proceedings ncr days 2003 v7 - vliz · both presentations and posters are incorporated in these...

NCR-days 2003

Dealing with Floods within Constraints

November 6 – 8

N. Douben &A.G. van Os (eds.)

July 2004

Publication of the Netherlands Centre for River StudiesNCR publication 24-2004

ISSN 1568-234X

NCR is a cooperation of the Universities of Delft, Utrecht, Nijmegen, Twente and Wageningen, UNESCO-IHEInstitute for Water Education, RIZA, ALTERRA, TNO-NITG and WL|Delft Hydraulics

- i - Proceedings NCR-days 2003

Preface

The Netherlands Centre for River studies (NCR) started its activities on October 8th, 1998. So, thisautumn of 2003 we celebrated our first lustrum.From 2000 onwards, NCR organises the so-called NCR-days once a year.

The NCR-days 2000 and 2001 were held in Soesterberg (“Kontakt der Kontinenten”) and Wageningen(“De Wageningse Berg”), respectively. The proceedings of these conferences were published in theNCR publication series (NCR publication 03-2001 and 07-2001, respectively). The NCR-days 2002took place on November 7th and 8th, 2002 in Conference Centre Jonkerbosch (Nijmegen). Theproceedings have been published very recently, of course also in the NCR publication series as NCRpublication 20-2003.

The NCR-days 2003, of which the publication at hand contains the proceedings, were organised byUNESCO-IHE and Conference Agency Routine on November 6th and 7th, 2003 in Resort MarinaOolderhuuske, Roermond. This resort is situated on a peninsula in the river Meuse. It boasts anumber of floating houses in Mediterranean style.

The organising committee took theevaluations of preceding NCR-daysseriously, which resulted in a distinctreduction of the amount ofpresentations in favour of timeavailable for bilateral or multilateraldiscussions and poster visits. Severalparticipants told us that theyappreciated this very much.Of course this meant that we had todisappoint a number of prospected oralpresenters. Fortunately most of themagreed to submit a poster instead.Both presentations and posters areincorporated in these proceedings ofthe NCR-days 2003.

We would like to thank Huub Savenije and Guy Beaujot of UNESCO-IHE, Tine and Marijn Verheij andLieke Janssen of the Conference Agency Routine and Jolien Mans of NCR for their help in organisingthe NCR-days 2003.

We are looking back on a very inspiring conference, an elucidating field trip and fruitful discussions.We hope these proceedings will inspire you to come (again) to the NCR-days 2004, to be organizedby ALTERRA.

Klaas-jan Douben, Ad van Os,President of the NCR-days 2003 Organising Committee Programming secretary of NCR

Proceedings NCR days 2003 - ii -

- iii - Proceedings NCR-days 2003

Contents

Preface ..................................................................................................................................... i

Abstract ................................................................................................................................... vi

Samenvatting ......................................................................................................................... vii

Introduction NCR-days 2003

NCR-days 2003; IntroductionN. Douben & A.G. van Os............................................................................................................. 1

Opening address NCR-days 2003L. Bijlsma ..................................................................................................................................... 3

Flood Management and Defence

Hydrological challenges in the Rhine- and Meuse basinsT.A. Bogaard, W.M.J. Luxemburg & M.J.M. de Wit ....................................................................... 7

Economic safety standards for dike ringsC.J.J. Eijgenraam......................................................................................................................... 10

Modelling of surface water – groundwater interaction on the Alzette river basin inLuxembourgF. Fenicia, G.P. Zhang, T.H.M. Rientjes, P. Reggiani, L. Pfister & H.H.G. Savenije....................... 12

Environmental impact of floodingE.E. van der Hoek, L.C.P.M. Stuyt, J.E.A. Reinders & M. de Muinck Keizer .................................. 15

Uncertainty assessment due to time series inputs using disaggregation for floodforecastingS. Maskey & R.K. Price ................................................................................................................ 18

Is the southern Dutch delta the solution to future flood water levels in the tidal river area?N. Slootjes.................................................................................................................................... 20

Dealing with hydraulic models from a user’s perspectiveM. van Roode............................................................................................................................... 22

Appropriate spatial sampling resolution of precipitation for flood forecastingX.H. Dong, M.J. Booij & C.M. Dohmen-Janssen ........................................................................... 24

Spatial Development and Land-use

Searching for the right problem - a case study in water management on the relationbetween information and decision makingM.J. Kolkman, M. Kok & A. van der Veen ..................................................................................... 27

Combining integrated river modelling and agent based social simulation for rivermanagement; The case study of the Grensmaas projectP. Valkering, J. Krywkow, J. Rotmans & A. van der Veen.............................................................. 30

Proceedings NCR days 2003 - iv -

Holocene fluvial dynamics in the Geul river catchmentJ.J.W. de Moor ............................................................................................................................. 32

The RIVER 21 concept: envisioning the future of international river basinsL.L.P.A. Santbergen & J.M. Verhallen........................................................................................... 34

Extreme floods in the Meuse river over the past century: aggravated by land-usechanges?M. Tu, M.J. Hall, P.J.M. de Laat & M.J.M. de Wit .......................................................................... 37

Monte Carlo method applied to a two-dimensional morphodynamic modelJ.J.P. Lambeek ............................................................................................................................ 40

Regional Floodplain Management

Impact of land use changes on breeding birds in floodplains along the rivers Rhine andMeuse in the NetherlandsR.S.E.W. Leuven, R.P.B. Foppen, A.M.G.R. Houlliere, H.J.R. Lenders & R.J.W. de Nooij............. 43

Developing integrated water system models from a groundwater perspective - case WaterBoard Regge and Dinkel groundwater modelJ.J.J.C. Snepvangers, B.M. Minnema, C.B.M. te Stroet, Y. van der Velde & M. Kuijper ................. 46

Mapping of vegetation characteristics using lidar and spectral remote sensingM. Straatsma, H. Middelkoop, E. Pebesma & C. Wesseling .......................................................... 48

Alien amphipod invasions in the river Rhine due to river connectivity: a case ofcompetition and mutual predationM.C. van Riel, G. van der Velde & A. bij de Vaate......................................................................... 51

Data-driven modelling in context to sediment transportB. Bhattacharya & C. Buraimo...................................................................................................... 54

Environmental impact assessment study for the Common Meuse (MER Grensmaas 2003)J. Onneweer, B. Peerbolte & D.G. Meijer...................................................................................... 56

Poster presentations

Vegetation dynamics of a meandering river (Allier, France)G.W. Geerling, A.M.J. Ragas & A.J.M. Smits................................................................................ 59

Soil, hillslope and network structure as an opportunity for smart catchment scalehydrological modelsP.W. Bogaart & P.A. Troch ........................................................................................................... 62

Integrated flood management strategies; A case study in the Lower Dong Nai – Sai Gonriver system, VietnamA.D. Nguyen & N. Douben............................................................................................................ 64

Integrated water resources management groupwork - case Mura river basinM.L. Mul & T.D. Schotanus........................................................................................................... 66

Effects of subsurface parameterisation on runoff generation in the Geer basinG.P. Zhang, F. Fenicia, T.H.M. Rientjes, P. Reggiani & H.H.G. Savenije....................................... 68

River nutrient transport under various economic development scenarios – thetransboundary Lake Peipsi/Chudskoe drainage basinD.S.J. Mourad, M. van der Perk, G.D. Gooch, E. Loigu, K. Piirimäe & P. Stålnacke ...................... 71

- v - Proceedings NCR-days 2003

Reduction of siltation in harboursS.A.H. van Schijndel..................................................................................................................... 74

Morphological behaviour around bifurcation points of the Dutch Rhine branchesL.J. Bolwidt & P. Jesse ................................................................................................................. 76

Which processes determine the downstream fining of bed sediment in the Dutch Rhinebranches?R.M. Frings .................................................................................................................................. 78

IJSSELKOP: modelling of 3D grain size distributions, one step closer to realityS.H.L.L. Gruijters, D. Maljers, J.G. Veldkamp, J. Gunnink, M.P.E. de Kleine, P. Jesse &L.J. Bolwidt................................................................................................................................... 80

Eurasian rivers in flood: application of the mixed distribution theory to runoff extremesW.M.J. Luxemburg, B.I. Gartsman & L.M. Korytny ........................................................................ 84

Geological modelling: how reliable are input parameters?D. Maljers & S.H.L.L. Gruijters...................................................................................................... 86

Delft – FEWS: Flood Early Warning SystemA.H. te Linde, A.H. Weerts & J. Schellekens ................................................................................. 88

Influence of non-linearity in the storage-discharge relationship on discharge droughtsE. Peters & H.A.J. van Lanen ....................................................................................................... 90

Spatial distribution of heavy metals in floodplainsA.M. Uijtdewilligen ........................................................................................................................ 92

Data-driven modelling for flood-related problemsD.P. Solomatine ........................................................................................................................... 94

Modelling sediment dispersion over floodplains using a particle tracking methodI. Thonon, K. de Jong, M. van der Perk & H. Middelkoop .............................................................. 96

The hillslope-storage Boussinesq model for variable bedrock slopeA.G.J. Hilberts, P.A. Troch, E.E. van Loon & C. Paniconi .............................................................. 98

Future Developments

NWO and WaterH. de Boois .................................................................................................................................. 101

Five years NCR, review and previewA.G. van Os ................................................................................................................................. 103

Authors index ........................................................................................................................ 106

NCR Supervisory Board and NCR Programme Committee ................................... 107

NCR Publications series .................................................................................................... 108

Colophon ................................................................................................................................. 111

Proceedings NCR days 2003 - vi -

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 in-depth 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 so-called annual NCR-days. The publication at hand contains the proceedings of the NCR-days, organised on November 6-7, 2003.

The proceedings of the NCR-days 2003 are sub-titled “Dealing with Floods within Constraints”. Theconcept of constraints is widely defined in this context, varying from increasing climate variability to theimplementation of tangible technical and organisational measures.

In general, the constraints are defined from a socio-economic point of view (sustainable maintenanceand increase of the protection level against flooding). Furthermore the importance of integrated riverfunctions, spatial interests (also on river basin level) and sustainability criteria is increasing within theframework of flood prevention. The complexity and time consumption of decision-making processesare increasing as a result of an intensified public participation and long lasting legislation procedures.

The three different themes of the NCR-days 2003, (i) Flood Management and Defence, (ii) SpatialDevelopment and Land-use, and (iii) Regional Floodplain Management, cover to a large extend theresearch which is performed in The Netherlands nowadays, regarding safety against flooding.Hydrological, morphological and ecological (riverine planning) problems are clarified with modellinganalysis and field survey studies. The progress of strategic policy studies and the development ofinstruments to enhance perceptions in decision-making processes were presented. Moreoverinnovative methods for mapping techniques and flood forecasting are not missing, as well asenvironmental impacts of floods and economical risk analysis.

- vii - Proceedings NCR-days 2003

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 NCR-dagen. Voorliggende publicatie bevat de “proceedings” van de NCR-dagen die gehouden werden op 6en 7 november 2003.

De proceedings van de NCR-dagen 2003 dragen de subtitel “Dealing with Floods within Constraints”,vrij vertaald “Hoogwaterbescherming binnen strikte randvoorwaarden”. Het begrip randvoorwaarden isin deze hoedanigheid breed gedefinieerd, variërend van toenemende klimaat variabiliteit totimplementatie van concrete technische en organisatorische maatregelen.

De randvoorwaarden worden over het algemeen gedefinieerd vanuit een sociaal-economisch oogpunt(duurzame handhaving en verbetering van het beschermingsniveau). Daarnaast nemen integratie vanfuncties, ruimtelijke belangen (ook op stroomgebied niveau) en duurzaamheidcriteria in toenemendemate een belangrijke plaats in binnen het krachtenveld van hoogwaterbescherming.Besluitvormingsprocessen worden complexer en tijdrovender van aard als gevolg van eenintensievere betrokkenheid van de burgerbevolking en langdurige vergunningsprocedures.

De drie verschillende thema’s van de NCR-dagen 2003, (i) Hoogwaterbescherming en beheer, (ii)Ruimtelijke ordening en landgebruik en (iii) Uiterwaard beheer, dekken een groot gedeelte van hethedendaagse onderzoek dat in Nederland wordt uitgevoerd met betrekking tot veiligheid tegenoverstroming. Hydrologische, morfologische en ecologische (inrichtings-)vraagstukken wordentoegelicht met modelmatige analyses en veldwerk studies. De voortgang van beleidsanalytischestudies en de ontwikkeling van instrumentaria ter bevordering van het inzicht inbesluitvormingsprocessen zijn gepresenteerd. Daarnaast zijn tevens vernieuwde kartering- enhoogwater voorspellingsmethoden belicht, alsmede milieueffecten van hoogwaters en economischerisico analyses.

Proceedings NCR days 2003 - viii -


- 1 - Proceedings NCR-days 2003

NCR-days 2003; Introduction

N. Douben1 & A.G. van Os2

1 UNESCO-IHE Institute for Water Education, P.O. Box 3015, 2601 DA Delft, The Netherlands2 Programming 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, UNESCO-IHE, ALTERRA,TNO-NITG, RIZA and WL | Delft Hydraulics.NCR’s goal is to build a joint knowledge baseon rivers in the Netherlands and to promoteco-operation 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.

• Research-orientated and educational co-operation: 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 theso-called NCR-days, 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 co-operation.

NCR organised these NCR-days for the fourthtime on November 6th and 7th, 2003 in ResortMarina Oolderhuuske in Roermond, theNetherlands.In the publication at hand the presentationsand posters presented are summarized.

The NCR-days of 2003 were, as far as thefeed back from the participants indicates sofar, a great success. The organisers clearlytook notice of the remarks made during theevaluation in 2002: there was ample time forthe poster sessions and bi- or multilateraldiscussions.

The statistics of the 2003 days are verysatisfying too: some 105 participantsdistributed evenly over the NCR partners(apart from ALTERRA) and other institutes andconsultancy agencies (Fig. 1). Also thepresentations and posters were reasonablydistributed over NCR partners and non-NCRparticipants (Fig. 2).In total 23 oral presentations were given and20 posters could be seen and discussed. Infact much more presentations were proposed,but the organisers had to limit the amount toapproximately 20 (plus 3 key notepresentations) to give the participantsopportunity for the poster sessions anddiscussions.

Figure 1. Number of participants per institute.

Number of participants per institute

0

5

10

15

20

RIZA

ALTERRA

IHE

KUNUU

TUDW

L UTNIT

GW

URRW

S

CONS

Other

s

Introduction NCR days 2003

Proceedings NCR days 2003 - 2 -

Nadine Slootjes Simone van Schijndel

Figure 2. Number of presentations and posters per institute.

To celebrate the first 5-year lustrumanniversary, the NCR Programme Committeedecided to establish the NCR-daysPresentation and Poster Awards. They bothconsist of a Certificate and the refunding of theparticipation costs for the NCR-days.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 NCR-days. We use the ‘Hyde Park

Corner approach’ where the primary posterauthors are given the opportunity in ‘two-minute-talks’ to give the participants anappetite to come and see the posters anddiscuss the content with the authors. Thisworked again very well.

The winners of the NCR-days Awards wereannounced at the end of the NCR-days.

The NCR-days Presentation Award 2003 (Fig.3) was won by Nadine Slootjes for herpresentation ‘Is the Southern Dutch Delta theSolution to Future Flood Water Levels in theTidal River Area?’ (see page 20).The NCR-days Poster Award 2003 (Fig. 4)was won by Simone van Schijndel for herposter ‘Reduction of siltation in harbours’ (seepage 74).

Figure 3. NCR-days Presentation Award 2003. Figure 4. NCR-days Poster Award 2003.

Number of presentations and posters per institute

0

2

4

6

RIZA

ALTERRA

IHE

KUN UUTUD

WL UT

NITG

WUR

RWS

CONS

Other

s

Presentations

Posters



Opening address NCR-days 2003

L. BijlsmaProject Director De Maaswerken, Ministry of Transport, Public Works and Water Management, Rijkswaterstaat, P.O. Box 1593,6201 BN Maastricht, The Netherlands

Summary“It’s not about accuracy, stupid, it’s aboutuncertainty”.

IntroductionThe river investment scheme “DeMaaswerken” is a comprehensive integratedscheme. Comprehensive given it’s size andintegrated given its objectives. The objectivescover flood protection, nature development,sand and gravel mining and inland navigation.The trigger to plan for this investment schemais in the past: the flood disasters of 1993 and1995. As a policy response, the objective setfor flood protection is a 1/250 protection levelfor settlements along the river. The riverstretch to be covered is the upstream sectionof the river Meuse, where the river flows in avalley. This section is located betweenMaastricht in the south and ‘s-Hertogenbosch,roughly 200 kilometres. Objectives of other

sectors of government: nature development,mining and inland navigation are integrated inone investment plan. The result of plandevelopment is an agreement to line upinvestments in nature development and miningactivities with flood protection. Theinvestments for upgrading the Rhine-Meuseinland navigation channel have been optimisedwith flood protection. Optimisation is especiallygained through combined planning for: (i) thesoil removal, soil transportation, upgrading andstorage scheme, (ii) the hydraulic andhydrological impact assessment, and (iii) multipurpose land use planning in the river valley.

The comprehensiveness of the investmentscheme is given in approximate figures inTable 1. The planning phase of “DeMaaswerken” is concluded in 2003 and theimplementation phase is under preparation.

Table 1. Investment scheme “De Maaswerken”.

Role of applied scienceThe comprehensive investment scheme - “DeMaaswerken” - is information intensive, in plandevelopment as well as in the implementationphase. Information from physical sciences(hydraulics, hydrology and morphology) mayhave some emphasis in the frame of “DeMaaswerken” objectives; also otherinformation is of importance. Information fromother natural sciences as biology, ecology andsocial and economic science is playing animportant role, in strategy and policydevelopment as well as the implementation. IfNCR’s objective is formulated as ‘to facilitatedecision making processes for sustainabledevelopment of rivers and river basins’ the

focus should be broader then physicalsciences only.There are many stakeholders involved indecision-making processes for flood protectionplan development as well as for itsimplementation. These processes aretherefore extremely communication sensitive.The plan development phase of “DeMaaswerken” took 6 years to conclude andaddresses multiple national and local planningschemes and environmental impactassessments. For its implementation manythousands licences will be needed. Alldecisions are to be built on a sound scientificfoundation that can sustain a longer timeframe, that can convince stakeholders and atthe end of the day the “Raad van State”: thenational court of appeal.

Flood protection Inland navigation

• Public investment € 550 million• Private investment € 600 million• Ready 2015 – 2018• Embankments• River deepening – widening• Nature development 1600 ha• Sand and gravel mining 60 million m3

• Public investment € 500 million• Private investment € 0 million• Ready 2015 – 2018• Lock enlargement, replacement• Channel improvement• Bridges

Introduction NCR days 2003


Figure 1. Stakeholders of flood protection information.

The stakeholders for flood protectioninformation are given in Fig. 1 and includedecision makers at national and local level, theriver manager, public investors, privateinvestors and the public at large.

Experiences in developing and implementing“De Maaswerken” scheme learn that manystakeholders have different perspectives,expectations and interpretations about whatapplied science can deliver. The notion is thatthe prediction of extreme events, given arecurrent frequency, is by definition uncertain;however, stakeholders have differentexpectations, as summarized in Fig. 2.

The decision maker can handle uncertainty,however hates instability, meaning that adecision should sustain his or her period ofgovernance. The policymaker is investing inapplied science in the perception that the

uncertainty may be reduced. The rivermanager has to compare development optionsfor land use for licensing reasons and isjudging in millimetres rather then centimetres.The public investor, as “De Maaswerken”, hasaccounted for uncertainty and inaccuracy inplan development and the design of theindividual measures. The private investorhates uncertainty and needs absolute figures‘for high water free’ development. Finally, thepublic at large is demanding absolute safetyand is blaming government for all risks. Theaspirations and expectations of stakeholders,as summarized in Fig. 2, demonstrate acommunication failure, hindering decision-making processes. These stakeholders are theend users of applied science for floodprotection. The notion is that applied science,represented here today by NCR, needs toaddress this communication failure.

Figure 2. Communication failure with stakeholders in flood protection information.

decision makernational

local

policy makernational

local

river managerlicencing

maintenance

public investorDe Maaswerken

private investordeveloperfinancierinsurar

societypublicgroupspress

flood protectioninformation

decission maker

uncertainty okbut stable

policy maker

less uncertaintyis better

river manager

relative mmabsolute cm

public investorDe Maaswerken

accuracy 1 a 2 dm short / long term

private investordeveloper

absolute numbersno uncertainty

societypublic

absolute safetyno risk

flood protectionapplied science

tomorrow prediction is better

science



NCR’s challengeAccuracy of predictions of extreme events is ofimportance and my plea is not to neglect theissue. However, model inaccuracy is only oneof the elements that contribute to uncertainty inthe prediction. Other contributors are amongstothers: boundary conditions, river behaviour,land use change, climate change, processschematisations and parameterisation. Theend result is, by definition, uncertain anduncertainty lies roughly in the order ofmagnitude of 0.5 to 1 meter for 1/250 to1/1,250 exceeding frequencies (95% reliabilityinterval).

On the other hand, the perception andexpectation of NCR’s end users of appliedscience information is scattered and biasedfrom reality. In fact there is a communicationfailure amongst stakeholders on floodprotection issues. This situation is hinderingthe sound development of flood protectionpolicies and its implementation. Therefore ashift is needed in NCR’s focus from accuracyissues to uncertainty issues. The question how

society - NCR’s end users - best can deal withuncertainty should be on the foreground.

Consequently, I would recommend thefollowing shift in NCR’s programmatic focus:• Given the communication failure in society

concerning: how to deal with uncertainties inflood protection; there is a need for a shift inapplied science for river management.

• This shift should move in the direction ofinformation and education of safety and riskconcepts - including its uncertainties - insociety at large.

• In order to do so, there is a need for betterintegration of science and information,including economic and social sciences.

In order to do so, there is a need for betterintegration of science and information,including economic and social sciences.

I would like to congratulate NCR with itslustrum and wish the organization all the bestin helping to address societal challenges in thefield of flood protection.



Hydrological challenges in the Rhine- and Meuse basins

T.A. Bogaard1, W.M.J. Luxemburg2 & M.J.M. de Wit3

1 Department Physical Geography, Faculty of Geosciences, Utrecht University, P.O. Box 80115, 3508 TC Utrecht, TheNetherlands; [email protected]

2 Faculty of Civil Engineering and Geosciences (CiTG), section Hydrology & Ecology, Delft University of Technology, P.O. Box5048, 2600 GA Delft, The Netherlands; [email protected]

3 Institute for Inland Water Management and Waste Water Treatment, P.O. Box 9072, 6800 ED Arnhem, The Netherlands;[email protected]

IntroductionSociety is facing more and more socio-economic problems as result of floodoccurrences in Western Europe. One of theworld’s largest re-insurance companies,Munich Re, describes in its Annual Review ofNatural Catastrophe 2002 that floods accountfor 30% of all loss events and 42% of allfatalities due to natural catastrophes in 2002(Munich Re, 2003). The total economic loss by

floods was estimated at US$ 27 bn in thatyear. Furthermore, Munich Re (2003) rankedRotterdam-Amsterdam (“Randstad”)metropolis in the hazard index at position 18 ofthe 50 mega cities in the world. Although this isa first attempt to index hazard risk ofmetropolis, and in case of the “Randstad” thehazard is based on flood and wind, it showsthat there is a major economic and societalrisk, partly induced by floods.

Figure 1. Statistics of natural hazards in 2002, according to the Munich Re (2003).

NCR initiatives: ‘HydrologicalTriangle’ and ‘Genesis of Flood’themeFor Dutch water management, knowledge ofthe hydrological processes in the Rhine andMeuse basins upstream of the Netherlands isessential. Both for flood forecasting as for longterm spatial planning. In 2003 at least tworeconnaissance of hydrological research were

made and published; the Dutch ‘foresightcommittee on Hydrology’ (KNAW, 2003) andGeosciences, the future (IUGG, 2003). Bothreports were written after several discussionsand meetings. Although different from set-upboth reports come up with similar scientificbottlenecks and challenges for hydrology: 1)incomplete understanding of the hydrologicalprocesses; 2) incomplete theory and data sets



for modelling; 3) data integration and modelcalibration; 4) scale issues.NCR has stimulated two initiatives. First of all,a NCR hydrological working group(Hydrological Triangle) was initiated by RIZA inorder to stimulate and tune the hydrologicalresearch in the Rhine and Meuse basins withinthe Netherlands. A number of Dutchuniversities, technical institutes, andgovernmental institutes participate in thisworking group. Main aims of the group are tobring together skills, knowledge and manpowerof different research groups and stimulate datasharing and collaboration and define thoseresearch questions that are both relevant forwater management and challenging forscientists.Furthermore, NCR defined the ‘Genesis ofFloods’ theme, which aims at studying thegenesis of floods, or more broadly, dischargegeneration in upstream areas. The ‘Genesis ofFlood’ theme now resides under theHydrological Triangle. The overall objective of‘Genesis of Floods’ is to improve ourunderstanding of the spatial and temporalcontrols of discharge generation underchanging hydrometeorological behaviour andland use management. The researchquestions have been grouped into: i) dischargeforecasting, ii) determination of designdischarge, iii) influence of climate change, andiv) influence of land use change. Someexamples will be given of ongoing projects onthese topics. Besides some research topicswill be listed that have potential for further jointstudy.

ExamplesIt is generally accepted that continuation of theengineering approach of controlling the naturalriver system is not sustainable. More and moreit is discussed that river systems should getmore room to allow for their natural behaviour.The reasoning is that a river stream can(temporarily) store more water and diminishesflood related problems along a river. Besidesstorage of discharge in the river system itself,the discussion also focuses on retardation ofrainfall in the upstream area of the rivercatchments. Land use has changedenormously during the last century, especiallywith urbanisation and an intensification of

agriculture and forestry. The overall believe isthat these activities result in an accelerateddrainage of the upstream areas. The WWF hasinitiated a discussion on combining naturerestoration and water storage with their report‘Mountains of water’ (WWF, 2000). This‘enhancing the natural sponginess of riverbasins’ could provide an enjoyable network ofrivers, water and nature, solve floods anddroughts and stimulate business in tourismand recreation. As the public was told thatdeforestation, soil drainage, and sealing ofsurfaces was the cause for the increased floodfrequency and magnitude, why should doingthe reversed not have the opposite effect?However, the Munich Re (2003) report is veryclear about the effects of river restoration:River restoration measures make sense andare very welcome; but their effectiveness inextreme cases is often overestimated ormisrepresented. As a rule, they are incapableof preventing really catastrophic floods and inmany cases will not even bring about anysignificant reduction. The volumes of waterthat amass in extreme events are simply toohuge.De Wit et al. (2003) focused on the Meuse anddiscussed the limited effects of reforestation,re-meandering, increased inundation offloodplains, nature recovery and uncontrolledretention on the peak discharge of the Meuseat Borgharen.But this is all based on the extreme events,when storage is already or otherwise quicklyused. What are the effects of the proposedmeasures on floods of lower return periodsand lower magnitudes? It is challenging to tryto quantify in what cases the water storagecould be beneficial for flood prevention, or toquantify the criticism on the WWF report.

In the field of process knowledge thechallenges are to be found in the temporal andspatial dynamics of soil moisture and theinteraction between the hydrological cycle andterrestrial ecosystems. Another research itemis the quantification of groundwater effects ondischarge generation in the Ardennes. Anexample of this research is given in Fig. 2.Here one sees the groundwater dynamics inonly 6 months of a small catchment inLuxemburg.



Figure 2. Fluctuation of piezometric levelHuelerbach catchment Apr. 2003 - Sep. 2003.

Challenging new model concepts like there arethe Hillslope-storage Boussinesq formulationand the Representative Elementary Watershedconcept are brought forward in the last fewyears that also describe the questions of waterstorage and water flow, but tackle theproblems with the Richards equation based,distributed, models.On the operational side of the dischargeforecast there is a clear trend from modellingtowards data processing: from point datatowards spatial time series and from oneforecast towards several forecasts anduncertainty. This is a logical trend, more andmore attention should be given to all availabledata sources. This asks for powerful dataassimilation tools to integrate observations andmodel predictions. Also the post-processingneeds attention as model outputs shifts fromone output towards output ensembles, whichshould lead to insight and quantification of theuncertainty of the model results.The scale issues relate to all hydrologicalsciences and include the research of dominantprocesses and interaction between processes

that occur at different spatial scales. How tocome from process knowledge to a floodforecasting?

In the above, the NCR initiatives of theHydrological Triangle and Genesis of Floodwere presented and discussed. The overallaim is bring together skills, knowledge andmanpower of different research groups andstimulate data-sharing and collaboration anddefine those research questions that are bothrelevant for water management andchallenging for scientists. The objectives of theNCR theme Genesis of Floods fit perfectlywithin this network. The meetings that wereorganised by the Hydrological Triangle canalready be called a success. Different researchgroups were brought together and differentresearch topics have been addressed anddiscussed. Initiatives have seen daylight like afield excursion to the Ardennes to look forcollaborative field research site in theframework of the Genesis of Flood theme. Forfurther information the reader is referred to theauthors or to the NCR web page:http://www.ncr-web.org

ReferencesInternational Union of Geodesy and Geophysics

(IUGG), 2003.Geoscience: The Future. FinalReport of the IUGG Working Group Geosciences:The Future. http://www.iugg.org/geosciences.pdf

Royal Netherlands Academy of Arts and Sciences(KNAW), 2003. Hydrology: a Vital Component ofEarth System Science; Preliminary ForesightStudy on Hydrological Science.http://www.knaw.nl/verkenningen/hydrologie.html

Munich Re, 2003. Annual Review: naturalcatastrophes 2002. Municher Rückversicherungs-Gesellschaft, München. http://www.munichre.com

Wit, de M., W. van Deursen, J. Goudriaan, U. Boot,A. Wijbenga & W. Van Leussen, 2003. Integratedoutlook for the river Meuse (IVM); a look upstreamBorgharen. In: Leuven, R.S.E.W., A.G. van Os &P.H. Nienhuis (Eds.). Proceedings NCR days2002: Current themes in Dutch river research.NCR publication 20-2003. Netherlands Centre forRiver Studies, Delft, The Netherlands.

World Wide Fund for Nature (WWF), 2000.Mountains of Water. Zeist, The Netherlands.http://www.wnf.nl

Transsection P11 - P15

285.5

286.0

286.5

287.0

287.5

288.0

288.5

0 5 10 15 20 25 30 35 40 45 50Distance [m]

Gro

un

dw

ater

leve

l[m

]

soil surface23-apr-0325-apr-0328-apr-0329-apr-038-mei-0312-mei-0314-mei-0319-mei-0321-mei-0328-mei-036-jun-0317-jun-033-jul-0317-jul-0329-jul-037-aug-0322-aug-0328-aug-0328-aug-034-sep-03

P12P13

P11P14

P15 Transsection P11 - P15

285.5

286.0

286.5

287.0

287.5

288.0

288.5

0 5 10 15 20 25 30 35 40 45 50Distance [m]

Gro

un

dw

ater

leve

l[m

]

soil surface23-apr-0325-apr-0328-apr-0329-apr-038-mei-0312-mei-0314-mei-0319-mei-0321-mei-0328-mei-036-jun-0317-jun-033-jul-0317-jul-0329-jul-037-aug-0322-aug-0328-aug-0328-aug-034-sep-03

P12P13

P11P14

P15



Economic safety standards for dike rings

C.J.J. EijgenraamCPB Netherlands Bureau for Economic Policy Analysis, P.O. Box 80510, 2508 GM The Hague, The Netherlands; [email protected]

AbstractAfter the flood disaster in 1953 Prof. D. vanDantzig was asked to solve the economicdecision problem concerning the optimalheight of dikes. His formula with a fixed upperbound for the exceedance probability(Econometrica, 1956) is still in use today incost benefit analysis of the optimal size offlood protection measures, as the height ofdikes. However, this formula is wrong. In thereal optimal safety strategy not theexceedance probability but the expected yearlyloss by flooding is the central variable. Undersome conditions it is optimal to keep thisexpected loss in the future within a constantinterval. Therefore, when the potential damageincreases by economic growth, the floodingprobability has to decline in the course of timein order to keep the expected loss between thefixed boundaries. The paper gives the formulasfor the optimal boundaries. One is that the rateof return at the moment of investment (FYRR)has to be zero (or positive). In case of apositive rate of growth of the expected damagethe net present value (NPV) of a safetyinvestment, which passes the FYRR criterion,will be very positive. Therefore, the well-knownNPV criterion in cost benefit analysis of asingle project is not the right criterion forinvesting in this type of projects.

IntroductionWhat is – from an economic point of view - theoptimal strategy for protecting a dike-ringagainst flooding? In designing such a strategythe social costs of investing in water defenceshave to be compared with the social benefits ofavoiding damage by flooding. In the end thechoice of safety standards is a politicaldecision. In the Netherlands the Act on theWater Defences gives for different types ofdike-rings a standard for the maximumexceedance probability of the design waterlevel a dike-ring must sustain. This approach isin accordance with the formula of Van Dantzig(1956) concerning the optimal height of dikes.However, this formula is wrong because theinfluence of economic growth in the form ofincreasing potential damage in the course oftime is not properly taken into account.

Deriving optimal safety standardsFor the simplest case of one dike-ring, onefailure mechanism: overflow, one type ofdefence: dikes, and social costs and benefitswhich all can be monetarized, the correctreasoning is as follows.

The expected yearly loss by flooding St issupposed to increase by β% a year as a resultof economic growth and rising frequencies ofhigh water levels caused among others byclimate change. To cope with these systematicchanges more than one defence action ut isneeded in the future. Because the existence ofa rate of discount δ we like to postponeinvestment costs I(u) as much as possible.This asks for actions closely in line with theneed for better protection and therefore pointsto frequent and small investments. But a lot ofthe investment costs are fixed costs, notdependent on the size of the action. In order tokeep the costs per centimetre heightening low,we should do rather big investments at once.In doing so, the protection level will be highjust after an investment resulting in a lowexpected yearly loss by flooding S+, but safetywill gradually decline afterwards. When acertain low level of safety with a high-expectedloss by flooding s- is reached, we decide that anew action is profitable and we invest again.Taken together the strategy consists ofkeeping the expected yearly loss by flooding St

within an interval [S+, s-] by repetitiveinvestment in the heightening of dikes. Therelative size of the fixed costs plays animportant role in determining the width of theinterval.

Under certain conditions the future will exactlylook the same every time the system is backon the re-order level s-. Then we will also takethe same optimal decision again to bring theexpected loss back to the S+ level. So in thiscase the interval [S+, s-] will be constant overtime. The most important conditions are:• All (growth) parameters are constant in time;• δ > β;• Investment costs I(u) are not dependent on

the height of the existing dike and notdependent on time, the derivative Iu > c > 0en I(ε) >> 0, so both the marginal and thefixed costs are clearly positive.



The assumptions on the investment costsseem not to far from reality. And aboutdevelopments in the far future we know solittle, that keeping the growth rates constant ina calculation is not a problem, provided the farfuture doesn’t have a great influence on theresults. It is the second condition δ > β thatmakes sure that only the near future isimportant, but also that the future will alwayslook the same. This condition is not alwaysfulfilled in practice, but it is too complicated tohandle this problem in this paper.

With fixed boundaries the size of theinvestments u and the time betweeninvestments D will also be fixed. The totalcosts of damage by flooding and investment inprotection measures can then be written asfollows, starting from a situation that thesystem is on the re-order level s-:

( ) ( ) 1)( 1)(11 −−−−+ −

+−

−= DD

hulp euIeSK δβδ

βδ

Total costs are these costs plus the costs inthe waiting period till s- has been reached.

There are two independent instruments, whichcan be used to minimize total costs: theinvestment size u and the waiting time.Differentiating total costs to these twovariables gives two optimality conditions. Thefirst, looking at u, is:

+

−−

+−

= −−+hulp

Du KseSI

βδδ

βθ

βδθ δ 11

The marginal costs of investment must beequal to the marginal diminishing of the lossdirectly after investment, corrected for theeffects of lengthening the interval betweeninvestments. A marginal period with high-expected damage costs is added at the end of

the interval and the total costs in furtherperiods are a bit shifted to the future. Since thedecision to invest has been taken already, it isonly marginal costs that matter in thedetermination of u.

( )uISs δ=− +−

The second optimality condition looks at thetiming of the first investment:The second criterion is the well-known FirstYear Rate of Return (FYRR). The benefits ofthe jump in safety at the moment of investmentmust be equal (or bigger) than the totalinvestment costs on a yearly basis. Becausethis criterion involves the decision to invest ornot, all costs are relevant. Together with thedefinition equations relating S+, s-, u, D (andwaiting time), this system of equations can besolved.

ConclusionsNot the exceedance probability but theexpected yearly loss by flooding is the centralvariable in an optimal safety strategy.Under certain conditions it is optimal to keepthe expected yearly loss within a fixed interval.Because the yearly costs of investment staythe same, but the benefits of protectionincrease by β% a year, the net present value(NPV) will be very positive for a safetyinvestment which passes the FYRR criterion.Therefore, the well-known NPV criterion in costbenefit analysis of a single project is not theright criterion for investing in this type ofprojects.

ReferencesVan Dantzig, D., 1956. Economic decision problems

for flood prevention. Econometrica. 24, pp. 276-287.



Modelling of surface water – groundwater interaction onthe Alzette river basin in Luxembourg

F. Fenicia1, G.P. Zhang1, T.H.M. Rientjes1, P. Reggiani2, L. Pfister3 & H.H.G. Savenije1


2 WL | Delft Hydraulics, P.O. Box 177, 2600 MH Delft, The Netherlands3 Gabriel Lippmann Institute, 162a avenue de la Faïencerie, L-1511 Luxembourg, Luxembourg

AbstractDetailed field measurements of water tablelevels and river stages operated on the Alzetteriver basin in Luxembourg have pointed outthat the runoff response of the catchment dueto rainfall is strongly dependent on thegroundwater level behaviour and consequentlyon the water storage of the catchment. This isdue to the soil structure and geology of thearea, which allow the formation of fastpreferential flow that quickly reaches theaquifer and rapidly increases the near-streamsaturated area.In order to simulate such situation anintegrated approach that allows the simulationof coupled groundwater and surface waterflows is necessary; for this purpose the‘Representative Elementary Watershed’(REW) model is selected.The model considers the total watershed areaas subdivided in a number of sub-watershedsaccording to the Strahler stream order system.Each sub-watershed is considered as a spatialunit, and is divided in various zones accordingto the different physical processes observed.Only averaged quantities are considered withineach zone, and mass exchange processes areevaluated between the various zones and thevarious sub-watersheds.In particular the mass exchange between thesaturated zone and the channel reaches issimulated, allowing an interaction betweengroundwater and surface water.Calibration of the model is performed bymeans of piezographic measures as well aschannel flow data available from the Centre deRecherche Public Gabriel Lippmann (CRP-GL)of Luxembourg.

IntroductionSince 1995 a dense hydro-climatologicalobservation network comprising rain gauges,stream gauges and piezographs operates onthe Alzette catchment. By analysis of the datasome insights on the flood-generatingprocesses in the Alzette catchment has beengained. In particular, channel flow data andpiezographic measures show that thecatchment runoff response to a rainfall event is

strongly dependent on the observed watertable level.The objective of this study is to simulate watertable variations and the interaction betweendominant hydrological processes that affectthe hydrograph generation.The REW approach is particularly suitable forsimulating hydrological processes interactions,and for this reason it is selected for this study.

Study areaThe Alzette River basin is located almostentirely in Luxemburg, and has an area of1,176 km2. For this study the Hesperange sub-catchment in the south Alzette is selected.The Hesperange sub-catchment covers anarea of 268 km2, its elevation ranges between260 and 450 m a.s.l. and average slope is3.8°. Watershed geometrical properties areextracted from a digital elevation model with agrid size of 50 m.Soil cover is mainly represented by pasturesand forests, while the geology of the studyarea is characterized by Mesozoic rocks,dominated by marls and schists.

Model descriptionThe REW model is a physically based, semi-distributed, deterministic model, based on theapplication of balance equations directly at thesub-catchment scale.

Figure 1. Hesperange sub-catchment sub-divided inREW’s.



The river network is generated directly fromthe DEM of the area, and the watershed isdiscredited into a number of sub-watersheds orREW’s depending on the Strahler order that isselected (Fig. 1).

In the approach balance laws of mass andmomentum are spatially averaged over aREW, and constitutive relations (such asDarcy’s law, Chézy formula, and the SaintVenant equation) are derived directly at theREW scale.Each REW therefore is a spatially lumped unitthat is represented by spatially averagedparameters and variables.The REW model, compared to otherhydrological models, presents the advantagesto posses a low number of parameters, and torequire small computational efforts.

Processes interactionEach REW is subdivided in five zones basedon the major hydrological processes that takeplace within an REW, their different geometriesand time scales. The five zones are: saturatedzone (below the water table), unsaturated zone(above the water table), channel reach,concentrated overland flow (soil surfacecorresponding to the unsaturated zone),saturated overland flow (soil surfacecorresponding to the saturated zone).

Figure 2. REW cross section.

Fig. 2 shows a cross section of an REW basedon the five zones above mentioned, where themass exchange eij is allowed for the zones iand j. The position of the water tabledetermines the extension of the various zones.

Model sensitivity and calibrationA model sensitivity analysis is performed inorder get some insights in the model behaviourand to help model calibration. A sensitivityanalysis varying one parameter at time whilekeeping the others constant is performed. Thechange in some hydrograph properties is

investigated, and in particular the variation inpeak discharge, time to reach the peak andtime to reach equilibrium is defined. The sameanalysis is executed on the simulated watertable level. The calibration procedure is carriedout by trial and errors, trying to maximize theobjective function given by the Coefficient ofDetermination (R2).

Results and discussionObserved water table levels at a piezographicstation nearby the catchment outlet andsimulated counter parts for REW1 are shownin Fig. 3. The R2 for the entire simulationperiod is 0.80, showing that there is a good fitbetween observed and simulated averagedwater table levels.

Figure 3. Water table level simulation.

The hydrograph simulation of Fig. 4 shows aR2 of 0.64. Even though the high peaks as wellas the base flow trend is quite well simulated,some hydrograph peaks for low water tablevalues are not observed in reality. This is themain reason why the R2 is not higher.

Figure 4. Hydrograph simulation.

The recession limbs of the hydrograph couldbe simulated more accurately, but thisobjective was beyond the intentions of thisstudy. The REW model in its currentconfiguration does not consider processes with



time scales in between the ‘very fast’ directrunoff coming from the saturated overland flowzone and the ‘very slow’ groundwater flow fromthe saturated zone.

Future research and conclusionsThis application allows us to define thesubjects for future research.The first objective is improving the water tablelevel – saturated overland flow zone relation byimproving the manner topography is simulated.When considered that peak discharges aremainly dependent on the extension of thisarea, the role of topography is evident. Futureresearch will then focus on the derivation ofthe soil profile for each REW from geometricalproperties directly obtainable from the DEM ofthe study area.A second subject deals about the simulation ofthe recession limbs of the hydrograph. Thistask will probably require the implementation ofother components that consider additionalhydrological processes, such as rapidgroundwater flow, or perched subsurface flow.We conclude that the REW model for thisapplication has given satisfactory results, andit is a promising tool to simulate interactinghydrological processes.This work shows that catchment runoff is wellsimulated after model calibration by means ofpiezometric observations, indicating that thewater table level represented by the selected

piezometer graph is a good indicator of thecatchment behaviour.

AcknowledgmentsThe “Antonio Ruberti” foundation is thanked forits support of this study.

ReferencesReggiani, P., S.M. Hassanizadeh, M. Sivapalan &

W.G. Gray, 1999. A unifying framework forwatershed thermodynamics: constitutiverelationships. Adv. Water Resources. 23, pp. 15-39.

Reggiani, P. & T. Rientjes, 2003. TheRepresentative Elementary Watershed approachas an alternative modelling blueprint: applicationto a natural basin. Water Resources Research(submitted).

Reggiani, P., M. Sivapalan & S.M. Hassanizadeh,1998. A unifying framework for watershedthermodynamics: balance equations for mass,momentum, energy and entropy, and the secondlaw of thermodynamics. Adv. Water Resources.22, pp. 367-398.

Reggiani, P., M. Sivapalan & S.M. Hassanizadeh,2000. Conservation equations governing hill sloperesponses: exploring the physical basis of waterbalance. Water Resources Res. 36, pp. 1845-1864.

Rientjes, T., 1999. Physically Based Rainfall-Runoffmodelling applying a Geographical InformationSystem. Communications of the WaterManagement and Sanitary Engineering divisionReport no 83. Delft University of Technology,Delft, The Netherlands, pp. 74.



Environmental impact of flooding

E.E. van der Hoek1, L.C.P.M. Stuyt2, J.E.A. Reinders3 & M. de Muinck Keizer4

1 GeoDelft, P.O. Box 69, 2600 AB Delft, The Netherlands; [email protected] Alterra-Wageningen UR, P.O. Box 47, 6700 AA Wageningen, The Netherlands3 TNO-MEP, P.O. Box 342, 7300 AH Apeldoorn, The Netherlands4 Delphiro, Rotterdamseweg 183C, 2629 HD Delft, The Netherlands

In 2001 a Delft Cluster research project startedto quantify the consequences of a floodresulting from a river dike breach in theNetherlands. In the past much attention hasbeen paid to casualties, economical lossesand losses to buildings. One of the goals of theDelft Cluster project is to quantify theenvironmental consequences of the release ofpollutants during flooding. Only after thisinformation is known a realistic assessment ofthe relative importance of environmentaldamage can be made as compared toeconomic losses, structural damage andcasualties.

First an extensive literature study has beenperformed into the environmentalconsequences of floods that have occurred inindustrialised countries. Results particularemphasis on the relative importance ofaspects like water height, water velocity,

contaminated suspended sediments and typesof pollutants.In the Delft Cluster study a conceptualframework was defined, based on thesequence of events during flooding. This chainreaction is depicted and explained below(Table 1):• A dike breach releases water;• The scour of the water flow or the high water

level can damage or destroy source objects(constructions, houses, industrial complexesand farms);

• The damaged source objects may fail andsubsequently release chemicals and micro-organisms into the water;

• Pollutants will be dispersed through air,water and suspended matter;

• The polluted water and suspended matterwill affect people and ecosystems in theinundated area.

Table 1. Sequence of events.

To asses the environmental impacts as well asother impacts socio-economic, casualties anddamage to infrastructure and buildings a casestudy was used In this study a dike breach ofthe river Rhine near the city of Krimpen wassimulated and this causes inundation of theregion Rotterdam to The Hague.

The following simplifying assumptions weremade:• Toxic substances are stored in various types

of containers in-, and outside buildings, inthe soil and in cars; parameter Cs (g/m2);The volumes of pollutants that will bereleased from the source objects areestimated on the basis of expert knowledgeand information from the literature study.

• Two criteria for the release of toxicsubstances into the flooding waters wereused, labelled ‘v’ and ‘h’:

! High flow condition (‘v’): release willoccur once the water velocity v > 2m/s;

! High water level condition (‘h’): releasewill occur once the water height h > 1m.

• Substances, dissolved in water (Cw) (g/m3)are subject to transport, supply from the soil,sedimentation and decay; the decay issimulated as a first order process.

In the sediment- and flow models release,dispersion and sedimentation of the variouspollutants are simulated using a dedicated,integrated modelling tool (Sobek-Delwaqmodel of Delft Hydraulics) and also by ariverine sedimentation model. The maximumextent of the inundated area was 654 km2,storing 1.37*109 m3 of water. The ‘h’ criterionwas exceeded in a very large area, and isresponsible for the releases of most pollutants.The ‘v’ criterion was reached in a small areanear the dike breach only.

Dykebreach

and flood→

Sourceobject fails →

Pollutantis released →

Pollutantis

dispersed→

Pollutantaffectstarget

→Qualification and

quantification of theimpact

Initiating mechanism Dispersion mechanism Consequencemechanism



Both the Emerged Retention Areas (ERA) andthe Delwaq models were used to simulatetransport and sedimentation of suspendedparticles. Given 0.284 kg/m3 of suspendedmaterial in the river, approximately 388*106 kgof sediment (i.e. a uniform sediment layer of0.35 mm) was deposited in the flooded area(ρsediment ≈1,700 kg/m3). The sedimentationpattern is depicted in Fig. 1 and 2. The areaaround the dike breach shows a concentriczone were no deposition occurs because ofhigh stream velocities. Up to 6 kilometres fromthe dike breach, the layer thickness variesbetween 2-8 mm, in the lower parts of the areafrom 0.4-2 mm, elsewhere < 0.4 mm.

Figure 1. Deposition of sediment during theinundation stage, according to the ERA model[kg/m2].

Figure 2. Deposition of sediment during the settlingstage, according to the ERA model [kg/m2].

The results of the simulations show that thefloodwaters contain amounts of benzene,Perchlorethylen (PERC; DNAPL) andpesticides well in excess of Dutch Interventionvalues. These compounds accumulate at theborders of the flooded area. Clearly, oneshould be careful with the intake of surfacewater for drinking purposes during this time. Aslong as the flood lasts, most components arelikely to decay naturally to sub toxic levels.Simultaneously, growth of algae is likely,causing oxygen depletion and growth of toxicorganisms like blue algae. The deposition of

Figure 3. Concentrations of PAH’s, releasedfollowing the ‘v’-criterion, 10 days after the dikebreach.

polluted sediment may cause health risks andadverse effects on ecosystems.

As example the results for Poly-aromatricHydrocarbons (PAH) are shown. PAH’s willadsorb to and settle with floating particles. Thedeposition PAH’s as simulated with the Delwaqmodel is shown in Fig. 3 and 4. PAH’s,imported with the incoming Rhine water areresponsible for the bulk of the contamination.Erosion of (contaminated) particles within theflooded area with the Delwaq model wasseverely underestimated due to conceptualshortcomings of this model. This shortcomingof Delwaq resulted in an underestimation ofthe release of land based material, which canbe easily observed in Fig. 3, which only showsmaterial that was simulated to be releasedrelatively close to the dike breach. It can beexpected that PAH’s. (and other substances)will be deposited further away from the dikebreach as well. The contamination levels of thesediment slightly exceed the Dutchintervention value for PAK10, i.e. 40 mg/kg drymatter for a ‘standard’ soil.

Figure 4. Concentrations of PAH’s, present after 10days, either deposited after release subject to the 'h'criterion, or as unreleased source.



Fig. 4 clearly illustrates that a significant land-based source of PAH’s is formed by cars:motorways can clearly be identified.

Finally the impact of the pollutants on thethreatened objects in the area is determinedand cleanup costs are estimated. The totalclean-up cost are estimated to be M€ 350-550when a 10 cm layer of ground is removed is atall places were sediment is deposited. If not allsediment is removed a reduction of the landuse potential for agriculture due to Cd- and Zn-contamination is assessed using the Dutch‘LAC-alert values’. Mixing agriculture land mayeliminate possible risks in arable and pastureland to values below LAC-values.

Assuming that crop damage is 100% after 72hours continuous inundation, which is the caseat all agricultural fields in this simulation, thedamage was calculated by multiplying theproduce (€/ha) with the associated acreage(ha), yielding a total damage due to flooding ofM€ 853 (worst case scenario).

The model needs still improvement. Notable isthat small but numerous sources such as carscan give high amounts of toxic chemicals inwater and sediment during and after the flood,while large sources such as chemical plantsonly give problems near a dyke breach at highwater velocity.



Uncertainty assessment due to time series inputs usingdisaggregation for flood forecasting

S. Maskey & R.K. PriceUNESCO-IHE Institute for Water Education, P.O. Box 3015, 2601 DA Delft, The Netherlands; [email protected]

IntroductionThe disaggregation (or reconstruction) of data,especially time series data, observed at lowerfrequency into higher frequency is a well-recognised problem in water systemsmodelling. Common examples are the timeseries of quantities like precipitation and riverdischarges. For an example, precipitations areobserved in terms of accumulated sums for acertain period or are forecasted in terms ofuniform rates of precipitation depth per unittime. Normally, the accumulated sum isuniformly distributed throughout the period forwhich it is observed. When such data are usedfor flood forecasting, e.g. using rainfall-runoffmodels, the assumption of the uniformdistribution may introduce significantuncertainty in the estimated flows especiallywhen the period is large compared to thecatchment reaction time. This paper presents amethodology to assess the uncertainty in theforecasts due to the uncertainty in time seriesinputs using temporal disaggregation into subperiods. The characteristics of thismethodology are: (i) it is independent to thestructure of the forecasting model, and (ii) itcan be applied in the framework of the MonteCarlo (MC) method as well as the fuzzyExtension Principle (EP).

MethodologyThis methodology can be described into threesteps:1. Uncertainty in the magnitude of the input;2. Temporal disaggregation of the input;3. Propagation of the input uncertainty.

The uncertainty in the input quantity, whichcould be measured or forecasted, is explicitlytaken into account in this methodology.Depending on the framework used (MC or EP)for its implementation, the input uncertainty isrepresented by a Probability Density Function(PDF) or by a Membership Function (MF).The application of temporal disaggregation foruncertainty assessment is the main feature ofthis methodology. It consists in dividing thetemporal period over which the accumulatedinput is known into a fixed number of

sub periods and random disaggregation of theaccumulated sum over sub periods. Thedisaggregated input should aggregate up tothe accumulated sum. For simplicity,disaggregation coefficients bi,j (i = 1, …, m; j =1, …, n), are defined such that thedisaggregated signals are given by xi,j = Xibi,j.Where, n is the number of sub basin, m is thenumber of sub period and Xi is theaccumulated input quantity over the period forthe sub basin i. The coefficients bi,j are subjectto the constraints: 0 ≤ bi,j ≤ 1 (for all i,j) andbi,1 + … + bi,n = 1 (for all i). Allowing thecoefficients bi,j to vary over the sub periodsaccounts for different possible temporaldistributions over various sub basins, whereasvarying the coefficients over sub basinsaccounts for the spatial variations of the rainfallfield.Finally, the disaggregated signal is used as aninput to the forecasting model and theuncertainty is propagated using either the MCmethod or the EP. In the former approach, thedisaggregation needs to be repeated for eachvalue of the accumulated quantity obtainedfrom its PDF. In this case, the model outputuncertainty is also represented by a PDF. Thelater approach has been implemented by theα-cut method. This method requires anoptimisation algorithm to find the minimum andmaximum of the model outputs. In this study aGenetic Algorithm (GA) is used as anoptimiser. In this approach, the outputuncertainty is represent in the form of a MF.

ResultThe methodology is applied to a rainfall-runoffmodel of Klodzko catchment (Poland) takingthe precipitation as an uncertain input. Asample result obtained from the fuzzy EPmethod is presented in Fig. 1. In this examplethe time period is taken as 3 hours, which isdivided into three sub periods of one houreach. Fig. 1 shows the difference betweenupper bound and lower bound of the outputdischarges at different forecast hours with andwithout disaggregation. It is clearly seen that inall forecasts the output uncertainty is muchlarger with than without disaggregation.



Figure 1. Uncertainty range in forecast discharges (QUB − QLB) for α-cut level 1.

ConclusionThe results show that the output uncertaintydue to the uncertainty in the temporal andspatial distributions can be significantlydominant over the uncertainty due to theuncertain magnitude of precipitation. Thissuggests that using space- and time-averagedprecipitation over the catchment may lead toerroneous forecasts. The results also show the

great potential of the fuzzy extension principlecombined with a genetic algorithm foruncertainty propagation. The application of thismethod, however, requires a careful analysisof (i) the methods for the generation ofdisaggregation coefficients, and (ii) theselection of the appropriate number of subperiods.

0

50

100

150

42 54 66 78 90 102

Time (hours)

QU

B-

QL

B(m

3 /s)

Without disaggregation

With disaggregation



Is the southern Dutch delta the solution to future floodwater levels in the tidal river area?

N. SlootjesInstitute for Inland Water Management and Waste Water Treatment, P.O. Box 52, 3300 AB Dordrecht, The Netherlands;[email protected]

AbstractClimate change will lead to higher designdischarges en sea level rising. This isespecially important for the tidal river area ofthe river Rhine and Meuse in the Netherlands.This area will deal more often with high waterlevels and maybe even with flooding whennothing is done to increase the safety. Duringstorms at sea, the storm surge barriers areclosed to protect the Dutch inland and thewater levels in the tidal river area will risebecause of the river discharge. At that specificmoment it will be a measure to led water intothe southern Dutch Delta (from HollandschDiep to Volkerak-Zoommeer andOosterschelde). According to this study thewater levels in the tidal river area will decreasemaximum about 11 cm. Other studies showactually a decrease in water level with amaximum of 30 cm. It seems that thisdifference is caused by the different approachof the leaking discharge of the Oosterscheldestorm surge barrier.

IntroductionFlood protection in the Netherlands is basedon a maximum tolerable risk of flooding, whichis set at a probability of 1/1,250 per year.Based on this safety standard, the so-calleddesign discharge has been defined. This is thepeak flow that is exceeded on average onceevery 1,250 year. Climate change may lead toan increase of the design discharge and alsomay lead to a sea level rise. Depending on therate and magnitude of the assumed climatescenario the design discharge may increase 5to 20 % for the river Rhine and Meuse over thenext century relative to the present dayconditions. Sea level may rise up to 60 cm bythe year 2100. Increasing river discharges incombination with sea level rise is a futureproblem for the Dutch estuary. The chancesthat storm at sea and peak river discharges willtake place at the same time will increase.During storms at sea, the storm surge barriersare closed to protect the Dutch inland. Butwhat will happen if there is a peak riverdischarge at the same time? The western partof Holland will fill like a bathtub…Raising the river dikes is no longer considereda desired option, because increasing flood

water levels may lead to greater inundationdepth of the hinterland and can cause dikefailure. A possible solution to this problem is tolead water into the southern Dutch deltathrough the lake Volkerak-Zoommeer.The objective of this study is to investigate:• The effect at the water levels in the tidal

river area of this measure;• How often this event will take place;• What the water levels at the southern Dutch

delta will be and especially at Volkerak-Zoommeer.

Study areaFig. 1 shows the study area. The arrows showthe direction of the transition of water into thesouthern Dutch delta.

Figure 1. Study area.

MethodFor the model calculations the 1D hydraulicmodel SOBEK is used. This model containsthe rivers Rhine and Meuse from the border ofthe Netherlands to the North Sea including thesouthern Dutch delta. Model conditions for thecalculations are a design discharge of 18,000m3/s for the river Rhine and 4,600 m3/s for theriver Meuse. The sea level rise is 60 cm andthe calculations are made for 6 differentstorms. Transition of water into the southernDutch delta will be when the water level at theVolkerak-weir reaches 2.58 m above sea level.

ResultsThe effect on the water level in the tidal riverarea caused by the transition of water into thesouthern Dutch delta is showed in Table 1.Also a comparison is made to the preceding“Spankracht” study.

NCR-daysPresentationAward 2003



Table 1. Effect on water level in tidal river area caused by the transition of water into the southern Dutch delta.

The transition of water into the southern Dutchdelta during storm and high river discharge willtake place around every 32 years up to 2100.The chance that water levels at Volkerak-Zoommeer will be higher than 1 m is once in45 year and the chance that water levels willbe higher than 2 m is once in 700 years.

DiscussionThe effect on the water level in the tidal riverarea according to this study seems to be lesseffective than according to the “Spankracht”study. This is probably caused by a differentapproach of the leaking discharge of theOosterschelde storm surge barrier. The“Spankracht” study uses a leaking dischargethat results in an increase of water levels at theOosterschelde of about 5 cm/hr. This study

uses a certain leaking discharge calculatedwith the difference in water level at both side ofthe storm surge barrier.

ConclusionThe maximum effect at the water level in thetidal river area is 11 cm.The chance that transition will be needed isevery 32 yearsThe chance that water levels will be higherthan 1 m and 2 m is respectively 1/45 year and1/700 year at the Volkerak-Zoommeer.

Further research has to be done to the leakingdischarge of the Oosterschelde storm surgebarrier, the effect on ecology in the southernDutch delta and the draining problems for thesurrounding areas of the Volkerak-Zoommeer.

River Branch Location Reduction water level Reduction water level

Study Volkerak-Zoommeer “Spankracht” study

Nieuwe Maas Rotterdam 0.00 0.04

Oude Maas Dordrecht 0.04 0.21

Hollandsch Diep Rak-Noord 0.11 0.39



Dealing with hydraulic models from a user’s perspective

M. van RoodeRijkswaterstaat Directorate Limburg, P.O. Box 25, 6200 MA Maastricht, The Netherlands; [email protected]

AbstractThe regional Rijkswaterstaat Directorate inLimburg uses hydraulic models for severalpurposes, such as flood forecasting and theprediction of hydraulic effects of watermanagement projects. These models consistof three main components that are regularlyupdated: the software, the river schematisationand the boundary conditions. For somepurposes a frequent updating of thesecomponents is desired, but for other purposesit is disturbing. In order to tackle thiscontroversy and to clarify the process andresponsibilities related to model development,Rijkswaterstaat Limburg developed a modelmanagement plan.

IntroductionI represent the Rijkswaterstaat LimburgDirectorate in the field of water management. Iam not a researcher, but a client. For our workwe use hydraulic models that may have beendeveloped by some of you, researchers. Ibelieve it is a good idea that I, your client,inform you about the challenges we face whenusing these models.

Where do we apply hydraulicmodels?At Rijkswaterstaat in Limburg we use hydraulicmodels to predict discharges, flow velocities,flow directions and water levels for differentpurposes/clients. We use the Sobek (onedimensional) model for flood forecasting andmaster plan and feasibility studies, when quickresults are needed. When more precision isrequired, e.g. for determining design floodlevels (so called “normwaterstanden”), licencerelated effect predictions (so called “WBR-calculations”) and detailed design studies, weuse the Waqua (two dimensional) model.

ResponsibilitiesThe hydraulic models basically consist of threecomponents: the software, the riverschematisation (mainly geometry, hydraulicweirs and roughness values) and the boundaryconditions (e.g. upstream and lateral inflow,downstream water levels). It is not ourresponsibility to develop the hydraulic modelsoftware, this is for our counterparts, theresearchers. RIKZ is responsible for theWaqua software development and RIZA for

Sobek. We receive new versionsautomatically, sometimes several times a year.The boundary conditions are determined everyfive years by the DWW, anotherRijkswaterstaat research institute. Ourdirectorate is responsible for providing theinformation on river geometry (changes).

ChallengesAlthough the above mentioned responsibilitiesseem clear, we have spent a lot of time andenergy on discussing which model (version) touse, questioning the model output quality andthe status of output results and on initiating ad-hoc improvement actions.Another question is how to deal withcontradictory requirements with regard toupdating the model components. For floodforecasting we generally wish to use the mostrecent updated model, believing that thisprovides the most reliable predictions. But forpredictions or calculations that have a legalstatus (such as the determination of designflood levels or licence related effectpredictions), frequent updating of software,schematisation and/or boundary conditions isdisturbing. Frequent updating may lead todifferent model output, which may beconfusing, not credible and/or which may havenegative legal consequences. In a courtroomwe are not able to convince the judges that ourassessments are consistent, verifiable andindependent from time. Every request for alicence should be dealt with in a similar matter,as the law describes. Therefore, for officialpurposes a certain constancy in model outputis required.So the challenge was to make our internalmodel development process clear, to exactlydefine what we need and to plan accordingly.

The model management plan(“modellenbeheerplan”)In order to tackle the problems mentionedabove, Rijkswaterstaat Limburg developed amodel management plan (van Roode, 2002).The main topic is the policy on updatingsoftware, schematisations and boundaryconditions.The basis of the model management plan isthe division of our hydraulic model applicationsinto three categories:



1. Quick results and up-to-date (e.g. floodforecasting, master and feasibility studies);

2. Detailed and constant (e.g. design floodlevels, licence related predictions);

3. Detailed and up-to-date (e.g. designstudies).

In short, the model management planprescribes the Sobek model for category 1.This involves an annual installation of thelatest Sobek software and an annualevaluation of the river geometry changes(which may lead to updating the riverschematisation).For category 2 (for official purposes) we usethe Waqua model, updated 5-yearly. Weintend to use the Waqua HR model (softwareand schematisation) developed 5-yearly by theDWW for the official determination of theboundary conditions. After the HR model ishanded over to us, we install it, test it, andsubdivide the overall Meuse schematisationinto several smaller sub-schematisations.Apart from the official Waqua software version,we may install newer versions for non-officialpurposes to cover category 3. An updatedWaqua schematisation will then be retrievedthrough the annual Sobek updating process.

There is no rule without an exception, and thisis also the case for the model managementplan. In exceptional cases we may update theofficial Waqua software and/or schematisationbefore the 5 year period expires. This can onlybe done if the update has a large positiveimpact on model outcome and only aftersuccessful re-calibration of the model. Forfurther details I refer to the model managementplan.

ImplementationThe plan was not made for the bookshelf and itappears very helpful in discussions and inplanning processes. However, at times it is achallenge for the RWS organisation to stickwith the plan! Let me give you some examples

of new challenges we face that are notcovered in the plan:• What to do when the official Waqua HR

software contains bugs?• What to do when the official Waqua HR

schematisation appears not well calibrated?• Can we stick with the 5-year principle when

we are asked to provide design flood levels,knowing that they will change?

• What to do when our official design floodlevels disagree with those used and issuedby Directorate Maaswerken?

• When are calibration results acceptable?

Above questions have in some cases lead tore-thinking and fine-tuning of the modelmanagement plan.

ConclusionThe model management plan has createdmuch awareness within our organisation. It is avery good basis for discussion and makes(financial) planning and decision making mucheasier. However, not all aspects are entirelycovered. Therefore, the plan needs some fine-tuning and will be updated periodically.For you, researchers, I would like to give yousome – in my view - interesting finalconclusions:• We will not automatically install new

software versions!• We need clear info concerning new software

versions!• The 5-yearly HR models need to be good!

Reactions on this presentation are verywelcome; it can only make our plan better.

ReferencesRoode, M. van, 2002. Modellenbeheerplan,

handboek voor het beheer van hydraulischemodellen voor RWS Directie Limburg, versie 1.0(in Dutch). Rijkswaterstaat Limburg Directorate,Maastricht, The Netherlands.



Appropriate spatial sampling resolution of precipitation forflood forecasting

X.H. Dong, M.J. Booij & C.M. Dohmen-JanssenFaculty of Engineering Technology, Department of Water Engineering & Management, University of Twente, P.O. Box 217,7500 AE Enschede, The Netherlands; [email protected]

IntroductionFor the implementation of lumped conceptualmodels in flood forecasting, point precipitationtime series records need to be aggregated intoregionally averaged time series. Therefore aquestion arises: how many rainfall gaugesinside a specific river basin are needed toprovide sufficient precipitation records for therainfall-runoff models?The variability of rainfall recorded at a singlespot is generally much larger than that of thedischarge recorded at the outlet of the basinconsidered. This is quite understandableconsidering the damping effect of the rainfall-runoff transformation process. For a givenphysics-based hydrological model, the greaterthe diversity between the input (rainfall) andthe output (discharge), the greater difficulty willbe encountered in establishing the mappingfrom the input to the output. Therefore anymeasures, which can help to decrease thevariability of the input, or increase the similaritybetween the input rainfall, time series and theoutput discharge time series will promote theperformance of a calibrated physical model.The operation of aggregating the point-sampled rainfall time series into regional-averaged time series has the expectedsmoothing-out effect. One important aspectthat has to be addressed here is that thepurpose of the appropriate spatial sampling ofrainfall is not to master the regime of therainfall events as detailed as possible, but toserve the practice of flood forecasting.Therefore the appropriate spatial sampling isdefined as that which can possibly lead toaccurate forecasting of discharge in the riverchannel.The problem is attacked in two steps. The firstone is based solely on the statistical analysisof recorded rainfall and discharge data, tryingto identify the appropriate amount of raingages for flood forecasting. The second step isto verify the results obtained in the previousstep by running a physical hydrological model:HBV, which is developed by the SwedishMeteorological and Hydrological Institute(SMHI, 2003). Qingjiang river basin in China isadopted as the case studied in this research,and the area upstream of the Yuxiakou flowstation is considered here, as shown in Fig.1.

Figure 1. Qingjiang river basin.

Both rainfall and discharge data are measuredat 6 hours time interval.

Statistical analysis on rainfall-discharge dataVariances (square of the standard deviation) ofrainfall time series give an indication of thevariability of rainfall at a location or of a region.The recorded rainfall data on the neighbouringrain gages can be treated as mutuallycorrelated random variables if they are notseparated over long distance. The variance ofthe sum of these individual time series (whichis one method of estimating the aerial rainfall)is given by (Osborn & Hulme, 1997):

][ )1(122n

rnin sS −+= (1)

where 2is is the mean of the station variances;

s2n is the variance of the average of the

combination of n station time series; r is themean inter-station correlation between all pairsof stations within the basin which is calculatedfrom Fig. 2; n is the amount of rain gagesincluded in the aggregation.

As can be seen from equation 1, the variancewill be reduced with an increasing number ofrain stations n involved in averaging. This isproved by the curves in Fig. 3, from which canbe seen that the variance is reducinghyperbolically as n increases, but after acertain threshold it levels off, which implies thatthe effect of smoothing on the variability(variance reduction) is no longer significantwhen n is greater than about 10.

#

#

#

#

#

#

#

#

#

#

#

#

##

#

#

##

#

#

# #

#

##

##

#

##

#

#

#

#

#

#

#

#

#

#

#

#

#

#

#

$

$ $Yuxiakou



Figure 2. Correlation between pairs of point rainfalltime series versus their separation distance for allpossible combinations of station pairs in Qingjiangriver basin.

Figure 3. Variance of rainfall time series obtained byaveraging n station time series together for differentn.

With the decreasing variability of the rainfall, itis expected that the similarity between therainfall time series and discharge time serieswill be increased, which will benefit to therainfall-runoff modelling. This similarity ismeasured with the maximum cross correlationbetween rainfall time series (with different n)and discharge time series as shown in Fig. 4.

Figure 4. Maximum cross correlation (at time lag 3)versus different values of n.

The use of the concept of maximum crosscorrelation stems from the fact that thecorrelation between two time series differsaccording to different time lags. For the areaupstream to Yuxiakou the hydrologicalresponse time can be identified as 3 timeunites (18 hours) (see Fig. 5).

It is assumed that with the averaged rainfalltime series becoming more similar to thedischarge time series, as indicated by theincreasing values of correlation coefficients inFig. 4, the performance of a given hydrologicalmodel will be increased. Fig. 4 reveals that,when n is increased to 10, the correlationbetween averaged rainfall time series anddischarge time series hardly increases anymore, therefore 10 will be identified as theappropriate number of rain gages thatdetermines the appropriate spatial resolution ofrainfall sampling for the area upstream toYuxioakou in Qingjiang river. This result isverified by performing real flood forecastingwith HBV model.

Figure 5. Cross correlation between averagedrainfall time series (with different n) and dischargetime series for different time lags.

Verification with HBV modelThe area upstream to the Yuxiakou flowstation is treated as one sub basin as seen inFig. 1. The one-sub basin HBV model iscalibrated with rainfall, evaporation anddischarge data ranging from 1989 to 1996.Then it is validated with data from 1997 to1999. The effect of the amount of rainfallstations on the forecasting results is shown inFig. 6, where Nash-Sutcliffe efficiencycoefficient R2 and relative accumulateddifference between computed and recordeddischarge are used as the criteria to judge theperformance of forecasting. The validationresults compare favourably to the onesobtained in the preceding step.

0 50 100 150 200 2500

0.2

0.4

0.6

0.8

1

distance (km)

corr

elat

ion

coef

ficie

nt

0 5 10 15 20 255

10

15

20

25

30

number of rain gages

varia

nce

(mm2 /

(6hr

)

MeanMaximumMinimum

0 5 10 15 20 250.4

0.45

0.5

0.55

0.6

0.65


max

lag

cros

sco

rrel

atio

n

MeanMaximumMinimum

0 2 4 6 8 100.3

0.35

0.4

0.45

0.5

0.55

0.6

0.65

0.7

← 26 gages

time lag

corr

elat

ion

coef

ficie

nt

← 9 gages

← 3 gages

← 1 gage



Figure 6. Verification using HBV model.

ConclusionsAmong 27 rain gages used in the studied area,10 are found to be the effective amount forflood forecasting. This is deduced first from theobserved rainfall and runoff data, bycalculating the variance reduction effect of allcombinations of rain gages from one to the

total amount of the cluster, and theircorrelation with discharge time series. Theresult is further verified by running rainfall-runoff model-HBV, with only one combinationfor a certain number of rain gages. Althoughthe latter verification process is not as robustas the preceding statistical method (which isnot practically possible for running hydrologicalmodels), it provides a reasonable prove on thestatistical result from a different point of view.

ReferencesSwedish Meteorological and Hydrological Institute

(SMHI), 2003. Integrated Hydrological ModellingSystem (IHMS); Manual version 4.5. Norrköping,Sweden.

Osborn, T.J. & M. Hulme, 1997. Development of arelationship between station and grid-box raindayfrequencies for climate model evaluation. Journal ofClimate. 10, pp. 1885-1908.

0.6

0.65

0.7

0.75

0.8

0.85

0.9

0.95

1 3 6 9 12 15 18 21 24 27


R2

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

rela

tive

accu

mu

late

dd

iffe

ren

ce

R2 | RelAccDiff |



Searching for the right problem - a case study in watermanagement on the relation between information anddecision making

M.J. Kolkman, M. Kok & A. van der VeenSchool of Civil Engineering, Department of Integrated Assessment, University of Twente, P.O. Box 217, 7500 AE Enschede,The Netherlands; [email protected]

AbstractThe solution of complex, unstructuredproblems in integrated water management isfaced with policy controversy and dispute,unused en misused knowledge, project delayand failure, and decline of public trust ingovernmental decisions. Concept mapping is atechnique to analyse these difficulties on afundamental cognitive level, which can revealexperiences, perceptions, assumptions,knowledge and subjective beliefs ofstakeholders, and can stimulatecommunication and learning.

IntroductionIntegrated assessment consists of gathering,synthesising, interpreting, and communicatingknowledge from various expert domains anddisciplines, to help responsible policy actorsthink about problems and evaluate possibleactions. But in the decision-making discoursescientific information is not taken for granted,can be explained in different ways, and is ‘justanother’ element in the policy making process.And values, assumptions and limitations thatare inherent in scientific models are often notcommunicated explicitly. Furthermoreparadigm differences between actors fromdifferent stakeholder groups, between policyactors and scientists, and between scientistsfrom different discipline groups influence theuse and interpretation of information. But theassumptions and limitations present within aparadigm are seldom openly communicated.These difficulties result in lack of informationand insight on policy alternatives, lack ofexchange of information and communication,and lack of co-operation.Current theories about the relation betweenscience and policy give the followingrecommendations on the issue of knowledgeproduction and knowledge use: 1) knowledgemust not be produced from one singledominant paradigm, but from the whole rangeof paradigms that are present in the policyarena; 2) open debate is needed concerningchoices and basic assumptions, which underliethe production of knowledge; 3) debate shouldalso include non-scientific stakeholders fromthe policy arena, the intensity of this

communication depending on the complexity ofthe problem.The present research develops a newmethodology that may support integratedassessment in the light of the difficulties andrecommendations mentioned above.

Theoretical basis: Mental modelsWe start from the observation that in complex,multifunctional and multidisciplinary problemsthe meaning of information is sociallyconstruct, and guided by different frames ofperception (Funtowicz & Ravetz, 1994; Schon& Rein, 1994). Frames are the structures ofbelief, perception and appreciation underlyingpolicy positions. The frames held by the actorsdetermine what they see as being in theirinterests. Frames are grounded in theinstitutions that sponsor them. Framedifferences cause communication barriers thatprevent mutual learning and understanding.Policy controversies are seen as disputesbetween conflicting frames. It is within theframes that information is judged andsynthesised into a problem solution.Instead of analysing frames, the presentresearch follows Courtney (2001) andanalyses the mental models that underlieframes. A frame contains actors’ knowledge,assumptions, interests, values and beliefs. Butit is the mental model that determines whatdata the actor perceives in the real world (as a‘filter’ through which the problem situation isobserved), and what knowledge the actorderives from it. Therefore the perspective fromwhich alternative problem solutions aredeliberated en decided upon is ultimatelybased on an actor’s mental model. Differentmental models of the problem situation, andmismatch of decision data with the mentalmodels, will result in different opinions on theproblem solution, and in this way constitute thebasis of many problems in the policy cycle.Ambiguity and confusion exists in the definitionof ‘mental model’ between the manydisciplines that use this term. Doyle (1998)argues that it should be used to refer to only asmall subset of the wide variety of mentalphenomena to which it is often associated, andproposes the following definition (:17):



“A mental model of a dynamic system is arelatively enduring and accessible, butlimited, internal conceptual representation ofan external system whose structuremaintains the perceived structure of thatsystem”.

A mental model includes not only knowledgebut also information about interconnection andorganization of that knowledge (in nodes andlinks). A mental model does not include ends(goals), means (strategies, tactics, policylevers) and connections between them (themeans-ends model) – these are ‘inputs’ for themental model. ‘Running’ a mental model isequivalent to propagating information throughthe conceptual structure. The ‘model output’ isused to plan actions, explain and predictexternal events.

Elicitation of a mental modelNovak (1984) used the term ‘concept map’, todenote the external representation of themental model. This map is the researcher’sconceptualisation of a subjects’ mental model.All fields of research indicate that elicitation ofmental models will reveal the experiences,perceptions, assumptions, knowledge andsubjective beliefs that a ‘model user’ operatesto reach his conclusion about some issue.Concept mapping assess tacit knowledge,broadens the narrow understanding of aproblem by confronting one stakeholders’ mapwith the map of others, makes aware ofalternative perspectives on the problem,encourages negotiation and helps to reducedestructive conflict. The basic idea is toexternalise a person’s knowledge andconsequently make it discussable. This isprecisely how concept mapping may link to theneeds signalled by many authors in the field ofintegrated problem solving.Several tools are available to support conceptmap generation. The present research usesthe IHMC CmapTools from the University ofWest Florida (Coffey, 2000). This toolfacilitates generation and manipulation ofconcept maps, and allows map sharing overthe Internet. It has been used to supportlearning processes and expert systemdevelopment

Application of concept mappingThe Zwolle storm surge barrier is used as acase study to investigate the practicalapplicability of the methodological conceptsdescribed above. The research looks forconfirmation of the theory. The research startswith analysis of available documents, e.g. theenvironmental impact assessment report

(mer), and associated reports such as“Richtlijnadvies” and “Toetsingsadvies” of theEIA-commission, as well as research reportsand publication. Documents from the earlierphases of the project are available to monitorthe development of the conceptual model intime. Media messages are used to detailstakeholders’ concepts. Based on this analysisa first version of stakeholders’ conceptualmaps will be constructed. Based on thisselected actors will be interviewed to validateand refine the maps.

ConclusionThe complex, multifunctional andmultidisciplinary nature of problems causes alarge range of mental models to spring intoexistence. When all actors are not adequatelyinvolved early in the problem solution process,to share each others mental models, the (oftenimplicitly) developed mental model could beinsufficient to legitimise the preferred solution,and incomplete or even wrong knowledgecould have been produced in the project orselected for inclusion in the project report.Comparison of mental models, decisionprocess structure and actual use of knowledgewill reveal potential points of conflict, whichthen could be dealt with. Concept maps canalso identify blind spots in knowledge, givescientists clues they need to produceknowledge that fits into the frames of thediverse stakeholders in order that knowledgethey produce can be of use to thestakeholders, enlarge insight in possible anddesirable problem solutions, and supportcommunication between actors. Applyingconcept-mapping techniques in the earlyphase of decision-making for these purposesthus could improve the problem solving anddecision making process.A main advantage of the analysis of mentalmodels above the analysis of frames is theunchallenged institutional and normativeposition of the actors, because conceptmapping does not doubt the validity of anactor’s frame, but merely wants it illuminate itby focusing on the information used within aframe. Of course, this can be the starting pointof a learning process or critical dispute.

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

knowledge management in inquiringorganizations: toward a new decision-makingparadigm for DSS. J. Decision Support Systems.31, pp. 17–38.

Doyle, J.K. & D.N. Ford, 1998. Mental modelsconcepts for system dynamics research. Systemdynamics review; Journal of the SystemDynamics Society. 14(1), pp. 3-30.



Novak, J.D. & D.B. Gowin, 1984. Learning How toLearn. Cambridge University Press, Cambridge,England.

Coffey, J.W. & A.J. Cañas, 2000. A LearningEnvironment Organizer for AsynchronousDistance Learning Systems. Proceedings 12th

IASTED International Conference Parallel andDistributed Computing and Systems. Las Vegas,US.

Funtowicz, S.O. & J.R. Ravetz, 1994. Uncertainty,complexity and post-normal science.Environmental Toxicology and Chemistry. 13(12),pp. 1881-1885.

Schön, D.A. & M. Rein, 1994. Frame Reflection -Toward the Resolution of Intractable PolicyControversies. Basic Books, New York.

Grensmaas laag water



Combining integrated river modelling and agent basedsocial simulation for river management; The case study ofthe Grensmaas project

P. Valkering1, J. Krywkow1,2, J. Rotmans1 & A. van der Veen1,2

1 International Centre for Integrative Studies (ICIS), University of Maastricht, P.O. Box 616, 6200 MD Maastricht, TheNetherlands; [email protected]

2 School of Civil Engineering, Department of Integrated Assessment, University of Twente, P.O. Box 217, 7500 AE Enschede,The Netherlands

AbstractIn this paper we present a coupled IntegratedRiver Model – Agent Based Social Simulationmodel (IRM-ABSS) for river management. Themodels represent the case of the ongoing riverengineering project “Grensmaas”. In the ABSSmodel stakeholders are represented ascomputer agents negotiating a rivermanagement strategy. Their negotiatingbehaviour is derived from the so-called Theoryof Reasoned Action. The Integrated RiverModel represents stakeholder knowledge bydescribing possible long-term impacts of rivermanagement options such as broadening,floodplain lowering, and dike building. Thecomputer agents are allowed to specify valuesfor a set of ‘uncertain parameters’ in the IRMfor representing subjective stakeholderknowledge. We show how the coupled modelframework can aid to assess the robustness ofriver management strategies, both with respectto environmental uncertainty and societalsupport.

IntroductionRiver management is a typical example of acomplex problem involving a variety ofstakeholder interests and fundamentalenvironmental uncertainties. The Dutchgovernment aims to take the different interestsand views explicitly into account by allowingstakeholders to participate in the planningprocess. The aim of our research is to analysestakeholder influence on the decision-makingprocess, and its possible consequences for theriver system, using Integrated Assessmentmodelling and Agent Based Social Simulation.

MethodsThere are numerous mutually comparabletheories of individual and collectivestakeholder behaviour, stemming from thefields of social and individual psychology,economics, and artificial intelligence. In thispaper we apply the so-called Theory ofReasoned Action (Ajzen & Fishbein, 1980).According to this theory the individual'sintention to perform a behaviour is determined

by its attitude (personal desire) towardsperforming the behaviour and a subjectivenorm, see Fig. 1.

Figure 1. Relations among beliefs, attitude,subjective norm, intention, and behaviour accordingto the Theory of Reasoned Action (After Ajzen &Fishbein, 1980).

The Integrated River Model concept isdepicted in Fig. 2. The model concept isimplemented in computer code as a cross-section model of the Meuse at Borgharen. Themodules are based on basic principles of riverengineering (Jansen, 1994), groundwaterdynamics (Strack, 1989), nature development(Maaswerken, 1998), etc.

Figure 2. The concept of the Integrated River Model.

The concept of model coupling can bedescribed as follows (Krywkow, 2002). The‘behavioural beliefs’ of agents (see Fig. 1) arerepresented as their ‘perspectives onuncertainty’: value settings for uncertain IRMparameters related to climate change,hydraulic roughness, morphologicaldevelopments, and costs and benefits. Theagents feed these settings into the IRM to



Table 1. Egocentric strategies for the salient stakeholders of the Grensmaas project. The strategy variables applyto a river cross-section.

calculate values for a number of decision-making criteria for a given river managementstrategy. These values form the basis of theiroutcome evaluation of the river managementstrategy.

ResultsAs a first step towards model experiments wehave determined ‘egocentric’ river engineeringstrategies for the salient stakeholders of theGrensmaas project, see Table 1.The egocentric strategy of an agent is obtainedby maximizing over its attitude, given itsperspective on uncertainty, while neglectingsubjective norms. Obviously, no one of theegocentric strategies receives broad societalsupport (see Fig. 3), the strategy of the policymaker being still the most accepted option.

Figure 3. Stakeholder attitude towards the differentegocentric strategies. A value of 1 indicates a highlyfavourable position towards the strategy, 0 indicatesa neutral position, and –1 indicates a highlyunfavourable position.

In particular, the positions of the natureorganization and gravel extractor deviate.Their disagreement, however, does not resultfrom conflicting goals, but rather from adifference in uncertainty perspective. This isrepresented clearly by Fig. 4, which shows ahuge difference in the estimations of costs andbenefits of river engineering measuresbetween the gravel extractors and the otherparties involved.

ConclusionWe have shown an ‘egocentric’ model ofstakeholder behaviour coupled to a schematicriver model. This methodology seemspromising to assess robustness and societalsupport of a broad range of river managementoptions and reveals underlying motives forstakeholder disagreement. Furthermore, themethod may be used to elicit stakeholder goalsand perspectives on uncertainty, and enhancecommunication. Further developments willinclude participatory model use, and thedevelopment of more advanced models ofcollective action and negotiation to assessplausible outcomes of operational rivermanagement.

Figure 4. Decision-making criteria for the egocentricstrategy of the nature organization according todifferent perspectives on uncertainty. The valuesare scaled with respect to a target value.

ReferencesAjzen, I. & M. Fishbein, 1980. Understanding

Attitudes and Predicting Social Behaviour.Prentice-Hall, London.

Jansen, P.P. (Ed.), 1994. Principles of RiverEngineering: The non-tidal alluvial river. DelftseUitgevers Maatschappij, Delft, The Netherlands.

Krywkow, J., P. Valkering, A. van der Veen & J.Rotmans, 2002. Agent-based and IntegratedAssessment Modelling for Incorporating socialDynamics in the Management of the Meuse in theDutch Province of Limburg. Proc. 1st Biennialmeeting of the IEMSS, Lugano, Switzerland.

Maaswerken, 1998. MER Grensmaas, Deelrapport4: Natuur (in Dutch). Maastricht, The Netherlands.

Strack, O.D.L., 1989. Groundwater Mechanics.Prentice Hall, London.

Strategy variables / Stakeholders Citizen Gravel extr. Policymaker Nature org. Farmer

Summer bed deepening (m) 0 4 0 0 0

Summer bed broadening (m) 250 125 250 250 0

Length of floodplain lowering (m) 0 250 500 500 0

Dike building (m) 1 0 1 0 0

Clay shield surface (m) 0 2 0 0 0

Nature area (m) 250 0 750 750 0



Holocene fluvial dynamics in the Geul river catchment

J.J.W. de MoorDepartment of Quaternary Geology and Geomorphology, Faculty of Earth and Life Sciences, Vrije Universiteit, De Boelelaan1085, 1081 HV, Amsterdam, The Netherlands; [email protected]

IntroductionThe small, pristine, tributary valleys of theMaas River at the northern edge of theArdennes offer excellent conditions foragriculture, drinking water extraction andnature protection. Water managers and naturedevelopers agree that a reconciliation ofcommercial use and conservation of theseriver valleys is important. In order to support asustainable development of the valuabletributary valleys of the Maas River, we mustunderstand the past and ongoing processes inthe river valleys and assess the impact of themain factors that influence those processes.The aim of this research is to reconstruct theHolocene evolution of the Geul River valleyand to determine, quantify and model theimpact of changes in climate and land-use onfluvial processes and catchment developmentfor the Geul River. This will be done by fieldand modelling studies. In this paper wepresent the first results from field studies onthe fluvial dynamics of the Geul River.

The Geul River catchmentThe Geul River originates in eastern Belgium(near the German border) and flows into theMaas River west of Meerssen. Its length is 56km and the catchment area is about 350 km2

(Fig. 1). The gradient decreases from 0.02 mm-1 in the source area to 0.0015 m m-1 nearthe confluence with the Maas River. Theaverage discharge near Meerssen is 3.4 m3s-1

with occasional peak discharges of about 45m3s-1.

Figure 1. The Geul River catchment.

Present-day fluvial processes are dominatedby lateral erosion and sedimentation on point-bars (Fig. 2). These processes are especiallyactive during peak discharges, when the riveris capable of transporting coarse sedimentsand several meters of lateral erosion canhappen during a short high discharge event.

Figure 2. The present-day Geul River withsedimentation on the point-bars and steep cut-banks.

Methods and resultsSeven detailed cross-valley transects werecored between Gulpen and the Dutch-Belgianborder (Fig. 1). The drillings were performedusing an “edelman” hand auger. Sedimentswere described every 10 cm until a gravellayer was reached. The cross-sections showseveral distinct fluvial/sedimentological units(Fig. 3, the “Schoutenhof” transect):• A basal gravel unit of Pleistocene origin

(Van de Westeringh, 1980), which has beenreworked by the Geul throughout theHolocene.

• A medium to coarse loamy sand layer,sometimes containing coarse organicmaterial (branches, leaves): these are point-bar deposits or sometimes channel-filldeposits. Also finer point-bar deposits arefound (loams).

• Organic loam and silt loam sediments arechannel-fill and/or backswamp sediments.

• The upper unit in the valley consists of finegrained (silt loam) floodplain deposits.

• Silt loam with small fragments of brick andcharcoal: this is colluvium and is depositedon the transition of the valley bottom to thevalley slope, often in the form of alluvial fans(not indicated in the “Schoutenhof” transect).

The Netherlands

Belgium

Kelmis

Plombieres

Gulpen

Valkenburg

Meuse River

Epen

N

0 2.5 5 km

“Schoutenhof”transect



Figure 3. The “Schoutenhof” transect.

DiscussionThe cross-valley transects show a uniformdistribution of Holocene fluvial deposits in theriver valley. The presence of point-bar andchannel-fill deposits throughout the wholevalley indicates a widespread lateral migrationof the Geul River as well as reworking of thesediments by the river. Lateral erosion mainlytakes place during high discharge events.Stam (2002) has shown that this lateralerosion can be as fast as 30 meters in 25years. A future increase in high dischargeevents will probably accelerate the lateralerosion rates and thereby intensify thereworking of the sediments. This processclearly will affect landowners, who are indanger of loosing their arable land due to therapid lateral migration.The point-bar deposits are covered by 1.5 – 2m of floodplain deposits. The main source forthe fine floodplain sediment is loess from thevalley slopes. The presence of these fine,loess-like, deposits in the river valley supportsthe idea that intense deforestation in historictimes caused severe erosion of loess from thevalley slopes (e.g. van de Westeringh, 1980).

The loess was transported as colluvium to theriver valley, where sedimentation andreworking of the sediments took place. Due tothis massive input of fine-grained sediments,the Geul River probably changed from abraided river to a meandering river. Thischange clearly illustrates that land-use (i.e.deforestation) has a big impact on the river.The results obtained from the field studies willbe used as input for the modelling of historicand future scenarios of valley development.This combined approach of field and modellingstudies will provide a detailed insight in thedevelopment of a small tributary of the MaasRiver and its sensitivity to changes in climateand land-use.

ReferencesStam, M.H., 2002. Effects of land-use and

precipitation changes on floodplain sedimentationin the nineteenth and twentieth centuries (GeulRiver, The Netherlands). Spec. Publs. Int. Ass.Sediment. 32, pp. 251-267.

van de Westeringh, W., 1980. Soils and theirgeology in the Geul Valley. MededelingenLandbouwhogeschool Wageningen. 80, pp. 1-25.

110

111 m+ NAP

109

108

107

111 m+ NAP

110

109

108

107

22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1

EW

Silt loam

Medium to coarse loamy sand

Gravel

Loam

Humic silt loam and loam

25 m

Geul River



The RIVER 21 concept: envisioning the future ofinternational river basins

L.L.P.A. Santbergen & J.M. VerhallenHydrology and Quantitative Water Management Group / Integrated Water Management Team, Wageningen University, NieuweKanaal 11, 6709 PA Wageningen, The Netherlands; [email protected]

AbstractThe RIVER 21 concept, as developed byuniversities in France, Belgium and theNetherlands, is a tool for teaching vision-building processes on international riverissues. The concept is based on a combinationof traditional knowledge transmission andinteractive, outdoor learning activities in aninternational context. In a two weeks project,staff and students experience important stepsof a decision making process in a multi-stakeholder context with absence of centralauthority. The concept inspires students to bevisionary thinkers and learn how to deal withuncertainties of the future. Vision buildingenables stakeholders to share information andto reach a common understanding of stakesand goals. It can be a tool for planners bylooking at an entire river basin system andstructuring problem solving. Finally, visionbuilding is important for politics: a sharedvision makes it easier to hold stakeholdersaccountable.

IntroductionDecision making processes on internationalriver issues are highly influenced by the hydro-geographic context (upstream-downstreamrelationships), the absences of centralauthority, the presence of multi-levelnegotiation games (multinational, bi-national,intra-national, inter- and intra-organisational),socio-economic characteristics, powerbalance, institutional and cultural differences(Clevering, 2002; Santbergen, 2000; Meijerink,1999).Experiences from the Scheldt River Basin,shared by France, Belgium and theNetherlands, learn that the InternationalScheldt Commission (installed in 1998) mainlydeals with unstructured, wicked problems inwhich no consensus on values and knowledgehas been reached and ill-structured problemsin which consensus only on knowledge (andnot on values) exists (applied after de Bruijn &ten Heuvelhof, 2002; Table 1).Related to the Scheldt river basin, Meijerink(1999) speaks about a pluricentric perspectiveof decision making in which interdependentstakeholders play games in multi-level

Table 1. Four types of policy problems (de Bruijn &ten Heuvelhof, 2002).

networks, driven by self-interest andmaintaining autonomy.Above all, a lack of political ambition of theriparian states on a shared (and supported)long term vision on sustainable developmentand management of the Scheldt river basin,seems to hinder progress, more than theimpact of historical grown distrust, languagebarriers and cultural differences (Santbergen,2000). Or to quote a former Dutch chairman ofone of the working groups: “Now, when welook back at the first five years of theInternational Scheldt Commission, I think wewill have to admit that we have been too blindto our common interests”.

Expectations are that the European WaterFramework Directive, aiming at river basinmanagement plans for all European riverbasins, will cause a window of opportunities fortransboundary river basin commissions like theInternational Scheldt Commission. The riverScheldt pilot project on testing the guidance’son the implementation of the Water FrameworkDirective will improve the internationalcooperation on integrated water managementand will result in a shared internationalmanagement plan for the entire river Scheldtdistrict (Scaldit, 2003). According touniversities in the Scheldt river basin, such ariver basin management plan should be basedon a shared long-term vision of allstakeholders involved. The first step inachieving future cooperation is to trainstudents and young professionals in‘transboundary river basin thinking’ (Ruijgh-vander Ploeg & Verhallen, 2002). Therefore, theENGREF Montpellier Center of the NationalSchool of Water Management and Forestry,the University of Ghent, the University of

No consensuson values

Consensus onvalues

No consensuson knowledge

Unstructuredproblems

Moderatelystructuredproblems

Consensus onknowledge

Ill-structuredproblems

Structuredproblems



Antwerp, Delft University of Technology andWageningen University and Research Centre,developed a concept in which students andstaff undergo a two weeks vision-buildingprocess together. This concept, RIVER 21(aiming at envisioning the future of the world’srivers in the 21st century), has been developedand applied in the Scheldt river basin (2000,2001, 2002 and 2003) and partly in the Tiszariver basin (shared by Romania, Ukraine,Hungary and Yugoslavia; 2002).

Materials and methods

Figure 1. The River 21 concept (in: Ruijgh-van derPloeg & Verhallen, 2002).

Fig. 1 summarizes the RIVER 21 concept andincludes the following steps:• Teaching principles of integrated water

management, policy analysis and systemsanalysis (at each individual university).

• Preparations in national teams: literaturereview by staff and students, interviews withstakeholders.

• Joint excursions and meetings withstakeholders in the river basin: in the field,students are introduced to the major water-related issues of an Europeantransboundary river and asked to explorethe future of water management in the basinin a systematic manner, together with theirEuropean peers.

• Systems analysis in international sub-groups aiming at formulating shared goalsfor several aspects of river basinmanagement.

• Integration of systems analysis for thedifferent aspects in a plenary session.

• Scenario analysis in international sub-groups aiming at identification of drivingforces and possible futures.

• Design of desirable futures for the differentaspects in international sub-groups.

• Negotiation and decision-making on oneshared vision in a plenary session.

• Presentation of the vision to the involvedstakeholders of the riparian states.

ResultsLessons learned so far are:• The RIVER 21 concept is a good instrument

to learn how to deal with cultural diversityand to express oneself in a not nativetongue.

• The RIVER 21 concept offers possibilities todevelop negotiating skills; to experienceone’s own strengths or weaknesses innegotiation and the gap between one-sidestatements and common interests.

• A lot of participants were not familiar withvision building and learned that the systemsanalysis and vision building language is nota common language. For example: ways oflearning at universities in Hungaria andRomania are different than in theNetherlands.

• The multi-disciplinary and visionaryperspective was new for most participantsand the usefulness of the underlyingsystems analysis was acknowledged as atool to structure available information and tocome to joint fact-finding.

• There is not one central way to integratedriver basin management. Involvement ofdifferent disciplines doesn’t automaticallylead to an integration of knowledge from α,β and γ sciences.

• By applying the RIVER 21 concept, views,issues and interests of the different ripariancountries become clear; students bring innew and fresh ideas but also are tempted todefend their own countries interests.

DiscussionThe RIVER 21 concept is no new method or ablueprint, but a concept in which existingmethods like systems and scenario analysisare combined in a multi-stakeholder context.The essence of the approach is thatstakeholders undergo an entire processtogether, from problem definition to envisioningthe future and defining actions. In this sense,the concept can be seen as an active form ofpublic participation as mentioned by theEuropean Water Framework Directive (article14). Although only completely tested byuniversity staff, students and youngprofessionals in the Scheldt river basin, theconcept can be applied in other transboundaryriver basins as well as by planners, decision-makers, scientists and other stakeholders atdifferent institutional, spatial and temporalscales and at strategic, normative and/or

1 shared vision

Dissemination and debate

A sharedframework of

reference

Shared goalsfor 4 aspects of

water management

Desirable futuresfor 4 aspects

Possible futures

Ideas of the vision-building team members

Excursion throughentire river basin

Preparation innational teams

Problem analysis

Negotiation

Systems analysis

Identification ofdriving forces

3 days

1 day

2 day

1 day

1 day

Design of desirablefuture

1 day

1 week

3 daysWriting and presentation

Expert

consulatation

Expert

consulatation

Stakeholder

consulatation



operational levels. Students and youngprofessionals are the water managers oftomorrow. Unfortunately, universities do nothave much of interactive, transboundaryprogrammes in regular their bachelor-mastercurricula. The problem of how to secure theseintensive courses needs to be addressedurgently.

Conclusions and recommendationsTransboundary water management issues canbe tackled by considering the river basin as aunit including everyone affected, and, whiletemporarily ignoring boundaries, discoveringthe issues and possible solutions for the future.Building a vision challenges us to be creativeand allows us to dream. According to thestudents, vision building enables stakeholdersto share information and learn from each otherin order to reach a common understanding ofthe stakes and goals. Vision building can be atool for planners: it structures problem solvingwhen spatial scales are large and time scalesare long. It demands that planners look at theentire system. Vision building is important forpoliticians; a shared vision makes it easier tohold stakeholders accountable. According tothe staff, the concept can help delegationleaders of international river basincommissions at the highest levels to overcomebusiness as usual, if they are willing andpolitical supporting the creation of a sharedvision by actively participating in the process.The concept can be further developed andimproved in transboundary river basins withdifferent hydro-geographic, political-institutional, cultural and socio-economic

contexts. In 2004, the fifth Scheldt edition willtake place involving university staff, studentsand stakeholders from all riparian states(including Brussels and Wallonia). Plans aredeveloped to apply (and improve) the conceptin former Soviet states and accessioncountries to the European Union.

ReferencesBruijn, H. de & E. ten Heuvelhof, 2002. Policy

analysis and decision making in a network: how toimprove the quality of analysis and the impact ondecision making. Impact assessment and ProjectAppraisal. 20(4), pp. 232-242.

Clevering, H.-J., 2002. Indicators of transboundaryriver conflicts. A comparative analysis on conflictindicators for ten large river basins. MSc. thesisIntegrated Water Management (number 54).Wageningen University and Research Centre,Wageningen, The Netherlands.

Meijerink, S.V., 1999. Conflict and cooperation onthe Scheldt River Basin. A case study of decisionmaking on international Scheldt issues between1967 and 1997. Kluwer Academic Publishers,Rotterdam, The Netherlands.

Peteghem, M. van et al., 2003. Scaldit. Een projectin de schoot van de ISC met steun van InterregIIIB NEW: 2003-2005. Een internationaalactieprogramma voor een schoner en veiligerstroomgebieddistrict van de Schelde (in Dutch).Vlaamse Milieu Maatschappij, Aalst, Belgium.

Ruijgh-van der Ploeg, T. & J.M. Verhallen, 2002.Envisioning the future of transboundary riverbasins: with case studies from the Scheldt riverbasin. Wageningen University and ResearchCentre, Wageningen, The Netherlands.

Santbergen, L.L.P.A., 2000. Het internationaleSchelde overleg: culturen maken het verschil (inDutch). In: Bos, C.F. (Ed.) Zeelandboek, deel 4.Yerseke, The Netherlands, pp. 165-177.



Extreme floods in the Meuse basin over the past century:aggravated by land-use change?

M. Tu1, M.J. Hall1, P.J.M. de Laat1 & M.J.M. de Wit2

1 UNESCO-IHE Institute for Water Education, P.O. Box 3015, 2601 DA Delft, The Netherlands; [email protected] Institute for Inland Water Management and Waste Water Treatment, P.O. Box 9072, 6800 ED Arnhem, The Netherlands

IntroductionIn the recent decade, Western Europe hasexperienced frequent floods. Concern is risingamong the public about the effects of climatevariability and land-use changes. The Meuseriver is one of two largest rivers in TheNetherlands and experienced severe floods in1993/1994, 1995, 2002 and 2003. It is oftenassumed that rapid land-use changes in thebasin since the 1950’s have aggravated thefloods in the river. This possibility has providedthe motivation to analyse the flood peaks in theMeuse river and causal extreme precipitationin the basin during the past century. Thehistorical land-use changes were also tracedout. The present paper summarises thepreliminary results.

Data and methodsStudy sites and dataThe Meuse river is roughly classified as rain-fed, with a total drainage area of about 33,000km2 (see Fig. 1). Extreme floods in the riveroften occur in the winter season. Borgharenstation (with a drainage area of about 21,260km2) is normative in The Netherlands and hasa long record of daily discharge (1911-2002)for statistical analysis. Because downstream ofLiège (Belgium) several canals extract asignificant percentage of water from the Meuseriver, the discharge of the ‘undivided’ Meuse atMonsin (Belgium) has been reconstructed(1911-2000) for analysis too (RWS, personalcommunication). Other selected gaugingstations in the tributaries include: Membre inthe Semois, Stah in the Roer (Rur), Nekum inthe Jeker and Meerssen in the Geul (see Fig.1). Membre station has a relatively long recorddating from 1929, while the other stations havethe reconstructed records starting from the1950’s.

The long records of daily precipitation (1911-2002) from seven Belgian stations werecollected (see Fig. 1). The daily values forthese stations were simply averaged (notweighted) to obtain the estimates of relativedaily values of aerial precipitation in the basinupstream of Borgharen, although the Ardenneshas the largest annual precipitation in thebasin.

Figure 1. The Meuse basin and locations of theselected stations.

Hydro-meteorological variablesExtensive examination concentrated on theflood peaks and volumes at Monsin, and theextreme precipitation depths in the basin. Thehydrological years (November to October) andhydrological winter (November to April) andsummer (May to October) half-years wereconsidered separately. The derivedhydrological variables for Monsin includeannual maximum k-day moving averagedischarge (k = 1, 3, 5, 7, 10, 15 and 30)(denoted as AMAXD, AMAX3D, …,AMAX30D) as well as maximum k-day movingaverage discharge for both the winter half-year(denoted as WMAXD, WMAX3D, …,WMAX30D) and summer half-year (denoted asSMAXD, SMAX3D, …, SMAX15D). For othergauging stations, only AMAXD, WMAXD andSMAXD are included. For the aerialprecipitation, annual extreme k-dayprecipitation depth (denoted as AEXTP,



AEXT3P, …, AEXT30P) was derived as wellas extreme k-day precipitation depth for boththe winter half-year (denoted as WEXTP,WEXT3P, …, WEXT30P) and summer half-year (denoted as SEXTP, SEXT3P, …,SEXT15P). The water year is designated bythe ending year.

Statistical methodsSpearman’s rank correlation method wasapplied to evaluate the absence of linear trendin the time series. Then, change-point analysiswas performed by both the Pettitt test (Pettitt,1979) and the SNHT test (Alexandersson,1986; Alexandersson & Moberg, 1997). Aprobability of 80% was used to judge thesignificance of change-points in the Pettitt test,while the 90% level was given preference inthe SNHT test. A change-point year in theseries refers to the first year after the change.

Land-use changesOver the past centuries, agriculturaldevelopment has brought about a radicalchange of land covers in the Meuse basin,generally from forests to agriculture andpastures. However, during the 20th century, theforest area has been stable in Wallonia,Belgium, where most of the Ardennes riverscontribute to the extreme floods in the Meuseriver, and even shows a slight increase datingfrom the year 1984. Of the forest area, thepercentage of the conifer production forest hasobviously increased, e.g. from 14% in 1895 to50% in 1984, while the percentage of thebroad-leaved forest has dropped (DGRNE,

2000). Lorraine-Meuse in the upper Frenchpart also exhibited a slight increase in forestedarea. After the late 1940s, intensification ofagricultural land and rapid urbanisation havecontributed the most important land-usechanges in the middle and lower part of thebasin. The period 1950-1960 is often seen asa turning point corresponding to theindustrialisation of farming and to considerableurban development and to the expansion oftransport infrastructure. In Wallonia, the arableagricultural area had obviously decreasedbetween the 1950’s and the 1970’s, slightlyreduced during the 1980’s, but thereaftershowed a marginal increase. The extent ofurbanisation has increased from 10.5% in1980 to 12.7% in 1996.

Test resultsFlood peaks and volumesNo significant upward trend was found for theflood peaks and volumes at Monsin over thepast century. Since 1984, AMAXD andWMAXD at both Borgharen and Monsin havesignificantly increased. However, the floodvolumes at Monsin with increasing k-values (k≥ 5) did not show such abrupt change. For thetributaries, the Semois and the Roer exhibiteda significant increase in AMAXD and WMAXDafter the end of the 1970’s, while the Jekershowed such an obvious increase only inSMAXD. Fig. 2 shows the change-point resultsfor WMAXD in these rivers. As an exceptionalcase, a decrease after 1971 was observed forthe flood peaks in the Geul.

Figure 2. Change-point results for the winter maximum daily discharge (WMAXD) in the Meuse basin.

0

1000

2000

3000

4000

1910 1925 1940 1955 1970 1985 2000

Disc

harg

e (m

3 /s)

Borgharen, the Meuse

1339

1771

19840

50

100

150

200

1950 1960 1970 1980 1990 2000

Dis

char

ge (m

3 /s)

Stah, the Roer

66

88

1980

0

100

200

300

400

500

1925 1935 1945 1955 1965 1975 1985 1995 2005

Disc

harg

e (m

3 /s)

Membre, the Semois

153

244

1979

0

10

20

30

1950 1960 1970 1980 1990 2000

Disc

harg

e (m

3 /s)

Nekum, the Jeker

8

10

1981 (Insignificant)



Extreme precipitation depthsIn the long term, most winter extremeprecipitation depths in the basin (k ≤ 10)showed an upward trend with high probabilities(> 90%) or were marginally significant (only forWEXT5P). Except for AEXTP and AEXT30P,most annual extreme series exhibited asignificant increase after 1980, which was alsofound for SEXT5P, SEXT7P, SEXT10P andWEXT7P over the shorter period post-1950.Winter extreme series with k-values ≤ 10, withthe exception of WEXT5P, showed an abruptincrease in the 1930’s.

Antecedent precipitation depthsFurther investigation was made on the effectsof the k-day precipitation depths (denoted asANTP, ANT3P, …, ANT30P) in the basin priorto and on the day of WMAXD at Monsin. Asignificant increase since 1984 was alsoidentified for ANT3P, ANT5P, ANT7P andANT10P over the full period, of which ANT7P,closely followed by ANT10P, showed strongercorrelation (r = 0.771) with WMAXD at Monsin.Statistical tests were also carried out for therunoff coefficients for Monsin during differentantecedent k-days (k = 3, 5, 7, 10), which werederived based on the Rational Method. Nosignificant change-points were shown in theseseries.

Discussion and conclusionsThe test results indicate that the flood peaks inthe Meuse river have significantly increasedsince the 1980’s. On a global scale, the largestchange in terms of land area, and arguablyalso in terms of hydrological effects, is fromafforestation and deforestation. However, theforest area in the upstream Meuse basin haschanged little over the past century and mostchanges in the forest type, agricultural landand urbanisation occurred before the 1980’s.Therefore, the global tendency for the land-usechanges in the upstream basin cannotconvincingly explain the increased flood peaksin the main river after the 1980’s. At the largerbasin scale, the influence of land-use changes

in upland catchments on the major floodsdownstream may hardly be detected. Theincreased flood peaks in the Meuse river sincethe 1980’s are affected more by change in theantecedent (e.g. 7-day) precipitation in thebasin. Precipitation-induced increase in floodpeaks was also observed in the selectedtributaries except for the Geul, where thedecreased flood peaks may possibly beattributable to storage effects after 30 years ofnature development. Therefore, the apparentchanges in frequency and magnitude of floodsin the Meuse river over the last two decadescan broadly be ascribed to climatic variability.Ongoing work will investigate the relationbetween the European AtmosphericCirculation Patterns and the precipitationchange in the Meuse basin.

AcknowledgementsThis research is sponsored by the Institute forInland Water Management and Waste WaterTreatment (RIZA), The Netherlands. Data wereprovided by RWS, Roer and Overmaas WaterBoard in The Netherlands, KMI and MET-SETHYF in Belgium and DWD, Eifel – RurWater Board in Germany.

ReferencesAlexandersson, H., 1986. A Homogeneity Test

Applied to Precipitation Data. Journal ofClimatology. 6, pp. 661-675.

Alexandersson, H. & A. Moberg, 1997.Homogenization of Swedish Temperature Data:Part I Homogeneity Test for Linear Trends.International Journal of Climatology. 17, pp. 25-34.

Pettitt, A.N., 1979. A Non-parametric Approach tothe Change-point Problem, Applied Statistics.28(2), pp. 126-135.

Ministère de la Région wallonne, DGRNE, 2000.Etat de l'Environnement Wallon 2000:L'Environnement Wallon à l'aube du XXIe Siècle(in French).http://environnement.wallonie.be/eew2000/gen/framegen.htm



Monte Carlo method applied to a two-dimensionalmorphodynamic model

J.J.P. LambeekWL | Delft Hydraulics, P.O. Box 177, 2600 MH Delft, The Netherlands; [email protected]

IntroductionFor a reach of the river Rhine close to theDutch-German border a two-dimensionalmorphodynamic model has been applied inorder to investigate the morphological impactof possible measures to increase thenavigability (WL | Delft Hydraulics, 2003). Thisinvestigation used a single hydrograph thatwas repeated for several years in order toinvestigate the long-term morphological impactof structural interventions. This approachincludes the discharge variation during only asingle year and is therefore considered verymuch deterministic. The actual variation of thedischarge can be much more significant and isnot included in this approach (see Fig. 1).

Figure 1. Historical Rhine discharge data incomparison with the hydrograph used in earliermodel investigations.

Van der Klis (2003) analysed the uncertainty ofdischarge fluctuations and the impact on theriverbed development. She examined theapplicability of Monte Carlo simulations toquantify the uncertainty of a one-dimensionalmorphological numerical model. Sheconcluded that the Monte Carlo method is thebest method to quantify the impact of riverdischarge uncertainties on the riverbedmorphology. The Monte Carlo method makes itpossible to analyse the morphological andhydraulic output of simulations with normalstatistical methods.The objective of the present research is toinvestigate whether the application of thisMonte Carlo method to a two-dimensionalmorphodynamic model would provideadditional useful information that is consistent

with prototype observations. For this reason,the number of simulations has been limited toonly ten.

Case StudyIn the investigation a detailed two-dimensionalmorphological Delft3D model has been usedthat was developed by Baur et al. (2002) (Fig.2).

Figure 2. Morphological Delft3D model.

The model schematisation includes the mainchannel and floodplains over a length of 42km. The average cell length and width in themain channel are 150 m and 35 m,respectively.The boundary conditions for the modelsimulations consisted of the water level at thedownstream boundary and the dischargedistribution over the grid cells of the upstreamboundary. In case of a Monte Carlo simulationthis implies that for every discharge value to beused a separate set of water level anddischarge distribution values is required.The historical data have been converted to seewhich set of discrete discharge values wouldhave similar characteristics as the historicaldataset and still would be workable as input forthe Delft3D simulations.For now, it was concluded that discretisationwith steps of 500 m3/s would represent theactual discharge values well enough for a firstapproach.

Monte Carlo approachThe principle of Monte Carlo simulation is torun a deterministic model many times. Eachmodel run is driven by a different realisation of



the input time series (synthesized on the basisof randomly generated parameters). Theoutput of all model runs is used to determinethe statistical properties of the bed levels, likethe expected bed level position, its varianceand the percentile values. In case of non-linearsystems, the statistical mean of themorphological responses is not equal to themorphological prediction based on the ‘mean’value of each model input (Gardner & O’Neil,1983).The random discharge generator is based on amethod derived by Duits (1997). The seasonaldependency and the correlation betweensuccessive river discharges are considered inthis method. The basic data represents 55years of daily measured Waal discharges atthe location Lobith. A statistical description hasbeen derived in five steps:• Each year is split into 36 periods of 10 days;• The measured daily discharges are

averaged over these 10-day periods;• For corresponding 10-day periods through

the years, a shifted lognormal distributionfunction is estimated;

• The correlation between discharges insuccessive periods is estimated. Duitsshows that the discharge in 10-day period idepends on that in the two preceding 10-dayperiods, i –1 and i -2.

• Construction of a multivariate lognormaldistribution function per 10-day period.

This method has been previously applied tothe same data by Van Vuren et al. (2002) andVan der Klis (2003). With this statistical model,discharge hydrographs have been synthesizedby random sampling from these multivariatelognormal distribution functions. The set ofresults is used to determine the statisticalproperties of the predictions.The averaging process of the Monte Carlosampling adjusts the discharge values, so thatthe sediment transport generated by theaveraged discharge data for the period of 10days equals the sediment transport generatedby the daily average discharge data.In fact a few hundred simulations are requiredfor a good statistical analysis of themorphodynamic behaviour of the riverbed (Vander Klis, 2003). The objective of the presentresearch, however, is an initial test to see theapplicability of a Monte Carlo simulation with atwo-dimensional morphological model.Therefore the number of simulations is limited.

SimulationsDuring the simulations, the hydraulic andmorphological output is stored with time stepsof five days. Two types of analyses can be

carried out with the output data of thesimulations. The first analysis considers theimpact of discharge fluctuations on themorphodynamic behaviour of the riverbed, bystatistically analysing the output of the bedelevation data.The second analysis considers the impact ofdischarge fluctuations on the navigability. Inthis case the bed elevation output for each 5days time step has been merged in a newmodel schematisation with which steady statesimulations have been carried out under OLR-conditions. For each model grid the fulfilmentof the OLR-criteria (water depth of 2.8 m andchannel width of 170 m) can be analysedstatistically to investigate the impact of themorphological development under varyingdischarge conditions on the navigability.

ResultsThe morphological sensitivity to dischargefluctuations for the present situation has beenanalysed with figures of the maximum bedlevel difference (maximum minus minimumbed level) and with the standard deviation ofthe bed elevation during the Monte Carlosimulation. For each grid cell, the statisticaloutput for each of the presented parameters isbased on 2,190 values.

Figure 3. Absolute maximum bed level differenceduring Monte Carlo simulation.

The magnitude of the morphodynamicvariations under varying discharge conditionsis presented in Fig. 3 and 4. Fig. 3 presentsthe absolute maximum bed level differencesfor a certain model reach. Fig. 4 presents the



standard deviation of the bed elevation ofanother model reach. High values imply thatthe local riverbed (elevation) is highly sensitiveto discharge fluctuations.

Figure 4. Standard deviation of the bed elevation asan indicator for the morphodynamic behaviour.

The impact of the morphodynamic behaviouron the navigability has been analysed,focussing on the navigation channel depth andwidth. The navigation channel width has beenanalysed with the standard deviation, and themaximum and minimum width for a minimumwater depth of 2.8 m. Fig. 5 presents thefrequency that the water depth during OLR-conditions is smaller than the OLR-criteria of2.8 m. High values indicate that the OLRcriterium is frequently not fulfilled.

Figure 5. Frequency that the water depth duringOLR-conditions is smaller than the OLR-criteria of2.8 m.

With the statistical information it is possible todetermine the chance that the OLR-criteria willnot be fulfilled and that the navigability in acertain reach will face difficulties. Thisinformation could be used to make a decisionwhether, where and when dredging should becarried out to guarantee navigability.

The results of the Monte Carlo simulation areconsistent with prototype observations. Thereaches where navigability problems occur inthe prototype are clearly represented in theresults of the Monte Carlo simulations.

ConclusionsThe application of a Monte Carlo simulation toa two-dimensional morphological model isconsidered to be a promising method to obtaininsight in the morphodynamic behaviour of theriverbed.The results of this non-deterministic modellingapproach clearly provides insight into theimpact of discharge fluctuations on themorphodynamic behaviour of the riverbed.This additional information is consideredvaluable when insight is required about theimpact of the morphodynamic behaviour onother functions of the river.The impact of discharge variations on themorphodynamic behaviour is clearly visible inthe results of the Monte Carlo simulation. Theresults appear to be consistent with prototypeobservations.

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

Zusammenarbeit bei der Planung vonRegulierungsmaßnahmen am Niederrhein:Durchführung flussmorphologischer Simulationen(in German). HANSA International MaritimeJournal. 10, pp. 51-56.

Duits, M., 1997. Baggeroptimalisatie in rivierbochten(in Dutch). MSc. Thesis. Delft University ofTechnology, Delft, The Netherlands.

Gardner, R.H. & R.V. O’Neill, 1983. Parameteruncertainty and model predictions: a review ofMonte Carlo results. In: Beck, M.B. & G. vanStraten, Uncertainty and forecasting of waterquality. Springer-Verlag, Berlin.

Klis, H. van der, 2003. Uncertainty analysis appliedto numerical models of river bed morphology. PhDThesis, Delft University of Technology, Delft, TheNetherlands.

Vuren, S. van, H. van der Klis & H.J. de Vriend,2002. Large-scale floodplain lowering along theriver Waal: a stochastic prediction ofmorphological impacts. River Flow 2002;Proceedings International Conference on FluvialHydraulics, Louvain-la-Neuve, Belgium. A.A.Balkema, Rotterdam, The Netherlands.

WL | Delft Hydraulics, 2003. Grensproject Bovenrijn/Grenzproject Niederrhein (in Dutch); Report no. 3,Investigation of individual engineering measureswith 2D morphological model (part 2). WL | DelftHydraulics Project Q2496. Delft, The Netherlands.



Impact of land use changes on breeding birds infloodplains along the rivers Rhine and Meuse in theNetherlands

R.S.E.W. Leuven1, R.P.B. Foppen2, A.M.G.R. Houlliere1, H.J.R. Lenders1 & R.J.W. de Nooij11 Department of Environmental Studies, Faculty of Science, Mathematics and Computing Science, University of Nijmegen,P.O. Box 9010, 6500 GL Nijmegen, The Netherlands; [email protected]

2 SOVON, Dutch Centre for Field Ornithology, Rijksstraatweg 178, 6573 DG Beek-Ubbergen, The Netherlands

IntroductionOver the last decades the habitat availabilityand quality for breeding bird species infloodplains along the rivers Rhine and Meusesignificantly changed as a result of physicalreconstruction measures and land usechanges aiming at flood defence andecological rehabilitation (e.g. excavation offloodplains, removal of levees, digging of sidechannels and transition of agricultural land innature). This paper presents spatial patternsand trends in breeding bird diversity on variouslevels of scale, i.e. river branches, floodplainareas and riverine ecotopes. In addition,breeding bird composition of rehabilitatedfloodplains is analysed. The relations betweenspecies richness and several landscapeecological characteristics (e.g. surface area ofvarious riverine ecotopes) are described.

Material and methodsData on presence and abundance of breedingbirds in floodplains were obtained from birdwatch organisations, floodplain managers and(scientific) literature (Erhart & Bekhuis, 1996;Faunawerkgroep Gelderse Poort, 2002 &SOVON Vogelonderzoek Nederland, 2003).Relations between cumulative speciesrichness and surface area of several riverine

ecotopes were analysed using BiodiversityProfessional Beta 1 (McAleece, 1997). Thesurface area of ecotopes per grid or floodplainwas calculated with Arcview 3.2, using theriver ecotope map of Rijkswaterstaat (1997).Canonical Correspondence Analysis (CCA)was performed with Canoco for windows(Jongman, 1995).

ResultsOver the period 1973-2000 more than 70percent of the grids surveyed along the riversMeuse and Rhine in the Netherlands showed adecrease in the number of red-listed breedingbird species (Fig. 1). This particularly holds forbird species that are characteristic of reedmarshes and wet grasslands. However, thediversity of characteristic species of riverineecotopes such as lakes, side channels,pioneer vegetation and softwood forest tendedto increase. Similar trends were observed atthe regional scale. For instance, in theinternational wetland “Gelderse Poort” adecrease was observed in distribution of 21out of 35 red-listed breeding bird species overthe period 1975-2000 (Erhart & Bekhuis, 1996& Faunawerkgroep Gelderse Poort, 2002).Rehabilitated floodplains showed a strongincrease of breeding bird diversity.

Figure 1. Changes in number of red-listed breeding bird species between 1973-1977 and 1998-2000 atlassurveys of 5 x 5 km grids in Dutch river district, only probable and confirmed breeding records included

(Data: SOVON Vogelonderzoek Nederland, 2002).

0

10

20

30

40

50

60

< -18 -18--11 -10- -6 -5 - -2 -1 - 1 2 - 5 > 5

Changes in number of species

Gri

ds

(%) IJssel

Nederrijn

Waal

Maas



Figure 2. Cumulative species richness (R) of protected and endangered bird species in ecological rehabilitatedfloodplains along the rivers Meuse and Rhine.

In most study sites the number of bird speciesquickly reached a meta-stable equilibrium.Graphs of cumulative species richness indicaterelatively high species turnover rates (Fig. 2).Fig. 3 shows the relation between cumulativespecies richness of breeding birds and surfacearea of reed marshes in floodplains of theGelderse Poort. The species pool of reedmarshes saturated at about 35 ha. Table 1summarises data on minimum surface areasrequired for saturation of species pools ofvarious riverine ecotopes.

Table 1. Surface area of various riverine ecotopes infloodplains of the Gelderse Poort required forsaturation of the species pool.

DiscussionThe number of endangered and protectedbreeding bird species in the Dutch river districtstrongly decreased over the period 1973-2000,due to deterioration and fragmentation of their

habitats (e.g. as a result of land use changesand physical reconstruction of floodplains).Ecological rehabilitation of floodplains hadpositive effects on species diversity ofbreeding birds. However, species turnoverrates in rehabilitated sites were high owing tovegetation succession. Some protected andendangered species were also negativelyaffected by floodplain rehabilitation (e.g.species associated with wet grasslands suchas Black-tailed Godwit, Common Redshank).Decrease of these species was caused bysuccession of wet grasslands into herbaceousvegetation, shrubs and soft wood forest.Preliminary results of multivariate analysesrevealed that bird diversity was stronglycorrelated with habitat availability (surfacearea) and landscape heterogeneity (Houlliere,2003). Other studies also showed thatdistribution of species associated with reed-dominated marshes (e.g. Great Bittern andGreat Reed Warbler) was strongly limited byhabitat availability and fragmentation (Foppen,2001). Habitat quality (e.g. sediment and soilcontamination) limits the distribution of birds ofprey, such as Little Owl and Kestrel (Kooistraet al., 2001). Abundance of herbivorouswaterfowl was significantly correlated withnutrient contents in river water, flooding regimeand winter temperature (Schaap, 2003).

Ecotopetype

Total surfacearea

available (ha)

Surface arearequired for species

saturation (ha)

Open water 1613 680

Grassland 409 83

Forest 288 200

Reed marsh 96 34

Bio-safe species

0

5

10

15

20

25

30

35

40

1986 1988 1990 1992 1994 1996 1998 2000

years

R

Ewijkse Plaat Meinerswijk

Koningsteen Blauwe Kamer

Kleine Weerd Hochter Bampd

Ossenwaard Dilkensweerd

Bemmelsche en Gendtsche Polder Klompenwaard

Rijnstrangengebied Meginhardweg



Figure 3. Breeding bird diversity – surface area relations for reed marshes in river floodplains of the GeldersePoort (Blue line: arranged from small to large ecotope patches; Red line: average of 50 random samplings).

In spite of numerous ornithological surveys inthe Dutch river district, it appeared to be verydifficult to construct a consistent database onlong-term landscape dynamics in relation tothe distribution of breeding birds. Cumulativespecies richness – habitat surface area graphsare very useful for guidelines for floodplainreconstruction. Future research will focus ondevelopment and validation of GIS-basedmodels for the evaluation of effects of riverinelandscape dynamics on presence, abundanceand population viability of (breeding) birds.These models must be suitable for integrationin decision support systems for rivermanagement and will particularly be used forthe evaluation of strategies for cyclic floodplainrejuvenation and multiple spatial planning.Remote sensing techniques and additionalfield surveys will have to be performed toacquire consistent data on spatial patterns andtrends of (breeding) birds in relation tolandscape ecological factors (e.g. availability,quality and configuration of habitat, land useand flooding regime).

ConclusionsRecently, a strong decline in the number of redlisted breeding birds has been observed in theDutch river district due to habitat deteriorationand fragmentation. Ecological rehabilitationyields positive effects on bird diversity in riverfloodplains. However, breeding birdcommunities still show high species turn overrate, due to vegetation succession. Negativeeffects of rehabilitation measures have beenobserved for species associated with wetgrasslands. GIS-based models for ecologicalimpact assessment of multiple spatial planningand cyclic floodplain rejuvenation strategiesrequire validation of species-habitat-ecotopealgorithms.

ReferencesErhart, F.C. & J.F. Bekhuis, 1996. Broedvogels van

de Gelderse Poort 1989-94 (in Dutch). Vogelwerk-groep Arnhem e.o., Vogelwerkgroep Rijk vanNijmegen e.o. & NABU-NaturschutzstationKranenburg. Arnhem, The Netherlands.

Faunawerkgroep Gelderse Poort, 2002. Vogels inde Gelderse Poort, deel 1: broedvogels 1960-2000 (in Dutch and German). VogelwerkgroepRijk van Nijmegen e.o., KartierergemeinschaftSalmorth, Vogelwerkgroep Arnhem e.o., NABU-Naturschutzstation Kranenburg, Naturschutz-station im Kreis Kleve e.v., Provincie Gelderland& SOVON Vogelonderzoek Nederland, BeekUbbergen, The Netherlands.

Foppen, R.P.B., 2001. Bridging gaps in fragmentedmarshland. Applying landscape ecology for birdconservation. Thesis University of Wageningen.Alterra Scientific Contributions 4. Alterra GreenWorld Research, Wageningen, The Netherlands.

Houlliere, A.M.G.R., 2003. Breeding birds inecological rehabilitated floodplains along largerivers in the Netherlands: spatial and temporaldevelopments. Reports Environmental Studies235. University of Nijmegen, The Netherlands.

Jongman, R.H.G., C.J.F. ter Braak & O.F.R. vanTongeren, 1995. Data analysis in community andlandscape ecology. Cambridge Univ. Press, UK.

Kooistra, L., R.S.E.W. Leuven, R. Wehrens, L.M.C.Buydens & P.H. Nienhuis, 2001. A procedure forincorporating spatial variability in ecological riskassessment of Dutch river floodplains.Environmental Management. 28(3), pp. 359-373.

McAleece, N., 1997. Biodiversity Professional Beta1. The Scottish Association for Marine Scienceand the Natural History Museum, London.

Schaap, M., 2003. Herbivore watervogels langsgrote rivieren. Het effect van bemestingsgraad,waterstand en temperatuur op de aantallengrasetende watervogels (in Dutch). SOVONReport 2003/3. SOVON VogelonderzoekNederland, Beek Ubbergen, The Netherlands.

SOVON Vogelonderzoek Nederland, 2003. Atlasvan de Nederlandse broedvogels 1998-2000 (inDutch with English summary). NederlandseFauna 5. KNNV Uitgeverij & EuropeanInvertebrate Survey, Leiden, The Netherlands.

0

2

4

6

8

10

0 20 40 60 80 100 120

cumulative surface area (ha)

R

0

2

4

6

8

10

0 20 40 60 80 100 120

cumulative surface area (ha)

R



Developing integrated water system models from agroundwater perspective - case Water Board Regge andDinkel groundwater model

J.J.J.C. Snepvangers, B.M. Minnema, C.B.M. te Stroet, Y. van der Velde & M. KuijperNetherlands Institute of Applied Geosciences-TNO, Groundwater Department, P.O. Box 80015, 3508 TA Utrecht, TheNetherlands; [email protected]

Water management in the Netherlands isfocused on the water system as a whole(integrated water management). Managementtools need therefore be available for integratedwater system management, for example toprovide detailed information for GGOR (targetgroundwater and surface water regime).However, the models on which managementtools can rely, are far from integrated. Surfacewater and groundwater modelling are differentworlds, most importantly due to the temporaland spatial discretisation differences.Groundwater flows slowly throughout thewhole subsurface (3D), while surface waterflows much faster usually in predefinedchannels (1D/2D). In many surface watermodels the groundwater interaction ismodelled with an average infiltration ordrainage flux, in which groundwater dynamicsare not included. In groundwater models, onthe other hand, the surface water componentconsists usually of stationary river stages asstatic boundary conditions.

This study focuses on integrated water systemmodel development in order to fulfil the need ofregional water managers for detailed, regionalmodels. As in the Netherlands discharge isstrongly dominated by groundwater, to startfrom a groundwater perspective is a naturalchoice. The case study of the Water BoardRegge en Dinkel groundwater model will beused to exemplify the current status of ourresearch.

The water board Regge en Dinkel (WRD) hasasked the groundwater department of TNO-NITG to develop a state-of-the-art groundwatermodel to support their (ground)watermanagement. The regional WRD groundwatermodel has a 100 * 100 m horizontaldiscretisation and 3 model layers (quasi 3D).Four different model variants exist in relation tothe temporal discretisation: 1) stationary, 2) 1-year duration (representative meteorologicalyear) with 14-day time step, 3) 10-yearduration (1991-2000) with 1-month time step,4) 1-year duration (1998) with 1-day time step.In the case of the research presented here, thelatter (1998-model) is of special interest.

The WRD groundwater model is developed inMODFLOW. The latest developments ingroundwater modelling are incorporated in theWRD model. With respect to our intents todevelop an integrated water system model, themost important development is a newtechnique to obtain the stages of smallchannels and ditches.

Usually in surface water modelling only about20% of the watercourses is explicitly taken intoaccount. The remaining 80% is modelled withlumped parameters. By doing so, the non-linearity of the water system receives too littleattention. Modelling of groundwater dischargebecomes a ‘fitting’ procedure, as the exactmoments that channel reaches contribute tothe catchment discharge are unknown.Besides this, local convexity between smallerchannels is not taken into account. For GGOR,for example, this local information is of greatimportance.

In the WRD model the stages of all channelsare incorporated. For the larger channels(canals and channels managed by the waterboard) measured stages were available. Thestages of the smaller channels are retrievedfrom elevation information. For this purpose,we developed the AHN filtering techniqueconsisting of three steps. In step 1 noise andnon-hydrological structures are removed usingimage analysis techniques. In step 2 we scaleup from 5 * 5 m to 25 * 25 m. During this stepat known locations of channels and ditches(from topographic information; Top10Vecor)we select the 10% percentile during upscalinginstead of the median value, in order to detectthe water level instead of the average surfacelevel. In step 3 the water levels are smoothedto prevent ‘bumpy’ levels within individualreaches.

The advances of using stages of all channelsin the groundwater model can be seen in Fig. 1and 2. Fig. 1 shows the groundwater levels inWRD at a winter and summer moment. Clearlyin winter more channels are active in the watersystem than in summer, where there is muchless spatial detail in the groundwater levels.



Figure 1. Groundwater levels in summer (left) and winter (right) in Water Board Regge en Dinkel (colour scale:white: 0 m. orange: 60 m.).

Fig. 2 shows modelled discharge versusmeasured discharge for a sub-catchment inthe WRD area. The resemblance is striking,especially considering that no calibration orfitting to discharge measurement was usedduring the modelling process. Imperfectionscan be seen especially in the falling limb ofdischarge peaks, where the model does notlower fast enough. Most probably, this is dueto the fact that no routing and surface storage(re-infiltration) is taken into account.

In order to improve the model and especiallythe discharge prediction, there are severalpossibilities. First of all, routing will be

incorporated in the model. Promisingexperiments are being carried out with thestream-package of MODFLOW. Furthermore,the description of the unsaturated zone andsurface processes as infiltration and surfacestorage will be improved in the model. Last butnot least, we have started with non-steadycalibration on head and dischargemeasurement, especially to decrease theuncertainty in the resistance values of thechannel bottom. In this way, we expect tofurther expand our groundwater modellingapproach to an integrated water systemmodelling approach.

Figure 2. Discharge prediction and measurements for a sub catchment of the Regge en Dinkel area.

catchment 12

0

50000

100000

150000

200000

250000

300000

350000

26-d

ec-9

7

9-ja

n-9

8

23-j

an-9

8

6-fe

b-9

8

20-f

eb-9

8

6-m

rt-9

8

20-m

rt-9

8

3-ap

r-98

17-a

pr-

98

1-m

ei-9

8

15-m

ei-9

8

29-m

ei-9

8

12-j

un

-98

26-j

un

-98

10-j

ul-

98

24-j

ul-

98

7-au

g-98

21-a

ug-

98

4-se

p-9

8

18-s

ep-9

8

2-ok

t-98

16-o

kt-

98

30-o

kt-

98

13-n

ov-9

8

27-n

ov-9

8

11-d

ec-9

8

25-d

ec-9

8

Dis

char

gein

m3/

day

measurements

model



Mapping of vegetation characteristics using lidar andspectral remote sensing

M. Straatsma1, H. Middelkoop1, E. Pebesma1 & C. Wesseling2

1 Department Physical Geography, Faculty of Geosciences, Utrecht University, P.O. Box 80115, 3508 TC Utrecht, TheNetherlands; [email protected]

2 PCRaster environmental software, Van Swindenstraat 97, 3514 XP Utrecht, The Netherlands

AbstractThe spatial distribution of vegetationcharacteristics is an important factor infloodplain hydraulics. In the Netherlands, thesecharacteristics are currently mapped based onmanual interpretation of aerial photographs.Lidar (LIght Detection And Ranging) provides afast and detailed alternative. The aim was tocompare the structure of the lidar derived pointcloud with field measurements of vegetationheight and density. We collected lidar data atdifferent flying heights and with different gains,focussing on low vegetation, which is thedominant type in Dutch floodplains.Regression equations for vegetation heightand density are given.

IntroductionThe distribution of vegetation structuralcharacteristics and inherent roughness is themain knowledge gap in hydraulic modelling offloodplains (Cobby et al., 2001). Vegetationslows down water flow, thereby creating higherwater levels and increasing the flood risk.Dynamic management of floodplains inducessuccession of floodplain vegetation. This leadsto a high spatio-temporal variation of vege-tation structural characteristics, i.e. vegetationheight and density. To provide hydraulicmodellers with input, the spatial distribution ofvegetation characteristics is needed. To catchup with succession the method of vegetationmapping has to be detailed and fast.Lidar (LIght Direction and Ranging) is such adetailed and fast technique for vegetationmapping. Lidar scans the surface under ahelicopter and generates a cloud of data pointsin a local coordinate system. Some of thesepoints represent the ground surface, others thevegetation, see Fig. 1. Two parametersinfluence the distribution of the points in thedata cloud: (1) the flying height and (2) thegain (Wehr and Lohr, 1999). Gain defines theamount of amplification of the return signal atthe receiver in the helicopter. Low flying heightand a high gain increase the intensity andtherefore increase the chance of detecting thetop of the vegetation. However, no methodexisted to map vegetation characteristics fromlidar data.

Figure 1. 3D Lidar and field image of herbaceousvegetation. The lidar data represent a box of 15 by15 by 2 m.

Our main question is: can vegetation heightand density of low vegetation be predictedusing lidar for homogeneous areas. Wefocussed on low vegetation, as this is thedominant vegetation height in Dutchfloodplains. We subdivided into three subquestions:1. What is the effect of data segmentation in

ground and vegetation points on theprediction of vegetation characteristics?

2. What is the effect of flying height and gainon the prediction of vegetationcharacteristics?

3. What do semi-variograms tell about thespatial distribution of vegetationcharacteristics?

Data and data processingIn a study at the Gamerense Waard floodplain,the following data is gathered:• Summer 2002: True color and color infrared

images at a 20 cm resolution;• March 26, 2003:

! FLI-MAP Lidar data flown at 70 m height,normal gain, 30 points/m2;

! FLI-MAP Lidar data flown at 50 m height,maximum gain, 20 points/m2;

! Field reference of vegetation height anddensity data at 47 locations in 15 by 15m blocks.



Table 1. The effect of data segmentation on prediction of vegetation height and density.Abbreviations: Hv = Vegetation height (m), Dv = Vegetation density (m/m2),

SE = standard error of the estimate, Norm. SE = SE divided by the average (-),Sk = Skewness, Kurt = Kurtosis, CV = Coefficient of variation, SD = Standard deviation.

FLI-MAP is the lidar system of Fugro-Inpark.Lidar data processing consisted of:1. Selection of data points within field plots

using a quadtree indexing to facilitateprocessing of large volumes of data;

2. Data segmentation in ground and vegetationpoints;

3. Calculation of 25 statistical parameters foreach field location based on:a) Segmented and unsegmented data;b) Data collected at high altitude and

normal gain and low altitude and highgain.

In subsequent multiple linear regressionanalyses, vegetation height and density asmeasured in the field were regressed to thelidar derived statistical parameters. The resultswere checked for co linearity.

Multiple linear regression equationsPredictions of vegetation height fromsegmented and unsegmented both data havean explained variance of 0.82, but thepredictors differ, see Table 1. The field dataalso showed two types of outliers, see Fig. 2:(1) a cluster of short grasses with a heightlower then 10 cm and (2) one outlier with avegetation height 80% higher the secondhighest point. The explained variance with theoutliers left out dropped to 55 and 40% forunsegmented and segmented datarespectively.

Figure 2. Log-log scatter plots of field and predictedvegetation heights using equation in table 1. Lidaroverestimates vegetation heights of very shortvegetation.

The explained variance for both vegetationheight and density are higher for the datacollected at a low flying height and a high gain,while the predictors are the same (Table 2).Semi-variograms for segmented data fit to anugget model, whereas for unsegmented datathey fit to an exponential model. The sill of thesemi-variogram of the unsegmented data is20% of the nugget of the segmented data.

DiscussionData segmentation affects the parameters thatpredict the vegetation height. No difference inexplained variance was found for the twomethods. For unsegmented data theskewness, kurtosis and 95 percentile are thestrongest predictors. All describe the tail of thevertical distribution.

Table 2. The effect of flying height and gain on prediction of vegetation height and density,based on segmented data (for abbreviations, see table 1).

Regression equation adj. R2 SE

Hv = 0.022+1.05D95+0.21Sk-0.0097Kurt 0.82 (0.55) 0.22 (m)Unsegmenteddata Dv = -0.479-0.521D99.8+2.23D80+0.0075CV 0.56 0.0595 (m/m2)

Hv = -0.338+2.53D70+0.00645CV 0.82 (0.40) 0.24 (m)Segmented data

DV = -0.123+1.5D50-0.216D99-0.414D99.8+0.3Range+0.0043CV 0.62 0.056 (m/m2)

Regression equation adj. R2 SE

Hv = 0.612+8.5D70-5.3SD 0.41 0.26 (m)High altitude, normal gain

Dv = 0.126-0.0165Sk+5.3*10-4Kurt 0.073 0.045 (m/m2)

Hv = 0.33+2.52D70-1.64SD 0.80 0.15 (m)Low altitude, high gain

Dv = 0.212-0.169Sk+0.282Kurt 0.75 0.023 (m/m2)

0.01

0.1

1

10

0.1 1 10

Lidar (m)

Fie

ld(m

)SegmentedUnsegmented



For segmented data, the 70 percentile and thecoefficient of variation are the strongestpredictors. This is as expected as the variationin the lidar point cloud increases as thevegetation height increases. The vegetationdensity is predicted using many differentvegetation parameters. No clear difference ispresent in the regression equations betweensegmented and unsegmented data.

The cluster of data points with low vegetationheights (see Fig. 2) is caused by thesegmentation method. The method classifiestoo many points as vegetation. Differentmethods of data segmentation exist (Sithole &Vosselman, 2003). It is unlikely though thatany method can detect vegetation heights of 2to 3 cm since even non-vegetated areas havea 20 cm wide band of scatter which allrepresent the ground surface.

The effects of low flying and a high gain arevery positive. The explained variances for thepredicted vegetation height and density bothincrease for a lower flying height and a highgain. These relations are based on elevenpoints. The increase is significant. Theindividual effects of low flying or a higher gaincould not be established.Semi-variogram parameters do not show anyrelation to vegetation height or density. Webelieve this would only show with a pointdensity of 500 points/m2, because in that casethe point spacing of lidar points representingthe vegetation would coincide with a typicalstem spacing of herbaceous vegetation.

ConclusionWe have made a first step to provide hydraulicmodellers with input of vegetationcharacteristics of floodplains based on lidardata. Contrary to previous studies (Asselman,2002) we found strong correlations betweenlidar and field data. For homogeneous areas,the relations in Table 2 should be used toderive vegetation height and density from lidardata. These relations are valid for segmenteddata and should be collected by flying at 50 mhigh and with a maximum gain. The relationsare calibrated for a vegetation height rangebetween 39 and 149 cm and vegetationdensity between 0.016 and 0.155 m/m2.Spectral remote sensing could further improvethe mapping by eliminating the lowestvegetations from further processing.

ReferencesAsselman, N.E.M., 2002. Laser altimetry and

hydraulic roughness of vegetation. WL | DelftHydraulics report Q3139. WL | Delft Hydraulics,Delft, The Netherlands.

Cobby, D.M., D.C. Mason & I.J. Davenport, 2001.Image processing of airborne scanning laseraltimetry data for improved river flood modelling.ISPRS Journal of Photogrammetry and RemoteSensing. 56, pp. 121-138.

Wehr, A. & U. Lohr, 1999. Airborne laser scanning –an introduction and overview. ISPRS Journal ofPhotogrammetry and Remote Sensing. 54, pp.62-82.

Sithole, G. & G. Vosselman, 2003. Comparison offiltering algorithms. ISPRS International Archivesof Photogrammetry, Remote Sensing and SpatialInformation sciences. Volume XXXIV, Part 3/W13,pp. 71-78.



Alien amphipod invasions in the river Rhine due to riverconnectivity: a case of competition and mutual predation

M.C. van Riel1, G. van der Velde1 & A. bij de Vaate2

1 University of Nijmegen, Department of Animal Ecology and Ecophysiology, P.O. Box 9010, 6500 GL Nijmegen, TheNetherlands; [email protected]

2 Institute for Inland Water Management and Waste Water Treatment, P.O. Box 17, 8200 AA Lelystad, The Netherlands

IntroductionEach river basin harbours a characteristic floraand fauna, including species that remainendemic due to biogeographical barriersisolating the populations in the watersheds.Linking river systems by constructingwaterways leads to mixing of species and tonew interactions between species in the foodweb (Bij de Vaate et al., 2002; Van der Veldeet al., 2002). Competition between invasiveand indigenous species can cause a decreasein the densities of indigenous species andchanges in the community structure, and caneven change the functioning of food webs. Theoutcome of competition depends on thebehavioural, physiological and morphologicaltraits of the species involved, in relation toenvironmental conditions and anthropogenicdisturbances (Van der Velde et al., 2002;Wijnhoven et al., 2003). This means that theimpact of an invader on existing food web isdifficult to predict.In Europe, a network of canals connects nearlyall large rivers, so that invasions starting ineastern and southern Europe can penetratethrough the canals towards other rivers. One ofthe latest connections to be constructed is theMain-Danube canal, which was officiallyopened in 1992. This canal facilitatesinvasions by Ponto-Caspian species, that is,species originating from the area of the BlackSea, Asov Sea and Caspian Sea, towards theriver Rhine.As a result, the number of Ponto-Caspianinvaders in the river Rhine has greatlyincreased. The most successful of thesePonto-Caspian invaders have been amphipods(Crustacea). They have had a great impact onthe Rhine food web as they have becomedominant in numbers and biomass within avery short time (Van der Velde et al., 2000 &2002; Haas et al., 2002). Subsequentamphipod invasions by gammarids andcorophiids have made dominance patternsshift among the various invasive amphipodspecies. The indigenous species Gammaruspulex decreased in numbers after the North-American species Gammarus tigrinus, whichwas introduced as fish food, invaded the riverin 1982. In 1989, the Ponto-Caspian mudshrimp Chelicorophium curvispinum entered

the Rhine through the Mittelland Canal andbecame dominant in number and biomass,covering the hard substrates with muddytubes. Another Ponto-Caspian amphipod,Echinogammarus ischnus, made its waythrough the Mittelland Canal to the Rhine in1989. The Main-Danube canal enabled one ofthe most abundant of the current invaderspecies, the stone-dwelling amphipodDikerogammarus villosus, to enter the Rhine.This species is currently the largest amphipodspecies in the Rhine and is a strong,omnivorous predator. Since D. villosus enteredthe Rhine in 1995, the densities of othermacroinvertebrates, including Gammarustigrinus and Chelicorophium curvispinum, havedeclined, which may be due to predation by D.villosus on all kinds of macroinvertebrates,including other amphipod species, andcompetition for food and space with othergammarids.Our research questions were: (a) doescompetition among invasive gammaridsexplain the replacement of these species in theriver Rhine and (b) if so, by which mechanismdoes this competition take place?

MethodsInterspecific competition for substrate by theexotic gammarid species D. villosus, G.tigrinus and E. ischnus was tested inmicrocosm experiments in the laboratory. Thesubstrates used in these experiments reflectedthose occurring in the river Rhine and provided

Figure 1. Substrata used in the experimenta) groyne stone, b) stones, c) gravel, d) sand.

A

DD

BB

CC



the species with various opportunities for anddegrees of shelter (Fig. 1). Shifts in substratepreference after the addition of anotherspecies and interspecific predation wereassumed to indicate interspecific competition.

Collecting Gammarid speciesDikerogammarus villosus and Echino-gammarus ischnus were collected from stonesubstrates in the river Waal, the main Rhinebranch in the Netherlands, near Nijmegen(5°48’E, 51°51’N). Gammarus tigrinus wascollected from the IJsselmeer lake. Allspecimens were kept separately, in aeratedbasins (40 * 40 * 50 cm) at 15°C, with a 9/15hours dark/light regime before being releasedinto the experimental aquaria. The gammaridswere fed chironomids during captivity.

Experimental designThe experiments were carried out in a climateroom at 15°C with a 9/15 hours dark/lightregime (2 36W/840TLD-lamps). Aquaria (25 *25 * 30 cm) were filled with Rhine water andaerated. Four different types of Rhinesubstrate – groyne stone, stones, gravel andsand (Fig. 1) – were put into cups (Ø 11.5 cm,6.5 cm height), which were placed in eachaquarium at random. Fifty gammarids of thesame species were allowed to choosebetween substrates. After 24 hours, the cupswith substrates were collected and thegammarids inside each cup were counted andtheir body lengths measured from rostrum tilltelson. In addition, we established the numberof gammarids that had been consumed. Theexperiment was repeated for combinations ofG. tigrinus with D. villosus and E. ischnus withD. villosus. In these mixed experiments, 50individuals of G. tigrinus or E. ischnus wereallowed to hide for 2 hours before 50individuals of D. villosus were added.

ResultsIn the absence of other gammarid species, allspecies preferred stones (Fig. 1b). In thepresence of D. villosus, G. tigrinus shifted fromits preferred stone habitat to gravel and sand.D. villosus, however, remained at its preferredstone substrate once it has established itself

Table 1. Interspecific and intraspecific predation(mean percentage of individuals consumed ± SE;each experiment repeated 14 times). - = no data.

Figure 2: Substrate preference of (a) Gammarustigrinus, (b) Dikerogammarus villosus and (c)Echinogammarus ischnus in the absence and thepresence of a second Rhine invader (DV = D.villosus, EIS = E. ischnus, GT = G. tigrinus).

and displaced G. tigrinus (Fig. 2a,b). Theseresults indicate direct competition for substratebetween the two invasive species, with D.villosus being the strongest competitor.Competition between D. villosus and G.tigrinus was characterized by a high level ofpredation on G. tigrinus (Table 1).As neither D. villosus nor E. ischnus changedtheir substrate preference in each other’spresence (Fig. 2b,c), and their mutualpredation did not increase (Table 1), a directcompetition between these Ponto-Caspianinvaders is unlikely.

ConclusionsConnecting river systems leads to mixing andinteractions among species. Interspecificcompetition for substrate and mutual predationdetermines the colonization success of agammarid species in the ecosystem.Increasing the heterogeneity of the ecosystem

Species G. tigrinusconsumed

D. villosusconsumed

E. ischnusconsumed

G. tigrinus 0.38 ± 1.41 2.00 ± 0.60 -

D. villosus 26.80± 4.66 5.57 ± 1.41 7.43 ± 1.32

E. ischnus - 5.14 ± 2.03 4.14 ± 0.85

a) G. tigrinus

0

5

10

15

20

25

30

35

40

groyne stone stones gravel sand

substrate

%in

div

idu

als

0

5

10

15

20

25

30

35GT

GT+DV

b) D. villosus

0

5

10

15

20

25

30

35

40


substrate%

ind

ivid

ual

s

DV

DV+GT

DV+EIS

c) E. ischnus

0

5

10

15

20

25


substrate

%in

div

idu

als

EIS

EIS+DV



provides chances for weak competitors tosurvive. Once rivers have become connected,and exotic species have successfully colonizeda new ecosystem, these species becomeimportant residents, often dominating in termsof numbers and biomass. They shouldtherefore be taken into account in water qualityassessments. Continuous biomonitoringprovides insight into the development ofpopulations of macroinvertebrates in theRhine, allowing new invaders to be discovered.Ecological field and experimental studies arenecessary to examine the ecology of thesespecies and to predict the impact of theseinvaders.

ReferencesBij de Vaate, A., K. Jadzdzewski, H.A.M. Ketelaars,

S. Gollasch & G. van der Velde, 2002.Geographical patterns in range extension ofPonto-Caspian macroinvertebrate species inEurope. Can. J. Fish. Aquat. Sci. 59, pp. 1159-1174.

Haas, G., M. Brunke & B. Streit, 2002. Fast turnoverin dominance of exotic species in the Rhine Riverdetermines biodiversity and ecosystemfunctioning: an affair between amphipods andmussels. In: E. Leppäkoski, S. Gollasch & S.Olenin (eds.), Invasive aquatic species of Europe.Distribution, impacts and management. KluwerAcademic Publishers, Dordrecht, TheNetherlands, pp. 426-432

Van der Velde, G., S. Rajagopal, B. Kelleher, I.B.Muskó & A. bij de Vaate, 2000. Ecological impactof crustacean invaders: general considerationsand examples from the Rhine River. CrustaceanIssues. 12, pp. 3-33

Van der Velde, G., I. Nagelkerken, S. Rajagopal &A. bij de Vaate, 2002. Invasions by alien speciesin inland freshwater bodies in western Europe:The Rhine Delta. In: E. Leppäkoski, S. Gollasch &S. Olenin (eds.), Invasive aquatic species ofEurope. Distribution, impacts and management.Kluwer Academic Publishers, Dordrecht, TheNetherlands, pp. 360-372.

Wijnhoven, S., M.C. van Riel & G. van der Velde,2003. Exotic and indigenous freshwater gammaridspecies: physiological tolerance to watertemperature in relation to ionic content of thewater. Aquat. Ecol. 37, pp. 151-158.



Data-driven modelling in context to sediment transport

B. Bhattacharya1 & C. Buraimo2

1 UNESCO-IHE Institute for Water Education, P.O. Box 3015, 2601 DA Delft, The Netherlands; [email protected] National Directorate for Water Affairs, P.O. Box 1611, Maputo, Mozambique.

AbstractArtificial neural networks (ANN) have beenused to develop a sediment transport modelusing measured data for predicting totaltransport rates. During the initial modellingattempts with a relatively small dataset it hasbeen observed that the predictive accuracy ofthe ANN model is better than that of themodels of van Rijn and Engelund-Hansen.

IntroductionIt is essential to have a reasonable estimate ofsediment transport in context to several watermanagement issues. The transport mechanismis utterly complex and the availabledeterministic transport models have beendeveloped with simplifying assumptions thatoften leads to large-scale prediction errors. Analternative may be a data-driven modellingapproach that is particularly useful in modellingprocesses about which adequate knowledge ofthe physics is not available. In the presentpaper a methodology for developing data-driven models of sediment transport usingartificial neural networks (ANN) is presented.The predictive accuracy of the ANN model iscompared with that of the models of van Rijnand Engelund-Hansen.

ModellingThe parameters controlling a transport processcan be described as:

),,,,,,,( µρρ wsgSDhuftq = (1)

where, u = velocity, h = depth of flow, D =particle size (mainly D50), S = energy slope, g= acceleration due to gravity, ρs = density ofsediment, ρw = density of water, µ = kinematicviscosity and qt = total transport rate. Based onthis information the input variables to themodel were chosen as: {u, h, D, S}, leavingaside the other nearly constant variables. Theoutput of the model was chosen as qt.

Data from a series of laboratory experimentsconducted at the USWES (1935) wasconsidered. Data points with a specific gravity2.65 were considered. Frequency distributionof the output variable (qt) revealed a non-uniform distribution of data. Most data-drivenmodelling methods perform well when the datahas a distribution close to the normal. A

(natural) log transformation of qt showed adistribution much closer to the normal andtherefore, was adopted. The distributions ofthe input variables were close to the normaland so the inputs were not transformed. Thetraining and testing datasets were formed with2/3rd and 1/3rd of the total data pointsrespectively. In data-driven modelling it isimportant to maintain (nearly) equal statisticaldistribution in the training and testing data.Data exploration revealed that selection of twoconsecutive data points for training followed byone data point for testing would ensure closerdistribution between the training and testingdataset and accordingly, was chosen. In totalthere were 513 and 255 data points for trainingand testing respectively.

A Multi-Layered Perceptron network trained bythe back-propagation algorithm was used forthe ANN modelling. The hyperbolic tangentfunction (bounded between –1 and +1) wasused for the hidden layer with linear transferfunctions at the output layer. The number ofnodes in the hidden layer was set as anoptimisation parameter and was found as 7.

Preliminary resultsThe transport rates predicted by the ANNmodel on the testing dataset were found to bereasonably accurate (Fig. 1). A large scatteroutside the lines of 2-times and 0.5-timesmeasured transport rates were observed in therates predicted by the models of van Rijn (Fig.2) and Engelund-Hansen (Fig. 3) whereas incase of the ANN model the predicted rates

Figure 1. A comparison of the total transport ratespredicted by the ANN model with the measuredtransport rates.

1.E-08

1.E-07

1.E-06

1.E-05

1.E-04

1.E-08 1.E-07 1.E-06 1.E-05 1.E-04Measured transport rate (m2/s)

Com

pute

dtra

nspo

rtra

te(A

NN

)(m2 /s

)

2.0 times measured transport rates




Table 1. Training and testing statistics of the different models.(RMSE: Root mean square error, NRMSE: Normalized root mean square error, r: Coefficient of correlation).

Figure 2. A comparison of the total transport ratespredicted by the van Rijn model with the measuredtransport rates.

were mostly within these lines. Table 1presents the superiority of the ANN model.

ConclusionsThe predictive accuracy of the data-drivenmodel built using an ANN was found to bebetter than that of the models of van Rijn andEngelund-Hansen. However, the dataset usedwas small and did not cover all transportranges. Also the model was not tested withany field data. More research is needed toexplore the effectiveness of a data-drivenmodel in context to sediment transport.

Figure 3. A comparison of the total transport ratespredicted by the Engelund-Hansen model with themeasured transport rates.

Currently, a huge dataset comprising of flumeas well as field data covering many transportand flow ranges is being used to build data-driven models for predicting transport rates.

ReferencesUnited States Army Corps of Engineers, 1935.Studies of River Bed Materials and Their Movementwith Special Reference to the Lower MississippiRiver, Paper 17. U.S. Waterways ExperimentStation, Vicksburg, Mississippi.

Training Testing

RMSE10-7 m2/s

NRMSE r RMSE10-7 m2/s

NRMSE r

Van Rijn 50.8 0.935 0.63 51.5 0.924 0.66

Engelund-Hansen 28.8 0.799 0.71 31.8 0.815 0.68

ANN 12.1 0.623 0.85 11.9 0.623 0.85

1.E-08

1.E-07

1.E-06

1.E-05

1.E-04


Com

pute

dtra

nspo

rtra

te(R

ijn)(

m2 /s)

0.5 times measured transportrates


1.E-08

1.E-07

1.E-06

1.E-05

1.E-04


Com

pute

dtra

nspo

rtra

te(E

-H)(

m2 /s

)0.5 times measured transport rate

2.0 times measured transport rate



Environmental impact assessment study for the CommonMeuse (MER Grensmaas 2003)

J. Onneweer1, B. Peerbolte2 & D.G. Meijer3

1 Projectbureau De Maaswerken, P.O. Box 1593, 6201 BN Maastricht, The Netherlands2 Royal Haskoning, Barbarossastraat 35, 6522 DK Nijmegen, The Netherlands3 Meander Consultancy and Research, Bisonspoor 342A, 3605 JW Maarssen, The Netherlands; [email protected]

BackgroundThe Common Meuse Project is one of themost radical river management projects ofmodern history. Over an almost continuouslength of 40 km, ecological rehabilitationprojects, river widening measures and floodchannels are planned to be realized in thecoming 15 years (Fig. 1). The Common Meuseis the border between Belgium and theNetherlands. The objectives of this project areflood defence, nature rehabilitation and gravelmining. In this newly created ecological areawith forests, brushwood and gravel islands,traditional species of animals and plants, whichhave vanished in the past century, areexpected to return.

The Project Organization “De Maaswerken”assigned the Royal Haskoning and MeanderConsultancy and Research combination for theEnvironmental Impact Assessment (EIA) Study2003 (parts: ‘River Management’ and ‘RiverMorphology’). The combination executed thesestudies, wrote the EIA Reports andparticipated in the presentation of the results tothe citizens.

River Management StudyThe River Management Study comprised theresearch of hydraulic effects of the intendedriver measures. Therefore a 2-dimensionalmodelling system (WAQUA) was applied (Fig.2). The model covers 247 km of the MeuseRiver: from Eijsden (Belgian-Dutch border) toKeizersveer.

For this study six model variants, representingfuture scenarios, were composed(Autonomous Development, PreferredAlternative and four design variants) andapproximately 200 hydraulic simulations(stationary discharges and dynamic floodwaves) were carried out for different sets ofboundary conditions.

The model results were used to assess theeffects of the measures on water levels andflow velocities under different conditions (i.e.daily, bankfull, design and extreme conditions).

Figure 1. Overview of the project area (PreferredAlternative).

Locally the intended measures showed a waterlevel reducing effect of more than a meter. Themodel results supported other parts of the EIAStudy of the Common Meuse, such asMorphology, Ground Water, and Water Qualityas well.

Morphological StudyMorphological desk studyThe Morphological Study was split in two parts.The first part was a desk study, in which the



short-term effects (5 to 10 years) of theCommon Meuse Project on the rivermorphology were assessed by using the 2-dimensional computational results (WAQUA).

Figure 2. The computational grid (WAQUA) of theMeuse (embankments in red)

By analysing the actual and critical shearstresses (under different conditions)morphological tendencies were signalised.Important themes of study were bed stability,bank stability, the position of the river axis andmorphodynamics in general. The bed andbank stability is a very important subject ofanalysis in the ‘bottlenecks’: the original(relatively) narrow river channel between twosuccessive widened locations. An importantaspect of the river morphology in the CommonMeuse is the armouring process of theriverbed, caused by selective sedimenttransport. The riverbed sediment in this river isstrongly graded. The results of this analysissupported the design of bed and bankdefences; required lengths and size of the rockmaterial could be estimated. Other subjectswere the sequence of realisation of the works

and the risk of silt precipitation in the stagnantzones.

Morphological modelling (SOBEK-graded)The second part of the Morphological Studywas supplementary to the first part, and dealtwith long-term (up to 100 years) predictions ofthe riverbed development for different futurescenarios. For this purpose the 1-dimensionalhydraulic and morphological modelling systemSOBEK-graded was applied. The programtakes the gradation of the riverbed material aswell as vertical differentiation of the bedmaterial (in layers) into account. Bothcharacteristics are relevant: the riverbedmaterial consists of a mixture of sand andcoarse gravel, and the geological profile in thisarea shows significant vertical differences. Onseveral locations layers of coarse gravel coverlayers of fine sand.

In total, six model variants were composed:• Autonomous Development, for the

prediction of the morphologicaldevelopments without execution of theCommon Meuse Project;

• Preferred Alternative, for the prediction ofthe morphological developments afterexecution of the Common Meuse Project;

• Two design variants, in which the depths ofthe widening measures and flood channelswere designed 0,5 meter less and more withregard to the Preferred Alternative,respectively;

• Two historical reference situations: 1978and 1987, used for the morphologicalcalibration of the model, the extensivelymonitored time span of 1978-1995 wassimulated for this purpose.

For the 100 years’ simulations four dischargescenarios were considered (historicaldischarges, based on the hydrograph of 1911-1998; high discharges, based on the eighties;low discharges, based on the seventies;average discharges, based on the fifties). Thesensitivity of the riverbed development for thedischarge regime is significant.

The simulations for the Autonomous Develop-ment show a continuous riverbed erosion (Fig.3).



Figure 3. Riverbed development (Autonomous Development, historical discharge scenario).

At some point, the fine-sand layers will bereached, and the riverbed erosion willaccelerate. The most important result of thestudy is the conclusion that beside severallocal erosion and sedimentation sites, the

Common Meuse Project will stop this ongoingerosion and stabilize the global riverbed, asthe reduced flow velocities in the widened riverchannel will interrupt the chain of passingsediment transport (Fig. 4).

Figure 4. Riverbed development (Preferred Alternative, historical discharge scenario).

10

15

20

25

30

35

40

45

0 10 20 30 40 50 60 70

Meuse (rkm)

bed

leve

l(m

+NA

P)

20152040206520902095river measuresbed protection

Bos

sche

rvel

d

Bor

ghar

en

Itter

en

Aan

deM

aas

Mee

rs

Maa

sban

d

Urm

ond

Nat

tenh

oven

Gre

venb

icht

Koe

wei

de

Vis

sers

wee

rt

Roo

ster

en

wei

rB

org

har

en

wei

rL

inn

e

Kot

em

Ber

g

Obb

icht

10

15

20

25

30

35

40

45

0 10 20 30 40 50 60 70

Meuse (km)

bed

leve

l(m

+NA

P)

top of fine-sand layer20152040206520902095

wei

rB

org

har

en

wei

rL

inn

e

exposure of fine-sand layer

Poster Presentations


Vegetation dynamics of a meandering river (Allier, France)

G.W. Geerling1,2, A.M.J. Ragas1 & A.J.M. Smits1,2

1 Department of Environmental Studies, University of Nijmegen, P.O. Box 9010, 6500 GL Nijmegen, The Netherlands;[email protected]

2 Section Nature Conservation of Stream Corridors, University of Nijmegen, P.O. Box 9010, 6500 GL Nijmegen, TheNetherlands.

IntroductionNew riverine nature management strategiesfocus on managing river stretches rather thanindividual floodplains. They combine natureand safety management. For safetymanagement the discharge capacity of fixedrivers has to be maintained at a certain level.The floodplain vegetation relates to thedischarge capacity through its hydraulicresistance. Process knowledge of vegetationpatterns and their dynamics in a natural settingseemed necessary for the design of a goodmanagement strategy. The meanderingprocess of a river creates a diversity oflandscape elements in different stages ofecological succession. By erosion in theoutside bend older succession stages areremoved while at the same time new pioneerstages are formed on the point bar in theinside bend. In the research at hand the mainresearch questions are:• Is there a relation between vegetation

dynamics and scale? Is vegetation typedistribution stable on larger scale?

• How does the meandering process influencethe vegetation distribution? What arerejuvenation frequencies?

• What is the age distribution of successionstages along a meandering river?

Material & MethodThe Allier is a gravel river south of the cityMoulins in the centre of France, Fig. 1.

Figure 1. Oblique aerial photograph showing theresearch area.

It has braided and meandering sections. Thesummer discharge is 20 - 50 m3s-1 and peakdischarge can be up to 1,000 m3s-1.

To be able to study the vegetation types on theriver stretch scale (several floodplains) a GISsystem was used. The use of aerialphotographs made it possible to reconstructthe history of this part of the Allier.The photo series covers the years 1954, 1960,1967, 1978, 1985 and 2000. The photos are ofscale 1:20,000. Products of the mappingprocess are maps as shown in Fig. 2 (year2000 map). The legend on the right shows themapped vegetation types.

Figure 2. A produced map of year 2000.

ResultsA map of floodplain age is created byoverlaying in GIS the bare and water types ofthe different years, see Fig. 3. It shows thespatial distribution of stages of vegetationdevelopment, important for ecological diversity.

Changes seen in the landscape (Fig. 4) arequantified in change matrices.

Table 1 shows the matrix for the 1967-1978transformation. In these matrices therejuvenation can be read as the turnover inwater & bare soil.

The surface area of vegetation types duringthe years is shown in Fig. 5. Especially theearly succession stages grass and bare gravelare dynamic.

Legenda:

Vegetatiestructuurtypen

Cultuurbos

Cultuurgrasland

Gesloten bosvegetatie

Open bosvegetatie

Gesloten struweelvegetatie

Open struweelvegetatie

Ruigtevegetatie

Graslandvegetatie

Pioniervegetatie

Kaal

Water

Strang

0 1 20.5

Kilometers



Table 1. Example change matrix for the years 1967-1978. The turnover is given in percentage area per vegetationtype. The absolute area is given between brackets in the first column.

Figure 3. The age distribution of Allier floodplains.

Discussion & ConclusionsAlong a natural river one will find patches ofvegetation and side channels of various agesand in different stages of succession.Important is to notice that the spatialheterogeneity on the small scale may look likechaos, but seen on the larger scale the systemis quite stable. Continuous succession anddisturbance ensure that no place remains

Figure 5. Percentage coverage of vegetation typesof all the researched years.

Figure 4. Detail of map sequence showing thedevelopment of one point bar.

stable for long, although one will always findpioneer stages, full grown softwood forests orclosed side channels on the river stretch scale.In the ideal situation, the overall vegetationcomposition (and hydraulic resistance) isconstant or fluctuating between upper andlower boundaries, although it might beirregularly fluctuating on the floodplain scale.As a consequence, the scale on whichmanagement by rejuvenation will work is thatof a river stretch because only then enoughnatural elements are present for a dynamic,stable and ecologically rich system.

Future researchThe results presented here are preliminary. Toobtain satisfying answers to the raisedquestions the data will be further analysed. Anarticle will be prepared in spring 2004.

AcknowledgementsThis project is part of a PhD thesis. The PhDproject is funded by the ministry of publicworks and transport, survey department, Delft.Contact person for this project is RegineBrügelmann.I would like to thank my students Sanne,Daniel en Maarten for working soenthusiastically on this subject. Furthermore Iwould like to thank Janrik van den Berg (UU)and Martin Baptist (TUD) and their students forall the discussions and fun we had doing

Sum of 67 - 78 Naar

Van Cultuur Gesloten Bos Gesloten Struweel Grasland Open Bos Open Struweel Pionier Ruigte Strang Water & Kaal Totaal

Cultuur (1023,3 ha.) 96,2 0,1 0,0 0,2 0,0 0,0 0,1 0,5 1,0 1,8 100Gesloten Bos (46,7 ha.) 5,2 59,8 13,6 2,8 2,0 1,0 0,0 5,8 0,2 9,7 100Gesloten Struweel (58,4 ha.) 10,5 13,1 47,0 5,3 1,3 0,5 0,0 4,4 0,0 17,9 100Grasland (170,5 ha.) 5,0 2,5 6,2 54,5 1,0 4,3 0,5 2,0 0,1 23,9 100Open Bos (4,9 ha) 4,2 49,4 1,1 12,6 3,2 4,0 0,0 0,0 0,0 25,5 100Open Struweel (21,9 ha.) 0,0 4,1 36,0 19,6 3,6 12,6 0,1 3,9 0,0 20,1 100Pionier (14,3 ha.) 1,7 2,6 2,3 36,7 4,4 0,8 2,7 0,0 0,0 48,7 100Ruigte (11,5 ha.) 41,1 10,1 11,4 10,7 0,3 11,1 0,2 3,0 0,0 12,0 100Strang (1,4 ha.) 5,9 0,0 0,8 38,6 0,0 0,0 0,0 0,3 35,2 19,2 100Water & Kaal (186,5 ha.) 0,3 1,7 2,7 20,1 0,3 0,5 3,6 5,1 2,5 63,2 100

0 1,000 2,000500

Meters

Legenda:

2000198519781967

19541960

Ouder

OuderdomUiterwaarden

0 1,000 2,000500

Meters

Legenda:

2000198519781967

19541960

Ouder

OuderdomUiterwaarden

Percentage surface area of vegetationtypes

0,0

5,0

10,0

15,0

20,0

25,0

30,0

35,0

40,0

1954 1960 1967 1978 1985 2000

Years

%o

fp

oin

tbar

area


Open bosvegetatie



Ruigtevegetatie

Graslandvegetatie

Pioniervegetatie

Kaal

Stilstaand water

Stromend water

Legenda:


Cultuurbos

Cultuurgrasland


Open bosvegetatie



Ruigtevegetatie

Graslandvegetatie

Pioniervegetatie

Kaal

Water

Strang

1967

1985

1978

2000

Legenda:


Cultuurbos

Cultuurgrasland


Open bosvegetatie



Ruigtevegetatie

Graslandvegetatie

Pioniervegetatie

Kaal

Water

Strang

1967

1985

1978

2000



fieldwork along the Allier river in 2002 and2003!

NoteThe Allier is researched by co-operating NCRinstitutes: M. Baptist (river modelling; DelftUniversity of Technology) and J. van den Berg(river morphology; Utrecht University).

ReferencesDuel, H., M.J. Baptist & W.E. Penning (Eds.), 2001.

Cyclic Floodplain Rejuvenation. A new strategybased on floodplain measures for both flood riskmanagement and enhancement of the biodiversityof the river Rhine. IRMA-SPONGE programPublication 14-2001. Netherlands Centre for RiverStudies, Delft, The Netherlands.

Geerling, G.W., J. van den Berg, M. Baptist, A.M.J.Ragas & A.J.M. Smits, in prep. Vegetationdynamics of the Allier (France) within the contextof dynamic river management; cyclic rejuvenationof floodplains. University of Nijmegen, Nijmegen,The Netherlands.

Liefhebber, D. & M. Breedveld, 2003. Vegetationand Geomorfo-dynamics of the Allier river.Studying the Allier river using historicalphotographs of 1954 to 2000 (in Dutch). StudentReport 236. University of Nijmegen, Nijmegen,The Netherlands.

Smits, A.J.M., P.H. Nienhuis & R.S.E.W. Leuven(eds.), 2000. New approaches to rivermanagement. Backhuys Publishers, Leiden, TheNetherlands.



Soil, hillslope and network structure as an opportunity forsmart catchment scale hydrological models

P.W. Bogaart & P.A. TrochHydrology and Quantitative Water Management Group, Wageningen University, Nieuwe Kanaal 11, 6709 PA Wageningen, TheNetherlands; [email protected]

IntroductionIt is now widely recognised that ‘landscapeproperties’ such as hillslope and catchmentmorphology and soil layer geometry andhydraulic properties are first order controls ofthe hillslope and catchment hydrologicalresponse. The spatial structure of theseproperties is not random in time or space, butis the result of processes like hillslope erosionand soil formation. We aim at understandingthe spatio-temporal structure of hydrologicalrelevant landscape properties by means oflandscape evolution modelling. Two groups ofproperties can be recognised:‘geomorphological’ parameters like slopegradient, hillslope width, soil cover thicknessand the slope type distribution, and ‘hydraulic’parameters such as hydraulic conductivity andeffective porosity. Here we discuss only thefirst, geomorphological, group of parameters.

The Network ScaleThe distribution of hillslope form types(divergent, straight, convergent) is controlledby the spatial structure of the channel networkwithin a catchment. Here we investigate abovedistribution by making use of simulated‘artificial’ channel networks that obey to somegeomorphological rate laws. Within thesenetworks, individual hillslope types can bemapped and their statistics calculated. Theprocedure works as follows:• First, starting from an outlet node new

branches are added from random directions,resulting in a random walk network (Fig. 1a).Under steady-state conditions, fluvialdowncutting rate and tectonic uplift ratebalance each other. By solving equations forthese for ‘graded’ slope gradient, values forelevation can be assigned to each grid cell.The iterative addition of an additionalconstraint that flow from every grid cell istowards the steepest-descent neighbourresults in a realistic topography and flowpattern (Fig. 1b). (This step is based onwork published by Howard (1971)).

• The number and locations of channelbranches entering or leaving grid nodesdefine the number and size of individualhillslopes (Fig. 1c).

• Mapping the hillslope types (Fig. 1d) helpsin understanding the spatial patterns ofthese types. Although the overall distributionis rather uniform, different hillslope types arebound to certain positions. Because eachhillslope type has a different hydrologicresponse, the catchment hydrograph is acomposite of the individual hillslopehydrographs, weighted by their relativefrequency. Initial hydrological modellingshows that the tail of the composite unithydrographs is due to the contribution ofconvergent hillslopes. An implication of thisis that low-flow conditions originate inconvergent, near source topographic areas.

Figure 1. Procedure for mapping individual hillslopetypes.

The Hillslope ScaleOn this scale, the hillslope morphology and thespatial distribution of soil thickness areinvestigated by means of landscape evolutionmodelling techniques, in combination withrandom-walk network techniques describedabove. The procedure operates as follows:• A ‘virgin’-landscape can be constructed by

first crafting a random walk / steepestdescent network (similar as for ‘the networkscale’ above).

• Then this topography is smoothed by a longperiod of diffusive processes like soil creepand shallow mass wasting. Theseprocesses are modelled with a non-lineardiffusion model: it is assumed that soil cover



is non-limiting. The resulting topography isshown in Fig. 2a. Next, soil formation rate ismodelled as an exponential decrease withsoil thickness. The resulting soil thicknessmap is shown in Fig. 2b. It is assumed herethat the channel network contracts, and firstorder channels fill in. This step draws frompublished landscape evolution-modellingtechniques such as published by Tucker &Slingerland (1997).

• For each hillslope, data on elevation, slopegradient and soil thickness can be collected.Here it is assumed that a channel head ispresent at 20 grid cells downhill of the lower-right corner of the map. (Fig. 2c). Hillslopestatistics like collected above can serve asinput for hydrological models such as thehillslope-storage Boussinesq model (seealso the contribution by Hilberts in theseproceedings).

Figure 2. Procedure of landscape evolutionmodelling techniques.

ConclusionThe use of random-walk network andgeomorphological landscape evolution modelsenable the study of the spatial distribution ofhydrological relevant landscape properties.New generations of ‘smart' semi-distributedhydrological models, such as the hydraulicgroundwater theory based `hillslope-storageBoussinesq (hsB) model, (see Troch et al. (inpress) and the contribution by Hilberts (inthese proceedings) may exploit thisunderstanding to parameterise the hydrologicalmodel.

ReferencesHoward, A.D., 1971. Simulation Model of Stream

Capture. Geological Society of America Bulletin.82, pp. 1355-1376.

Troch, P.A., C. Paniconi & E.E. van Loon, in prep.The hillslope-storage Boussinesq model forsubsurface flow and variable source areas alongcomplex hillslopes: 1. Formulation andcharacteristic response. Water ResourcesResearch (in press).

Tucker, G.E. & R. Slingerland, 1997. Drainage basinresponses to climate change. Water ResourcesResearch. 33, pp. 2031-2047.



Integrated flood management strategies; A case study inthe Lower Dong Nai – Sai Gon river system, Vietnam

A.D. Nguyen & N. DoubenUNESCO-IHE Institute for Water Education, P.O. Box 3015, 2601 DA Delft, The Netherlands; [email protected]

IntroductionThe lower Dong Nai – Sai Gon river system issituated in the southwest of Vietnam and hasthe country’s highest economic developmentrate. It houses Vietnam’s largest populationcentre (Ho Chi Minh City) and has the largestconcentration of industrial output. At the sametime, the river basin continues to diversify itsagricultural sector with products ranging frombasic staples like rice and maize to rawmaterials for the local industry (Claudia et al.,2001). As a typical flat and low-lying transitionarea, it is flood prone and faces increasingdamages as a result of re-occurring floodevents, both from rivers as well as stormsurges from the sea (Fig. 1).

Figure 1. Vietnam and the Lower Dong Nai – SaiGon river system.

In general, the existing flood defencestructures are small and situated in a scatteredorder over the river basin (JICA, 1996). Thedike systems are discontinuous and contain arelatively low safety standard (SIWRP, 2001).Two large reservoirs have been developed inthe recent past (Dau Tieng and Tri An).

Table 1. Historical flood damages (JICA, 1996 &SIWRP, 2001).

In the (recent) past the area experienced someexceptional floods, with return periods rangingfrom 1 – 25%, which brought about seriousdamages to private houses and properties,public facilities, agricultural crops and livestock(Table 1). During the flood season, thedrainage possibilities of different rural andurban areas are severely affected.

Flood management strategiesIn order to reduce flood damages and impacts,strategies have been developed with a 1-Dhydrodynamic model (SOBEK-RURAL).Different (steady state) scenarios for futureflood situations have been taken into account,considering both upstream runoff as well astidal dynamics.

Flood management strategies have to complywith relevant aspects of social-economic,physical and ecological conditions in the lowerDong Nai – Sai Gon river system. Therefore,various starting points for different geographicareas have been defined (Fig. 2) (Petry, 2002):• ‘Keep the floods away from people’;• ‘Keep people away from floods’;• ‘Accept floods and clean up afterwards’.

After analysing flood risks, potential damagesand losses for future flood situations (based onhistoric GIS data), flood managementstrategies have been derived, primarily basedon the combination and complementarily ofstructural and non-structural measures, takingeffectiveness and technical and socio-economic feasibility into consideration (Greenet al., 2000).

Figure 2. Spatial distribution of starting points fordevelopment of flood management strategies.

DamagesYear(freq.)

Houses(#)

Road(km)

Paddycrops (ha)

Cereals(ha)

1952(1%) 7,598 13 31,270 120

1978(10%) 1,834 20.5 8,110 203

2000(25%) 1,100 32 6,500 100

Lower Dong Nai – Sai Gon river system

Ho Chi Minh City



Non-structural approaches, as defined by ICID(1999), comprise activities, which are plannedto eliminate or mitigate adverse effects offlooding without involving the construction offlow-modifying structures (i.e. land-useplanning, preparedness and emergency andrelief).

ResultsIn total eight different strategies have beendefined, based on three conceptualapproaches:• Different principles of structural measures

(discharge control, water level control andconfinement);

• Compatibility between structural and non-structural measures, and;

• Spatial distribution and feasibility of variousmeasures.

The developed strategies have been assessedand compared with a specific set of criteria,derived from studies like Hooijer et al. (2002)and Silva et al. (2001) (i.e. costs, excavationworks, land-use, environmental impacts,implementation time, etc.) in order todetermine a preferred strategy, which could besuitable for implementation.

The strategies can be roughly divided intothree main categories: favourable, lessfavourable and non-favourable. The advise onfavourable strategies is composed by takingexisting environmental conditions (ecologically,socially and economically) of the study areainto consideration. The ‘storage & confinement’strategy (reservoirs, detention, dikes andembankments, and various non-structuralmeasures) seems to be the most promisingsolution to mitigate floods in the lower DongNai – Sai Gon river system (Fig. 3).

Figure 3. Spatial distribution of measures in the‘storage & confinement’ strategy.

DiscussionThe results indicate that the implementation ofthe ‘storage & confinement’ strategy willreduce peak water levels significantly as wellas the consequences of flooding, hencepotential losses and damages. Although aconsiderable reduction of flood impacts will beachieved, pre-established protection levels stillare not fully met in many urban andeconomically important areas. Additionalmeasures (e.g. drainage works, local dike andembankment systems) need to be studied andimplemented as well.As a result of lacking environmental systemdata, together with non-accurate hydrologicaldata, the inter-relationships between land andwater can be considered as the most importantuncertainty of the study. Therefore, further(implementation) studies should consider thebasin as a whole, not only in terms of thegeographical and functional interdependenciesinvolved, but also the inter-relationshipsbetween land and water as well as economicand human development.

ReferencesClaudia, R., C.C. Nguyen & V.H. Nguyen, 2001.

Water Allocation and Use in the Dong Nai RiverBasin in the Context of Water Sector Reforms.Workshop IWRM, Manglang, Indonesia.

Green, C.H., D.J. Parker & S.M. Tunstall, 2000.Assessment of Flood Control and ManagementOptions. WCD Thematic reviews. WorldCommission on Dams Secretariat, South Africa.

Hooijer, A., F. Klijn, J. Kwadijk & B. Pedroli, 2002.Towards Sustainable Flood Risk Management inthe Rhine and Meuse River basins, Main resultsof the IRMA SPONGE research program. NCR-Publication 18-2002. Netherlands Centre for Riverstudies, Delft, The Netherlands.

International Commission on Irrigation and Drainage(ICID), 1999. Manual on Non-structuralApproaches to Flood Management. New Delhi,India.

JICA, 1996. The Master Plan Study on Dong NaiRiver and Surrounding Basins Water ResourcesDevelopment. Nippon Koei Co. Ltd., Tokyo,Japan.

Petry, B., 2002. Coping with floods: complementarilyof structural and non-structural measures.Proceedings 2nd International Symposium onFlood Defence. Beijing, China.

Silva, W., F. Klijn & J. Dijkman, 2001. Room for theRhine branches in the Netherlands, what theresearch has taught us. Institute for Inland WaterManagement and Waste Water Treatment & WL |Delft Hydraulics, Report no’s 2001.031 & R3294.Arnhem/Delft, The Netherlands.

Sub-Institute for Water Resources Planning(SIWRP), 2001. Water Resources Plan forSouthern Focal Economic Area (in Vietnamese).Ho Chi Minh City, Vietnam.



Integrated water resources management groupwork - caseMura river basin

T.D. Schotanus & M.L. MulUNESCO-IHE Institute for Water Education, Department Management and Institutions, P.O. Box 3015, 2601 DA Delft, TheNetherlands; [email protected] & [email protected]

AbstractThe Mura groupwork is an integrating exerciseat the end of the curriculum of the WaterManagement programme at UNESCO-IHE.The groupwork concentrates around thefictitious Mura river basin. Teams are selectedbased on creating multi-disciplinary teams thathave to develop an Integrated WaterResources Management plan for this riverbasin. In four weeks the participants use theirexperience and acquired knowledge indeveloping this plan.

IntroductionThe groupwork on the Mura river basin is apart of the curriculum of the WaterManagement (WM) programme at UNESCO-IHE Delft. After nine months of acquiringknowledge in many different fields, such asintegrated water resource management,modelling, environmental planning, watershedmanagement and water law and institutions, allaspects are addressed and integrated in thisgroupwork.

ObjectivesThe main objective of the groupwork is toapply experience and understanding, insightand skills gained in preceding modules tocome up with environmentally andeconomically sustainable policy suggestions,strategic plan and practical solutions.Furthermore the participants have to learn todeal with and manage time, data and moneyconstraints in planning. They have to learnhow to work in a multi-disciplinary team andmake optimal use of it, with recognizing andstimulating good leadership qualities in selfand others.

The caseThe Mura river basin is a fictitious river basinsituated somewhere in the south-eastern Asiaregion in an equatorial island country calledAsiatica (Fig. 1). The Mura river basinillustrates many of the severe problems typicalfor the region and climate in water resourcemanagement. The basin experiences severeproblems such as floods, water qualityproblems (especially in the capital city Atra,due to lack of sanitation) and allocation

Figure 1. The Mura river basin.

problems between different water users, withthe most important water using activities beingdomestic and industrial use, agriculture,hydropower, ecology, recreation and water toflush the canals in the city of Atra (Fig. 2).

Integrated water resources management isessential to provide the basin’s inhabitants andsurrounding areas with an economic andenvironmentally sustainable way forward.

Figure 2. Problems in Mura river basin.



Set-upThe participants were divided into teams; themembers of each team were selected basedon background and affinity with the topics inthe groupwork, so as to create multi-disciplinary teams. The teams representedcompeting consultancy companies that wereinvited to develop a 20-year Integrated WaterResources Management Plan for the Murariver basin by the government of Asiatica. Theparticipants were provided with a lot ofinformation and data on the river basin (seenote) and its problems. They used this data toestimate the effects of their management plansand to come up with an environmentally andeconomically sustainable Integrated WaterResources Management Plan. Staff membersof UNESCO-IHE acted as the governmentrepresentatives who set-out the invitation fordeveloping the IWRM-plan and who assessedthe final plans. They also acted as expertconsultants who helped to structure theprocess, assist the teams and provideadditional information when necessary.

Groupwork 2003In July 2003 three teams of 10 participantsstarted with this groupwork, it took place in foursequential weeks, full time.

The first week the participants presented aproblem analysis of the river basin and aninception report. At the end of the secondweek they presented their strategy for the riverbasin. The last two weeks were dedicated todeveloping the IWRM plan for the Mura riverbasin. At the end of the last week theypresented and submitted the final report (seeFig. 3 for an example of a presentation). Allgroups presented their ideas and plans with alot of enthusiasm, which resulted in successfulcompletion of the groupwork.

ConclusionsThe groupwork turned out to be a goodintegrating exercise at the end of the WaterManagement programme at UNESCO-IHE.The participants were encouraged to addressand integrate all aspect of IWRM in developingan IWRM-plan for the Mura river basin. In theend the exercise was much appreciated by theparticipants.

NotesQuantitative data adapted from WL | DelftHydraulics (1986), Cisedane-Cimanuk IntegratedWater Resources Development (BTA-155). Usedwith permission.

Figure 3. Integrated River Basin Strategy.



Effects of subsurface parameterisation on runoffgeneration in the Geer basin

G.P. Zhang1, F. Fenicia1, T.H.M. Rientjes1, P. Reggiani2 & H.H.G. Savenije1


2 WL | Delft Hydraulics, P.O. Box 177, 2600 MH Delft, The Netherlands

AbstractA rainfall-runoff model has been developed forthe Geer catchment Basin based on theRepresentative Elementary Watershed (REW)approach. To test the model, the sensitivityanalysis has been carried out with respect tothe effects of the subsurface parameterisationon the runoff generation processes. Inaddition, a new approach for representing therelation between topography and variablesource area within a REW is proposed.Simulation results show that REW approach isappropriate for investigating rainfall runoffrelation.

IntroductionThe Representative Elementary Watershed(REW) approach has been introduced byReggiani et al. (1998, 1999, 2000 & 2001) andfully described by Reggiani & Rientjes (2003).By this approach, a catchment is discretizedinto a number of sub-watersheds, calledREWs, according to a specified Strahler order.Each REW consists of 5 model zones in whichwater flows are simulated based on a couplingprocedure of mass conservation andmomentum balance equations. Simulatedflows are (1) unsaturated zone flow, (2)saturated zone flow, (3) saturation overlandflow, (4) concentrated overland flow, and (5)channel flow, as shown in Fig. 1. Theapproach is classified as semi-distributedphysically based.

Figure 1. Schematised cross-sectional profile of aREW and the flow process.

The saturated zone plays a key role since thesize of dynamically varying saturated overlandflow source areas is function of the water tabledepth. Also groundwater flow across REW

boundaries and interactions between thegroundwater and river are simulated in thiszone. On the source areas, rainfall andexfiltration water are directly transformed intorunoff that largely determines the peaks ofhydrographs. In the model applied here, theconcentrated overland flow module has not yetbeen implemented. It only serves as aninterface between atmospheric zone andunsaturated zone to capture rainfall andfacilitate infiltration.Based on the work by Zhang et al. (2002),some improvements are introduced to themodel: the average bottom elevation of eachREW is defined with respect to its own surfaceelevation; parameter values of saturatedhydraulic conductivity and soil porosity in theupper and lower layer (Kus, Kss, εu, εL), aredissimilar respectively. Especially, therepresentation of the saturated overland flowarea fraction, ωo is redefined so that thetopography of REWs (or sub-catchments) canbe incorporated. In this work, detailednumerical simulations are carried out toanalyse the runoff behaviour with variousmodel parameterisations. Also effects ofdifferent catchment discretisation on runoffprocesses are studied.

Parameterised model EquationsInfiltration: ( )

+

Λ= u

cu

u

usuucu shy

Kie

2

1,min

ωρωρ (1)

Evaporation: uup

uwg see ρω−= (2)

Percolation: uuz

ususe ωρνβ±= (3)

Seepage to river: ( )hshrPlK

er

rrsrsr −Λ

±=ρ

(4)

Exfiltration to surface: ( )hshoK

eo

s

ossso −

Λ±=

γωρ

cos(5)

Groundwater flow: ( ) ( )[ ]sjsjsisiijij yyge ςςρα +−+±= (6)

Overland flow: ( ) 3/22/11 oyJn

=ν (7)

Channel flow:2/1

8

=

p

mJ

f

grχν (8)

Datum

Average surface

Average bottom

yu

ys

zs

zr

Z

eus

eso

ecu

esr

euwg

Average water table



REW geometry (area fraction of flow zones):

1≈≈+ sco ωωω (9)

ou ωω −=1 (10)

α

ωtg

rsurf

rss

sfo

zz

zzyA

−−+

= (11)

Numerical simulationsModel simulations are carried out for a genericcatchment based on the Geer basintopography. The basin is a sub-basin of theriver Meuse in Belgium and covers about 490km2. It is characterized by a deep groundwatersystem where the aquifer extends beyond thesurface boundaries of the catchment.Groundwater abstraction through pumpingwells and drainage galleries are present in thebasin and are simulated based on theavailable data. In the model subsurfaceproperties are assumed to be uniformlydistributed over the whole catchment.At first, model sensitivity towards the modelparameters is analysed. For this analysis 5numerical simulations are performed, of which4 cases are based on 2nd order stream networkdiscretisation (73 REWs, Fig. 2) and 1 on 3rd

order discretisation (17 REWs, Fig. 2).

Figure 2. Discretisation of the Geer River basin.

To each simulation a 5-year constant effectiverainfall series is set as the model forcing input.Parameter sets of the simulations for eachcase are listed in the Table 1. Results arepresented in Fig. 3.

Table 1. Parameter sets for the sensitivity analysis.

Figure 3. Sensitivity analysis for river discharges.

Secondly, model simulations are carried outwith observed rainfall/evaporation time seriescovering the period 1993-1997. Results arepresented in Fig. 4 and Fig. 5.

Figure 4. Comparison of simulated and measureddischarges at outlet of Geer river.

Figure 5. Cumulative discharges (1993-1997) at theoutlet of the Geer river.

The model performance is evaluated by meansof total water volume error, correlationcoefficient (R), root mean square error (RMSE)and Nash-Sutcliffe efficiency (NSE) that areshown in Table 2.

Table 2. Statistical analysis for the modelperformance.

DiscussionsResults of sensitivity analysis (Fig. 3) showthat the model is sensitive to subsurfaceparameters’ variations. It is shown that the sizeof saturated overland flow areas (sourceareas) varies consistently subject to differentsubsurface parameterisation. The saturatedoverland flow runoff generation mechanism,however, is well represented. In view of thedifferences between the results for Case IIIand Case V, the effects of catchment

Case I εεεε=0.35, Ks=1.0m/d, 73REWs; εεεε and Ks

same for Uzone and Szone

Case II ε=0.35, Ks=2.0m/d, 73REWs; ε and Ks

same for Uzone and Szone

Case III εu=0.45, εL=0.3, Ksu=1.5m/d, Kss=1.0m/d,73REWs

Case IV εu=0.45, εL=0.3, Ksu=3.0m/d, Kss=2.0m/d,73REWs

Case V εu=0.45, εL=0.3, Ksu=1.5m/d, Kss=1.0m/d,17REWs

Water volumeerror (%)

R RMSE(m3/s)

NSE

17 REWs 2.9 0.705 0.691 0.474

73 REWs 4.0 0.702 0.684 0.485



discretisation on runoff generation can beobserved:• It is clearly shown (Fig.4) that rainfall runoff

responses in the catchment are wellsimulated. In most of the simulation period,base flow is well simulated while some ofthe simulated peak discharges match themeasured ones reasonably well. Lag time ofpeak discharges is generally well matchedalthough some mismatches can beobserved.

• From Fig. 5 and Table1, water volumeerrors indicate that the water balance wasperfectly computed by the model. The modelrepresents the Geer basin well even thoughNSE coefficient is not yet satisfactory. Themodel can reach similar performanceindices applying two different catchmentdiscretisation.

Conclusions• The model is able to simulate the real world

runoff production mechanisms and waterbalance behaviour of the catchment.

• The model requires fewer parameters andless input data compared to fully distributed,physically based models (see Rientjes,1999) while yet the physics underlyingwatershed runoff dynamics is maintained.

• The REW approach is an appropriateapproach for catchment-scale hydrologicalstudies.

AcknowledgementThis research is funded by Delft Cluster project“Oppervlaktewater hydrologie”.

ReferencesReggiani, P., M. Sivapalan & S.M. Hassanizadeh,

1998. A unifying framework of watershedthermodynamics: balance equations for mass,momentum, energy and entropy and the secondlaw of thermodynamics. Advances in WaterResources. 22(4), pp. 367-368.

Reggiani, P., S.M. Hassanizadeh, M. Sivapalan &W.G. Gray, 1999. A unifying framework ofwatershed thermodynamics: Constitutiverelationships. Advances in Water Resources.23(1), pp. 15-39.

Reggiani, P., M. Sivapalan & S.M. Hassanizadeh,2000. Conservation equations governing hillsloperesponses: Physical basis of water balance.Water Resources Research. 38(7), pp. 1845-1863.

Reggiani, P., M. Sivapalan, S.M. Hassanizadeh &W.G. Gray, 2001. Coupled equations for massand momentum balance in a stream network:theoretical derivation and computationalexperiments. Proc. The Royal Society, A(2001)457, pp. 157-189.

Reggiani, P. & T. Rientjes, 2003. The representativeelementary watershed (REW) approach asalternative modelling blueprint: application to anatural basin. Water Resources Research.(submitted).

Rientjes, T., 1999. Physically Based Rainfall-runoffmodelling applying a Geographical InformationSystem. Communications of the WaterManagement and Sanitary Engineering DivisionReport no 83, Delft University of Technology.Delft, The Netherlands, pp. 74.

Zhang, G.P., P. Reggiani, T.H.M. Rientjes & S.M.Hassanizadeh, 2002. Modelling Rainfall-RunoffRelation by Representative ElementaryWatershed Approach. In: Leuven, R.S.E.W., A.G.van Os & P.H. Nienhuis (Eds.), ProceedingsNCR-days 2002: Current themes in Dutch riverresearch. Netherlands Centre for River studies,NCR-publication 20-2003, Delft, The Netherlands.



River nutrient transport under various economicdevelopment scenarios – the transboundary LakePeipsi/Chudskoe drainage basin

D.S.J. Mourad1, M. van der Perk1, G.D. Gooch2, E. Loigu3, K. Piirimäe3 & P. Stålnacke4

1 Department of Physical Geography, Faculty of Geosciences, Utrecht University, P.O. Box 80115, 3508 TC Utrecht, TheNetherlands; [email protected]

2 Department of Management and Economics, Linköping University, 58183 Linköping, Sweden3 Department of Environmental Engineering, Tallinn Technical University, Järvevana tee 5, 19086 Tallinn, Estonia4 Jordforsk (Norwegian Centre for Environmental Research), Frederik A. Dahls vei 20, 1432 Ås, Norway

The Lake Peipsi/Chudskoe drainagebasinLake Peipsi is a shallow lowland lake on theborder of Estonia and Russia, and in size thefifth largest lake of Europe (3,555 km2). Itsdrainage basin (Fig. 1 and 2) is shared byRussia (67 %), Estonia (27 %) and Latvia (7%).

Figure 1. Study area.

Figure 2. Landscape near Tartu.

Nutrient input and eutrophicationproblemsAlthough the lake is relatively undisturbed, it isvulnerable for eutrophication problems,because it is shallow and the ecological

balance is easily disturbed by changes innutrient input from the drainage basincombined with internal lake processes.Eutrophication causes algal blooms anddecreases in fish stock (Fig. 3 and 4).

Figure 3. Local fishermen fishing on the ice (photo:P. Unt).

Figure 4. Lake Peipsi/Chudskoe suffering fromEutrophication (summer 2002).

What happens in the future? –definition of five scenariosWithin the next twenty years, the area isundergoing rapid economical changes thathave influence on lake nutrient input. The EUWater Framework Directive (WFD) calls for anintegrated catchment approach andinformation about past, present and futurenutrient loads. This information helpsmitigating high loads.

Five scenarios for future development weredeveloped, on basis of the two variables



Figure 5. Scenario development: two axes and fourscenarios plus 1.

economic growth and international cooperationin the region (Fig. 5).

Putting this information in a nutrienttransport modelFirst, a large-scale GIS-based nutrientemissions and transport model (Fig. 6) was setup for the 1985 - 1999 period for whichcalibration data were available. After that, themodel period was extended to include thefuture scenarios (until 2019). Economicdevelopment scenarios, usually consisting ofqualitative storylines about expected changes,were translated into quantitative model inputchanges. Examples are changes in livestockand crop harvest, population amount, andwastewater treatment efficiency. Future TotalNitrogen (Ntot) and Total Phosphorus (Ptot)loads for the 2015 - 2019 period werecalculated.

Figure 6. Nutrient emissions and transport model.

Future load predictions andmanagement recommendations

Fig. 7 and 8 show the predicted Ntot and Ptotloads into the lake for the past and the fivefuture scenarios. For Ntot, only the Target/Fastdevelopment scenario delivers a loadcomparable to the communist period. Theother scenarios result in lower loads. For Ptot,loads are generally low in all scenarios. Fig. 9(an example from the 1995-1999 period)indicates that the share of point sources intotal loads is much higher for Ptot than forNtot. This explains the different future loaddevelopment for both compounds. Focussingon agriculture is the best way to mitigate Ntotloads, while a decrease of Ptot loads to thelake can be achieved by better waste watertreatment.

Figure 7. Simulated Ntot loads (T/yr) for 1985 - 1999 and five future scenarios.

Figure 8. Simulated Ptot loads (T/yr) 1985-1999 and five future scenarios.



Figure 9. 1995 - 1999: Share of point/diffuse sources loads for various regions in the drainage basin. The darkcolour is the point source contribution, the light colours represent diffuse sources.

LiteratureMourad, D.S.J., M. van der Perk, G.D. Gooch, E.

Loigu, K. Piirimäe & P. Stålnacke, 2003. GIS-based quantification of future nutrient loads intoLake Peipsi/Chudskoe using qualitative regionaldevelopment scenarios. Water Science andTechnology (submitted).



Reduction of siltation in harbours

S.A.H. van SchijndelWL | Delft Hydraulics, Inland Water Systems, P.O. Box 177, 2600 MH Delft, The Netherlands;[email protected]

IntroductionMany river and lake harbours suffer fromsignificant siltation, leading to frequentdredging activities to maintain the necessarywater depth. Stricter legislation, classifying thesediments as chemical waste, raises the costsfor the removal of these sedimentsenormously. In some cases this even leads topermanent closure of the smaller, recreationalharbours. Besides the smaller scale harboursalso the large commercial ports throughout theworld suffer from siltation problems. Theselarge economic consequences have initiatednumerous studies on the processes of siltationand on measures to reduce siltation in varioussituations.

Siltation mechanismsIn order to determine the siltation rate in aharbour it is important to distinguish theprocesses that bring the sediment into theharbour. In general the sediment is transportedin suspension and by exchange through theharbour entrance it is brought into the harbourwhere it is deposited because of relatively lowflow velocities. Fig. 1 shows in a schematicoverview which parameters are of importanceand which exchange mechanisms can occur.

Detailed analysis of the different exchangeprocesses has lead to the development of anumber of silt reducing measures. Whichexchange process is dominant is dependenton the water system. At the upstream part ofthe river and in lakes exchange through

Figure 1. Schematic overview of importantparameters and exchange mechanisms.

horizontal entrainment is almost always themost important mechanism.

Tidal effects and density currents generally donot occur in these situations. Theseproceedings describe the results of two casestudies where a river situation without tidaleffects and a lake system were studied todevelop silt-reducing measures.

Rivers, a scale model approachIn the harbour of Roermond at the river Meusehorizontal entrainment through the mixing layeris the dominant mechanism. Withmeasurements on the exchange of warm waterin a scale model this mechanism wasextensively studied leading to effectivesolutions for the reduction of siltation. Varyingfrom simple measures that can be designedwith a desk study to measures that requireextensive scale model research.

In a river situation without tidal effectsmeasures are focussed on reducing thehorizontal exchange through:1. Reduction of the cross section of the

harbour;2. Decrease in the strength of the mixing

layer;3. Modification of the configuration of the

harbour entrance;4. Diversion of the mixing layer from the

harbour entrance.

Figure 2. Schematic overview of scale model withmeasures.

NCR-daysPoster

Award 2003



The scale model study showed that thesiltation in a harbour could be reduced up to80% with a combination of a curved sill in theharbour entrance and a pile sheet at theupstream side of the entrance. Fig. 2 showsthis combination of principles 1 and 2.

The pile sheet influences the mixing layer infront of the harbour in such a way that thegradient in flow velocity becomes smaller andthe exchange of water is reduced. The sillreduces the cross section and, because of itsshape, more or less isolates the primary eddyin the harbour entrance, thus reducing thestrength of the secondary eddy and thereforethe exchange. Diversion of the mixing layer bymeans of a Current Deflecting Wall alsoproved to be very effective. A CurrentDeflecting Wall has been built in the Köhlfleetharbour of Hamburg and measurements in thefield have shown a reduction of the rate ofsiltation with 40%.

Lakes, a numerical model approachThe problem of siltation in lake systems hasbeen studied by means of a three-dimensionalnumerical model (Delft3D) of the LoosdrechtLakes. This model is well suited to describetransport processes of silt and the complexcirculating flows in both horizontal and verticaldirection that are caused by wind and waves.

The study focussed on the situation of Mantenharbour in the south west corner of the lake.Fig. 3 shows a picture of Manten harbour

where the exchange of silt through the harbourentrance is clearly visible. The study hasshown that in lake systems the siltation is notdependent on the exchange between lake andharbour but only on the trap efficiency. So,accumulation of very fine organic sedimentsthat are mixed well over the vertical onlyoccurs in sheltered areas.

In a lake system measures to reduce siltationfocus on the reduction of the trap efficiency inthe harbour. This can be realised by usingopen constructions as much as possible toprevent the development of dead water zones.To minimize siltation in new harbours theyshould not be developed in corners orsheltered areas of a lake.

Figure 3. Manten harbour at Loosdrecht Lakes(courtesy: Mr. Veenendaal).



Morphological behaviour around bifurcation points of theDutch Rhine branches

L.J. Bolwidt & P. JesseInstitute for Inland Water Management and Waste Water Treatment, P.O. Box 9072, 6800 ED Arnhem, The Netherlands;[email protected]

AbstractBifurcation points are important for thediversion of water and sediment. In this studythe morphological behaviour of two of the mainbifurcation points in the Netherlands will beinvestigated. Research proposals have beenset up and the first measurements took place.At the first coming peak discharge, the rest ofthe measurements will be performed.

IntroductionThere are three main bifurcation points in theRhine branches in the Netherlands. At theseriver bifurcations not only water is divided overthe two branches, but also the sediment loadin the water. The division of sediment is notnecessarily equal to the water discharge.The upstream sediment supply andcomposition influence local sedimentationpatterns.Little is known about the morphologicalbehaviour around these bifurcation points,while at these points you could tune the waterand sediment fluxes.To understand the processes aroundbifurcations a measuring programme has beenset up. Due to the fact that little is known aboutsediment division in the field there are alsodifficulties to calibrate morphological models.In the Netherlands the three main bifurcationpoints are (Fig. 1):• Pannerdensche kop (Boven Rijn,

Pannerdensch kanaal, Waal);• Merwede kop (Boven, Beneden en Nieuwe

Merwede);• IJsselkop (Pannerdensch kanaal, IJssel and

Nederrijn).

In the 90’s research has been performedaround the Pannerdensche Kop. At thismoment two research proposals have been setup for the Merwede Kop (Bolwidt & ten Brinke,2000) and the IJsselkop (Bolwidt & Jesse,2002).

This research is performed under authority ofthe Directorates Eastern Netherlands (DON)and Southern-Holland (DZH) ofRijkswaterstaat, in cooperation with Universityof Utrecht, TNO-NITG and MEDUSA.

Figure 1. The IJsselkop bifurcation during lowdischarges, on the left the Nederrijn, on the right theIJssel, August 2003 (Photo: Bert Boekhoven).

MethodsFor both locations an integral approach ismade with measurements on:Subsoil:Goal: to determine the structure and

composition of the subsoil. Besides thelocation of coarse and fine material, theresulting database gives information ofthe thickness of the active layer. Thislayer has been in motion during pasthigh discharges.

With: Vibrocore (grain size) and Seismicsystem (subsoil layers)

Sediment transport:Goal: To determine the amount and

composition of sediment load. Divisionover the bifurcation is taken intoaccount.

With: Delft Nile Sampler DNS (Bed loadtransport) and “Akoestisch ZandTransport Meter” AZTM (Suspendedtransport)

Bed forms:Goal: To determine the spatial and temporal

variation in bed load transport and grainsize of the top layer. Measurements aredirected at river dunes, which develop athigh discharges.

With: Multibeam (bed level) and MEDUSA(grain size top layer) from a travellingship.



Water level and discharge:Goal: To understand the driving force behind

the morphological processes you needto have detailed information ondischarge and water levels at allbranches around the bifurcations.

With: ADCP and water level recorders.

Different conditions-differentapproachMost morphological changes of the riverbedwill manifest during high discharges. Thereforemost measurements will be done during thisrelative short period. Due to localcircumstances, there will be differences inhydrological and grain size and therefore eachbifurcation will have a different additionalapproach.

Merwede KopFor the Merwede bifurcation point the tidaleffect is influencing the morphologicalbehaviour, which makes the situation complex.At the beginning of a discharge peak, 13-hoursmeasurements will be performed to determinethe effect of the tide at the sediment transport.

IJsselkopDue to the damming of the Nederrijn at lowdischarges the IJssel gets a higher discharge.The effect on the morphological behaviour willbe investigated at low discharges.

Results till nowLast year (2002) the measurements of thesubsoil around the IJsselkop have beenperformed by TNO-NITG (Gruijters et al.,2003), as well as one set of multibeammeasurements to determine the dune patternsin this area at high discharges.In fall of this year (2003) the low dischargemeasurements will be performed.

Around the Merwedes, so far a test campaignhas been made to give an impression of theeffect of the tides. At the end of the project allthe different parts of the studies will beintegrated in one report.The university of Utrecht has written a PhDproposal, in which the result of thesemeasurements will be used to understand theprocess of downstream fining (Frings, in prep).

Conclusions and future researchAt the first coming discharge peak themeasurements will start to determine themorphological behaviour during differentdischarges. In the end it will result in betterunderstanding of the water and sedimentbehaviour around these two bifurcation points.Together with the knowledge of thePannerdensche kop, the behaviour of theDutch Rhine river system can be betterunderstood in its present function, andpredictions can be made about theconsequences of interventions in the riversystem.

ReferencesBolwidt, L.J. & W.B.M. ten Brinke, 2000. Meetplan

Sedimenttransport Splitsingspunt Merwedes (inDutch). RIZA memo WSR/2000-030. Institute forInland Water Management and Waste WaterTreatment, Arnhem, The Netherlands.

Bolwidt, L.J. & P. Jesse, 2002. Meet- enanalyseplan: Onderzoek morfologische processensplitsingspunt IJsselkop (in Dutch). RIZAWSR/2002-15. Institute for Inland WaterManagement and Waste Water Treatment,Arnhem, The Netherlands.

Gruijters, S.H.L.L., J. Gunnink, J.J.M. Hettelaar,M.P.E. de Kleine, D. Maljers & J.G. Veldkamp,2003. Katering ondergrond IJsselkop, Fase 3,eindrapport (in Dutch). TNO-NITG report NITG03-120-B. Utrecht, The Netherlands.

Frings, R.M. (in prep). Downstream fining, aliterature review. Utrecht University, TheNetherlands.



Which processes determine the downstream fining of bedsediment in the Dutch Rhine branches?

R.M. FringsDepartment of Physical Geography, Utrecht University, P.O. Box 80115, 3508 TC Utrecht, The Netherlands;[email protected]

AbstractThe impact of abrasion, selective transport andsediment exchange processes on thedownstream fining of bed sediments in theDutch Rhine branches is evaluated. Abrasionhas probably a small influence near theGerman border, but is negligible in thedownstream part of the Rhine branches.Selective transport is assumed to be theprimary downstream fining mechanism. Thedegree of selective transport is probablyenhanced by vertical and horizontal sortingprocesses at the riverbed and by the presenceof suspended load transport. Among thesediment exchange processes, beddegradation is assumed to have the largestinfluence on downstream fining. At a smallerscale, the sediment distribution at riverbifurcations is important. Floodplainsedimentation and dredging probably haveonly a minor influence.

IntroductionMany rivers are characterized by adownstream decrease in bed grain size. Theprocesses causing this downstream finingtrend have extensively been studied in gravelbed rivers. Downstream fining processes insand bed rivers and sand-gravel bed rivershave received much less attention.The aim of the ongoing research is to identifythe primary causes of downstream fining in thelower reach of the river Rhine, which is a sand-gravel bed river near the German border and asand bed river near its mouth (Fig. 1).

Three possible influences on downstreamfining are considered: abrasion, selectivetransport and sediment exchange processes.

Figure 1. Fine bed sediments exposed alongsidethe river IJssel during the low-flow period of august2003.

AbrasionAbrasion, the mechanical breakdown ofindividual grains is usually assumed to benegligible in sand bed rivers, because of thelow shear stress, the small grain size, the highdegree of grain roundness and the virtualabsence of non-durable grains (e.g. Kuenen,1956). For the largest part of the Dutch Rhinebranches, therefore, the downstream fining ofbed sediment cannot be explained byabrasion. It is possible, however, that abrasionhas a small, but significant influence on thedownstream fining trend in the Rhine branchesnear the German border. In this area theriverbed consists of a sand-gravel mixture andBradley (1970) observed that sand grains insand-gravel mixtures are splitted and crushedby the impact of gravel grains. Kodama (1994)also observed that the abrasion rate of finegrains increased in the presence of coarsegrains.

Selective transport: suspension andsorting processesWhile gravel bed rivers are dominated by bedload transport, sand-gravel bed rivers andsand bed rivers also have a considerablesuspended load transport. This divisionbetween bed load transport and suspendedload transport has a major influence ondownstream fining. This is because coarsegrains are only taken into suspension at veryhigh bed shear stresses. Therefore thecomposition of the fast moving suspended loadlayer is much finer than the composition of theslowly moving bed load layer. As a result finegrains are preferentially transporteddownstream. Selective transport, however, notonly results from the division between bed loadtransport and suspended load transport. Thebed load transport and suspended loadtransport itself are also size-selective,contributing significantly to the downstreamfining trend. The size-selectivity of the bed loadtransport is strongly determined by vertical andhorizontal sorting processes at the riverbed.Two important vertical sorting processes aredune sorting (Kleinhans, 2002) and armouring(Parker & Klingeman, 1982). Dune sortingconcentrates coarse grains in deep bed layers(Fig. 2), which are only mobile at high



Figure 2. Vertical sorting by river dunes (afterKleinhans, 2002).

discharges. This decreases the mobility of thecoarse grains and increases the degree ofselective transport. Armouring has theopposite effect: coarse grains are concentratedat the bed surface where they are easilyentrained, thus decreasing the mobilitydifference between coarse and fine grains.Horizontal sorting processes may lead to thedevelopment of coarse and fine patches on thebed surface. It is assumed that within thosepatches grains of all sizes are nearly equallymobile, but because fine patches move fasterthan coarse patches, the overall bed transportwill be size-selective (Paola & Seal, 1995).The size-selectivity of the suspended loadtransport is the result of the large settlingvelocity of the large suspended grains.Therefore they are only present in thelowermost part of the suspended load layer,while fine grains are present throughout thesuspended load layer. Because the flowvelocity in the lower part of the suspended loadlayer is relatively low, the average velocity ofcoarse suspended grains is smaller than theaverage velocity of fine suspended grains,implying selective transport (Deigaard, 1980).

Sediment exchange processesThe systematic downstream fining trendcaused by abrasion and selective transportcan be disturbed by the supply of material witha different grain size into the river, or by thesize-selective withdrawal of material from theriver. In the case of the river Rhine fourpossibly important processes have beenidentified: floodplain sedimentation, dredging,bed degradation and sediment distribution atriver bifurcations.Floodplain sedimentation and dredging lead toan extraction of fine material from the riverbed.This has a diminishing influence on thedownstream fining rate. The influence isassumed to be relatively small because of thesmall amounts of sediment involved.Bed degradation involves a supply of materialof different origin into the river. Taking intoaccount the large bed level decrease that tookplace in the Rhine branches during the lastcentury, this must have had a significantinfluence on the bed grain size.The influence of river bifurcations, like theIJsselkop and the Merwedekop, on thedownstream fining trend is also very large. In

the meander bend upstream of a riverbifurcation the sediment load becomes sortedhorizontally through the process of bendsorting. Fine grains are concentrated in theinner bend, coarse grains in the outer bend.Therefore a river branch splitting from the mainchannel in the outer bend will receive acoarser sediment load than the river branchsplitting from the main channel in the innerbend. The downstream fining trend thus willdiffer between the two branches.

ConclusionSelective transport is considered to be theprimary cause of downstream fining in theDutch Rhine branches. The degree of selectivetransport may be enhanced by vertical andhorizontal sorting processes at the riverbedand by the presence of suspended loadtransport. Apart from selective transport, alsothe continuous bed degradation in the DutchRhine branches has probably exerted a largeinfluence on the present bed grain size. At alower scale the sediment distribution at riverbifurcations may be important.The exact influence of the several processesmentioned in this paper on the downstreamfining of bed sediment in the Dutch Rhinebranches is still unclear. This will be thesubject of study in the next three years.

AcknowledgementsThis project was made possible by financialsupport of RIZA. Maarten Kleinhans (UtrechtUniversity) is thanked for his comments on thedraft paper.

ReferencesBradley, W.C., 1970. Effect of weathering on

abrasion of granitic gravel, Colorado River(Texas). Geological Society of America Bulletin.81, pp. 61-80.

Deigaard, R., 1980. Longitudinal and transversesorting of grain sizes in alluvial rivers. Seriespaper 26. Institute for Hydrodynamic andHydraulic Engineering. Technical University ofDenmark, Lungby.

Kleinhans, M.G., 2002. Sorting out sand and gravel:sediment transport and deposition in sand-gravelbed rivers. Netherlands Geographical Studies293. Royal Dutch Geographical Society. Utrecht,The Netherlands, pp. 317.

Kuenen, P.H., 1956. Experimental abrasion ofpebbles, 2. Rolling by current. Journal of Geology.64, pp. 336-368.

Paola, C. & R. Seal, 1995. Grain size patchiness asa cause of selective deposition and downstreamfining. Water Resources Research. 31(5), pp.1395-1407.

Parker, G. & P.C. Klingeman, 1982. On why gravelbed streams are paved. Water ResourcesResearch. 18(5), pp. 1409-1423.



IJSSELKOP: modelling of 3D grain size distributions, onestep closer to reality

S.H.L.L. Gruijters1, D. Maljers1, J.G. Veldkamp1, J. Gunnink1, M.P.E. de Kleine1, P. Jesse2 &L.J. Bolwidt2

1 Netherlands Institute of Applied Geosciences-TNO, Geo-Infrastructure Department, P.O. Box 80015, 3508 TA Utrecht, TheNetherlands; [email protected]

2 Institute for Inland Water Management and Waste Water Treatment, P.O. Box 9072, 6800 ED Arnhem, The Netherlands

AbstractIn 2001 the Netherlands Institute of AppliedGeo-sciences TNO studied the subsoil at thebifurcation Pannerdensche Kop (Gruijters et.al., 2001), aiming at the construction of a 3Dgrain size distribution model of the top 5 m ofthe riverbed. This model is used as input formorphological models to predict themorphological stability of the bifurcation andthe long-term sediment transport. FromOctober 2002 until October 2003 a similarstudy was performed at the bifurcationIJsselkop and the river IJssel up to 10 kmdownstream of the bifurcation (Gruijters et. al.,2003). Both projects, which were funded byRIZA, used vibrocore drillings and seismicsurveys to construct a model of the geologicallayers in the subsoil. For the IJsselkopbifurcation a different sampling scheme wasused. This scheme resulted in an improvedunderstanding of the short-range variation ingrain size distributions, resulting inconsiderably less smoothing of the data afterinterpolation. Also a new technique wasdeveloped which estimates a full 'synthetic'grain size distribution (GSD) from theparameters, which are estimated for everysample in the vibrocore (silt content, gravelcontent and sand median). The combination ofthese data with the measured GSD stronglyenhances the spatial variation in theinterpolation result, and prevents sand (orgravel) layers that were not sieved and arebounded by gravel (or sand) layers fromdisappearing.

IntroductionThe Pannerdensche Kop project (Gruijters et.al., 2001) showed that the sampling schemeused was not suitable to detect the spatialvariation in GSD on a scale of 10 - 200 m.Furthermore the interpolation methodsmoothed the actual variation in GSDconsiderably. For the IJsselkop project(Gruijters et. al., 2003) again a combination ofvibrocore borings, measured GSD and highresolution seismics was used to construct athree dimensional database of the subsoil

(resolution 25 x 25 x 0.2 m) containing detailedgrain size information.

Materials and methodsFor the IJsselkop project 125 vibrocore boringswere carried out. These borings were placed at400 m intervals along 3 lines, a line in thecentre of the river and two lines near theborders. The first vibrocore at the centre line isplaced 200 m downstream of the first twovirocores located at the left and right border. Atthree locations in the centre line, five vibro-cores are placed at 10, 25, 50, 100 and 200 m.At TNO-NITG the vibrocore samples werephotographed. All lithological layers visible inthe vibrocore drillings were described in detail(Bosch, 2000), including estimations of silt andgravel content, and the median sand grainsize. In total 1796 layers were described, from625 layers samples were taken to measure theGSD by sieving. Both the descriptions as thephotos were combined with the seismic data toconstruct a 3D model of the geological layersin the subsoil at the IJsselkop (Fig. 1).

Because sieving all the layers is to expensive,a selection of 5 samples for every vibrocorewas made, leaving an average of 10 samplesper boring not selected. Therefore a methodwas developed to calculate a GSD using theestimated parameters from the vibrocoredescriptions.

Figure 1. Combination of the seismic data and vibro-core data (nr. 40B0576). The clay peat layer in thefirst meter of the vibrocore results in a clear reflectorin the seismic data (yellow brackets). The blue lineindicates the top of the active layer (i.e. sand/gravelthat is transported in dunes during high discharges),the thin red line is the top of the KreftenheyeFormation.

40B0576

40B0576

Clay/peat

200 meter

3.5 meter

Disturbance bypassing ship



Figure 2. Construction of synthetic GSD from measured silt content, sand median and gravel content for asand sample.

At first, each measured GSD was matched bya synthetic GSD using an ATAN function andthree measured parameters (silt content,gravel content and sand median). The ATANfunction was corrected for each fraction toaccommodate the less symmetric shape of themeasured GSD (Fig. 2). Secondly thiscorrected ATAN function was used to calculatea full GSD from the estimated silt, gravelcontent and sand median for the 1,171samples that were not sieved.The estimated silt content and sand mediandiffer systematically from the measured ones.Therefore, before using the estimatedparameters to calculate a synthetic GSD, asecond correction was applied.

ResultsUsing the new sampling scheme, both theshort-range (up to 250 m) as the long-range(up to 1,700 m) variability could be determined.Both variograms were combined for theinterpolation, resulting in a relative nugget of10%. In combination with a well-chosen searchneighbourhood within the range of thedetected spatial correlation, unrealisticaveraging of the grain size data wasprevented.

Both the measured GSD and synthetic GSDwere used in a 3D Kriging interpolation(ISATIS, 2002 & EDS, 2003) to create a 3Dmodel of the first 5 m of the subsoil at theIJsselkop. At each grid cell of 25 x 25 x 0.2 ma full GSD was calculated. The effect of usingthe synthetic GSD is illustrated in Fig. 3.

Figure 3. Results of the 3D interpolation of thefraction > 2 mm, without (above) and with (below)synthetic GSD. Using the synthetic GSD clearlyresults in a more differentiated model with moreexplicit changes in lithology.

ConclusionsFrom the IJsselkop study the followingconclusions arise:• Expert judgement on the selection of

variogram-parameters and the search

0

10

20

30

40

50

60

70

80

90

100

0.001 0.01 0.1 1 10 100grain size [mm]

<si

eve

[m/m

%]

0

10

20

30

40

50

60

70

80

90

100

>si

eve

[m/m

%]

GRAVELSANDSILTCLAY

zgmgfugzgmgmfzfuf

synthetic GSD

measured GSD

Silt content

Sand median

Gravel content



neighbourhood will prevent unrealisticaveraging in Kriging;

• The use of three groups of 5 vibrocores atshort distances in the centre line facilitatethe determination of the short range spatialvariability in GDS;

• Estimated silt content, sand median andgravel content, can be used for calculating asynthetic GSD. However, empirical fitting ofthe applied ATAN function is necessary.

• Using synthetic GSD results in a moredifferentiated model with more explicitchanges in lithology. It prevents sand (andgravel) layers that are not sieved and arebounded by gravel (or sand) layers todisappear after interpolation.

The translation from estimated grain sizeparameters from soil descriptions to full GSDcan be used for similar sediments at otherlocations. Taking a limited number of sievedGSD to fine-tune the needed corrections of the

ATAN function will lower the necessarylaboratory budget.

ReferencesGruijters, S.H.L.L., J. Gunnink, J.J.M. Hettelaar,

M.P.E. de Kleine, D. Maljers & J.G. Veldkamp,2003. Katering ondergrond Ijsselkop, Fase 3,eindrapport (in Dutch). TNO-NITG report NITG03-120-B. Utrecht, The Netherlands.

Guijters, S.H.L.L., J.G. Veldkamp, J. Gunnink &J.H.A. Bosch, 2001. De lithologische ensedimentologische opbouw van de ondergrondvan de Pannerdensche Kop (in Dutch). TNO-NITG report NITG 01-166-B. Utrecht, TheNetherlands.

ISATIS, 2002. Geovariances et Ecole des mines deParis (in French).

EDS, 2003. GoCad version 2.07, Earth DecisionSciences.

Bosch, J.H.A., 2000. Standaard boorbeschrijfmethode, versie 5.1 (in Dutch). TNO-NITG reportNITG 00-141-A. Utrecht, The Netherlands.



Eurasian rivers in flood: application of the mixeddistribution theory to runoff extremes

W.M.J. Luxemburg1, B.I. Gartsman2 & L.M. Korytny3


2 Pacific Institute of Geography FEB RAS, Vladivostok, Russia3 Institute of Geography SB RAS, Irkutsk, Russia

IntroductionThe theory of mixed distributions for frequencyanalysis can be applied when there is reasonto believe that the assumption of homogeneityof the data series is not valid. In hydrology thiscan be the case for annual extreme rainfallevents or annual extreme runoff events. Thereason for inhomogeneity can be that themechanisms causing the extremes aredifferent. As a result the occurrences shouldnot be treated as if they were part of the totalpopulation, as they belong to differentpopulations.

Method and resultsTo apply the theory of mixed distributions theoccurrences are grouped in sub-sets accordingto the mechanisms causing the extremes. Foreach of the sub-sets the statistical distributionis obtained. The probability of exceedance forthe mixed distribution writes:

)(*)|(.......)(*)|(

)(*)|()(

22

11

nn

mixed

SPSXxPSPSXxP

SPSXxPXxP

>++>+>=>

Where:

)( XxPmixed > = Probability of exceedanceaccording to the mixed distribution

)|( nSXxP > = Conditional probability ofexceedance according to the sub-set

)( nSP = The probability of having anextreme from a certain subset

Many Eurasian rivers show a clear distinctionbetween spring and summer runoff extremes.These extremes can be linked to the obviousdifferent mechanisms of snowmelt in springcausing floods and rain events causing floodsin summer. However, for several stationsdistinction into two sub-sets in this wayresulted in an unsatisfactory fit of thefrequency distribution. A strong deviation froma straight line of the sub-set of summerextremes gave rise to the idea that thesummer extremes might have been caused bydifferent mechanisms. In particular the highsummer values seem to belong to a differentsub-set. Physically a different rain causing

mechanism, such as cyclones or typhoonscould explain this effect.

Although the real cause of the distinction in‘summer high’ and ‘summer low’ extremesremains subject for investigations, distinction inthese two sub-sets was bluntly applied.The sub-sets were separated from the pointwere an inflection in the frequency distributionoccurs. Despite the high degree of objectivityin defining the two sub-sets for summerextremes the results are extremelyremarkable.

This procedure was applied in different regionsin the Far East of the Eurasian territory forseries with a clear distinction in spring andsummer extremes, i.e. Siberia, the upper Amurbasin and near the Japanese sea coast. Allthese series showed inflection points in thesummer subset when not split up. Thiseventually led to the definition of three sub-sets for all the series.

ConclusionsThe analysis was carried out to investigate thepotential of the mixed distribution theory inmodelling the frequency distribution of runoffextremes. So far this turned out to be fruitful.At present a more comprehensive investigationon a larger territory is carried out.Nevertheless, based on the location of theobservation points of the series analysed sofar, it can be concluded that the applicability ofthe method is not bounded to a particularregion only, as it successfully applies todifferent regions.Finally some conclusions on the usefulness ofapplying the mixed distribution theory can bemade. The mixed distribution theory showsthat frequency distributions not necessarilyneed to plot as a straight line on distributionpaper when composed of sub-sets offrequency distributions. When the sub-sets offrequency distributions can be satisfactorilydefined this will improve the fit of the floodpeak distributions and particularly theprediction of magnitude of rare flood events.



In addition application of the theory of mixeddistributions provides the opportunity to betteranalyse the effect of climate change on flood

frequency distributions, as it is known thatclimate change effect the different floodcausing mechanisms in a different way.

Gumbel mixed distributionUnda

0

50

100

150

200

250

300

350

400

-2 0 2 4 6Reduced variate y

An

nu

alM

ax.

Dis

char

ge

[m^

3/s]

Spring Summerhigh Summerlow Data Mixed

Gumbel mixed distributionMargaritovo

-500

0

500

1000

1500

2000

2500

-2 -1 0 1 2 3 4 5

Reduced variate y

An

nu

alM

ax.

Dis

char

ge

[m^

3/s]


Gumbel mixed distributionBOLSHAYIA

0.00

50.00

100.00

150.00

200.00

250.00

300.00

350.00

400.00

450.00

-4 -2 0 2 4 6

Reduced variate y

An

nu

alM

ax.

Dis

char

ge

[m^

3/s]


Gumbel mixed distributionNijnay

-50.00

0.00

50.00

100.00

150.00

200.00

250.00

300.00

350.00

400.00

450.00

-2 0 2 4 6

Reduced variate y

An

nu

alM

ax.

Dis

char

ge

[m^

3/s]




Geological modelling: how reliable are input parameters?

D. Maljers & S.H.L.L. GruijtersNetherlands Institute of Applied Geosciences-TNO, Geo-Infrastructure Department, P.O. Box 80015, 3508 TA Utrecht, TheNetherlands; [email protected]

AbstractThe main goal of the research presented hereis to quantify the reliability of grain size dataused in modelling of the subsoil.Preliminary results are given regarding theuncertainties of the used data, i.e. the gradingcurves and the used grain size parameters.Data used to quantify these uncertainties hasalready been used in modelling of the subsoil(for example the IJsselkop project, carried outfor RIZA) or has been obtained as part of aninternal NITG project.Parameters studied include lithology, D50(sand median) and differences in grain sizebetween laser diffraction techniques and sieveanalysis.The results show that uncertainties inestimating grain size parameters cannot beneglected.

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 reliability of themeasurements used in a geological model iscrucial for the overall reliability of theconstructed model. For morphological models,insight in the variation of the grain sizedistribution within a layer is dominant.The main goal of the current research is toquantify the reliability of the used grain sizedata. The Netherlands Institute of AppliedGeosciences-TNO carries out the aboveoutlined research. For this study data is usedfrom morphological studies at thePannerdensche Kop, Bovenrijn-Niederrheinand IJsselkop, as well as data gathered byTNO-NITG.

Data analysisAnalysed material mainly consists of fine tocoarse river sediment with a low gradingfactor.Three kinds of analysis have been performed:• Lithology analysis: lithology estimated

(extracted from DINO1) vs. lithology

1 DINO: “Databank Informatie Nederlandse Ondergrond”,database containing all layer descriptions made bydescribers TNO-NITG.

measured (sieves as well as laserdiffraction, 1,581 samples);

• D50 sand analysis: D50_sand estimated vs.D50_sand measured (only performed onsand samples, 1,190 samples);

• Laser diffraction techniques vs. sieveanalysis: comparison of fractions andD50_sand (124 samples, resulting in 248analysis).

ResultsLithology analysisIn 13% of the samples the estimated lithologydeviates from the lithology measured.• 80% of the incorrect estimated sand

samples are gravel samples with 50-70%sand admixture;

• 50% of the incorrect estimated gravelsamples are sand samples with 15-30%gravel admixture;

• Adding a buffer of 10% around the 30%gravel limit (< 30% = sand, > 30% = gravel)explains only 8.3% of the total number ofincorrectly classified samples.

Table 1. Estimations of lithology and grain sizeanalysis of the same samples (in % of 1,581samples). In red, samples estimated as sand, whichare gravel (4.7%), and vice versa (7.8%).

D50 sand analysis• 48% of the sand samples have the same

D50 class estimated and measured;• 44% of the samples have been

underestimated compared to grain sizeanalysis of which 35% is by one D50 class;

• 8% of the samples have been overestimatedcompared to grain size analysis.

DINO-estimations

Data based on grainsize analysis

Clay / Loam Sand Gravel

Clay / Loam 0.1 0.06 -

Sand 0.06 46.6 7.8

Gravel 0.06 4.7 40.6



Figure 1. Differences between estimation of sandmedian class by describers TNO-NITG (from DINO)and grain size analysis of sand median class. X axisis expressed as total difference in sand medianclasses. For example: estimated median class: veryfine sand, measured median class: very coarsesand, total difference in median classes is 3. With:-2/-1 estimated sand median class (from DINO)

coarser than measured sand median class0 no difference1/2/3/4 measured sand median class (from DINO)

coarser than estimated sand median class

Laser diffraction techniques vs. sieveanalysis• Laser diffraction techniques and sieve

analyses give slightly different grain sizeanalyses. Coarse fractions (500 – 2,000 µm)are larger with laser analysis, fractions <500 µm are smaller, compared to sieveanalysis.

• D50Malvern is on average 56 µm coarser thanD50sieve. There is good correlation betweenthe two, with R2 = 0.86.

• Samples analysed with the Malvern (up to 2mm) of which the rest of the sample issieved (from 2 mm up to 180 mm) cannot beconnected without showing a clear break atthe connection point (2 mm). This ‘break’ ismore clear with coarse samples (gravelsamples). The ‘break’ is inherent to thedifferences between the sieve method andthe Malvern, the latter uses laser diffractiontechniques from which the grain size isdeduced.

Conclusions• Lithology estimations made by TNO-NITG

describers are good and can thus be used inmodelling of the subsoil. Incorrectestimations (only 13% of the samples) canbe explained by the percentage of

admixtures and difficulties in estimatingthese correctly.

• D50 sand is difficult to estimate, estimationresults in 35% of the samples in anunderestimation by one sand median classcompared to measurements.

• Laser diffraction techniques and sieveanalyses give slightly different grain sizeanalyses. Coarse fractions (500 - 2000 µm)are larger with laser analysis, fractions <500 µm are smaller, compared to sieveanalysis.

• D50Malvern is on average 56 µm coarser thanD50sieve. There is good correlation betweenthe two, with R2 = 0.86.

• Samples analysed with the Malvern (up to 2mm) of which the rest of the sample issieved (from 2 mm up to 180 mm) cannot beconnected without showing a clear break atthe connection point (2 mm).

The results and conclusions outlined aboveshould always be kept in mind whenestimations of grain size parameters are usedin an underground model.

ReferencesGruijters, S.H.L.L. & L.W.A. Zwang, 2001.

Handleiding programma OPBRENGST (in Dutch).Publicatiereeks Grondstoffen, nr. 2001/01, DWW-report W-DWW-2001-015.

Guijters, S.H.L.L., J.G. Veldkamp, J. Gunnink &J.H.A. Bosch, 2001a. De lithologische ensedimentologische opbouw van de ondergrondvan de Pannerdensche Kop (in Dutch). TNO-NITG report NITG 01-166-B. Utrecht, TheNetherlands.

Gruijters, S.H.L.L., J.G. Veldkamp, J. Gunnink &J.H.A. Bosch, 2001b. Kartering van delithologische en sedimentologische opbouw vande ondergrond van de Bovenrijn-Niederrhein (inDutch). TNO-NITG report NITG 01-167-B.Utrecht, The Netherlands.

Gruijters, S.H.L.L., J. Gunnink, J.J.M. Hettelaar,M.P.E. de Kleine, D. Maljers & J.G. Veldkamp,2003. Katering ondergrond Ijsselkop, Fase 3,eindrapport (in Dutch). TNO-NITG report NITG03-120-B. Utrecht, The Netherlands.

Nederlands Normalisatie Instituut (NNI), 1980. NEN2560, Nederlandse norm controlezeven,draadzeven en plaatzeven met ronde en vierkantegaten (in Dutch). Delft, The Netherlands.

Nederlands Normalisatie Instituut (NNI), 1990. NEN5104, Classificatie van onverhardegrondmonsters (in Dutch). Delft, The Netherlands.

Difference between estimation and measurement of sand median class

0.08 0.17 0.50

7.23

48.15

34.96

7.73

0.84 0.340

10

20

30

40

50

60

-4 -3 -2 -1 0 1 2 3 4

Difference in sand median classes

%o

fto

tals

et



Delft – FEWS: Flood Early Warning System

A.H. te Linde, A.H. Weerts & J. SchellekensWL | Delft Hydraulics, P.O. Box 177, 2600 MH Delft, The Netherlands; [email protected]

IntroductionThe growing sensitivity of decision-makers withrespect to potential damage caused byinundation, calls upon increasinglysophisticated and accurate techniques topredict such type of risk. A flood early warningsystem provides information on the currentstate of the water system within a polder, ariver catchment or an estuary and forecasts

the flood flow and water levels at specifiedlocations for a range of lead times within acertain accuracy band. The main philosophybehind the development of Delft-FEWS is thedesign of the system as an open platform,which allows potential end-users to incorporatetheir own, already tested and validatedprediction tools, such as routing routines orhydrological models, into the FEWS systems.

Figure 1. User interface.

Figure 2. Radar.

What is Delft - FEWS?• User interface around hydrological and

hydraulic models• Runs model forecasts• Data integration, validation and transfer can

be automated to a desired level defined bythe user

• Uses grid based weather predictions likeHIRLAM

• Serial or spatial interpolation using variousinterpolation methods

• Runs the models in two steps, in the first,the models are initialised and dataassimilation techniques are used to improvemodel performance. In the second step themodels are run for the forecast period.



Figure 3. Catch-average. Figure 4. Catch-zones.

• Allows the user to analyse previousforecasts (also called hindcasting) andbuilds up a database with statistics offorecast accuracy for quality management

• Sends warnings to end users (by fax orinternet)

Application• FEWS for the complete Rhine basin of

Rhine, operational for RIZA and FOGW(Federal Office for Water and Geology,Switzerland).

• For the Environment Agency in GreatBritain, Delft-FEWS is being developed forthe main part of England.

Figure 5. Delft-FEWS in Great Britain, regionscurrently being developed (red) and regions inpreparations (blue).

• FEWS Sudan, FEWS Pakistan.• FEWS was used for the investigation of the

possible consequences of the 2002 Elbeflood on the Rhine basin in case the stormswhere to occur there

• Determination of the design discharges ofthe Rhine basin

• FEWS Regional, WS Hunze en Aa’s acts asa demo case to show the usefulness ofFEWS as a decision support tool in theNetherlands

Development and research• Improved data assimilation techniques

(Kalman Filtering) and methods for modeloptimisation.

• Methods for determining the aggregateduncertainties of meteorological forecastsand hydrological forecasts will beinvestigated.

• Automated model calibration• The use of weather RADAR images

Figure 6. The Kalman Filtering algorithm as appliedto the Orlice (tributary of river Elbe) is seen to leadto adequate predictions.



Influence of non-linearity in the storage-dischargerelationship on discharge droughts

E. Peters & H.A.J. van LanenSub-department Water Resources, Wageningen University, Nieuwe Kanaal 11, 6709 PA Wageningen, The Netherlands;[email protected]

During drought, most stream flow is derivedfrom stored sources, primarily groundwater. Inmany catchments the relation betweengroundwater storage and discharge is non-linear, for example because the aquifer isunsaturated, because of variations in theconductivity of the aquifer or because ofdecreasing drainage density as a result ofdecreasing groundwater levels. This non-linearity increases the persistence and severityof droughts according to Eltahir & Yeh (1999).In this paper it is examined whether and hownon-linearity in the storage-dischargerelationship increases drought duration andseverity.

The influence of non-linearity was examined bysimulating long time series (10 * 1000 years) ofoutflow from reservoirs with increasing non-linearity. As input to the non-linear reservoirsrecharge was simulated for two catchments:the sub-humid Pang catchment (UK) and thesemi-arid Upper-Guadiana catchment (Spain).For the Pang catchment, first 38 years ofobserved precipitation and evapotranspirationwas resampled to 10 * 1000 years usingNearest Neighbour resampling. Subsequently,recharge was calculated using a simplebucket-type model. For a more elaboratedescription see Peters (2003). For the Upper-Guadiana catchment first 58 years of rechargewas simulated with a spatially distributedmodel and that was subsequently resampledto 10 * 1000 years.

The groundwater discharge is not simulated forthese specific catchments, but for a range ofreservoirs with different storagecharacteristics. Each reservoir is characterisedby a reservoir coefficient j, which is expressedin days. Here only the results for a reservoircoefficient of 80~d will be presented. For amore full discussion see Peters (2003). For thesimulation of the non-linear reservoirs themethod developed by van de Griend et al.(2002) was used. The relation between thestorage S and the discharge q is given by:

For each time step de reservoir coefficient j isrecalculated as a function of the currentdischarge q. Using this formulation, van deGriend et al. (2002) derived the followingrecursive solution:

∫∆+

∆−

∆

−+=∆+tt

t

t rrRt

tqettq d)(2

)()( 2βββ

where t is time (d), R is the recharge (mm/d),

bqj )(

1=β and b is a parameter determining

the non-linearity (0 < b ≤ 1, where b = 1 is thelinear case).

From the time series of groundwater rechargeand discharge, droughts were derived usingthe threshold level approach, which meansthat a drought is defined as an excursionbelow a constant, predefined threshold. This isillustrated in Fig. 1.

Figure 1. Definition of drought deficit and durationusing the threshold level approach.

Fig. 2 shows the frequency distribution of thedroughts using Weibull plotting positions forthe recharge from the two catchments. Fromcomparing the drought duration for the linearand non-linear reservoir (Fig. 2c and 2d), it isclear that the drought duration (and thus thepersistence of droughts) did indeed increaseunder all circumstances. The drought severity,however, did not uniformly increase. For therecharge from the Pang catchment the droughtdeficit in fact decreased as a result of non-linearity for droughts with a return period lowerthan 20 years.

qqjS )(=



Figure 2. a) Drought deficit for the Pang catchment, b) drought duration for the Pang catchment, c) drought deficitfor the Upper-Guadiana catchment and d) drought duration for the Upper-Guadiana catchment. All reservoirs

have a reservoir coefficient or equivalent reservoir coefficient of 80 d.

ReferencesEltahir, E.A.B. & P.J.-F. Yeh, 1999. On the

asymmetric response of aquifer water level tofloods and droughts in Illinois. Water ResourcesResearch. 35(4), pp. 1199-1217.

Peters, E., 2003. Propagation of drought throughgroundwater systems – illustrated in the Pang(UK) and Upper-Guadiana (ES) catchments. PhDThesis Wageningen University. Wageningen, TheNetherlands.

Van de Griend, A.A., J.J. de Vries & E. Seyhan,2002. Groundwater discharge from areas with avariable specific drainage resistance. J. of Hydrol.259, pp. 203-220.



Spatial distribution of heavy metals in floodplains

A.M. UijtdewilligenDepartment of Innovation and Environmental Sciences, Faculty of Geosciences, Utrecht University, P.O. Box 80115, 3508 TCUtrecht, The Netherlands; [email protected]

AbstractIn the Netherlands, the floodplains are beingreconstructed to cope with the future higherdischarges. By doing so, the patterns ofsediment deposition change. Since thesesediments are contaminated with heavymetals, we studied the patterns of sedimentdeposition and the spatial distribution of heavymetals in Dutch floodplains along the RiverRhine.The most important factors for sedimentdeposition are: 1) distance to the river and 2)the topography within the floodplain.Concerning the distance to the river, mainlysand is deposited within a distance of 100meters from the river. Therefore, the heavymetal concentrations within these 100 metersare relatively low. There is no significantdifference between different inundationperiods. Regarding the influence oftopography, the cadmium and copperconcentrations of the higher and the lowerparts of the floodplain are different: in thehigher parts the cadmium and copperconcentrations are lower than in the lowerparts.

IntroductionIn the framework of the policy document“Ruimte voor de rivier”, several floodplainsalong the Rhine River will be reconstructed.This reconstruction will affect sedimentdeposition on these floodplains. Theassessment of changes in sediment depositionis important, since the sediment attachesheavy metals. Due to biological uptake, heavymetals accumulate in floodplain vegetation,spread through the food chain, therebycausing a risk for the entire ecosystem. To beable to map this ecosystem risk, we studiedthe spatial distribution of sediment-boundheavy metal deposition on Dutch floodplainsduring inundation periods.

Material and methodsSediment deposition was measured withsediment traps (Fig. 1) in several floodplains(i.e., near Brummen, Zwolle, Amerongen,Neerijnen en Druten) along the River Rhine.The amount of heavy metals found in thesediment traps is determined and used ingeographical and statistical analyses, likecorrelation and cluster analyses.

Figure 1. Sediment trap in floodplain.

ResultsThe most important factors for sedimentdeposition seem to be: 1) distance to the riverand 2) floodplain topography. Within a distanceof 100 meters from the river, sediment isdeposited that contains a concentration ofcadmium and copper of respectively at most3.8 and 39 mg/kg dry weight (Fig. 2). Behindthese 100 meters sediment is deposited withcadmium and copper concentrations aboverespectively 3.8 and 39 mg/kg.

Figure 2. Relationship between distance to the riverin the floodplain and the cadmium concentration ofthe sediment for all investigated floodplains anddifferent inundation periods in 2001 and 2002.

Because we used different inundation periods,we checked for a possible influence of theinundation periods on the relationship betweencontamination and distance-to-river. Fig. 3shows three inundation periods in theAfferdensche en Deestsche Waarden. Only asmall part of the floodplain was inundated in



Figure 3. Cadmium concentration of the sedimentagainst distance to the river in the Afferdensche enDeestsche Waarden for three inundation periods in2001 and 2002.

December. There seems to be no significantdifference in heavy metal concentrationsbetween inundation periods.

Regarding the influence of topography, thecadmium and copper concentrations of theupper and the lower parts of the floodplain aresignificantly different (p = 0.000). In the lowerparts the cadmium and copper concentrationsare higher than approximately 4 and 40 mg/kgdry weight; in the upper parts theconcentrations are lower (Fig. 4).

Figure 4. The cadmium concentrations in the lowerand upper parts of the floodplains.

DiscussionWithin the first 100 meters from the river theflow velocity is high so mainly sand isdeposited in the floodplain. Heavy metals aremainly bound to clay, less to sand. Thisexplains the lower concentrations of cadmiumand copper near the river. In depressions thewater remains longer, which gives the smallclay particles time to settle. This leads to ahigher heavy metal concentration in the lowerparts than in the upper parts.



Data-driven modelling for flood-related problems

D.P. SolomatineUNESCO-IHE Institute for Water Education, P.O. Box 3015, 2601 DA Delft, The Netherlands; [email protected]

Traditionally flood management is based onbehaviour-driven, or physically based(simulation) models following the equationsdescribing the behaviour of water bodies.Since recently models built on the basis oflarge amounts of collected data are gainingpopularity. This modelling approach we will calldata-driven modelling; it borrows methods fromvarious areas related to computationalintelligence - machine learning, data mining,soft computing etc. The well-known model ofthis kind is a linear regression model. A non-linear model that is becoming popular is anartificial neural network (ANN).Hydroinformatics section at UNESCO-IHE inDelft is applying data-driven methods in anumber of civil engineering problems, includingflood-related problems. An approach that isbeing mainly used now is that a hierarchical‘mixture’ of models is built, each responsible

Figure 1. Huai River area (I, II, III, IV, V represent 5sub-catchment areas).

for particular hydro meteorological conditionsand ‘specialised’ in predicting low or highflows. Fig. 1 and 2 show the application of theM5 model tree method in prediction of runoff ofthe Huai River in China. The input variablesincluded a number of past measurements ofrainfall and runoff and prediction was made for1 day ahead.

Fig. 3 shows the increased performance of aspecialised model for high flows (which in itsturn includes a hierarchy of several specialisedmodels).

The list of such applications includes: usingdecision trees in classifying flood conditionsand water levels in the coastal zone dependingon the hydro meteorological data, usingartificial neural networks (ANN) and fuzzy rule-based systems for building controllers for real-time control of Dutch regional water systems,using ANNs and M5 model trees in rainfall-runoff modelling and flood prediction, usingchaos theory in predicting water levels in thecoastal zone, etc.

A number of research results and practicalapplications were developed in the frameworkof the Delft Cluster project ‘Data mining anddata-driven modelling’.

Figure 2. M5 and ANN models trained on the floodseason data (testing, fragment).

Figure 3. Performance of the data-driven model(M5) trained only on the high-flows data during the

flood season (samples with QXt-1 > 1000 m3/s). Anincrease of accuracy can be observed.

FS-m1-M5, module 1, training

0

1000

2000

3000

4000

5000

0 1000 2000 3000 4000 5000Observed discharge

Pre

dict

eddi

scha

rge

FS-m1-M5, module 1, testing

0

1000

2000

3000

4000

5000

0 1000 2000 3000 4000 5000

Observed discharge

Pre

dict

eddi

scha

rge

0

1000

2000

3000

4000

96-6-1 96-7-1 96-7-31 96-8-30

Time

Dis

char

ge(m

3/s)

OBS

FS-M5

FS-ANN



Symposium ‘Data mining and data drivenmodelling: neural, fuzzy and other methods insolving civil engineering problems’ organisedin April 2002 in IHE-Delft showed the growinginterest of practitioners to the novel methods offlood modelling. The latest experiences make itpossible to conclude that the data-drivenmethods could effectively complementphysically based simulation models. Moreinformation can be found onhttp://datamining.ihe.nl.

ReferencesBhattacharya, B. & D.P. Solomatine, 2002.

Application of artificial neural network inreconstructing stage-discharge relationship.”Proc. 4th Int. Conference on Hydroinformatics.Iowa, USA.

Data-driven modelling, 2003. http://datamining.ihe.nlDibike, Y.B. & D.P. Solomatine, 2001. River flow

forecasting using artificial neural network. Physicsand Chemistry of Earth (B). 26(1), pp. 1-7.

Solomatine, D.P. & L.A. Torres, 1996. Neuralnetwork approximation of a hydrodynamic modelin optimizing reservoir operation. Proc. 2nd Intern.Conference on Hydroinformatics, Zurich, pp. 201-206.

Solomatine, D.P. & K. Dulal, 2003. Model tree as analternative to neural network in rainfall-runoffmodeling. Hydrological Sci. J. 48(3), pp. 399-411.

Solomatine, D.P. & Y. Xue, 2003. M5 model treescompared to neural networks: application to floodforecasting in the upper reach of the Huai River inChina. ASCE Journal of Hydrological Engineering(accepted).



Modelling sediment dispersion over floodplains using aparticle tracking method

I. Thonon, K. de Jong, M. van der Perk & H. MiddelkoopCentre for Geo-ecological Research (ICG), Faculty of Geosciences, Utrecht University, P.O. Box 80115, 3508 TC Utrecht, TheNetherlands; [email protected]

AbstractTo monitor the dispersion of pollutedsediments over the Dutch floodplains, we havedeveloped a sediment-dispersion model inwhich we use a particle tracking method – theso-called ‘Method of Characteristics’ – toovercome numerical dispersion. This raster-based model uses input data such as initialsuspended sediment concentrations, waterlevels, flow velocities and diffusion coefficients.To check the model results, we took watersamples on a floodplain along the Waal Riverduring its inundation in Spring 2002. Wecompared the model results with themeasurements and it turned out that the modelpredicts the suspended sedimentconcentrations well in large parts of thefloodplain.

IntroductionScenario studies often use sedimentationmodels to study the deposition patterns ofsediment-associated pollutants on floodplains.However, since these models are often raster-based and work on a coarse grid scale,numerical dispersion can lead to erroneousresults. A particle tracking method – whichworks at the point scale – can overcomenumerical dispersion. Konikow & Bredehoeft(1978) have developed the Method ofCharacteristics (MoC), a particle trackingmethod to simulate solute transport ingroundwater. We modified this concept to beapplicable for surface water and implementedit in PCRaster (Wesseling et al. 1996).

Figure 1. Location of the study area, theAfferdensche and Deetsche Waarden near Druten.

The purpose of this study was to determine theperformance of the modified MoC model in thestudy area along the upper Waal River (Fig. 1).The RIZA hydrodynamic model WAQUAprovided the input data – such as water levels,x- and y- flow velocities – for the sedimentdispersion model. In addition, we usedempirical formulae to derive the longitudinaland transversal dispersion coefficients. Finally,to check the model output we gathered fielddata.

Materials and MethodsThe Afferdensche and Deestsche Waarden(ADW)floodplain is located along the WaalRiver, some 25 km downstream of Nijmegen(Fig. 1). Here we carried out our fieldmeasurements and we applied our model tothis area. During the inundation period fromlate February to mid-March 2002 we measuredsuspended sediment concentrations (SSCs) atthree times. Next, we derived all modelparameters, such as water height, flowvelocities and dispersivities, for the averageriver discharge between February 26 andMarch 2 (7,000 m3s–1 at Lobith) from thehydrodynamic model WAQUA and empiricalformulae.

ResultsLooking at the measurements, we found ageneral decrease in SSC from the upstreamentrances to the downstream part of thefloodplain. The model, however, indicated anorth-south gradient in SSCs with high valuesin the north western part of the floodplain andlow values in the southern part (Fig. 2).

Figure 2. Modelled suspended sedimentconcentrations in the model area. The box denotesthe same area as shown in Figure 3.



Discussion and conclusionsWe determined the performance of thesediment dispersion model for the study areaand found a good agreement betweenmeasured and modelled results in large partsof the study area (Fig. 3). At the locations inthe north-western part of the study area,however, we found a large difference: in thatarea the model predicts the SSC to be 16 to 21mg l–1 higher than the measurements indicate.This systematic difference points at an error inthe modelled influx of sediment through theinlet in the central north of the floodplain (Fig.3).

Figure 3. The percentage difference betweenobserved and predicted suspended sedimentconcentrations.

In conclusion, the above evaluation of modelperformance shows that the model wellpredicts the pattern of suspended sedimentconcentration of most of the ADW floodplain.Besides, it shows that the model is sensitive

for the water level and flow field. We believethe erroneous predictions in the northwest willdisappear when we use a more detailed digitalelevation model. All in all, the implementationof the Method of Characteristics in a 2D modelwas successful and can aid in a betterprediction of the dispersion of pollutedsediments over floodplains.

AcknowledgementsWe are greatly indebted to Menno Straatsmafor taking the water samples with his kayak.We also acknowledge the support from TonVisser and Gertjan Zwolsman (RIZA-WST inDordrecht), who carried out the WAQUAcalculations and provided the necessaryliterature, respectively. Besides, we thank EdHull for his comments on the draft versions ofthe paper.

ReferencesKonikow, L.F. & J.D. Bredehoeft, 1978. Computer

model of two-dimensional solute transport anddispersion in ground water. Techniques of water-resources investigations of the United StatesGeological Survey Book 7 (Chapter C2). U.S.Geological Survey, Reston, pp. 90.

Wesseling, C.G., D. Karsenberg, W.P.A. vanDeursen & P.A. Burrough, 1996. Integratingdynamic environmental models in GIS: thedevelopment of a dynamic modelling language.Transactions in GIS. 1(1), pp. 40-48.



The hillslope-storage Boussinesq model for variablebedrock slope

A.G.J. Hilberts1, P.A. Troch1, E.E. van Loon1 & C. Paniconi21 Hydrology and Quantitative Water Management Group, Wageningen University, Nieuwe Kanaal 11, 6709 PA Wageningen,The Netherlands; [email protected]

2 University of Quebec, CP 7500 Sainte-Foy (Quebec), GV1 4C7 Canada

AbstractThe recently introduced hillslope-storageBoussinesq (hsB) model is cast in ageneralized formulation enabling the model tohandle a locally varying bedrock slope. Thisgeneralization extends the analysis ofhydrological behaviour to hillslopes of arbitrarygeometrical shape, including hillslopes havingcurved profile shapes. The generalized hsBmodel performance for a free drainagescenario is evaluated by comparison to a fullthree-dimensional Richards equation (RE)based model. In addition, comparison of bothmodels to a storage based kinematic wave(KW) model enables us to assess the relativeimportance of diffusion processes for differenthillslope shapes, and to analyse the influenceof profile curvature on storage and flowpatterns specifically.

Model developmentThe mass balance equation for describingsubsurface flow along a unit-width hillslopereads:

Nx

qfh

t+

∂∂

−=∂∂

)( (1)

where h = h(x,t) is the elevation of thegroundwater table measured perpendicular tothe underlying impermeable layer, f isdrainable porosity, x is distance to the outletmeasured parallel to the impermeable layer, q= q(x,t) is subsurface flux along the hillslopebedrock, t is time, and N is a source term. Thisequation can be generalized by mapping thethree-dimensional hillslope shape onto a one-dimensional soil pore space (Fan & Bras,1998):

),()(),( txhxfwtxS = (2)

where S(x,t) is the actual storage, w(x) is thehillslope width, and h(x,t) is the water tableheight averaged over the width of the hillslope.Moreover we define subsurface flux as(Boussinesq, 1877]):

+∂∂

−= )(sin)(cos xixix

hkhq (3)

where k is the saturated hydraulic conductivity,i(x) is the local bedrock slope at x. Uponcombination of Eqs. 1, 2, and 3, we obtain ageneralized hsB equation that can handlespatially variable bedrock slope:

( ) ( ) fNwxikSxx

wSxi

f

kS

xt

Sf +

∂∂

+

∂

∂∂∂

=∂∂

)(sin/

)(cos (4)

Experiment set-upWe have applied the generalized hsB model,the KW model as described by (Troch et al.,2002), and the RE models as described by(Paniconi & Wood, 1993) to a set of ninecharacteristic hillslopes (see Fig. 1).

Figure 1. Nine characteristic hillslopes.

The nine characteristic hillslopes consist ofthree divergent, three straight and threeconvergent hillslopes with profile curvaturevarying from convex to straight to concave. For5 and 30% slopes free drainage experimentsare inter-compared, starting from 40%saturation. In order to generalize the results adimensional analysis is conducted. Adimensionless representation will give usscaled model results. We define the followingdimensionless variables:



d

txh

V

tQcum

L

x

fL

tki ),(,

)(,1, ==−== σφχτ

where L is hillslope length, Qcum(t) is cumulativeflow up to time t, V is the initial volume of waterstored in the hillslope, and d is soil depth. Thedimensionless variables define (in order)kinematic time, flow distance as a fraction ofthe total flow path, dimensionless flow, anddimensionless storage.

Results and conclusionFig. 2a, 2b, and 2c show the dimensionlessstorage patterns for the 5% bedrock slopesimulations.

Figure 2a. hsB 5% storage.

Figure 2b. RE 5% storage.

Figure 2c. KW 5%/30% storage.

Fig. 3a and 3b show the storage patterns forthe 30% bedrock slope case.

Figure 3a. hsB 30% storage.

Figure 3b. RE 30% storage.

Fig. 4 displays the dimensionless hydrographsfor all hillslopes (k = 1 m/h, 0.22 ≤ f ≤ 0.3).

In this work we have generalized the hsBmodel to variable bedrock slope and we haveconducted an intercomparison of a RE model,taken as a benchmark, with the hsB, and a KWmodel, in order to investigate within whichsettings the application of the KW and hsBmodel is valid.

Figure 4. Hydrographs 5%/30%, all models.



Generally one can conclude that (1) the hsBresults show a good match with the RE results.(2) In relation to the validity of applying a KWmodel it is limited to simulation settings inwhich hydraulic diffusion has a marginalimpact on storage. (3) The RE hydrographsappear to have a slight delay compared to thehsB results. This may be explained by thecapillarity effects of the soil.

ReferencesBoussinesq, J., 1877. Essaie sur la theorie des

eaux courantes (in French). Mem. Acad. Sci. Inst.Fra. 23(1), pp. 252-260.

Fan, Y. & R.L. Bras, 1998. Analytical solutions tohillslope subsurface storm flow and saturationoverland flow. Water Resour. Res. 34(4), pp. 921-927.

Paniconi, C. & E.F. Wood, 1993. A detailed modelfor simulation of catchment scale subsurfacehydrologic processes. Water Resour. Res. 29(6),pp. 1602-1620.

Troch, P.A., E.E. van Loon & A. Hilberts, 2002.Analytical solutions to a hillslope-storagekinematic wave equation for subsurface flow. Adv.Water Resour. 25, pp. 637-649.

Future Developments


NWO and Water

H. de BooisCoordinator Earth Sciences and Global Change Research, Council for Earth and Life Sciences (ALW), NWO, NetherlandsOrganisation for Scientific Research, P.O. Box 93138, 2509 AC The Hague, The Netherlands; [email protected]

NWO: the organisationThe Netherlands Organisation for ScientificResearch (NWO) promotes fundamental andapplied research of high quality. Aninternational orientation has high priority. Themeans to perform this mission are M€ 400 peryear from the Ministry of Education, Cultureand Science and about M€ 50 per yearadditional funds from other ministries andenterprises. Via several funding schemesNWO finances over 4,300 positions in scientificresearch, both at universities and at NWOinstitutes.

NWO has a variety of funding schemes.A line of personal grants exists for youngtalented postdocs up to well-establishedpostdocs with budgets from k€ 200 up to k€1,250 (“Vernieuwingsimpuls”). On top of this,annually four Spinoza-prises (M€ 1.5) areawarded to internationally renowned scientists.The research councils fund open programmesand a wide range of dedicated programmes,aimed at research on specific themes. NWOparticipates in many thematic Europeanresearch programmes, in particular in therecent so-called EUROCORES, which arecoordinated by the European ScienceFoundation. Moreover, NWO has fundingschemes for investments, for centre subsidiesand for travel, workshops, and internationalcoordination.

The mode of operation of NWO ischaracterised by three layers of appraisal ofresearch proposals, which are submitted forfunding. At the first level proposals arereviewed by independent peers, for ALWusually scientists from abroad. The reviewscan be commented by the applicants of theproposals. The next level is the judgment andprioritising of proposals by the programmecommittee, which is composed of senior Dutchscientists. Decisions on funding are made bythe Research Councils, Boards, or SteeringCommittees in case of co-funding by externalparties. By approximation, about 4,000 – 5,000external referees are annually involved in thereview process and more than 1,000 (mostly

Dutch) scientists are involved in committees,boards and councils.

The funding by NWO is mainly organisedaccording to clusters of disciplines in ResearchCouncils (“Gebiedsbesturen”). In addition twofoundations exists for research in a variety ofdisciplines: WOTRO for research in tropicaland developing countries and NCF forsupercomputing.Besides, NWO administers a couple ofresearch institutes: on physics (four institutes),sea research (NIOZ), space research andearth observation (SRON) and astronomy(ASTRON). In addition, NWO hostscoordinating offices for managing externalfunds for research on Genomics, Biopartner,ICT, and the European office for EDCTP(clinical trials of medication in Africa).

The NWO strategyIn the Strategic Plan 2002-2005 of NWO seveninterdisciplinary themes were presented, oneof them called ‘System Earth’. This theme actsas an umbrella over a number of issues relatedto the large-scale human influences on earthsystems, generally indicated as GlobalChange. This covers a wide variety ofresearch, e.g., on climate change, ondevelopment of hydrogen energy systems, onpolicy mechanisms under the Kyoto-protocol,on the use of tropical coastal zones, and onwater management in The Netherlands.

In particular the Research Council on Earth anLife Sciences has identified strategic themeson Global Change: biodiversity, the coupledbio-geosphere, climate variability, continent-ocean boundaries, water and coastal zones,bio- and geo-informatics and monitoring. Someof these themes are shared with WOTRO andthe Council on Exact Sciences (EW). Inaddition, several external parties join inrelevant funding programmes. In particular theMinistry VROM has committed funding for aseries of issues related with climate change,which are covered by different programmes.In the near future, the following relevantprogrammes are open for proposals (seeTable).

Future Developments


*: external funders

The NWO-programme WATER:rationale, themes and sub themesThe Water programme is a joint venture ofALW and WOTRO. The programme has awide, multidisciplinary scope, but the centralrationale is in global change and in particularclimate change in relation with water. The firstcall for M€ 3 closes 02-02-2004. A second callfor M€ 2 will be opened by the end of 2004.

The Water programme document names threethemes and several sub-themes:• Water and the earth system:

! The water cycle! Atmospheric water! Surface water! Soil and ground water! Terrestrial ecosystems and hydrology

• Global change and aquatic ecosystems:! Water quality and harmful algal blooms! Aquatic food webs and biodiversity! Waterborne diseases

• Water and society:! Water scarcity, governance and

contestation! Water productivity, regenerative capacity

and livelihoods around water uncertainty! Water management, networks and

control of floods and water supply

Within the context of global change and inparticular climate change, the scientificchallenges of the programme relate tocommon threads through all aspects:• Diversity, heterogeneity, variability;• Non-linearity, thresholds, boundaries;• Vulnerability, resilience and adaptation (in

The Netherlands, Europe and in developingcountries).

For information on the Water-programme:www.nwo.nl/alw and e-mail [email protected].

Funding Programme Deadline

ALW, EW, *VROM Climate variability (incl. C-cycle) 6-1-04

ALW, *NERC (UK), *NRF (Norway) RAPID Climate Change 15-12-03

ALW, WOTRO, *VROM Water 2-2-04

ALW, *BuZa, *LNV, *VROM Netherlands Polar Programme 15-10-04

ALW, *V&W, *VROM Earth Observation 17-11-03

MaGW, ALW, *VROM Vulnerability, Adaptation, Mitigation spring 2004

Future Developments


Five years NCR, review and preview

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

AbstractFive years ago NCR was founded by nineDutch research institutes to build a jointknowledge base on rivers and to promote co-operation between the most importantscientific institutes in the field of river studies inthe Netherlands. This co-operation was alsomeant to strengthen the national andinternational position of Dutch scientificresearch and education. In these five years,NCR succeeded in filling in its platformfunction through e.g. the organisation of theso-called yearly NCR-days and building astrong international image through amongstothers the leadership of the INTERREG-IICIRMA-SPONGE program.

IntroductionThe Netherlands Centre for River Studies(NCR), a co-operation of the major developersand users of expertise in the Netherlands inthe area of rivers, was founded on October 8,1998. So, this year, we are celebrating our firstlustrum.The extent of the research challenges for thecoming decades necessitates co-operation onnational and international levels: one singleresearch institute cannot perform theinterdisciplinary research needed nowadays onits own, especially where the institutionalstructure in a country more or less followsmethodological lines. In the case of riverresearch in the Netherlands, some institutesfocus on field monitoring, some on fieldcampaigns, some on numerical modelling andlaboratory experiments, some on behaviouranalysis, and some on uncertainty analyses.Co-operation between the various institutes,making use of the specific expertise and

infrastructure of each institute, is thereforeessential. And this is exactly what has beendone in the Netherlands with the establishmentof NCR in October 1998.

NCR partnersThe NCR partners of the first hour were:• TUD (Delft University of Technology),• UU (Utrecht University),• KUN (University of Nijmegen),• UT (University of Twente),• UNESCO-IHE,• RIZA (Institute for Inland Water

Management and Waste Water Treatment),• ALTERRA (Dutch centre of expertise on

rural areas, two separate institutes at thattime, now merged into one) and

• WL | Delft Hydraulics.

In 2000 they were joined by TNO-NITG(Netherlands Institute of Applied Geoscience)and at the beginning of 2003 by WU-CWK(Wageningen University, Centre for Water andClimate).

NCR FunctionsNCR 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; examples are the yearly NCR-days and the different workshops NCRorganises;

• research-orientated and educational co-operation: in which a real commitment of thepartners is reflected.

Figure 1. Signing of NCR co-operation agreement on October 8, 1998.

Future Developments


Main results of first 5 yearsPlatform functionFive specific parts of the platform function ofNCR can be distinguished:1. The NCR thematic days for young

researchers and the NCR professionalorientation

2. The co-organising role of variousInternational Conferences and Symposia

3. The NCR Internet site www.ncr-web.org4. The NCR publications series, and of course5. The NCR-days.

The thematic days and professional orientationhave been explained at the NCR-days of 2002by Reinier de Nooij and can be found in theproceedings of these days (NCR publication20-2003).One of the meetings NCR helped to organiseis the Symposium of Lowland RiverRehabilitation, the results of which weresummarised by Tom Buijse during these NCR-days. In the coming period NCR will also playa role in the organisation of the 7th INTERCOLInternational Wetlands Conference (July2004), the 3rd International Symposium onFlood Defence (May 2005) and the 8th

International Fluvial SedimentologyConference (August 2005).Information on these meetings can be foundon the NCR Internet site. This site has beendeveloped since 2000. It gives generalinformation on NCR and its partners and it hasthe possibility for NCR researchers to inserttheir own information on the site. We wouldlike the researchers to use this possibility moreoften. At the moment the site is visited 50times a day (as an average on a working day),but only 40% is from the Netherlands, so mostprobably only 50% (25 hits a day) are seriousvisitors.

The NCR publication series started in 2000too. Until now the series consists of 23publications. In the years 2000, 2002 and 2003three publications were published; 2001showed an outburst of 14 publications, due tothe final reporting of the various IRMA-SPONGE subprojects.

The NCR-days officially also started in 2000.But already in 1997 and 1998 so called pre-NCR-days were organized in Lunteren andDelft, where ongoing (university) research waspresented and in 1999 NCR organised aworkshop with presentations from non-university partners, which was reported in theNCR publication series (NCR 02-2000).The NCR-days from 2000 onwards, meant foryoung researchers to present their ongoing

river studies on two consecutive days, in orderto maximise the exchange of ideas andexperiences between the participants and toprovide the researchers a sounding board fortheir study approach and preliminary results,proved to be very successful. The participationincreased from some 80 to 105 (2003) andeven 125 (2002), with very enthusiastic feedback from the participants. Of course theproceedings of these days are alsoincorporated in the NCR publication series.To celebrate our first lustrum the NCR-daysPresentation and Poster Awards wereestablished.

Research co-operationThe research function of NCR wassummarised in Dutch in the “NCR Programma”in 2000 and in ‘Summary of NCR Programme,version 2001 – 2002’ in 2001.It is worth mentioning that NCR is not afunding facility. The partners have committedthemselves to put at least the equivalent ofsome € 100.000 each year into the co-operation, but the real added value of NCR liesin putting together projects that are financedwithin different research frameworks. Many ofthese frameworks provide ‘shared costfinancing’ (e.g. EU framework programme,Delft Cluster), meaning that only part of thetotal costs of the project is funded. Otherframeworks (e.g. NWO) only provide financingfor University research personnel orinstruments. NCR, being a research co-operation without research funds of its own,provides the platform to bring theseframeworks together.Senior researchers of the NCR-partnersprovide the capacity to analyse thepossibilities, to draft the proposals and tosupervise the research.

In 2002 the following research themes wereadopted:• Living (in harmony) with the River

(“(Over)leven met de rivier”)• Dynamic River Management (Cyclic

Rejuvenation)• Genesis of Floods, which is closely linked

with the Hydrological Triangle (see page19), and last but not least

• The Morphological Triangle(see proceedings NCR-days 2002).

A very important part in the NCR-role as far asresearch is concerned is the building of astrong international image throughinternational research co-operation. This aimof NCR received a boost by the invitation to

Future Developments


lead the EU sponsored (INTERREG-IIC)IRMA-SPONGE research program. Thisprogram ended in 2002. It was reviewed asvery good and successful. The most importantresults were published in the NCR publicationseries.

Future activitiesIn the coming period the NCR ProgrammingCommittee will focus on the updating of theNCR program along the lines of the themesmentioned above. Important funding sourcesfor this program will be, apart from the ownresources of the NCR partners, the DutchBSIK program (Delft Cluster, Living with Wateretc.), the Dutch NWO/ALW Water and LOICZprograms and the 6th framework programme ofthe EU (Global Change).For instance as a follow up of the IRMA-SPONGE program already two EU sponsoredprojects, ‘Freude am Fluss’ and ‘FLOODsite’are in the contract negotiation stage.

‘Freude am Fluss’ (INTERREG-IIIB) will be atri-national, multi-disciplinary, public/privateproject that looks beyond public’s ‘immediatereactions’ (NIMBY): it explores the deepervisions of the public concerning living alongrivers with their floods, seeks for the benefits(economic, social and cultural) of co-designingsolutions (of maximizing Freude am Fluss) andwill provide attractive examples of concretedesigns and plans for measures mitigating riskand maximizing pleasure.Co-ordinator is KUN. It has 12 partners from 3countries (NL, D, F). NCR partners are KUN,Delft Hydraulics, RIZA (RWS-DON and -DWW). Project duration is 5 years, the totalproject budget is approximately 7.7 Mio€ andthe EU contribution approximately 3.8 Mio€.

‘FLOODsite’ (FP6 Global Change) will be anIntegrated Project with 36 partner institutes.Co-ordinator is HRWallingford, NCR plays aco-co-ordinator role. NCR partners involvedare Delft Hydraulics, TUD, UT, WU andUNESCO-IHE. KUN, Alterra and possiblyTNO-NITG will participate throughsubcontracting. RIZA has a seat in theApplication and Implementation Board. Projectduration will be 5 years, the total budget14.133 Mio€ with an EC contribution of 9.68Mio€.

‘FLOODsite’ aims to provide an integratedframework for flood risk management fromoperational to planning time horizons (50 yearsand beyond) in:• Sustainable pre-flood measures

(infrastructure provision, planning andvulnerability reduction)

• Flood event management (early warning,evacuation and emergency response)

• Post-event activities (review andregeneration).

‘FLOODsite’ is subdivided in seven themes.

ConclusionsAfter five years of NCR we may conclude thatNCR succeeded in its aim to provide an openplatform for all people interested in scientificresearch and communication on River issuesand to promote co-operation between the mostimportant scientific institutes in the field of riverstudies in the Netherlands (and abroad).The NCR-days prove to be very successful.Thanks to the successful IRMA-SPONGEprogram the international image of NCR isstrong. It provides NCR with a major role infuture EU projects like the FP6 IntegratedProject ‘FLOODsite’.

Scope of NCR: river basin river branch floodplain ecotope


Authors index

Bhattacharya, B........................ 54Bijlsma, L. ................................ 3Bogaard, T.A............................ 7Bogaart, P.W............................ 62Bolwidt, L.J............................... 76, 80Booij, M.J. ................................ 24Boois, H. de ............................. 101Buraimo, C. .............................. 54Dohmen-Janssen, C.M............. 24Dong, X.H. ............................... 24Douben, N................................ 1, 64Eijgenraam, C.J.J. .................... 10Fenicia, F. ................................ 12, 68Foppen, R.P.B.......................... 43Frings, R.M. ............................. 78Gartsman, B.I. .......................... 84Geerling, G.W. ......................... 59Gooch, G.D. ............................. 71Gruijters, S.H.L.L...................... 80, 86Gunnink, J................................ 80Hall, M.J................................... 37Hilberts, A.G.J. ......................... 98Hoek, E.E. van der ................... 15Houlliere, A.M.G.R. .................. 43Jesse, P. .................................. 76, 80Jong, K. de............................... 96Kleine, M.P.E. de...................... 80Kok, M...................................... 27Kolkman, M.J. .......................... 27Korytny, L.M............................. 84Krywkow, J............................... 30Kuijper, M................................. 46Laat, P.J.M. de......................... 37Lambeek, J.J.P. ....................... 40Lanen, H.A.J. van..................... 90Lenders, H.J.R. ........................ 43Leuven, R.S.E.W...................... 43Linde, A.H. te ........................... 88Loigu, E.................................... 71Loon, E.E. van.......................... 98Luxemburg, W.M.J. .................. 7, 84Maljers, D................................. 80, 86Maskey, S. ............................... 18Meijer, D.G............................... 56Middelkoop, H. ......................... 48, 96Minnema, B.M. ......................... 46Moor, J.J.W. de ........................ 32Mourad, D.S.J. ......................... 71Muinck Keizer, M. de................ 15

Mul, M.L. .................................. 66Nguyen, A.D............................. 64Nooij, R.J.W. de ....................... 43Onneweer, J............................. 56Os, A.G. van ............................ 1, 103Paniconi, C............................... 98Pebesma, E. ............................ 48Peerbolte, B. ............................ 56Perk, M. van der....................... 71, 96Peters, E. ................................. 90Pfister, L................................... 12Piirimäe, K................................ 71Price, R.K................................. 18Ragas, A.M.J............................ 59Reggiani, P. ............................. 12, 68Reinders, J.E.A. ....................... 15Riel, M.C. van........................... 51Rientjes, T.H.M. ....................... 12, 68Roode, M. van.......................... 22Rotmans, J............................... 30Santbergen, L.L.P.A. ................ 34Savenije, H.H.G. ...................... 12, 68Schellekens, J. ......................... 88Schijndel, S.A.H. van................ 74Schotanus, T.D. ....................... 66Slootjes, N................................ 20Smits, A.J.M............................. 59Snepvangers, J.J.J.C. .............. 46Solomatine, D.P. ...................... 94Stålnacke, P............................. 71Straatsma, M............................ 48Stroet, C.B.M. te....................... 46Stuyt, L.C.P.M. ......................... 15Thonon, I.................................. 96Troch, P.A. ............................... 62, 98Tu, M........................................ 37Uijtdewilligen, A.M. ................... 92Vaate, A. bij de......................... 51Valkering, P.............................. 30Veen, A. van der ...................... 27, 30Velde, G. van der ..................... 51Velde, Y. van der...................... 46Veldkamp, J.G.......................... 80Verhallen, J.M. ......................... 34Weerts, A.H.............................. 88Wesseling, C............................ 48Wit, M.J.M. de .......................... 7, 37Zhang, G.P............................... 12, 68


NCR Supervisory Board NCR Programme Committee

December 2003 December 2003

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. P.H. Nienhuis dr. R. LeuvenUniversity Nijmegen University Nijmegenemail: [email protected] email: [email protected]

prof. dr. S.J.M.H. Hulscher dr. J.S. RibberinkUniversity Twente University Twenteemail: [email protected] Email: [email protected]

M.W. Blokland J.L.G. de SchutterUNESCO-IHE Institute for Water Education UNESCO-IHE Institute for Water Educationemail: [email protected] email: [email protected]

J. Ridder dr. M.J. van BrachtTNO-NITG TNO-NITGemail: [email protected] email: [email protected]

prof. dr. J. Bouma 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:

NCR-publication no:00-2000 “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)

01-2000 “NCR Programma, versie 1999 – 2000”,editors R. Leuven & A.G. van Os (in Dutch)

02-2000 “NCR workshop, de weg van maatschappelijke vraag naar onderzoek”,editors A.F. Wolters & E.C.L. Marteijn (in Dutch)

03-2001 “NCR dagen 2000, het begin van een nieuwe reeks”,editors A.F. Wolters, C.J. Sloff & E.C.L. Marteijn (partly in Dutch)

04-2001 “Umbrella Program IRMA-SPONGE, Background, Scope and Methodology”,editors A. Hooijer & A.G. van Os

05-2001 “Summary of NCR Programme, version 2001 – 2002”,editor A.G. van Os(also downloadable from the NCR Internet site)

06-2001 “The Netherlands Centre for River studies, a co-operation of the major developers andusers of expertise in the area of rivers”,editors A.G. van Os & H. Middelkoop

07-2001 “NCR-days 2001, from sediment transport, morphology and ecology to river basinmanagement”,editors E. Stouthamer & A.G. van Os

08-2001 “Land van levende rivieren: De Gelderse Poort”,Stichting Ark, ISSN 90 5011 150 5; € 27,50

09-2001 “Guidelines for rehabilitation and management of floodplains, ecology and safetycombined”,editors H.A.Wolters, M. Platteeuw & M.M.Schoor

10-2001 “Living with floods: resilience strategies for flood management and multiple land use inthe river Rhine basin”,editors M. Vis, F. Klijn & M. van Buuren

11-2001 “Development and application of BIO-SAFE, 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

12-2001 “Extension of the Flood Forecasting Model FloRIJN”,authors E. Sprokkereef, H. Buiteveld, M. Eberle & J. Kwadijk

13-2001 “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

13M-2001 “Interactive Flood Management and Landscape Planning in River Systems” - UserManual version 1.0.0

14-2001 “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

15-2001 “Intermeuse: the Meuse reconnected”,authors N. Geilen, B. Pedroli, K. van Looij & L.Krebs, H. Jochems, S. van Rooij & Th.van der Sluis

16-2001 “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

17-2002 IRMA-SPONGE, “Towards Sustainable Flood Risk Management in the Rhine andMeuse River Basins”, Proceedings of IRMA-SPONGE Final Working Conference,editors A.G. van Os & A. Hooijer


18-2002 IRMA-SPONGE, “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

18E-2002 IRMA-SPONGE, “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

18D-2002 IRMA-SPONGE, “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)

18NL-2002 IRMA-SPONGE, “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)

18F-2002 IRMA-SPONGE, “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)

19-2002 “Laseraltimetrie en Hydraulische Vegetatieruwheid van Uiterwaarden”,authors M.R. Ritzen & M.W. Straatsma (in Dutch)

20-2003 “NCR -days 2002, Current themes in Dutch river research”,editors R.S.E.W. Leuven, A.G. van Os & P.H. Nienhuis

21-2003 “Kansrijkdom voor rivierecotopen vanuit historisch-geomorfologisch perspectief -Rijntakken - Maas - Benedenrivieren”,authors H. Middelkoop, E. Stouthamer, M.M. Schoor, H.P. Wolfert & G.J. Maas (inDutch)

22-2003 “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. Greijdanus-Klaas