Transcript
Research paper Integrated modelling for flood risk mitigation in Romania:case study of the Timis–Bega river basin I. POPESCU, UNESCO IHE Institute for Water Education, Westvest 7, 2601 DA Delft, The Netherlands. E-mail:
[email protected] ( Author for correspondence )A. JONOSKI, UNESCO IHE Institute for Water Education, Westvest 7, 2601 DA Delft, The Netherlands S.J. VAN ANDEL, UNESCO IHE Institute for Water Education, Westvest 7, 2601 DA Delft, The Netherlands E. ONYARI, UNESCO IHE Institute for Water Education, Westvest 7, 2601 DA Delft, The Netherlands V.G. MOYA QUIROGA, UNESCO IHE Institute for Water Education, Westvest 7, 2601 DA Delft, The Netherlands ABSTRACT An integrated floodmodelling approach has beenapplied in a demonstrator ofa floodmanagement system, which was developed withinthe framework ofacollaborativeprojectbetweenRomaniaandtheNetherlands.Thedevelopeddemonstratorsystemhadtwoobjectives:(a)operationalwatermanage-ment under extreme conditions when actions have to be taken quickly; (b) off-line analysis and design of flood mitigation measures and alternatives.This article presents the applied approach and the achieved results for meeting the second objective. The pilot basin for the development of the systemwastheTimis–Begariverbasin,in whichtheriversTimis andBegawereconsideredjointly.Thesystemisbasedonmodellingthefloodgenerationandrouting processes by combined development and application of hydrological and hydrodynamic models. The modelling system HEC-HMS was usedfor the hydrological model, HEC-RAS for the one-dimensional hydrodynamic model and SOBEK for the two-dimensional (2D) model used for down-stream flood analysis and design of mitigation measures and alternatives. The 2D model includes alternatives of deliberate dike breaching as part of theanalysisofthesystem response.Theanalysispresentedis concentrated onaspecific floodeventthatoccurredin April 2005,whichoccurreddue todike breaches along the Timis river. The combination of models is first used for reconstruction of inundation patterns resulting during this flood event. Sub-sequently the models were used for testing flood mitigation alternatives of deliberate (planned) breaches of flood protection dikes located in the down-stream part of the Timis river at the same location where they had occurred during the 2005 flood event, but at different times with respect to the arrivalof the flood hydrograph. The demonstrated approach can enable decision-makers to analyse the behaviour of the physical system and design possible preventive and / or mitigation measures. Keywords: Flood modelling; flood mitigation; dike breaches; decision support system 1 Introduction Floods remain one of the most frequent and devastating naturalhazards worldwide. While existing forecasting and warningsystems have made a significant contribution to the reductionof losses due to floods, nevertheless there remains a consider-able potential for further prevention of losses by making useof technological advances for better integration of data andmodels and consequently design of better flood mitigationmeasures and issuing of more accurate forecasts and possiblewarnings. In on-line situations, flood modelling is central inforecasting and warning systems as it can help us to understandflood generation and identify the potential areas to be inun-dated. This allows issuing of targeted early warning to down-stream communities located in the floodplains which will beaffected. In off-line situations flood modelling also enableslong-term planning for flood damage reduction, which is com-monly carried out by using the models for evaluating variousflood mitigation measures in order to determine which alterna-tive will be economically and environmentally feasible, giventhe prevailing conditions. Models have different requirementsfor on-line and off-line applications. On-line models, for flood forecasting, require fast and accurate simulation of dis-charge peaks for a known water system. Often, capabilities of Received 5 January 2010. Accepted 29 July 2010.ISSN 1571-5124 print / ISSN 1814-2060 onlineDOI:10.1080/15715124.2010.512550http://www.informaworld.com269 Intl. J. River Basin Management Vol. 8, Nos. 3–4 (2010), pp. 269–280 # 2010 International Association for Hydro-Environment Engineering and Research hot-starts and data assimilation are a pre-requisite to fulfil theserequirements. Off-line models need to provide a physically based reliable simulation of the water system behaviour for awide range of conditions and for changes in the modelledwater system itself, to allow design and analysis of structuralflood management measures. The off-line modelling systemsneed to allow users to translate proposed measures intochanges in the model, e.g. flood plain adjustments. Integratedflood modelling is increasingly being demonstrated to be anecessity due to the complexity of the interactions amongdifferent components of the river and its floodplain. Variousflood modelling studies have been carried out that show howinteractions between rivers and floodplains can lead to accurateflood forecasting and prediction at critical points. Flood model-ling systems usually combine rainfall–runoff models with floodrouting models. The flood routing is carried out either byhydrologic routing approaches, which are used to obtain theflood peak by routing flood events between streamflowgauging stations, or by more complex one-dimensional (1D)hydrodynamic models which simulate flood propagation based on detailed channel geometry (Blackburn and Hicks2002). However, in areas with complex river flow conditions,especially in the presence of complex river–floodplain inter-actions, two-dimensional (2D) models need to be used for off-line spatial hydraulic analysis (Horrit and Bates 2002).The Netherlands has a long history of water management,during which significant knowledge and experience related tofloodprotectionhasbeenaccumulated.Inthiscountry,floodman-agementiscurrentlyapproachedasanintegralpartofwiderwater management and spatial planning processes. The practicalimplementation of this approach relies heavily on integratedflood modelling. The accumulated knowledge and experiencein the field of forecasting excessive rainfall events, predictingand controlling high river water levels as well as mitigation of floods, can be exported to other countries in Europe and the rest of the world. Similar to the Netherlands, Romania is a countrywhere rivers discharge into the sea. Romania is, however, amuch larger country also characterized by mountainous catch-ments, which are absent in the Netherlands. Romanian Waters(the National Agency responsible for overall water resourcesmanagement) complies with the legislation compatible with theEU regulations regarding water resources management and the preservation of aquatic ecosystems and water areas. In thisrespect, this agency is responsible for the ways in which surfaceand groundwater resources in the Romanian territory are used.The same agency is also responsible for flood management andcontrol.Inthisarea,theagencyiscurrentlydevelopingandimple-mentingnew,technologicallyadvanceddecisionsupportsystems(DSS) for flood forecasting and warning, as well as long-termflood risk planning and management. In these efforts, theagency is facing numerous challenges due to lack of modellingexperienceaswell asdataavailability anddataquality assurance.As part of these ongoing activities, within the framework of acollaborative project between Romania and the Netherlands, aflood forecasting system demonstrator has been developed,which can support operational water management under extreme conditions when rapid action has to be taken. Thissystem had a comparatively simpler (with short running time)modelling component (only rainfall–runoff model developedwith HEC-HMS), and the focus was on the on-line integrationof this component with meteorological and hydrologicaldata. Next to the on-line system, integrated flood modellingapproaches for off-line analysis and design of flood mitigationalternatives were implemented, which are the focus of this paper.Thiswas achieved bycombiningtheHEC-HMSmodelwith a1D hydrodynamic model developed with HEC-RAS and aSOBEK 1D-2D model for flood inundation modelling. In themodelling approach, the Timis and Bega rivers were considered jointly, since their joint hydrodynamic response are conditioned by the operation of existing hydraulic structures used for water transfer between the two rivers. The present paper describes theapproach taken in the off-line integrated flood modelling andthe usage of the models for analysis and design of possibleflood mitigation measure and alternatives. The focus is on theanalysis of a particular flood event that occurred in April 2005,when a large area close to the Romanian border with Serbiawas inundated as a consequence of dike breach failures alongtheriverTimis.Actualbreachesofthedikesoccurredduetostruc-tural failure (induced by poor maintenance), even though thewater levels were barely at the overtopping level. Initially, theintegrated modelling approach was applied for reconstructingthis particular flood event. Subsequently the same models wereused for analysis and testing of possible flood mitigation alterna-tives. Assuming that high flood water levels and dischargeswouldanyhow lead toeventual dike breaches due toovertopping(even if dikes are well maintained), alternatives have been testedin which deliberate (planned) dike breaches are carried out at thesamelocationsasthoseoccurringduringApril2005,butatdiffer-enttimingwithrespecttothearrivalofthefloodhydrograph.Thisapproach has not been extensively researched in the past and, astheresultsdemonstrate,itcanleadtothereductionoffloodimpact in terms of lower flood volumes and reduced area of inundation. 2 The Timis–Bega river system 2.1 Case study location In terms of flooding problems, one of the most vulnerableregions in Romania is in the west. Furthermore, many rivers inthis region are of a trans-boundary nature: their basins areeither in Romania and Serbia or in Romania and Hungary. Anyevent occurring in these rivers is advected downstream to theneighbouring country. A typical example of this situation is theriverTimis,which – intherecentpast – hascausedsevereflood-ing in the two neighbouring countries of Romania and Serbia. Inthe past, many dikes were constructed along this river for flood protection, which – in return – made the downstream flood propagation quicker, causing severe flood damage. 270 I. Popescu et al. The Timis–Bega basin is located in the south-west of Romania in the province of Banat (Figure 1). The climate is tem- perate continental, characterized by northerly cold winds in thewinter and moderate westerly winds coming from the Atlanticin the summer. The climate is influenced by the CarpathianMountains, located to the east and north of the basin, whichdiminish the climate influence of Eastern Europe. At altitudeshigher than 1000 m.a.s.l, heavy snowfalls are generated. InJanuary, which is the coldest month, the average temperaturesrange from 2 4 8 C to 0 8 C. Temperatures as high as 34 8 C have been recorded in the Danube Valley in July.ThemainriversintheTimis–Begabasindischargetheirwater in Serbia beyond the Romanian borders. Bilateral agreements between the two countries specify the discharge and water quality conditions downstream of the border (INHGA 2004).Theriver Begastarts inthemountains, and hasadraining areaof 2878 km 2 . At the end of its course, the river enters the lowBanat plain, an area that was frequently flooded in the past.Because of this flooding, a new 114 km-long Bega canal wasconstructed, parallel to the existing canal (Bega Veche (‘OldBega’) – 97 km). These two canals are connected downstreaminSerbianterritory.TheBegacanalrunsthroughthecityofTimi-soara and continues to the south-west. The river Timis is 359 kmlong, rising in the Semenic Mountains (the southern CarpathianMountains) and flowing through the Banat region and into theDanube in northern Serbia. The drainage area covers13,085 km 2 of which 8085 km 2 is in Romania and 5000 km 2 in Serbia. After entering the Banat plain, the river becomesslow and meandering, causing floods in rainy years. Especiallydevastating were the floods of 2005, when villages along theriver, in both Romania and Serbia, were severely damaged.2.2 Past flood events in the area Asaconsequenceoflargedischarges(exceeding1000m 3 / sinthehigh water seasons), and because of the morphology of the Timisand Bega basins, a series of technical measures were taken, someofwhichstartedalreadyduringtheeighteenthcentury(i.e.hydro-technical plants, deviation canals, dikes and polders). Thesedevelopments make up the so-called Timis–Bega System that aims to eliminate the risk of high floods. The most important measure was the establishment of a double connection betweenthe Timis and Bega rivers, consisting of two canals that allowgravitational flow diversion into one of the two rivers dependingon the flow conditions. Under normal flow conditions, water flowing from the Timis to the Bega ensures a minimum water supply for the city of Timisoara, while under high-water con-ditions for Bega, water is diverted from Bega to Timis ensuringflood protection for the city of Timisoara.Despite all of these measures, high water levels have stilloccurred. These high water levels caused the dike breaches andconsequent catastrophic floods of 1912, 2000 and 2005. Theflood of 2005 produced a flood covering an area of about 25,000 ha when there were two breaches in the right bank dikeof the river Timis.The biggest recorded floods produced in the Timis–Begariver catchment are those of May 1912, July 1966 and April2000, in addition to the flood of 2005. The flood of 25–30 Figure 1 Timis and Bega basins Integrated modelling for flood risk mitigation in Romania 271 May 1912 was the highest, with a maximum recomputed dis-charge of approximately 1600 m 3 / s, and estimated to have areturn period of 1 in 50 years (Hunter et al. 2007).For comparison purposes, the hydrographs of four recordedfloods, including the 2005 flood, are represented on the graphin Figure 2 and centred at the time of peak. The 2005 totalflood volume was three times greater (about 720 million m 3 )than that of the flood in 2000, though the peaks of these twoevents were very close. In 2005, as a result of the high flood peak, the embankment of the right-side bank of the river Timiswas barely overtopped, but still collapsed and multiple dike breaks were reported. The volume of inundation was approxi-mately 300–350 million m 3 . After the 2005 flood, a significant financial investment was made to restore the structural com- ponents of the various affected water dikes.In these conditions, the question naturally arises as to what measures need to be taken so that new flood events of thesame magnitude as the one in 2005 will not cause further damage in the future.The current flood forecasting and warning system in theTimis–Bega basin is based on empirical models which showthe relationships of the flow at a downstream point versus theone at an upstream station. During the flooding crisis in Banat in 2005, specifically with regard to the Timis and Bega River basins, there were three significant time periods with high rain-fall. Warnings of high water levels associated with these threehigh rainfall periods were issued, but the established proceduredidnotprovidesufficientlyaccurateforecasts,whichledtoflood-ing and consequently to severe flood damage. While the precipi-tation data have been quite reliable, lack of good rainfall–runoff model that could produce forecasts of upstream dischargestogether with the unreliable empirical approach used for floodrouting were the main reasons for the inaccurate forecasts.This event proved that better, more accurate models wererequired. It led to the conclusion that the elaboration of an inte-grated hydrological-hydraulic model for the Timis–Bega river basin is needed, comprising different flooding scenarios andindicating the spatial extension of the inundation, water depthsand velocities. 2.3 Availability of data for a flood event in the basin Before setting up an integrated model for the Timis–Begasystem, the analysis of the April 2005 flood which was carriedoutbytheRomanianBanatWaterBoardwastakenintoconsider-ation. This analysis was done on the basis of the dischargerecords at 20 gauging stations. The analysis checks the degreeof reliability of the data concerning the gauged and ratingcurve, i.e. the extrapolation discharges.The check was made for the total runoff hydrograph, thesurface runoff and the base runoff. This analysis reported unu-sually high estimates of the runoff coefficients a (surfacerunoff versus rainfall) (Stanescu and Drobot 2005).The analysis of the available observed data for the Timis– Begabasin forthe year 2005refers totheamountof precipitationand climatic conditions prior tothe flood period and tothe periodof the flood itself (14–22 April 2005) (Figure 3).The period covering two weeks before the flood event wascharacterized by comparatively small amounts of rainfall (10– 15 mm – recorded between 27 March and 1 April). However,the same period, especially the days between 8 and 13 April,was characterized by a sudden increase in temperature (4–6 8 Cdaily average in the hilly and mountain zones). These conditionscontributed to significant snow melt in the mountainous zones. It wasestimatedthat theincreaseinsoilmoistureduringthisperiodranged between 24 and 40 mm, and the larger part of this soilmoisture increase was because of snowmelt (compared withthe contribution from precipitation).By the time of the big precipitation events that eventuallycaused the flooding, the snow cover in the catchment has beenlimited to very small areas at very high altitudes. Under the alti-tude of 1000 m.a.s.l (corresponding to 96% of the entire area of the river basin) there was no snowpack. However, due to theabove-described conditions in the previous period, the wholecatchment was already quite wet when the big flood-causing pre-cipitation events came.This analysis led to the conclusion that the flood event itself was in fact entirely of pluvial nature and there was insignificant snowmelt contribution provided by the small mountainousareas,whichwasanyhowretainedintworeservoirsinthebasin(PoianaMarului at 650 m.a.s.l. and Trei Ape at 870 m.a.s.l.)During the flood event period (14–22 April 2005), significant rainfall occurred in three distinct time intervals separated by no-rainfall periods which varied between 30–45 h during 16–17April and 9.00–15.00 h on 21 April (Figure 4). Periods of time with reduced rainfall quantities continued after 22 April but they only fed the high discharges without contributing tothe increase of the water levels over the alarm threshold. Thecore of the biggest rainfall, in terms of quantity, in the floodzone ranged between 15 and 24 h.According to the data recorded at the pluvial stations, thecumulative rainfall (14–22 April) that caused the outstandingApril 2005 flood ranged between 60 and 221 mm. The smaller amounts, being recorded only around a small area, were in the Figure 2 Hydrographs of historical floods centred at their peak 272 I. Popescu et al.
