients werequantified along a vegetation gradient from theriver channel to the floodplain margin in fivenature reserves on four small rivers. Theamount of sedimentation during the floodseason of 20032004 was measured usingsediment traps. Grainsize and nutrientanalyses of the trapped sediment sampleswere carried out. The biomass of thevegetation is different for all investigated areasand varies between 900 g/m2 and 200 g/m2.Especially in the Kapperbult area on theDrentsche Aa, the biomass decreases withincreasing distance from the river. Measuredamounts of sediment in the Kapperbult areaare small and strongly decrease withincreasing distance from the river: 2.7 kg/m2

close to the river and 00.07 kg/m2 far from theriver. Likewise, nitrogen and phosphate inputthrough sedimentation also decrease withincreasing distance.

IntroductionFlooding or water retention in combination withnature development is not in all situationswithout risk for the vegetation (CommissieWaterbeheer 21e eeuw, 2000; Raad voor hetLandelijk Gebied, 2001). Especially vegetationin nutrientpoor conditions will have considerableharm from flooding with nutrientrich water. Thehypothesis is that differences in biomassproductivity are explained by differences insedimentation. Quantitatively, however, the inputof nutrients like nitrogen and phosphate byflooding is largely unknown (Sival et al., 2002).

The main question in our research is: whatis the relationship between input of nutrients bysedimentation and the productivity for differentvegetations along a gradient from the river to thefloodplain margin? Is the input from sedimentscomparable to the input from floodwater,atmospheric deposition, mineralization andgroundwater? We investigated five naturereserves on four small rivers in the Netherlands(Dommel, Drentsche Aa, Reest and OverijsselseVecht; Fig. 1). In this paper we will mainlypresent results from the Kapperbult area on theDrentsche Aa.

MethodsTo characterize the soil and vegetation, aninventory of both was included in the research.The standing crop of the vegetation wasmeasured in the summer by cutting thevegetation. After drying, the vegetation samplewas weighed and analyzed on N and P.Sediment traps were placed along vegetationgradients across the levee and floodbasin tothe floodplain margin, in each of the studyareas. After flooding, the traps were collectedand the trapped sediment was analyzed on: (1)quantity, (2) texture, (3) N and P content.Special attention was paid to spatial patternsof these variables in relation to floodplaingeomorphology.

ResultsThe biomass of the vegetation is different forall investigated areas and varies between 900

Figure 1. Location of the study sites

Figure 2. Vegetation biomass (g/m2) along a gradient fromthe river to the floodplain margin of all investigated sites

Ecology


g/m2 and 200 g/m2 (Fig. 2). With increasingdistance from the river the biomass decreasesand this effect is most pronounced in theKappersbult area on the Drentsche Aa. Lowproductive vegetation (<400g/m2), assumed tobe most sensitive to nutrient input by flooding,is present far from the river. A borehole crosssection at theKappersbult reveals an approximately 60cmthick clayey and peaty clay bed on underlyingpeat that fills the deep Drentsche Aapalaeovalley (Fig. 3). Further away from theDrentsche Aa the clayey bed rests on fine(loamy) sand representing a coversand ridgebordering the palaeovalley. These subsurfacedata suggest a recent increase in sedimentinput, although the absolute date of thischange is unknown. The composition of thetopsoil reflects the present sedimentaryprocesses. The impact of sedimentationhistory (and the resulting spatial variation insubsurface composition) on present vegetationproductivity is still under study.

In January 2004, most study areas wereflooded, with a maximum flooding duration ofone week. The amounts of sediment depositedin the Kappersbult area strongly decrease withincreasing distance from the river: 2.7 kg/m2

close to the river and 00.07 kg/m2 far from theriver. Absolute amounts of deposited clay andorganic matter, although being much lower,show a comparable spatial trend. Nitrogen andphosphate amounts also decrease with

increasing distance from the river (Fig. 4).Nitrogen input varies between 90 kg/ha closeto the river and 5 kg/ha far from the river (forobtaining an estimate of total input, anatmospheric deposition of 30 kg/h must beadded). Phosphate input varies between 45kg/ha close to the river and 1 kg/ha far fromthe river.

ConclusionsIn the Kappersbult area the texturalcomposition of the sediments that arepresently being deposited, matches thecomposition of the topsoil, indicating no recentchanges in the sedimentary processes. In thisarea the biomass of the vegetation seems todepend on the nutrient input from sediments:both significantly decrease with increasingdistance from the river. These results suggestthat increasing sedimentation, associated withincreased flooding/water retention, may causea change from lowproductive floodplaingrassland into highproductive floodplaingrassland. Generally, this process will involvea strong decrease in the amount of speciespresent in the vegetation.

AcknowledgementsThis study is funded by the Dutch Ministry ofAgriculture, Nature and Food Quality.

ReferencesCommissie Waterbeheer 21e eeuw, 2000. Waterbeleid

voor de 21ste eeuw. Advies van de commissieWaterbeheer 21ste eeuw. Den Haag, 54 p.

Raad voor het Landelijk Gebied, 2001. Bergen met beleid.Publicatie 01/4, Raad voor het Landelijk Gebied,Amersfoort, 48 p.

Sival, F.P., P.C. Jansen, B.S.J. Nijhof & A.H. Heidema,2002. Overstroming en vegetatie: literatuurstudie overde effecten van overstroming op zuurgraad envoedselrijkdom. Alterrarapport 335, Alterra,Wageningen, 66 p

Figure 3. Measured sedimentation and thecorresponding lithological borehole crosssection inthe Kappersbult area

Figure 4. Amount of sediment input of nitrogen (N)and phosphate (P) (kg/ha) along a gradient from theriver to the floodplain margin in the Kappersbult area

Ecology


Ecological impact of changes in groundwater withdrawal inriver forelandsM.W.A. de Haan 1, A.J.M. Jansen 2, J. Grijpstra 1 & C.J.S. Aggenbach 11 Kiwa Water Research, P.O. Box 1072, 3430 BB Nieuwegein, The Netherlands; [email protected] Vitens Watertechnologie, P.O. Box 400, 8901 BE Leeuwarden, The Netherlands

AbstractThe project ‘Buurtschap IJsselzone’ is aninitiative for rural development along the riverIJssel at Zwolle, The Netherlands. One of theobjectives of the project is to secure thepresent drinking water supply by thereallocation of abstraction wells of the drinkingwater production station Engelse Werk. Themodel NICHE was used to gain insight in theecological effects of the reallocation ofabstraction wells in the floodplains of theIJssel. Model output showed that not only riverwater dynamics affects habitats of plant andbirdlife directly, but groundwater hydrology aswell. The results of the assessment led to analternative design of abstraction wells, whichforms a sound basis for nature development inthe floodplain area.

IntroductionThe study area ‘Buurtschap IJsselzone’ issituated between the city of Zwolle and theriver IJssel, which is a distributary of the Rhine.River forelands (embanked floodplains) form amajor part of the study area and consist mainlyof nature reserves and agricultural meadows.

In the study area several major issuesrelated to spatial planning exist. In the riverforelands, more storage capacity for riverwater is required during periods of peakdischarge of the Rhine. This area is alsoprotected by the European Bird Directive andis part of the national ecological network of theNetherlands. Pollution of groundwater underthe city of Zwolle and the construction of a newrailway threaten the present drinking waterproduction station Engelse Werk, and a newdesign of abstraction wells is required toprevent further increase of treatment costs.Furthermore, socialeconomical functions suchas agriculture are declining. In order tocombine these issues in a multifunctionallandscape, an initiative for rural developmentin the ‘Buurtschap IJsselzone’ was started (DeKuijer et al., 2003).

For the objective to realise a sustainabledrinking water production without risks ofcontamination by pollution, a sustainabledevelopment of the catchment area isnecessary.

Figure 1. Predicted vegetation patterns of EleocharitoacicularisLimoselletum in the floodplainSchellerwaarden. A: reference situation, with naturedevelopment in current agricultural land.

Figure 1. Predicted vegetation patterns of EleocharitoacicularisLimoselletum in the floodplainSchellerwaarden. B: situation after reallocation ofdrinking water abstraction, with a new river sidechanneland nature development.

Ecology


Plans for reallocation of the presentgroundwater abstraction by the drinking watercompany Vitens are therefore interwoven withthe initiative for rural development.Supplementary values for Vitens are therealisation of a social basis for waterproduction and the prevention of agovernmental deadlock over the socioeconomical development and spatial planning.Since changes in groundwater abstraction mayaffect nature and environment, a detailedenvironmental impact assessment ofreallocation is required. This contributionfocuses on the assessment of the reallocationof abstraction wells on the occurrence ofvegetation and bird communities in the riverforelands.

Ecological impact assessmentThe river forelands in the study area containimportant ecological values. Ecological effectsare expected on a small scale (< 1 ha) andmay interfere with nature restoration projects.This requires an ecological impact assessmentmethod with high accuracy on this scale. Themethod needs to calculate effects of changesin groundwater level, changes in surface waterdynamics as well as changes in land use ondistribution patterns of plant communities andecological groups of breeding birds. The scaleof this output corresponds with the informationthat is needed in the planning and control ofnature management.

Aggenbach & Pelsma (in prep.)constructed a database (PREVIEW) with siteconditions of plant communities of riverforelands. From this database, parametersconcerning inundation with river water,groundwater level, soil characteristics andmanagement were taken into account. Most

data could be derived from availablehydrological and soil information. In order touse the model in river forelands, a calculationof inundation depth and inundation durationthroughout spring and summer was carried outby Blonk (2003). From these data, the siteconditions acidity and trophic state werecalculated with the model NICHE (Raterman etal., 2002).

The model used the calculated siteconditions to predict the potential occurrenceof plant communities. In order to generateinformation about the habitat suitability forbreeding birds, these results were combinedwith information about landscape structure anddisturbing factors such as infrastructure. Anavifaunadatabase (Sierdsema, 1995) wasused to translate these habitat factors topatterns of ecological bird communities.

ResultsAlthough several alternatives for reallocation ofthe drinking water abstraction were studied,only the results of the final mostenvironmentalfriendly alternative will bediscussed. In this alternative, hydrologicaleffects of the reallocation are restricted to theriver foreland currently in use for agriculture. Inorder to protect present nature reserves, thecatchment area of the groundwater extractionstation was reduced by proposing a river sidechannel in the agricultural floodplain.

Important hydrological parameters thatchanged due to reallocation were inundationduration with river water and lowestgroundwater tables. Due to the planned riversidechannel, the river water will flow freelyinto the floodplain. As a consequence lowlyingareas will be flooded easily. On the other hand,

Figure 2. Predicted vegetation patterns of RanunculoAlopecuretum geniculati equisetetosum palustris in thefloodplain Schellerwaarden. A: reference situation, withnature development in current agricultural land.

Figure 2. Predicted vegetation patterns of RanunculoAlopecuretum geniculati equisetetosum palustris in thefloodplain Schellerwaarden. B: situation after reallocation ofdrinking water abstraction, with a new river sidechanneland nature development.

Ecology


the groundwater table will decrease due to thereallocation of abstraction wells.

According to the model output, the effectsof these two hydrological changes onvegetation patterns appear to be very different.This is illustrated by the patterns of the pioneervegetation EleocharitoLimoselletum and thegrassland vegetation RanunculoAlopecuretumequisetetosum (Figs 1 and 2). Both plantcommunities occur in the low parts of riverforelands, often adjacent to each other.However, in the study area the EleocharitoLimoselletum showed a positive reaction toincreased inundation, whereas the RanunculoAlopecuretum equisetetosum showed anegative reaction to decreased groundwatertables. The model also calculated a differentreaction of breeding birds to the hydrologicalchanges. Habitats of typical meadow birds withground nests will decrease because of theincreased inundation duration, whereashabitats of marshland birds will move towardszones along the planned river sidechannel.

ConclusionsThe model output showed that not only thedynamics of inundation by river water affecthabitats of plant life and birdlife in riverforelands, but groundwater hydrology as well.Relative small changes in groundwater level infloodplains already affect plant communitiesrestricted to continuous high summer watertables.

Since the required site conditions of plantcommunities show great differences, and smallchanges in abiotic conditions may cause majorchanges in distribution patterns of valuableplant communities, ecological assessment ofhydrological impacts in river forelands requiresa model that distinguishes vegetation types oncommunity level.

The results of the ecological impactassessment in the project ‘BuurtschapIJsselzone’ led to an alternative design ofabstraction wells, which forms a sound basisfor nature development in the floodplain area.The planned construction of a river sidechannel will restrict effects of groundwaterabstraction to parts of the river foreland inagricultural use. At the same time, theconstruction of a river sidechannel opens upnew perspectives for nature development. Themodelinstrument appeared to be helpful infurther spatial planning of rural development inthe study area.

ReferencesAggenbach, C.J.S. & T.A.H.M. Pelsma, in press. Hydro

ecological assessment of vegetation of Dutch riverhabitats. Archiv für Hydrobiologie Suppl. LargeRivers.

Blonk, A.T., 2003. Bouw instationair grondwatermodel“Engelse Werk”. Tauw BV, Deventer.

De Kuijer, O., A.J.M. Jansen & H. de Graaf, 2003.Buurtschap IJsselzone biedt meervoudigperspectief. Nova Terra 3 (2), pp. 1419.

Raterman, B.W., M.W.A. de Haan, & A.F.M. Meuleman,2002. GIS in ecological impact assessment ofwetlands. ESRI User conference proceedings 2002(http://gis.esri.com/library/userconf/proc02/abstracts/a0622.html).

Sierdsema, H., 1995. Broedvogels en beheer: het gebruikvan broedvogelgegevens in het beheer van bos ennatuurterreinen. SBB/SOVON, Driebergen/BeekUbbergen.

Ecology


Cumulative effect assessment of physical reconstructionand landuse changes on riverine biodiversityR.S.E.W. Leuven, G.W. Geerling, S. Gerrits, H.J.R. Lenders & R.J.W. de NooijInstitute for Wetland and Water Research, Faculty of Science, Radboud University Nijmegen, P.O. Box 9010, 6500 GLNijmegen, The Netherlands; [email protected]

AbstractA GISbased scenario approach wascombined with the model BIOSAFE forcumulative effects assessment (CEA) ofphysical reconstruction and landuse changesin floodplains. A case study for floodplains inthe Middle Waal region visualises effects ofvarious scenarios for ecological rehabilitationand infrastructure facilities on the distributionof riverine ecotopes and on fish, herpetofauna(amphibians and reptiles) and breeding birds.In comparison with the actual situation, allscenarios have negative impact on breedingbirds. The saving scenario seems to favoursettlement of fish. The preserving scenarioappears to be most positive for herpetofauna.Species that will thrive most from futuredevelopments are those that preferhydrodynamic pioneer environments (e.g. sidechannels), whereas species that will encounterdifficulties are more related to relatively lowhydrodynamics (e.g. lakes and isolated riverchannels) or terrestrial habitats (pastures andhay lands).

IntroductionThe coming decades northwestern Europe’sriver basins will be significantly reconstructedfor the purpose of flood risk abatement,ecological rehabilitation and infrastructurefacilities. The physical reconstruction and landuse changes may offer opportunities toincrease biological diversity, but can alsoseriously endanger present natural values andpotentials (Lenders et al., 1998, 2001, De

Nooij et al., 2004). Integrated rivermanagement will require regional effectassessments of combinations of measures onbiodiversity (Leuven et al., 2002). This paperpresents a novel method for CEA andevaluates impacts of current spatial plans onriverine biodiversity in the Middle Waal region.

Material and methodsFor eleven floodplains in the Middle Waalregion (Fig. 1), a GISbased scenarioapproach was combined with the model BIOSAFE (Fig. 2). BIOSAFE is a valuation modelfor riverine biodiversity that focuses on specieslisted on Red Lists, the EU Habitats and Birdsdirectives, and the conventions of Bern andBonn (Lenders et al., 2001, De Nooij et al.,2004). Three scenarios project expectationsonto the riverine landscape in the year 2015(Leuven et al., 2002). The preserving scenariocomprises current plans for ecologicalrehabilitation in eight floodplains. The savingand utilising scenarios also includes currentplans for new infrastructure facilities (e.g.,several harbours, storages for contaminatedriver sediments and multimodal transportcentres). In the saving scenario, availablespace is maximally spared as a nonrenewableresource and ecological rehabilitation andinfrastructure facilities are implemented inaccordance with their ‘best’ spatialalternatives. In the utilising scenario, theavailable space is largely used forinfrastructure facilities and other human

Figure 1. The geographical location of the study area

Figure 2. GISbased procedure for CEA

Ecology


purposes. All spatial data were implementedand analysed in Arcview. GIS allowedcomparison of the future scenarios with thereference situation (year 1997) andcalculations of surface areas of riverineecotopes (input for BIOSAFE).

Results and discussion

Figure 3 presents an example of ecotopemaps of the reference situation and the threescenarios for the floodplains in the westernpart of the study area. The acreages of naturalecotopes, agricultural and builtup areas of theMiddle Waal region remarkably differ for overthe three scenarios (Fig. 4). In comparison withthe actual situation, all scenarios havenegative impact on breeding birds (Fig. 5).

The utilising scenario appears to have greaterimpact on breeding birds than the otherscenarios. The saving scenario seems tofavour settlement of fish, but occupies anintermediate position for herpetofauna andbreeding birds. The preserving scenarioappears to be most positive for herpetofauna.Species that will thrive most from futuredevelopments are those that preferhydrodynamic pioneer environments (e.g. sidechannels), whereas species that will encounterdifficulties are more related to relatively lowhydrodynamics (e.g. lakes and isolated river

channels) or terrestrial habitats (pastures andhay lands). Implementation of plans inaccordance with the preserving scenario yieldsan optimally balanced spectrum of species.However, it should be noticed that the savingand utilizing scenarios heavily rely oncompensating measures proposed forinfrastructure facilities, whose feasibility is stilldoubtful at present because of difficulties withland procurement. In addition, several speciesalso require habitat patches outside floodplainsin the course of their life cycle. The downsideof the abandonment of agriculture in thefloodplains is its intensification in thehinterland. This could prove to be quitedestructive for ecotopes (in terms of size orquality) for some species like the great reedwarbler (reed marshes) or the crested newt(isolated, high quality water bodies). If theoccurrence of these species is to besafeguarded, we cannot afford to sacrificevaluable nature areas in the hinterland as atradeoff for improving floodplains (Lenders etal., 1998). In order to minimise the cumulative effectsof infrastructure facilities and flood defencemeasures and to maximise the benefits ofecological rehabilitation, an integrated plan forphysical reconstruction and landuse changesof the study area as a whole should bedeveloped, in which all plans for individualprojects in floodplains are mutually attuned.Such a plan should provide conditions thatmight result in both a higher degree ofdefragmentation of riverine ecotopes andbetter opportunities for broad spectra ofspecies. It is only under this precondition thatphysical reconstruction of floodplains along themiddle reach of the river Waal may offer aprosperous future for protected andendangered species. BIOSAFE can be easily combined withGISbased scenario approaches for CEA ofriverine area developments and appears to bea useful tool for finetuning of spatial plans inearly phases of their planning process. Othermethods, e.g. detailed single species modelstaking into account more habitat demands ofthe species examined, are doubtlessly moresubtle and would lead to more accuratepredictions of settlement opportunities forbiodiversity. However, this would require moredetailed input of speciesspecific parametersas well as of landscape parameters in thetarget situation. Knowledge of the first type ofinput still appears to be lacking for manyriverine species, while data necessary for thesecond type of input can in most cases not bederived from floodplain reconstruction plans.

Figure 3. Ecotope maps for development of thefloodplains

Figure 4. Acreage of river ecotopes.

Ecology


Figure 5. BIOSAFE scores (TFI = Taxonomic group floodplain importance index).

Conclusions• BIOSAFE can be easily combined with

GISbased scenario approaches forriverine areas development.

• Physical reconstruction and landusechanges in floodplains along the MiddleWaal strongly affect the surface areas anddistribution patterns of riverine ecotopesand will have major effects on protectedand endangered species in this area.

• In spite of model and scenarioassumptions, the results facilitate thedebate about and the decisionmakingprocess for policy targets of integratedriver management.

ReferencesDe Nooij, R.J.W., H.J.R. Lenders, R.S.E.W. Leuven, G. de

Blust, N. Geilen, B. Goldschmidt, S. Muller, I.Poudevigne & P.H. Nienhuis, 2004. BIOSAFE:assessing the impacts of physical reconstruction onprotected and endangered species. River Researchand Applications 20(3), pp. 299313.

Lenders, H.J.R., R.S.E.W. Leuven, P.H. Nienhuis, R.J.W.de Nooij & S.A.M. van Rooij, 2001. BIOSAFE: Amethod for evaluation of biodiversity values on thebasis of political and legal criteria. Landscape andUrban Planning 55 (2), pp. 119135.

Lenders H.J.R., R.S.E.W. Leuven, P.H. Nienhuis, K.D.Oostinga & P.J.M. van den Heuvel, 1998. Ecologicalrehabilitation of floodplains along the middle reach ofthe river Waal: a prosperous future for fauna targetspecies? In: P.H. Nienhuis, R.S.E.W. Leuven &A.M.J. Ragas (eds.), New concepts for sustainablemanagement of river basins. Backhuys, Leiden, pp.115130.

Leuven, R.S.E.W., Y. Gerig, I. Poudevigne, G.W. Geerling,L. Kooistra & B.G.W. Aarts, 2002. Cumulative impactassessment of ecological rehabilitation andinfrastructure facilities in the middle reach of the riverWaal. In: R.S.E.W. Leuven, I. Poudevigne & R.M.Teeuw (eds.), Application of geographic informationsystems and remote sensing in river studies.Backhuys, Leiden. pp. 201216

Sediments, Hydraulics and Morphology


Inaccuracies in estimated grain size parameters and theirimplication on geological modelsD. Maljers, S.H.L.L. Gruijters & J.G. VeldkampNetherlands Institute of Applied Geoscience (NITG) TNO, P.O. Box 80015, 3508 TA Utrecht, The Netherlands;[email protected]

AbstractThe accuracy of the measurements of grainsize parameters used in a geological modelare crucial to the overall reliability of theconstructed model. The main goal of thecurrent research is to quantify the reliability ofestimated grain size data and to determine theimpact that these inaccuracies have on 3Dgeological models. Only comparisons of sandmedians will be presented in this text. Theanalysis shows that the sand median isunderestimated. The effect of the inaccuraciesin estimated sand medians on the 3Dinterpolation of grain size data is evaluated,using two methods, first a visual check and,second the calculation of the Shieldsparameter. The conclusion is that theinaccuracies of the sand median do not lead toany significant changes in whether or not thesediment is transported or changes in riversedimentation patterns.

IntroductionIn geological modelling, geostatistics andgeological knowledge are used to combine 1Dand 2D measurements to a 3D frame of thedifferent (geological) layers in the subsoil.Within these layers different parameters maybe estimated. The accuracy of themeasurements used in a geological model arecrucial to the overall reliability of theconstructed model. For morphological models,

insight into the variation of the grain sizeparameters within a layer is dominant. The main goal of the current research is toquantify the reliability of the used grain sizedata and to determine the impact that theseuncertainties have on geological models.The

data used in this study originated from variousprojects carried out for the Institute for InlandWater Management and Waste WaterTreatment (RIZA) during the years 20012003.Samples are taken from vibrocores and areanalysed by Fugro B.V. using the sievemethod (cf. NEN 2560). All of the sampleswere also described at TNONITG and thesedescriptions are all stored in the relationaldatabase DINO (Databank InformatieNederlandse Ondergrond); the database of allsubsoil data of the Netherlands used at TNONITG. In total, approximately 1500 sampleshave been used for analysis. First, theaccuracy of estimated grain size parameters(median grain size of the sand fractionbetween 63 and 2000 µm (D50), silt contentand gravel content) are compared with sieveresults. Only comparisons of sand medians willbe presented in this text. The effect of theseinaccuracies in estimated sand medians on the3D interpolation of grain size data is evaluatedby using a Monte Carlo procedure for a set ofsamples in the IJsselkop bifurcation in the

Figure 1. Scatter plot of estimated D50 values andassociated measured D50 values. In triangular pointsthe upper boundary of the bandwidth and in squaresthe lower boundary of the bandwidth. From within thisbandwidth the realizations of the sand median aretaken.

Figure 2. Triangular probability curve of the sandmedian. In this example an estimated sand median of400 µm and an associated bandwidth of 260770 µm;50 realizations of the sand median are taken from thisfrequency diagram



lower Rhine distributary system in theNetherlands. The inaccuracies in sandmedians are transformed into triangulardistributions. For every realization, the sandmedian, gravel and silt content are used tocalculate a synthetic grain size distribution(GSD). Together with measured grain sizedistributions these data are interpolated using3D kriging (cellsize 25x25x0.2 m).

Results of the analysis of grain sizeparametersOf the analysis of the three grain sizeparameters (sand median, silt and gravelcontent) only the results of the sand medianare presented in this paper. The sand median

is on average underestimated, which can beseen in Figure 1. More details of theinaccuracies and the way they are calculatedcan be found in Maljers & Gruijters (2004). Inthe next section the inaccuracies in the sandmedian are implemented in a geologicalmodel, and the implications on this model arediscussed.

Implementation and implications ofinaccuraciesThe implementation of the bandwidth found inanalyzing the sand median will be presentedbelow. In Fig. 1 the bandwidth of the sandmedian is shown (in triangular en squarepoints). This bandwidth means that forsamples of which the sand median isestimated as 400 µm, the actual sand mediancould be between 260 µm (lower boundary ofthe bandwidth) and 770 µm (upper boundary ofthe bandwidth). For modelling, 50 realizations of the sandmedian are calculated. The values are withinthe bandwidth and follow a (random) triangularprobability curve. This results in 50 differentrealizations of the sand median. For examplethe range of measured sand mediansbelonging to an estimated sand median of 400µm, is between 260 µm and 770 µm. The mainpoint will be around 400 µm (Fig. 2). The other two grain size parameters ofimportance in constructing synthetic GSDs arethe silt content and the gravel content. Siltcontent in this research has been set to 1% forall samples and during all realizations. Thegravel content has been estimated for allsamples and this value is used in modelling.This means that, during each realization theonly parameter that differs is the sand median.Details about constructing synthetic GSDs arefully discussed in the NCRpublication 24(Gruijters et al., 2004) and will therefore not bepresented in this paper. The implications on a geological model ofthe inaccuracies in the sand median can beshown in two ways. First of all a visual checkhas been made; second, the Shieldsparameter, which is a dimensionless measurefor the transport of sediment, has beencalculated for two realizations in combinationwith two river discharges (normal and high). For the visual check, the median of thewhole sample of the top of the Kreftenheyedeposits has been used instead of the top ofthe mobile pavement, because the active layerwas sampled in great detail and has thereforeonly measured GSDs. Therefore thebandwidth of the sand median and theassociated synthetic GSDs will not affect thetop of the mobile pavement. Only in details do

Figure 3a. The calculated Shields parameter ( ) fortwo realizations (a and b), only minor differencescan be seen, leading to no differences in whether ornot the sediment is transported in this example

cr=0.06; threshold of motion).

Figure 3b. The calculated Shields parameter ( ) fortwo realizations (a and b), only minor differencescan be seen, leading to no differences in whether ornot the sediment is transported in this example

cr=0.06; threshold of motion).



the two realizations of the top of theKreftenheye deposits differ in D50. The implementation of the bandwidth of theestimated sand median does not seem to haveany influence on river sedimentation patterns. The calculated Shields parameter, hasalso been calculated for the top of theKreftenheye deposits. When looked at theactual values for the Shields parameter for tworealizations during normal discharges, minordifferences can be seen (Fig. 3). Because theKreftenheye sediment is not very coarse, thethreshold of motion (which is set to 0.06) isalready exceeded for most cells; therefore theminor differences in the Shields parameter donot lead to any difference in whether or not thesediment is transported.

ConclusionsBased on the aforementioned methods it canbe concluded that the applied bandwidth to theestimated sand median does not influence thesedimentation patterns or whether or not the

sediment is transported. This means that themethod described in Gruijters et al. (2004) touse synthetic grain size distributions based onestimated parameters is indeed a step forwardin describing the natural variation in the subsoilof a riverbed.

AcknowledgementsThe authors would like to thank RIZA, RoyFrings and Maarten Kleinhans.

ReferencesGruijters, S.H.L.L., D. Maljers, J.G. Veldkamp, J. Gunnink,

M.P.E. de Kleine, P. Jesse, L.J. Bolwidt, 2004.IJSSELKOP: modelling of 3D grain size distributions,one step closer to reality. In: N. Douben & A.G. vanOs (eds.), NCRdays 2003; dealing with floods withinconstraints. NCRPublication 242004, NetherlandsCentre for River Studies, Delft, pp. 8082.

Maljers, D. & S.H.L.L. Gruijters, 2004. Geologicalmodelling: how realiable are input parameters? In: N.Douben & A.G. van Os (eds.), NCRdays 2003;dealing with floods within constraints. NCRPublication 242004, Netherlands Centre for RiverStudies, Delft, pp. 8687.





Supplylimited transport of bedload sediment at theIJsselkopR.M. Frings & M.G. KleinhansDepartment of Physical Geography, Faculty of Geosciences, Utrecht University, P.O. Box 80115, 3508 TC Utrecht, TheNetherlands; [email protected]

AbstractBedload transport calculations based onmultibeam echo soundings suggest that thesediment transport at the IJsselkop issuppressed by a limited supply oftransportable sediment. This probably resultsfrom bend sorting processes at the riverbifurcations Pannerdensche kop andIJsselkop.

IntroductionThe river bifurcations Pannerdensche Kop andIJsselkop determine the sediment and waterdistribution over the central part of theNetherlands (Fig. 1). A good understanding ofthis distribution process is crucial foroperational river management. It is known thatonly a small part of the bedload sediment thatarrives at the Pannerdensche Kop bifurcationis directed towards the Pannerdensch Kanaal(Kleinhans, 2002). Therefore, the IJsselkopbifurcation, which is situated at the end of thePannerdensch Kanaal, may be subject to alimited supply of transportable sediment. Thepurpose of this study was to determinewhether the bedload sediment transport at theIJsselkop is supplylimited.

MethodsWe used multibeam echo soundings incombination with a dune tracking technique todetermine the bedload transport rate at theIJsselkop. We thus assumed the sedimenttransport to be zero when dunes are absent,which is realistic according to direct transportmeasurements with a Delft Nile Sampler

(Frings, unpublished data). The multibeamecho soundings that we used were conductedduring discharge waves in November 2002and January 2004. The resolution of theseecho soundings enabled us to determine thetemporal, longitudinal and lateral variation insediment transport.

ResultsThe lateral variation in sediment transport waslarge, especially in the Pannerdensch Kanaal.The absence of dunes near the edges of theriver indicates that the sediment transport waslimited to a clearly demarcated zone in themiddle of the river (Fig. 2).

The temporal variability in sedimenttransport during the 2004 discharge wave isshown in Fig. 3, for three river sections. In thefirst section, the Pannerdensch Kanaal, themaximum sediment transport rate occurredwell before the peak discharge. In the othersections, both in the NederRijn, the maximumsediment transport rate occurred much later; inthe downstreammost section even two weeksafter the peak discharge. The longtermtemporal variability is also pronounced: in 2002the sediment transport was almost twice ashigh as in 2004 at the peak of the dischargewave, which was about 6500 m3/s at Lobith inboth years.

Figure 1. Rhine branches at the IJsselkop bifurcation,the Netherlands: 1 = Pannerdensch Kanaal, 2 =NederRijn, and 3 = IJssel

Figure 2. Bed image of the IJsselkop bifurcation,November 2002. A,B,C refer to the sections describedin Figure



Figure 4 shows the water and sedimentdistribution at the IJsselkop. The NederRijnreceived about 90 % of the sediment load inthe Pannerdensch Kanaal and 57 % of thewater discharge, while the third river branch,the IJssel, received an almost equal amount ofwater (43 %), but only 10 % of the sedimentload.

DiscussionAll results support the idea of supplylimitedtransport. Namely, in the case of redundanttransportable sediment, dunes would haveoccurred over the entire river width, while thesediment transport rates in 2002 and 2004would have been the same. Furthermore, themoment of maximum sediment transport wouldhave been equal for the Pannerdensch Kanaaland the NederRijn. The downstream shift inmoment of maximum sediment transport pointsat a sediment wave that moved from thePannerdensch Kanaal into the NederRijnduring the 2004 discharge wave. A sand wavehas also been observed at the PannerdenscheKop (Kleinhans, 2002), suggesting thatsediment waves are a structural phenomenonat bifurcation points.

The supplylimitation is probably larger inthe IJssel than in the other branches of theIJsselkop, because the IJssel receives hardlyany sediment. We expect this to be the resultof the interaction between sediment transportand bed grain size (Fig. 5), in the followingway.Bend sorting in themeander bendupstream of theIJsselkop causes thebed sediment at theentrance of the IJsselto be much coarserthan the bed sedimentat the entrance of theNederRijn. This coarsematerial probably isonly mobile at highdischarges, leading toa limited supply oftransportable sedimentinto the IJssel at lowdischarges and atintermediate dischargeslike those in 2002 and2004. The same process can be heldresponsible for the limited supply of sedimentinto the Pannerdensch Kanaal at thePannerdensche Kop. The supply limitation has severeimplications for the prediction of sedimenttransport rates, as theoretical transportpredictors all assume a redundant amount oftransportable sediment. Empirical predictionsare problematic too, because the supplylimitation is not constant in time. In 2002 thesediment transport was much higher than in2004, while the discharge was the same. Forreliable predictions of sediment transport at theIJsselkop, therefore, the sediment transporthistory needs to be taken into account.

AcknowledgementsWe would like to thank: Leonie Bolwidt (RIZA),Denise Maljers and Stephan Gruijters (NITGTNO), and the members of the MeetdienstDON.

ReferencesGruijters, S.H.L.L., J. Gunnink, J.J.M. Hettelaar, M.P.E. de

Kleine, D. Maljers & J.G. Veldkamp, 2003. Karteringondergrond IJsselkop; eindrapport. TNOrapportNITG 03120B. Nederlands Instituut voorToegepaste Geowetenschappen TNO, Utrecht.

Kleinhans, M.G., 2002. Sorting out sand and gravel:sediment transport and deposition in sandgravel bedrivers. Netherlands Geographical Studies 293,KNAG/Faculteit Ruimtelijke Wetenschappen,Universiteit Utrecht, Utrecht, 317 p.

Figure 3. Temporal variation in sediment transportduring the January 2004 discharge wave in three riversections (A, B, C). The location of the three riversections is shown in Figure 2. The moment of the peakdischarge is indicated with a dashed line

Figure 4. Water distribution (left) and sedimentdistribution (right) at the IJsselkop during the 2002 and2004 discharge waves. IJ: IJssel; NR: NederRijn

Figure 5. Grain size of the top of theriverbed at the IJsselkop (afterGruijters et al., 2003).



Morphological behaviour around bifurcation points;preliminary results of recent measurementsL.J. Bolwidt

1, P. Jesse

1& R.M. Frings

2

1 Institute for Inland Water Management and Waste Water Treatment (RIZA), P.O. Box 9072, 6800 ED Arnhem, TheNetherlands; [email protected] Department of Physical Geography, Faculty of Geosciences, Utrecht University, P.O. Box 80115, 3508 TC Utrecht, TheNetherlands

AbstractTo determine the morphological behaviour ofbifurcation points two measurementcampaigns were set up and performed inJanuary 2004 and September 2004. A varietyof measurements was performed successfully,resulting in a unique database on bifurcationpoints. These first results are promising.

IntroductionThe goal of this study is to determine themorphological behaviour of bifurcation points.Bifurcation points play a key role in the waterand sediment movement of a river system. Inthe Netherlands there are three mainbifurcation points: (1) Pannerdensche Kop(BovenRijn divides into Pannerdensch Kanaaland Waal); (2) Merwedekop (BovenMerwededivides into BenedenMerwede and NieuweMerwede); (3) IJsselkop (Pannerdenschkanaal divides into IJssel and NederRijn).

Measurement campaignIn January 2004 measurements on subsoil,sediment transport, active layer, water level,discharge, flow velocity and direction, wereperformed with 10 ships during highdischarges at the IJsselkop and theMerwedekop. The campaigns lasted twoweeks. In Fig. 1 the water levels can be seentogether with the moments of measurements.Comparable measurements at thePannerdensche Kop had been performedearlier, in 1998. In September 2004measurements on sediment transport anddischarges were performed at the IJsselkopduring low discharges to determine the effectof the weirs in the NederRijn. With theMEDUSA technique (Koomans, 2004) thevariation of the natural radioactivity of thesediment was determined, by dragging asensor that measures the natural radioactivityof the sediments, over the river bed. There is arelationship between the radioactivity and thegrain size. Three surveys at differentdischarges were carried out to determine thechange in grain size patterns of the top layer.In this way, it was possible to detect sortingprocesses and bed roughness patterns.

Preliminary resultsThe data of the IJsselkop have been analysedfirst. In Fig. 2, the results of a MEDUSA scanof the grain size patterns after the peakdischarge can be seen. In the PannerdenschKanaal the dune height reacts quickly onincreasing discharge, with a maximum of ca 50cm, reached at the peak discharge. Migrationspeed decreases with increasing dune heightand vice versa. The highest and longest duneswere formed in the Pannerdensch Kanaal andthe NederRijn. More results are described inthe paper of Frings & Kleinhans in this volume.

ConclusionsA variety of measurements was performedsuccessfully and the measurement resultsconstitute a unique database on themorphological behaviour of bifurcation points.The results from the IJsselkop show a cleardifferences between the three branches, withrespect to subsoil, sediment transport, grainsize and dune characteristics. TheMerwedekop measurement results arepresently under study. Both studies will resultin an integral report on the different data. From the studies around thePannerdensche Kop a clear view of thedistribution of water and sediment on thebifurcation point Pannerdensche Kop hasbeen obtained. The data helped to gain moreinsight into the morphological behaviour of thisbifurcation point. Much of these data wereused in two dissertations on the behaviour ofsand dunes and sorting processes ofsediments.

Figure 1. The water levels during the measurementcampaigns at the IJsselkop and at the Merwedekop(Werkendam).



CooperationThis research has been performed under theauthority of Rijkswaterstaat (DZH and DON),in cooperation with Utrecht University, TNONITG, MEDUSA Explorations BV and theMorphological Triangle of NCR.

ReferenceKoomans, R., 2004. Hoogwater op de IJsselkop, eind

rapport, 2003p029R3. MEDUSA Explorations BV,Groningen

Figure 2. Spatial distribution of bed material grain sizes near the IJsselkopafter peak discharge. The main route of the sediment runs from thePannerdensch Kanaal to the NederRijn. During high discharges erosiontakes place in the IJssel channel. The finer material is washed awayleaving coarser material in the top layer.



Channel roughness in 1D steady uniform flow: Manning orChézy?F. Huthoff 1,2 & D. Augustijn 11 Department of Water Engineering and Management, Faculty of Engineering Technology, University of Twente, P.O. Box 217,7500 AE Enschede, The Netherlands; [email protected] HKV consultants, P.O. Box 2120, 8203 AC Lelystad, The Netherlands

AbstractIn river flow applications, consensus on themost appropriate roughness descriptor has yetto be found. A disturbing observation is thatthe coefficients and formulae of Chézy, DarcyWeisbach, Manning, Strickler and WhiteColebrook are used rather arbitrarily, and thata widely accepted scientific justification islacking. The presented paper compares themost commonly used roughness parameters,and reflects on some arguments that are oftenused in favour of, or against, any of these.Some recent advances on the theoretical basisof different methods are put forward, andimplications for commonly used hydraulicmodelling packages are discussed.

Commonly used roughnessrelationsFor open channels, three different formulaeare commonly used to describe the relationbetween the mean flow field and channelresistance in the steady uniform case. Forcompletion, the DarcyWeisbach equation isincluded, but will not be further reflected upon,because of its functional equivalence to theChézy equation. The dates mentioned belowfor the Chézy, DarcyWeisbach and Manningformulae are from a historical overview byRouse & Ince (1957); the WhiteColebrookequation was published by Colebrook (1939).

1. Chézy (1769): RiCU = , where)/12log(18= NkRC “WhiteColebrook

(1939)”

2. DarcyWeisbach (1840’s50’s):RifgU /8=

3. Manning (1889): RiRnU 6/1)/1(= ,

where 25/= 6/1Skn “Strickler (1923)”

Where kN is the Nikuradse (1933) roughnessheight in the WhiteColebrook equation(adapted for hydraulically rough flow), and kSthe roughness height by Strickler. Originally,

the equations above are all empirical incharacter, giving each of them validity in atleast some specific situations. Since they aresupposed to describe the mean flow velocity(U) as a function of roughness (C, n or f) andgeometry (hydraulic radius R, slope i), thequestion arises which formula gives thesimplest, yet most widely applicable,representation in a (natural) open channel?

Presently, there is no consensus on this matterand conflicting arguments remain to be heard.It is often argued that the Chézy equation (orDarcyWeisbach for that matter) has a soundtheoretical basis (see for derivation Jansen,1979), which would justify its wider range ofuse. However, Gioia & Bombardelli (2002)have shown that similarity considerations offlow in the hydraulically rough regime lead tothe Manning equation, where n is a measure ofthe absolute roughness height (as in Strickler’srelation). In this respect it is not yet clear whichtheoretical foundation is most reliable, leavingthe issue unresolved. In Fig. 1 the twoformulations are compared with respect tochanging relative depth. The figure shows thatfundamental differences between the twoapproaches exist. How this affectsperformance of hydraulic models is discussedin the following sections.

Figure 1. Comparing the Manning equation (roughnessheight as in Strickler) with the Chézy equation(roughness height as in WhiteColebrook): although thetwo formulae are clearly not equal, they show similarbehaviour with respect to increasing relative roughness(hydraulic radius R/ roughness height k) because of thedecreasing gradient of the fractionfunction.



Calibration in a simple channelAccording to their respective theoreticalfoundations, both the Chézy and the Manningequation are only expected to be valid insituations where sidewall effects are negligible.As a result, situations with wide channels(depth<<width of channel) have to beconsidered. Although Fig. 1 shows that bothdescriptions may differ significantly, still bothcan be made to fit data by allowing therespective roughness heights in eachdescription to vary independently (assumedifferent roughness height definitions).Consider for example a simple channel ofwidth 100 m where a discharge of 300 m3/sresults in a water depth of 2.5 m (slope i = 0.01%). In this case the Chézy equation (withWhiteColebrook) yields a roughness height ofkN = 1.8 mm, while the Manning equation (withStrickler) gives kS = 3.1 mm. If these calibratedformulae are subjected to an increaseddischarge of 20% (360 m3/s) a water leveldifference of 1.3 cm results (~3% of total waterlevel rise, see Fig. 2). As could be expected,Fig. 2 shows that the two formulations follow asimilar behaviour when made to fit each otherat a specific (calibration) point.

Calibration in a composite channelIn a composite channel the overall roughnessmay be computed by using a divided channelmethod (e.g. Chow, 1957; Jansen, 1979; Yen,2002). This method assumes that flow in eachof the subsections is independent from othersection and that the total discharge equals thesum of sectiondischarges. The approach maybe expected to be valid if the different sectionsare wide, such that possible lateral transfermechanisms at the interfaces can beneglected. In Chow (1957) an indicative valueof 10 for the relative width (as compared towater depth) is given for a channel to beconsidered wide. Consequently, for twostage

channel this argument is taken to count foreach of the subsections. Fig. 3 shows a twostage channel with overall flow conditions thatresemble the previously mentioned case in asimple channel: the overall hydraulic radius is2.5 m, the total width is 100 m and thedischarge is 300 m3/s.

In order to calibrate the composite roughnessequations on these specific conditions,roughness parameters for each of thesubsections have to be determined. In thecase considered here, the following roughnessvalues are made to correspond to flowconditions: Manning’s n = 0.05 s/m1/3 in thefloodplain and n= 0.017 s/m1/3 in the mainsection (corresponding to C=20.0 m1/2/s andC=72.5 m1/2/s, respectively). Note thatcalibration is not performed on roughnessheights but on resistance coefficients C and n,as is often the case in practice. In realsituations the roughness in the main sectionsmay be determined independently at low waterlevels (flow below bankfull height). If, again,discharge is increased to 360 m3/s (+20%),different water levels result from using twodifferent methods. In this case the water leveldifference turns out to be about 5 cm (~11% oftotal water level rise), significantly higher thanin the simple channel case.

ConclusionsFor both the Manning and the Chézy equation,followup work provided a theoreticalfoundation. However, between the two, thereremains a fundamental difference with regardto the dependence on the hydraulic radius (Cdepends on R, while n does not). In a simplechannel the methods show equivalentbehaviour when used for extrapolating beyondthe level of the highest calibration point. In thespecific case presented, a water leveldifference of 1.3 cm resulted after a 20%discharge increase. While theory shows thatthe Chézy parameter is a measure of relativeroughness height, its value is often treated as

Figure 2. Difference in predicted water levels(~hydraulic radius R) at 20% discharge increase whenboth the Manning and Chézy equation are calibrated onQ = 300 m3/s and R = 2.5 m (see text for chosengeometry).

Figure 3. Crosssection of a twostage channel (A =crosssection subarea) showing relevant parameters fordivided channel method. If the roughness parameters (n,C) in subsections are calibrated to fit a discharge of 300m3/s (at a slope of 0.01%), then an increased dischargepredicts different water levels between the 2 calibrationsets (one based on addition of Chézy formulae UC insubsections, the other on Manning Um).



a constant in methods to determine acomposite roughness in more complex rivergeometries. This method seems more justifiedwhen using the Manning roughness parametersince theory predicts that this parameter isindeed a true measure of (absolute) wallroughness. Calibration of both Chézy andManning on specific flow conditionsconsequently gives a much larger discrepancyin predicted water level than in the simplechannel case, where calibration is performedon roughness height and thus theory isfollowed more closely. In the compositechannel case presented, the discrepancyamounted to 5 cm (higher water level in Chézyapproach) at a 20% discharge increase(equivalent composite characteristics as in thesimple channel case).In conclusion, for composite roughnessmethods it is advised to apply the Manningapproach for roughness calibration, in order togive a theoretically correct weight to theroughness of each subsection.

ReferencesChow, V.T., 1959. Openchannel hydraulics. McGrawHill,

New York.Colebrook, C.F., 1939. Turbulent flow in pipes, with

particular reference to the transition region betweenthe smooth and rough pipe laws, Journal of theInstitute of Civil Engineers (London), 11, pp. 133156.

Gioia, G. & F.A. Bombardelli, 2002. Scaling and similarityin rough channel flows. Physical Review Letters88(1), pp. 14501/14.

Jansen, P.Ph., 1979. Principles of river engineering.(Facsimile edition 1994) Delftse UitgeversMaatschappij, Delft.

Nikuradse J., 1933. Strömungsgesetze in Rauhen Rohren,Forschungsheft 362, vol. B., VDI Verlag, Berlin.

Rouse, H. & S. Ince, 1957. History of hydraulics. DoverPublications, New York.

Strickler, A., 1923. Beitrage zur Frage derGeschwindigkeitsformel und der Rauhigkeitszahlenfür Ströme, Kanäle und geschlossene Leitungen.Mitteilungen des eidgenossischen Amtes fürWasserwirtschaft, 16, Bern.

Yen, B.C., 2002. Open channel flow resistance. Journal ofHydraulic Engineering 128(1), pp. 2039





3D float tracking: insitu floodplain roughness estimationM.W. StraatsmaDepartment of Physical Geography, Faculty of Geosciences, Utrecht University, P.O. Box 80115, 3508 TC Utrecht, TheNetherlands; [email protected]

Abstract3D float tracking is a new method to quantifyhydrodynamic roughness of submergedfloodplain vegetation. In this method a floatingtripod is released on the inundated floodplainand tracked from shore by a tachymeterleading to a highly detailed representation ofthe water surface elevation along the flowpath. Simultaneously, an Acoustic DopplerCurrent Profiler collects flow velocity profilesand water depth. Preliminary roughnessvalues, based on backwatercurve modellingand the 1D equation for nonuniform flow,range less than one order of magnitude.

IntroductionHydrodynamic roughness of submergedvegetation is an important parameter for riverflow models. Roughness values are mostlybased on flume experiments, carried out withlow water depths and high water surfaceslopes (Carollo et al., 2002; Wilson & Horrit,2002). These experiments do not represent thehydrodynamic conditions of lower Rhinefloodplains well. For roughness estimationfrom hydrodynamic parameters detailedinformation is needed on the water surfaceslope, water depth and depth averaged flowvelocity (Van Rijn, 1994). The local watersurface slope has always been a difficultparameter to measure due to the smalldifferences in water level. Therefore the aim ofthis research is to collect accurate and detailedfield measurements of water surface slope,water depth and depthaveraged flow velocityand to determine the hydrodynamic roughnessof submerged floodplain vegetation. This canbe done with a new method: 3D float tracking.

Materials and methodsMeasurements were taken with a tripodfloating on the inundated floodplain (Fig. 1). AnAcoustic Doppler Current Profiler (ADCP)mounted underneath the float measured (1)the float velocity using the bottom track optionplus (2) deviations from the float velocity in thewater column and (3) the water depth. Resultswere averaged over 5 seconds to decreasenoise. Positioning is done with a shorebasedtachymeter that automatically tracks thereflector on top of the float. Positioning wasdone with a 2 Hz frequency. Changes inposition gave the float velocity and the water

surface slope. This resulted in detailedinformation for roughness calculation.

ResultsIn January 2004 the method was tested for thefirst time on two inundated floodplains: (1) the‘Groene rivier’ in Arnhem (Fig. 2) and (2) afloodplain on the river Waal upstream ofZaltbommel. Figure 2 shows the float data ofone run in the ‘Groene rivier’ floodplain inArnhem. Table 1 shows the key hydrodynamiccharacteristics of the two floodplains. Thewater surface slope in Arnhem was measuredmore accurately than on the river Waal due tothe absence of waves. ADCP measurementsshowed a fair amount of noise. However,averaged flow velocity profiles were consistentfrom one run to the other.

Figure 1. 3D impression of the 3D float trackingmethod.

Figure 2. Float data of one run in the ‘Groene Rivier’floodplain in Arnhem. Open dots (Z) represent individualwaterlevel measurements.



Preliminary roughness values (Nikuradse’sk) in Arnhem were 5 cm ± 3 cm, based on thepredictorcorrector method and 8 cm ± 10 cm,based on the locally solved 1D equation fornonuniform flow. Roughness values for theWaal floodplain were not calculated due to thelarge amount of scatter in the local watersurface slopes.

Discussion and conclusions3D float tracking can supply spatiallydistributed water surface elevations withunprecedented detail. Together with the flowvelocity and waterdepth measurements of theADCP it generates all necessary data forroughness calculations. However, theaccuracy of the method is limited by waveactivity.

Derived roughness values are within one orderof magnitude. Further improvements areexpected from more suitable window sizes forspatial filtering of water surface slope and localflow accelerations. The method can also beextended to roughness measurements ofhedges, groynes or minor embankments.

ReferencesCarollo, F.G., V. Ferro & D. Termini, 2002. Flow velocity

measurements in vegetated channels. Journal ofHydraulic Engineering 128(7), pp. 664673.

Van Rijn, L.C., 1994. Principles of fluid flow and surfacewaves in rivers, estuaries, seas and oceans. Aquapublications, Amsterdam.

Wilson, C.A.M.E. & M.S. Horrit, 2002. Measuring flowresistance of submerged grass. HydrologicalProcesses 16, pp. 25892598.

Table 1. Key hydrodynamic characteristics of the two floodplains.Floodplain WSS (cm/km) Water depth (m) Flow velocity (m/s)Arnhem 5.9 1.2 0.41Waal 8.8 2.5 0.74

WSS = water surface slope



Variations in roughness predictions (flume experiments)D. Noordam 1, A. Blom 1, H. van der Klis 2 & S.J.M.H. Hulscher 11 Department of Water Engineering and Management, Faculty of Engineering Technology, University of Twente, P.O. Box 217,7500 AE Enschede, The Netherlands; [email protected] WL | Delft Hydraulics, P.O. Box 177, 2600 MH Delft, The Netherlands.

AbstractData of flume experiments with bed forms areused to analyze and compare differentroughness predictors. In this study, thehydraulic roughness consists of grainroughness and form roughness. We predict thegrain roughness by means of the size of thesediment. The form roughness is predicted bythree approaches: Van Rijn (1984), Vanoni &Hwang (1967) and Engelund (1966). The totalroughness values (friction factors) arecompared with the roughness valuesaccording to the DarcyWeisbach equation.Results show that the different methods predictdifferent friction factors. In future researchuncertainties in the hydraulic roughness will betaken into account to determine their influenceon the computed water levels.

IntroductionIn the Netherlands, the heights and strengthsof dikes and other flood defense systems arebased on computed water levels which occurduring a certain extreme discharge, i.e. thedesign discharge. The uncertainty in thehydraulic roughness of the river bed is one ofthe main sources of uncertainty in thesecomputed water levels (Van der Klis, 2003).The purpose of the present research is tocompare different stateoftheart roughnesspredictors and examine the influence of theroughness predictor on water levels. We usethe same approach as Julien et al. (2002). Theoverall aim of this study is to gain knowledgeon the size and type of uncertainties in thehydraulic roughness and their influence oncomputed water levels.

Material and methodsFlume experiments were conducted by Blom etal. (2003) in the sand flume facility at WL|DelftHydraulics in the Netherlands (19972000).The experiments were performed under steadyuniform flow conditions and sediment from theWaal River (near the Pannerdensche Kop)was used. The experiments were aimed atconditions with bed forms. Their heights ( )and lengths ( ) were measured, as well as thehydraulic radius (R), flow depth (h), flowvelocity (u) and the energy slope (i). We derive

the friction factors by means of two differentmethods. The first method gives the referencevalues. It uses flow data and the DarcyWeisbach equation:

28ugRif = (1)

The second method for calculating theroughness is using a roughness predictor. Inthese experiments the only sources ofroughness are grain roughness f ' (caused bythe protrusion of grains from the bed into theflow) and form roughness f '' (created by thepressure differences over bed forms). The sumof grain and form roughness gives the totalroughness. To calculate the grain roughnesswe distinguish between a roughness height (k's) of d90 and 3d90. The value of the roughnessheight can be converted to a value for f ' withthe following relation (Van Rijn, 1993):

2

'12log24.0'

−

=

skRf (2)

For calculating the form roughness we studythree models. For the Van Rijn (1984)approach (3), a value for f''R is obtained byapplying equation (2) (using k''s instead of k's).

−∆= Λ

∆−25

11.1'' ek s (3)

The other two models are the Vanoni & Hwang(1967) approach:

2

3.2log3.3''−

−

∆Λ

=Rf VH (4)

and the Engelund (1966) approach:

hE e

Rf

∆−

Λ∆

=5.22

10'' (5)

Results and preliminaryconclusionsFigure 1 shows some results of thecalculations. The experiments T5, T7, T9 andT10 were conducted under different flowconditions, i.e. different discharges, velocitiesand slopes. All roughness predictors yield alarger friction factor than the DarcyWeisbachreference value. From other calculations it



Figure 1. Friction factors,black = grain roughness,white = form roughness,green = DarcyWeisbach roughness

appears that a difference of 0.05 in the frictionfactor (f) can lead to a 20 cm change inhydraulic radius (R), and thus a significantchange in water levels. The results give a firstimpression of the uncertain hydraulicroughness and show that variations in frictionfactors influence calculated water levels.

Further researchPlans for future research are first to choose themost appropriate roughness predictor (basedon the flume experiments). Then, we want toinclude uncertainties and perform a MonteCarlo analysis to examine the influence of theuncertain hydraulic roughness on water levels.Furthermore, we will examine what the resultsof the flume experiments mean for fieldsituations.

AcknowledgementsThis work is supported by the DutchTechnology Foundation STW, the appliedscience division of NWO, and the technologyprogramme of the Dutch Ministry of EconomicAffairs.

ReferencesBlom, A., J.S. Ribberink & H.J. de Vriend, 2003. Vertical

sorting in bed forms: flume experiments with a naturaland a trimodal sediment mixture. Water ResourcesResearch 39 (2), 1025, doi: 10.1029/2001WR001088.

Engelund, F., 1966. Hydraulic resistance of alluvialstreams. Journal of the Hydraulics Division 98 (HY2),pp. 315326.

Julien, P.Y., G.J. Klaassen, W.B.M. ten Brinke & A.W.E.Wilbers, 2002. Case study: bed resistance of RhineRiver during 1998 flood. Journal of HydraulicEngineering 128 (12), pp. 10421050.

Van der Klis, H., 2003. Uncertainty analysis applied tonumerical models of river bed morphology. Ph.D.thesis, Delft University of Technology, Delft.

Van Rijn, L.C., 1984. Sediment transport, part III: bedforms and alluvial roughness. Journal of HydraulicEngineering 110 (12), pp. 17331754.

Van Rijn, L.C., 1993. Principles of sediment transport inrivers, estuaries and coastal seas. AQUAPublications, Amsterdam.

Vanoni, V.A. & L.S. Hwang, 1967. Relation between bedforms and friction in streams. Journal of theHydraulics Division 93 (HY3), pp. 121144



Effect of main channel roughness on water levelsA.J. Paarlberg 1, C.M. DohmenJanssen 1, S.J.M.H. Hulscher 1 & A.P.P. Termes 21 Department of Water Engineering and Management, Faculty of Engineering Technology, University of Twente, P.O. Box 217,7500 AE Enschede, The Netherlands; [email protected] HKV Consultants, P.O. Box 2120, 8203 AC Lelystad, The Netherlands

IntroductionAs a result of changing discharge duringfloods, dunes may develop in the main channelof a river (Fig. 1) leading to a changing bedroughness. This roughness is dynamic sincedune dimensions and roughness lag thedischarge. In current practice, hydraulicmodels are tuned using the roughness of themain channel as a calibration factor; the realdynamics of roughness are not taken intoaccount. The aim of our research is toformulate an appropriate model concept fordynamic roughness during floods. This paperpresents an analysis of the effect of mainchannel roughness on water levels. Not thevariation of dunes is considered, but only theeffect of variations of their roughness. It is alsoanalyzed how the influence of main channelroughness depends on the geometry of thechannel.

MethodThe sensitivity analysis is performed inanalogy to Huthoff & Augustijn (2004), usingboth a numerical and an analytical approach.Floodplain roughness is set constant at k =0.25 m (Van Velzen, 2003). Calculations areperformed for a schematized channel of 100km as shown in Figure 2a. In the analytical approach, for simplicity nobackwater effects are included. The compositechannel roughness (Ccomp) can be expressedas:

( ) ( ) fmTcomp ChCDC ⋅⋅−+⋅⋅= 23

23

1 αα

Varied parameters are (Fig. 2): discharge: Q =500 – 10500 m3/s; depth of the main channel:DT = 3 – 5 m; ratio of main channel width to thetotal width of the crosssection: = Wm/WT =0.2 – 1; main channel bed roughness: k = 0.05– 0.65 m, the latter is the maximum valueobserved during a large flood in the Rhine in1995 (Julien et al., 2002); length of the areaover which the roughness is changed (in thenumerical approach): L = 0 – 50 km.

ResultsResults are presented by comparing with areference case in which the roughness heightk of the main channel is 0.35 m. As expected,an increased main channel roughness leads toincreased water levels (Fig. 3). At lower waterlevels (lower discharges), the influence of bedroughness is larger. For a floodplain width of400 m ( = 0.4 in this case) and a discharge of10500 m3/s (DT 9.9 m), the influence of mainchannel roughness on absolute waterleveldifference varies between 80 and +30 cm.The influence of a changing main channelroughness on the relative water level

Figure 1. Dunes at Pannerdensche Kop (Wilbers,2004).

Figure 2. Model schematization; (a) longitudinal, (b)crosssection.

Figure 3. Comparison of analytical (dashed) andnumerical (solid) computations for two differentdischarges. Waterlevel differences are scaled with thetotal depth in the reference case (i.e. km = 0.35 m).



increases, when the ratio of main channelwidth to total width ( ) increases (Fig. 4), sincemore water flows through the main channel.For a floodplain width of 400 meter, theinfluence of main channel roughness on thewater level varies between +40 and 100 cm.

Conclusions• The influence of main channel roughness

on absolute water levels is significant, andcan be up to 100 cm for a typical Dutchsituation ( = 0.4).

• The influence of main channel roughnesson relative waterlevel difference (dh/DT)decreases for increasing discharge andfloodplain width.

• Dependence on main channel depth issmall (not shown).

The numerical computations will be comparedwith 2D WAQUA computations in futureresearch.

AcknowledgementsThis work is supported by the DutchTechnology Foundation STW, the appliedscience division of NWO, and the technologyprogramme of the Dutch Ministry of EconomicAffairs.

ReferencesHuthoff, F. & D. Augustijn, 2004. Sensitivity analysis of

floodplain roughness in 1D flow. In: S.Y. Liong, K.K.Phoon & V. Babovic (eds.), Proceedings of the 6th

International Conference on Hydroinformatics, WorldScientific Publishing Company, Singapore, pp. 301308.

Julien, P.Y., G.J. Klaassen, W.B.M. ten Brinke & A.W.E.Wilbers, 2002. Case study: bed resistance of RhineRiver during 1998 flood. Journal of HydraulicEngineering 128 (12), pp. 10421050.

Van Velzen, E.H., P. Jesse, P. Cornelissen & H. Coops,2003. Stromingsweerstand vegetatie in uiterwaarden;handboek versie 12003, RIZA report 2003.028/029,RIZA, Arnhem.

Wilbers, A., 2004. The development and hydraulicroughness of subaqueous dunes. NetherlandsGeographical Studies 323, KNAG/FaculteitGeowetenschappen, Universiteit Utrecht, Utrecht,227 p.

Figure 4. Dependence of relative waterlevel differenceon main channel roughness and on ratio of mainchannel width to total width.



Effect of climate change on bedforms in the Rhine andconsequences for navigationL. Haitel, C.M. DohmenJanssen, S.J.M.H. Hulscher & A. BlomDepartment of Water Engineering and Management, Faculty of Engineering Technology, University of Twente, P.O. Box 217,7500 AE Enschede, The Netherlands; [email protected]

AbstractNavigation on the river Rhine is of greateconomic importance for the Netherlands. Lowriver discharges or the presence of river duneson the bed may restrict the water depthavailable for navigation. River dunes arebedforms that develop at high discharges, as aresult of the interaction between flow andsediment transport. Dunes might hindernavigation as their development shows adelayed response to changing flow conditions,because it takes time for a dune to form or todegrade. This means that the maximum duneheight is reached when the water depth isalready decreasing. Therefore, it is importantto know if river dunes will restrict the waterdepth significantly and whether climate changeinfluences the development of river dunes inthe Rhine. From the research it can beconcluded that dunes do not significantlyinfluence the hindrance of navigation, neithernow, nor in the future.

IntroductionMany events, like more frequent flooding andextreme drought are addressed to a risingaverage global temperature. Several scenarioshave been developed to forecast the possibleconsequences of this type of climate changefor the Rhine basin. One of them is the UKHIscenario that is used to determine theexpected discharges in the Rhine (Middelkoopet al., 2001). This scenario resulted in anincreasing discharge in the Rhine in winter anda decreasing discharge in summer. As thedevelopment of river dunes depends on thedischarge, it is important to know the effect onthe dune development, in order to give insightin the effect of dunes on navigation. Former research on river dunes in theRhine showed that dunes of about 0.8 m arepresent at a discharge of 7,000 m3/s. Riverdunes reach a height of about 1.6 m at anextreme discharge of 12,000 m3/s (Wilbers,2004). As a result of climate change, thesedune heights might increase.Next to that, erosion and sedimentationprocesses (Fig. 1), which play an importantrole in the dune development, take time.Consequently, river dune development

responds to changing flow conditions with atime lag. This means that the maximum duneheight is reached a few days after the peakdischarge, when the water depth is alreadydecreasing. This time lag could be relevant inthe hindrance of navigation by river dunes.

MethodIn order to determine the effect of climatechange on river dune development, the climatechange scenario UKHI (United KingdomMeteorological Office High Resolution GeneralCirculation Model) was used. This scenarioresults in factors that can be multiplied withrecorded discharges. Recorded discharges ofthree different years (‘base years’) withdifferent discharge patterns and peakdischarges were selected. Recorded andpredicted discharges are used as input for a1Dhydraulic model for unsteady flows, tocompute corresponding water depths. Thesewater depths were used in the calculation ofdune heights, with the method of Wilbers(2004). This method is based on the ideas ofAllen (1976) and has been specificallydeveloped for the Rhine branches. In order toquantify the hindrance of navigation, theloading capacity of vessels is calculated basedon the calculated water depths. This is donefor the water depth with dunes and withoutdunes present.

ResultsClimate change has a significant effect on thedevelopment of river dunes. Figure 2 presentsthe development of dune height in January forbase year 2003 and for the expected dune

Figure 1. Process of erosion and sedimentation (afterKnighton, 1998)



heights according to the UKHI scenarios for2050 and 2100. A dune height of 1.2 m iscalculated for a peak discharge of 9,000 m3/sin the base year.The UKHI 2100 scenario leads to a peakdischarge of 14,000 m3/s and thecorresponding maximum dune height is 1.7 m.

The time lag is clearly perceptible in the figure:the maximum dune height is reached later,while the peak discharge occurs on the sameday. The maximum water depth at peakdischarge is 14 m for the UKHI 2100 scenario.When the dune height is at its maximum, thewater depth has decreased to 12 m, which isstill large compared to the dune height of 1.7m. Despite the delayed response of riverdunes to changing discharges, the dischargedoes not decrease so quickly that the duneheight becomes substantial compared to thewater depth: after approximately one week thedune height is about 0.9 m, while the waterdepth is still about 9 m. It can be concludedthat river dunes do not restrict the water depthfor navigation during winter, because a lowdischarge does not occur very fast after a peakdischarge. In other words: dunes get enoughtime to decay. Figure 3 presents the loading capacity ofvessels in time to get a better insight in thehindrance of navigation. It is clear thatnavigation is restricted by low flows in summer,when the discharge is about 900 m3/s. Theeffect of climate change on the loadingcapacity is clearly perceptible; in Septemberthe loading capacity decreases from about75% to about 55%. The influence of riverdunes, however, appears to be insignificant:river dunes are calculated to be about 0.05 mhigh during summer. The restriction of theloading capacity by river dunes is only in theorder of 1% (in Fig. 3 ‘with dunes’ and ‘without

dunes’) for the base year (1991) as well as forthe UKHI 2100 scenario. Although climatechange is expected to lead to higher riverdunes, these higher river dunes do not furtherdecrease the loading capacity of vessels.

ConclusionsIt can be concluded that river dunedevelopment in the Rhine is stronglyinfluenced by changing discharges, as a resultof climate change. Higher discharges causehigher dunes and a larger time lag betweenpeak discharge and maximum dune height. Climate change does influence thehindrance of navigation during summer, due toa decrease of low discharges. However, duneheight compared to water depth will alwaysremain such that river dunes have nosignificant influence on the hindrance ofnavigation and this effect is not enlarged byclimate change.

ReferencesAllen, J., 1976. Bed forms and unsteady processes: some

concepts of classification and response illustrated bycommon oneway types. Earth Surface Processes 1,pp. 361374.

Knighton, A.D., 1998. Fluvial forms and processes; a newperspective, Arnold, London, 383 p.

Middelkoop, H., M. van Asselt, G. Können, S. van’tKlooster, M. Haasnoot, W.van Deursen, N. vanGemert, J. Kwadijk, H. Buiteveld, P. Valkering, J.Rotmans, 2001. Integrated water managementstrategies for the Rhine and Meuse basins in achanging environment. National ResearchProgramme on Global Air Pollution and ClimateChange (NRP), Bilthoven.

Wilbers, A., 2004. The development and hydraulicroughness of subaqueous dunes. NetherlandsGeographical Studies 323, KNAG/FaculteitGeowetenschappen, Universiteit Utrecht, Utrecht,227 p

Figure 2. River dune development in base year 2003 andfor the UKHI scenarios (vertical line represents the day ofthe peak discharge).

Figure 3. The effect of climate change and dune heighton loading capacity.



Navigability of the Niederrhein and Waal in theNetherlands; a stochastic approachS. van Vuren 1 & H.J. Barneveld 21 Section of Hydraulic Engineering, Faculty of Civil Engineering and Geosciences, Delft University of Technology, P.O. Box5048, 2600 GA Delft, The Netherlands; [email protected] HKV consultants, P.O. Box 2120, 8203 AC Lelystad, The Netherlands

IntroductionHalf of the cargo transport between the port ofRotterdam and Germany goes via the Rhine.Safe, efficient and profitable inland shippingrequires a deep and wide navigation channel,now and in the future. Navigability depends onmorphological and hydraulic conditions in theriver that exhibit spatial and temporalvariations, such as bed level and water levels.Whenever the navigation depth is less thanrequired, navigation is congested and/or shipsmay carry less cargo.

MethodsA 1dimensional morphodynamic SOBEKmodel of the Rhine (Jesse & Kroekenstoel,2001) is used for waterdepth predictions.These predictions, in combination with atheoretical correction for the transversal slopein river bends, are used to assess thenavigability of the Niederrhein and Waal forships of various drafts and make it possible toindicate nautical bottlenecks (the reach of theNiederrhein that is situated within theNetherlands is also known as the ‘BovenRijn’). The Rhine model is affected by variousuncertainties, including the modelschematisations, the specification of the modelinput (for example boundary conditions, initialconditions) and the model parameters. Van derKlis (2003) and Van Vuren et al. (2002) haveshown that the future discharge hydrograph isone of the important sources of uncertainty.

Monte Carlo simulation, applied to themodel, is utilised to quantify the uncertainty inthe water depth and thus navigability, given anuncertain river discharge. Monte Carlosimulation (Hammersly & Handscomb, 1964)involves a large number of model runs 400runs with statistically equivalent inputs. Foreach run a discharge time series of 10yearsduration is randomly generated using theBootstrapresampling technique (Efron, 1982).The outputs of these model runs, can beexpressed in terms of expected developmentand uncertainty of the navigability, whichprovides insight into the stochasticcharacteristics of the river's navigability.

The National Traffic and TransportationPlan gives guidelines with respect to thenavigation channel requirements. According tothis plan, during discharges above a thresholdvalue of 1020 m3/s at Lobith (where the Rhineenters the Netherlands), the navigationchannel in the Niederrhein and the Waal musthave a guaranteed width of 170 m and a depthof 2.8 m, respectively. This threshold value isexceeded during 95 % of the time. Theprobability that these requirements aresatisfied in the Dutch part of the Rhine for aperiod of 10 years is evaluated in this paper.The largest cargo ships in the Rhine have adraft of approximately 4 m.

Figure 1. Percentage of navigable time as a function ofthe ship draft for the Niederrhein and Waal in the periodbetween 2002 and 2012. Narrow bars representindividual simulations, whereas thick bars in other graphrepresent aggregates of all simulations.



Additionally, the navigability is assessed fordrafts ranging from 1.5 to 5 m.

ResultsThe navigability of the Niederrhein and theWaal for ships with a draft between 1.5 and 5m is statistically assessed on the basis of 400model runs. Each model run driven by one ofthe synthesised discharge time series resultsin one possible future morphological evolution.Figure 1 (left diagram) shows that thenavigable percentage in the 10year periodbetween 2002 and 2012 differs for eachsynthesised discharge time series. Using theresults of all model simulations, the statisticalcharacteristics of the navigable percentage arederived (Fig. 1; right diagram). For example,the percentage of navigable time for ships witha draft of 3 m at the Niederrhein and the Waal,averaged over a period of 10 years, is 84%.Figure 1 (right diagram) also shows that forthis draft there is a 90% probability that thepercentage of navigable time lies between78% and 91%.

The probability of not fulfilling thenavigation channel requirements of theNational Traffic and Transportation Plan forships at draft 2.8 m is of interest of both theriver manager and the users of the inlandwaterway. This indicates how well the rivermanager maintains the required navigationcondition. Figure 2 shows the cumulativedistribution function of the least availablenavigation depth over a period of 10 years forall simulations individually and the aggregateline of all simulations together. The figureindicates the probability of meeting therequirements, i.e. ships can navigate at draft ofat least 2.8 m, is 89.1 % (100 % minus 10.9 %). The figure shows that this percentage of

navigable time is at maximum 96.6 % and atminimum 77.9 % for the 400 model simulationsconsidered herein.

Figure 3 illustrates the percentage ofnavigable time as a function of the riverlocation for a draft of 2.8 and 4 m. Thisprovides insight into which locations are criticalto the navigability of the river. It appears thatlocations with strong spatial changes ingeometry, such as the bifurcationPannerdense Kop (km 867), the river bedprotection in the river bend near Nijmegen (km882 885), the variation in floodplain width inthe MiddenWaal and the sharp bend atsection Pannerdense KopNijmegen andSt.Andries, may evolve into navigationbottlenecks. Most of these bottlenecks becomemanifest in the dry period.

Conclusion and future researchThe foregoing showed that the water depth inthe Niederrhein and the Waal exhibit a strongspatial and temporal variation. The uncertaintyin the future discharge hydrograph, incombination with strong spatial variations inthe river geometry, leads to significantuncertainties in the predicted response. Somelocations could develop into nauticalbottlenecks, the removal of which may involvehigh costs.

The stochastic method that is proposed inthis paper could be used to assess the impactof various human intervention measures on theriver’s navigability. In addition to the expectedimpacts, the change in maintenancerequirements (and so costs) can be assessed.This is subject of further research. We haveused a 1dimenional model in this study. This

Figure 2. Cumulative probability distribution of theleast available depth for all simulations individually(black lines) and the aggregate line (white line)together.

Figure 3. Percentage of navigable time over a periodof 10 years as a function of the river location for shipsof 2.8 m and 4 m draft.



means that the impact of twodimensionalfeatures, such as alternate bars andtransverse bed slopes in bends, are notconsidered. Neither is the fact that largefloodplain areas along the Rhine branches arelocated alternately at the right and the left sideof the river, which under flood conditions maylead to strong 3D crossflows over the mainchannel. It is therefore recommended to repeatthis kind of analysis with a model capable ofdescribing these phenomena.

ReferencesEfron, B., 1982. The Jackknife, the Bootstrap, and other

resampling plans. Society for Industrial and AppliedMathematics. Philadelphia, 92 p.

Hammersly, J.M. & D.C. Handscomb, 1964. Monte Carlomethods. Methuen, Londen.

Jesse, P. & D.F. Kroekenstoel, 2001. 1D Morfologischmodel Rijntakken. RIZA rapport 2001.040, RIZA,Arnhem.

Van der Klis, H., 2003. Uncertainty analysis applied tonumerical models of river bed Morphology. Ph.D.thesis, Delft University of Technology, Delft.

Van Vuren, S., H. van der Klis, & H. de Vriend, 2002.Largescale floodplain lowering along the River Waal:a stochastic prediction of morphological impacts. In:D. Bousmar & Y. Zech. (eds.), Proceedings of RiverFlow 2002; volume 2, Balkema, Rotterdam, pp. 903912.





Stochastic modelling of twodimensional river morphologyH. van der Klis & H.R.A. JagersWL | Delft Hydraulics, P.O. Box 177, 2600 MH Delft, The Netherlands; [email protected]

AbstractUncertainty in the results of a 2D rivermorphological model of the Upper Rhine dueto the uncertainty in the river discharge isestimated by applying a Monte Carlosimulation. As a possible more efficientalternative, the applicability of the Quasi MonteCarlo method is studied. First results of theconvergence rate of statistical quantities of theriver bed are promising. Based on theseresults, further research on this subject hasbeen planned.

IntroductionThe importance of knowledge of uncertaintiesin model results is more and more recognisedby river managers. The quantification of theseuncertainties, however, is often difficult. In thisresearch we focus on a specific aspect of thisproblem, applied to a 2D river morphologicalmodel: quantifying the uncertainty in the modelresults due to uncertain inputs. Thus, we leaveuncertainties in the model structure and themodelling context out of consideration.

A (nonlinear) river morphological modelgenerally requires a Monte Carlolike approachto estimate the uncertainty in the model resultsdue to uncertain model input (Van der Klis,2003). Such an approach is robust and muchstatistical information can be obtained from theresults. An important disadvantage of themethod, however, is the number of modelsimulations required. In case of the large,computational intensive models we talk about,a standard Monte Carlo (MC) simulation ispractically impossible. Therefore, we searchfor alternative Monte Carlo approaches which

lead to drastic decrease in the requirednumber of simulations. As a case study, we choose an existing 2D river morphological model, namely a Delft3Dmodel of a reach of the Upper Rhine (Baur etal., 2002). The model schematisation includesfloodplains such that their effect during highdischarges is represented in the model (Fig.1). The model area has a length of 42 km.

A Monte Carlo simulationA MC simulation consists of a large number ofdeterministic simulations, of which theuncertain model input is randomly generatedaccording to prescribed probabilitydistributions. The output values constituterandom samples from the probabilitydistribution of the output. Standard statisticaltechniques can be used to estimate thestatistical properties of the model output andthe precision of the output distribution. We carried out 100 simulations, eachforced at the upper boundary by a randomlydrawn 3years discharge series. In order tosample discharge series, we derived astatistical description of the Rhine discharge atLobith. We based the statistical description ofthe Rhine discharge on daily discharge datafrom 1946 to 2000 following the methodpreviously applied by Van Vuren et al. (2002).

The MC simulation results in an estimate of theeffect of the uncertainty in the river dischargeon the river bed changes in the model. Figure2 shows an example of the information that

Figure 1. Overview of the morphological Delft3D model.

Figure 2. The 90% confidence interval of the shippingwidth (area between the two lines) based on a minimumwater depth of 2.8 m during OLRconditions.



can be obtained: the 90% confidence intervalof the shipping width in a part of the modelledriver reach. This type of information can help ariver manager to decide where problems mightoccur for shipping.

An alternative: Quasi Monte CarloThe sampling method used in a standard MCsimulation is rather inefficient: in order todouble the accuracy of the estimate of theoutput uncertainty, four times the number ofsamples is required. Many alternative samplingmethods have been developed to improve theefficiency of the Monte Carlo simulation. In thisstudy, we examine the applicability of the socalled Quasi Monte Carlo (QMC) approach toour case study.

The QMC method is the deterministicversion of the standard Monte Carlo method, inthe sense that the random samples in an MCsimulation are replaced by wellchosendeterministic points. Various methods havebeen developed to create sequences of thosedeterministic points. Often applied is the Sobolsequence, or LPτ sequence. Homma & Saltelli(1995) compared the efficiency of varioussampling methods and showed an evidentadvantage of using the LPτ sequence.

This gives us sufficient ground for examiningthe applicability of the method to a rivermorphological model. To test QMC to our case study, wesimplified the description of the dischargeseries following Chapter 4 in Van der Klis(2003). With this simplified description weperformed a QMC simulation and, forcomparison, a standard MC simulation. Figure3 illustrates the convergence rate of each ofthese methods. For two locations within themain channel of the modelled river reach theconvergence of the standard deviation of thebed level is shown. This figure shows arelatively fast convergence of the QMC results.

Conclusions and further researchThe first results presented here are promisingenough to further explore the possibilities ofQMC in our model. In this, we will try to applythe method to more advanced descriptions ofthe river discharge. Furthermore, we will studywhether QMC is applicable to other uncertainmodel parameters.

AcknowledgementThe model of the Upper Rhine and thedischarge measurements have been madeavailable by RWS, DON and RIZA.

ReferencesBaur, T., H. Havinga & D. Abel, 2002. Internationale

Zusammenarbeit bei derPlanung von Regulierungsmaßnahmen amNiederrhein: Durchführungflussmorphologischer Simulationen, HANSAInternational Maritime Journal10/2002, pp. 5156.

Homma, T. & A. Saltelli, 1995. Use of Sobol’sQuasirandom sequence generator for integration ofmodified uncertainty importance measure. Journal ofNuclear Science and Technology 32(11), pp. 11641173.

Van der Klis, H., 2003. Uncertainty analysis applied tonumerical models of river bed morphology. Ph.D.thesis, Delft University of Technology, Delft.

Van Vuren, S., H. van der Klis & H.J. de Vriend, 2002.Largescale floodplain lowering along the river Waal:a stochastic prediction of morphological impacts. In:D. Bousmar & Y. Zech (eds.), Proceedings of RiverFlow 2002, Volume 2, Balkema, Rotterdam, pp. 903–912.

Figure 3. Convergence rates of the standard deviationof the river bed level at two locations within the mainchannel, both for the random MC simulation and thequasi MC simulation.



Sediment density stratification and river channel patternsin the lower Yellow River, ChinaD.S. van MarenSection of Hydraulic Engineering, Faculty of Civil Engineering and Geosciences, Delft University of Technology, P.O. Box 5048,2600 GA Delft, The Netherlands; [email protected]

AbstractThe lower Yellow River is characterised by abraiding channel pattern, which changes into ameandering pattern in the downstreamdirection. Although the bedlevel gradient isimportant for this transition, the sedimentconcentration plays an additional role. At lowsediment concentrations the flow is subsaturated, leading to a meandering pattern. Athigher concentrations the turbulent structure ofthe flow is suppressed and the sedimentconcentration profile collapses, leading to abraiding channel pattern. At even highersediment concentrations, sediment is held insuspension by hindered settling, and thechannel pattern becomes increasinglymeandering.

IntroductionThe Chinese Yellow River carries hugeamounts of suspended sediment, especiallyduring hyperconcentrated floods whensediment concentrations exceed 100’s kg/m3.A large part of this sediment load is depositedin the lower Yellow River, leading to a rapidlyrising floodplain and therefore increasing floodrisks. In order to manage these siltationproblems a 3D morphodynamic model is beingdeveloped for the lower Yellow River withinDelft3D.However, the sediment concentration in theYellow River is so high that sediment densitystratification is important for sediment transport

mechanisms and river morphology. Therelationship between these stratificationprocesses and river channel patterns will bedescribed here shortly.

Vertical sediment densitystratificationAt low sediment concentrations, the downwardmotion of sediment particles is balanced by anet upward transport of sediment by turbulentmotions, resulting in a typical Rouse sedimentconcentration profile. However, at highconcentrations (order 1’s to 10’s kg/m3,depending on the grain size) these turbulentmotions are suppressed by the sedimentconcentration gradient (Winterwerp, 2001).Therefore the turbulent motions are no longerable to hold sediment in suspension and at acritical sediment concentration, the sedimentconcentration profile collapses: the flowchanges from subsaturated into supersaturated flow. At even higher sedimentconcentrations (order 10’s to 100’s kg/m3),sediment is additionally held in suspension byhindered settling. Therefore the sedimentconcentration profile is reestablished eventhough turbulence is low: subsaturatedhyperconcentrated flow (Winterwerp et al.,2003). This collapse and buildup of thevertical sediment concentration profile can besimulated with a 1DV version of Delft3D modelthat includes sediment density effects (Fig. 1).The sediment concentration at which thesediment concentration profile collapses or isreestablished is the saturation concentration.This saturation concentration initially increaseswith the (dimensionless) flow strength [u3/ghw0in which: u = depthaveraged velocity (m/s), g= gravitational acceleration (m/s2), h = waterdepth (m), w0 = sediment settling velocity inclear water (m/s)], but decreases whenhindered settling effects become important.This means that at very high sedimentconcentrations, the flow strength required tokeep sediment in suspension is low.Numerically modelled relations between flowstrength and saturation concentration matchessediment concentrations observed in theYellow River (Fig. 2).

Figure 1. Evolution of vertical sediment concentrationprofile during subsaturated (C = 30 kg/m3),supersaturated (C = 90 kg/m3), and hyperconcentratedsubsaturated (C = 500 kg/m3) flow conditions. Theflow velocity is 2 m/s and the grain size 40 microns.



River channel patternsThe upper part of the Yellow River is braiding,but changes into a meandering pattern in thedownstream direction. And although thistransition mainly results from a decreasing bedlevel gradient, field observations show that theYellow River becomes increasinglymeandering at concentrations below 30 kg/m3

and above 200 kg/m3, and braiding atintermediate concentrations (Xu, 2004).However, the reasons for this behaviour arenot yet fully understood. The supersaturatedflow conditions described in the previoussection are characterised by deposition,whereas the subsaturated flow conditions arecharacterised by bed erosion. This stronglysuggests that the vertical sediment densityeffects discussed in the previous section areimportant for the morphology of the YellowRiver. To verify this, 3D modelling experimentswere carried out to identify the effect ofhyperconcentration on the development ofriver morphology. Starting with an initially flatbut randomly perturbed bed, a braiding riverchannel develops during relatively lowsediment concentrations (upper panel in Fig.3).

However, when an additional wash loadfraction of 100 kg/m3 is included(hyperconcentrated flow), a meanderingchannel pattern begins to develop (lower panelin Fig. 3).

ConclusionsThe Yellow River channel pattern partlydepends on the sediment concentration with ameandering pattern at low and high sedimentconcentrations, but a braiding pattern atintermediate concentrations. This is caused bysediment density stratification, and can benumerically simulated with Delft3D.

ReferencesWinterwerp, J.C., 2001. Stratification effects by cohesive

and noncohesive sediment. Journal of GeophysicalResearch 106, pp. 22.55922.574.

Winterwerp, J.C., H.J. de Vriend & Z.B. Wang, 2003. Fluidsediment interactions in siltladen flow. Proceedingsof the First International Yellow River Forum on RiverBasin Management, Zhengzhou, China, 2124October 2003, Vol. II. Yellow River ConservancyPublishing House, Zhengzhou, pp. 351362.

Xu, J., 2004. Double thresholds in scourfill processes andsome implications in channel adjustment.Geomorphology 157, pp. 321330.

Figure 2. Relationship between dimensionless flowstrength u3/ghw0 (see text for explanation of symbols)and saturation concentration Cs in the Yellow river (leftpanel) and based on 1DV results (right panel). Flowvelocities during the 1DV simulations ranges from 0.5 to3 m/s, and the D50 between 20 and 40 microns. Figure 3. Channel development after approximately one

year in an initially flat bed using a slope of 2.2 · 104, adischarge of 5000 m3/s, and sediment with D50 = 65microns. Xaxis and Yaxis are in km and the water depthin m. The simulations in the upper panel are without awash load fraction whereas a wash load fraction withC=100 kg/m3 and D50 = 5 microns is included in the lowerpanel. Although this wash load fraction does not interactwith the river bed, it results in hyperconcentrated flow andconsequently a strongly incising meandering channel.



Causal relationships between climate change and naturalriver behaviour in the Rhine delta during the last 15,000yearsG. Erkens, W.Z. Hoek & E.A. KosterThe Netherlands Centre for GeoEcological research (ICG), Department of Physical Geography, Faculty of Geosciences,Utrecht University, P.O. Box 80.115, 3508 TC Utrecht, The Netherlands; [email protected]

AbstractThe Rhine delta in the Netherlands developedduring the last 15,000 years under influence oftectonics, sealevel rise and, most importantly,the sediment flux from the hinterland. Changesin this sediment flux since the last GlacialInterglacial transition, appear to be mainlyrelated to changes in climate, land use andvegetation in the upstream part of thecatchment, causing variations in sedimentdelivery from the German part of the basin(e.g. Berendsen et al., 1995, Vandenberghe,1995). This Ph.D.project (20042007) focuseson the relationship between upstreamsediment delivery in the Rhine drainage basinand downstream sedimentation in the delta asa result of vegetation changes since the end ofthe Weichselian.

IntroductionNatural river behaviour, i.e. sedimentationdynamics and fluvial style, in a delta area iscontrolled by several factors (Fig. 1). Allogenicinfluences on fluvial systems at drainage basinscale and on the time scales of 1,000 to100,000 years are tectonics, climate and sealevelchange. Although tectonics and sea leveldo have large effects on fluvial systems suchas the river Rhine, climate (temperature andprecipitation) ultimately controls river dischargeand sediment supply and, thereby, thedynamics and size of sediment fluxes in afluvial system. Beside climate change, landuse changes have a direct impact on thevegetation cover and hence, subsoil cohesionand effective runoff. All this results in variationsin discharge and sediment load of the riverRhine in time, and eventually, in variations indownstream sedimentation dynamics andfluvial style. Although a relationship between naturalchanges in climate, vegetation, and sedimentdelivery in the upstream area and theassociated sedimentation downstream in thedelta is obvious, this relationship has neverbeen quantified.With the current amount of data present in thewhole drainage basin, it is now timely to makethe link between upstream erosional phases

and downstream sedimentation in order toobtain a better insight in the evolution of thedelta and to determine causal relationshipsbetween climate change and natural riverbehaviour.

ApproachThe Netherlands is in the unique position ofhaving a nearcomplete capture of the Rhinesediment flux within the RhineMeuse delta forthe later part of the Holocene (Berendsen &Stouthamer, 2001). Decades ofpalaeogeographic research makes the Rhinedelta now one of the beststudied deltas in theworld. This research enabled a detailedpalaeogeographic reconstruction of the DutchRhineMeuse delta during the last 15,000years (Berendsen & Stouthamer, 2001). Withthe data from the extensive RhineMeuse deltadatabase, three large northsouth sections(length: ca. 20 50 km) were constructed.Together with two already existing sections(Törnqvist, 1993; Cohen, 2003) andcomplementary data, these sections will beused to obtain a detailed (100 m core spacing)Holocene stratigraphy, showing distinctdifferences in accumulation rates. A largenumber of radiocarbon dates provide time linesin these sections. Subsequently, depositedvolumes of Rhine sediment during the last15,000 years in the Rhine delta can be

Figure 1. Factors controlling sedimentationdynamics and fluvial style in a delta area



calculated per time slice, providingdepositional rates. During the last 15,000 years, climaticchange induced largescale changes invegetation in the German part of the Rhinedrainage basin. Several regional Germanstudies show a distinct human impact on thelandscape from 6400 cal yrs BP onward,leading to vegetation change and enhancedsoil erosion (e.g. Lang et al., 2003). Becausethere is a strong link between vegetation andfluvial dynamics (e.g. Dambeck & Thiemeyer,2002), vegetation changes are likely to havecaused variations in upstream sedimentdelivery in the Rhine drainage basin. Phasesof (increased) siliciclastic input in the Germanpart of the basin will be determined andcorrelated with phases of downstreamdeposition, in the Rhine delta. It is expectedthat changes in sediment discharge during thelast 15,000 years can be linked to climatechange and, for the last 5000 years, increasedhuman impact.

Perspectives and relevanceChanges in landuse and climate are likely toaffect rivers and their catchments during thenext centuries, altering flows of water andsediment, which will have a large impact onthe Rhine delta in the Netherlands. For theriver Rhine, modelling results suggest a morefrequent occurrence of abnormal low and highdischarges in the near future. In addition,suspended sediment concentrations in theriver are expected to increase due to anincreased production of sediment by soilerosion (Middelkoop, 1997). These future changes will besuperimposed on changes triggered in thepast. Therefore, a better understanding of(past) fluvial responses to landuse andclimate change is needed. Until now, mostresearch has been conducted in smallcatchments, while the response of largecatchments has a more comprehensiveimpact, which is especially relevant for theNetherlands.

The time period over which a large catchmentresponds to landuse or climate change ismuch longer than for small catchments andmuch longer than most instrumental timeseries. This means that palaeoenvironmentalreconstructions are essential, because onlythese reconstructions provide a record ofcatchment responses on century to millenniatime scales. Quantifying the link between upstreamerosional phases and downstreamsedimentation in the Rhine drainage basin, willimprove our insight in the evolution of thedelta, including the causal relationshipsbetween (future) climate and landuse changesand natural river behaviour.

ReferencesBerendsen, H., W. Hoek & E. Schorn, 1995. Late

Weichselian and Holocene river channel changes ofthe rivers Rhine and Meuse in the Netherlands (Landvan Maas en Waal). PaläoklimaforschungPaleoclimate Research 14, pp. 151172.

Berendsen, H.J.A. & E. Stouthamer, 2001.Palaeogeographical development of the RhineMeuse delta, The Netherlands. Van Gorcum, Assen,250 p.

Cohen, K.M., 2003. Differential subsidence within acoastal prism; Late Glacial – Holocene tectonics inthe RhineMeuse delta, The Netherlands.Netherlands Geographical Studies 316,KNAG/Faculteit Ruimtelijke Wetenschappen,Universiteit Utrecht, Utrecht, 176 p.

Dambeck, R. & H. Thiemeyer, 2002. Fluvial History of thenorthern Upper Rhine river (southwestern Germany)during the Lateglacial and Holocene Times.Quaternary International 93/94, pp. 5363.

Lang, A., H.R. Bork, R. Mäckel, N. Preston, J. Wunderlich& R. Dikau, 2003. Changes in sediment flux andstorage within a fluvial system: some examples formthe Rhine catchment. Hydrological Processes 14, pp.33213334.

Middelkoop, H., 1997. Embanked floodplains in theNetherlands; Ggeomorphological evolution overvarious timescales. Netherlands GeographicalStudies 224, KNAG/Faculteit RuimtelijkeWetenschappen, Universiteit Utrecht, Utrecht, 341 p.

Törnqvist, T.E., 1993. Fluvial sedimentary geology andchronology of the Holocene RhineMeuse delta, TheNetherlands. Netherlands Geographical Studies 166,KNAG/Faculteit Ruimtelijke Wetenschappen,Universiteit Utrecht, Utrecht, 176 p.

Vandenberge, J., 1995. Timescales, climate and riverdevelopment. Quaternary Science Reviews 14, pp.631638.



NCR Supervisory Board NCR Programme Committee

December 2004 December 2004

prof. dr. H.J. de Vriend, chairman prof. E. van Beek, chairmanWL | Delft Hydraulics Delft University of Technologyemail: [email protected] email: [email protected]

A.G. van Os, secretary A.G. van Os, secretaryprogramming secretary NCR programming secretary NCRemail: [email protected] email: [email protected]

A.R. van Bennekom G. BlomDG Rijkswaterstaat/RIZA DG Rijkswaterstaat/RIZAemail: [email protected] email: [email protected]

prof. dr. E.A. Koster dr. H. MiddelkoopUtrecht University Utrecht Universityemail: [email protected] email: [email protected].

prof. dr. M.J.F. Stive dr. C.J. SloffDelft University of Technology WL | Delft Hydraulicsemail: [email protected] Email: [email protected]

prof. dr. A.J. Hendriks dr. R. LeuvenUniversity Nijmegen University Nijmegenemail: [email protected] email: [email protected]

prof. dr. A.Y. Hoekstra dr. J.S. RibberinkUniversity Twente University Twenteemail: [email protected] Email: [email protected]

M.W. Blokland J.L.G. de SchutterUNESCOIHE Institute for Water Education UNESCOIHE Institute for Water Educationemail: m.blokland@unescoihe.org email: j.deschutter@unescoihe.org

dr. M.J. van Bracht I.L. RitsemaTNONITG TNONITGemail: [email protected] email: [email protected]

prof. dr. W.P. Cofino dr. H.P. WolfertALTERRA ALTERRAemail: [email protected] email: [email protected]

prof. dr. P.A. Troch dr. R. UijlenhoetUniversity Wageningen University Wageningenemail: [email protected] email: [email protected]


NCR Publications series

In this series the following publications were printed:

NCRpublication no:002000 “Delfstoffenwinning als motor voor rivierverruiming; kansen en bedreigingen”,

editors A.J.M. Smits & G.W. Geerling (in Dutch)(out of stock, but can be downloaded from the NCR Internet site)

012000 “NCR Programma, versie 1999 – 2000”,editors R. Leuven & A.G. van Os (in Dutch)

022000 “NCR workshop, de weg van maatschappelijke vraag naar onderzoek”,editors A.F. Wolters & E.C.L. Marteijn (in Dutch)

032001 “NCR dagen 2000, het begin van een nieuwe reeks”,editors A.F. Wolters, C.J. Sloff & E.C.L. Marteijn (partly in Dutch)

042001 “Umbrella Program IRMASPONGE, Background, Scope and Methodology”,editors A. Hooijer & A.G. van Os

052001 “Summary of NCR Programme, version 2001 – 2002”,editor A.G. van Os(also downloadable from the NCR Internet site)

062001 “The Netherlands Centre for River studies, a cooperation of the major developers andusers of expertise in the area of rivers”,editors A.G. van Os & H. Middelkoop

072001 “NCRdays 2001, from sediment transport, morphology and ecology to river basinmanagement”,editors E. Stouthamer & A.G. van Os

082001 “Land van levende rivieren: De Gelderse Poort”,Stichting Ark, ISSN 90 5011 150 5; € 27,50

092001 “Guidelines for rehabilitation and management of floodplains, ecology and safetycombined”,editors H.A.Wolters, M. Platteeuw & M.M.Schoor

102001 “Living with floods: resilience strategies for flood management and multiple land use inthe river Rhine basin”,editors M. Vis, F. Klijn & M. van Buuren

112001 “Development and application of BIOSAFE, a policy and legislation based model for theassessment of impacts of flood prevention measures on biodiversity in river basins”,authors R.J.W. de Nooij, D. Alard, G. de Blust, N. Geilen, B. Goldschmidt, V. Huesing,J.R. Lenders, R.S.E.W. Leuven, K. Lotterman, S. Muller, P.H. Nienhuis & I. Poudevigne

122001 “Extension of the Flood Forecasting Model FloRIJN”,authors E. Sprokkereef, H. Buiteveld, M. Eberle & J. Kwadijk

132001 “Interactive Flood Management and Landscape Planning in River Systems:Development of a Decision Support System and analysis of retention options along theLower Rhine river”,authors R.M.J. Schielen, C.A. Bons, P.J.A. Gijsbers & W.C. Knol

13M2001 “Interactive Flood Management and Landscape Planning in River Systems” UserManual version 1.0.0

142001 “Cyclic floodplain rejuvenation: a new strategy based on floodplain measures for bothflood risk management and enhancement of the biodiversity of the river Rhine”,editor H. Duel

152001 “Intermeuse: the Meuse reconnected”,authors N. Geilen, B. Pedroli, K. van Looij & L.Krebs, H. Jochems, S. van Rooij & Th.van der Sluis

162001 “Development of flood management strategies for the Rhine and Meuse basins in thecontext of integrated river management”,authors M.B.A. van Asselt, H. Middelkoop, S.A. van ‘t Klooster, W.P.A. van Deursen, M.Haasnoot, J.C.J. Kwadijk, H. Buiteveld, G.P. Können, J. Rotmans, N. van Gemert & P.Valkering

172002 IRMASPONGE, “Towards Sustainable Flood Risk Management in the Rhine andMeuse River Basins”, Proceedings of IRMASPONGE Final Working Conference,editors A.G. van Os & A. Hooijer


182002 IRMASPONGE, “Towards Sustainable Flood Risk Management in the Rhine andMeuse River Basins”, (incl. main results of the research project),editors A. Hooijer & A.G. van Os

18E2002 IRMASPONGE, “Towards Sustainable Flood Risk Management in the Rhine andMeuse River Basins”, main results of the research project,editors A. Hooijer, F. Klijn, J.C.J. Kwadijk & B. Pedroli

18D2002 IRMASPONGE, “Zu einem nachhaltigen Management des Hochwasserrisikos in denEinzugsgebieten von Rhein und Maas die wichtigsten Ergebnisse”,editors A.Hooijer, F. Klijn, J.C.J. Kwadijk & B. Pedroli (in German)

18NL2002 IRMASPONGE, “Naar een duurzaam hoogwater risico beheer voor het Rijn en MaasStroomgebied”, De belangrijkste conclusies van het onderzoeksprogramma,editors A. Hooijer, F. Klijn, J. Kwadijk & B. Pedroli (in Dutch)

18F2002 IRMASPONGE, “Vers une gestion durable des risques d’inondation dans les bassinsversants du Rhin et de la Meuse”,editors A.Hooijer, F. Klijn, J. Kwadijk & B. Pedroli (in French)

192002 “Laseraltimetrie en Hydraulische Vegetatieruwheid van Uiterwaarden”,authors M.R. Ritzen & M.W. Straatsma (in Dutch)

202003 “NCR days 2002, Current themes in Dutch river research”,editors R.S.E.W. Leuven, A.G. van Os & P.H. Nienhuis

212003 “Kansrijkdom voor rivierecotopen vanuit historischgeomorfologisch perspectief Rijntakken Maas Benedenrivieren”,authors H. Middelkoop, E. Stouthamer, M.M. Schoor, H.P. Wolfert & G.J. Maas (inDutch)

222003 “Lowland River Rehabilitation 2003: an international conference addressing theopportunities and constraints, costs and benefits to rehabilitate natural dynamics,landscapes and biodiversity in large regulated lowland rivers. Programme, abstracts andparticipants. Wageningen, September 29 – October 2, 2003”,editors A.D. Buijse, R.S.E.W. Leuven & M. GreijdanusKlaas

232004 “Om de toekomst van het rivierengebied”, gelegenheidsredactie N. Geilen, F. Klijn, S.A.M. van Rooij, C. Stegewerns en C.C. Vos; themanummer tijdschrift ‘Landschap’n.a.v. een studiedag georganiseerd door WLO en NCR.WLOsecretariaat, Postbus 80123, 3508 TC Utrecht, email [email protected]

242004 “NCR days 2003, Dealing with Floods within Constraints”,editors N. Douben, A.G. van Os

252004 “Summary of NCR Programme, version 2004 2005”, editor A.G. van Os



ColophonEditors:Bart Makakse (Alterra) and Ad van Os (NCR)

Design:Cover: KumQuat DordrechtLayout: Jolien Mans (NCR)

Print:JB&A Wateringen, The Netherlands

Number of prints:300

Keywords:NCR, Rivers, Research, Flood Management

To be cited as:B. Makaske & A.G. van Os (editors), 2004. Proceedings NCRdays 2004; Research for managingrivers; present and future issues.NCRpublication 262005.Netherlands Centre for River studies, Delft (ISSN 1568234X).

© Netherlands Centre for River studies, Delft, the NetherlandsISSN 1568234XAll rights reserved. No part of this book may be transplanted or reproduced in any form by print, photoprint, microfilm or any other means without the prior written permission of the publisher NetherlandsCentre for River studies, Delft, The Netherlands.


Date post:	21-Jun-2021
Category:	Documents
Upload:	others
View:	0 times
Download:	0 times

NCRdays€2004 · NCRdays€2004 Research€for€managing€rivers:...

Documents