Luận văn thạc sĩ: Model based river water quality assessment under current and future climate conditions

Bạn đang xem bản rút gọn của tài liệu. Xem và tải ngay bản đầy đủ của tài liệu tại đây (7.28 MB, 204 trang )

Trang 1<div class="page_container" data-page="1">

FACULTY OF ENGINEERING SCIENCE

MODEL BASED RIVER WATER QUALITYASSESSMENT UNDER CURRENT AND

FUTURE CLIMATE CONDITIONS

Thanh Thuy NguyenSupervisor: Dissertation presented in partialProf. dr. ir. Patrick Willems fulfilment of the requirements for thedegree of PhD in Engineering Science

August 2017

</div>Trang 2<div class="page_container" data-page="2">

Gir đạc it a reson on fre

Pls transi

</div>Trang 3<div class="page_container" data-page="3">

MODEL BASED RIVER WATER QUALITYASSESSMENT UNDER CURRENT ANDFUTURE CLIMATE CONDITIONS

Prof dc iW. Sansen, chairProf. iL Sets

Prof dei E. Toorman,Prof. de. ie A. van Grensven

(Ve Universite Bross)Assoc, Prof dt, Hl Ls Pham

(Water Resources Universiy, Hanoi, Viet)

August 2017

</div>Trang 4<div class="page_container" data-page="4">

Uigsgeven in gen ehecr, Thanh Thuy Ngwen, Kenedbph Arenberg 0 — bus 248,1.50 Hever Belg)

Alle rechten worden. Niet it deze agave mag worden vemenigvai!en/of‘openbaar gems worden deoe mill van de, fotokepe, cri, minh ofop wake endee wie ook seeder vònfgamidjte sche teemeg van de

Al dghe naensd, No part of the p.liodion maybe eproduced in any Em by pm,hotgpdnk miro, cso or ay oe cans without wren pemmdsion fons thepublishes

</div>Trang 5<div class="page_container" data-page="5">

Ai he 6 ieLangage i hy ityKame iy sss

</div>Trang 6<div class="page_container" data-page="6">

son fom ether people, The day I ch my PHD research ie not 0 fr T

First of all, would Hike to thank my promoter Prof Padddk Willems. Youbrought me t the river water quality modeling since my Master thesis. Afr myMaster, you recommended me to kesp working on the river water quality topicfor my PRD, Ar thar moment, I thought there wis aot much work ro do andeverything was very cleat. Duriag my PhD, I eaized that aothing is periAet thếgestion was how we ean mitigate the disadvaneages and evaluate thr effes.(On the way to find the answers, I have received your guidanes, comment andcncouragements. Especially, highly appreciate the rust and ficedom you gaveme to fill my reeateh in the USA forthe lst 15 years. I would like to thank“Thay Lai University for retchông mỹ position at the university dường mỹ PRD

1 would ike to thank Prof. he Smets and Prof, Enk Toorman for the Follow up‘of my research and foe providing comments and suggestions from the tart of myPRD research to the faa, Your advice helped me to consider the researchprublems in a more comprehensive xay In sđẩtion, T vosld ho ike to thankthe other members of my jury for sei feedback on my thesis ext. Prof, An vanGricasven, thank you for sharing your expertise to improve my farute cseuch,Prof. Thi Huong Lan Pham, thank you for making thế log journey to atend mypublic defence. Prof, Willy Santen and Prof. Hie Heynen, thank you forchhưởng my jun ao thank Prof. Yoram Rubin and Dr. Susan Hibbard fyoffering the opportunity to be a visting student at UC of Berkeley. 1 would bike

</div>Trang 7<div class="page_container" data-page="7">

"Thank you Ingeid Keupers for your excellent work on river water quimodding on which I could build further. Thank you Els Vạn Uytven for your<ouperation, and Vincent Wolfs for dlveing the InfoWerks RS dongle to meuring mỹ maternity leave. Tho would ike so thank all ur group members whosombmtel so the development of WETSPRO and the climate changeperturbation ool. More thanks go tall oF my ends who brosgh

Belgium and the

ny home tòân very clorfl culture environments

A special word of thanks goex to my parents, siblings, niec and ngphew: Youaways encouraged and supported me to finish my PHD.

1 aio would like to thank Phuong and Simba, who accompanied me since the‘very frst day of my PHD, We have shared the happy and sometimes though timetogether, even online or offline any sa proud of being with you, being your wifeand your mama,

Last bạt not leas, I would like to say thanks eo a very special peeson that L have‘admired since Iwas small. Your love, your Ife and sour work have beenfollowing me and lighting my life and my carer up.

ThuyLeuven, Ags 2017

</div>Trang 8<div class="page_container" data-page="8">

River water pollution is known sone of the major cmironmengl isues in theworld, In order to achieve the sustainable development goals i is crucial toprvvide quantitative information on iver water qualiy. Such information plays akey role in integrated river basin roanagement in genenl ane water qoarymanagement in pariedlar. Especially climate change with the increase intemperature, extreme flow conditions and changes of ecological systems ispredicted to cause watee quality degradation, Ie requtes poley makers o takeright remediation actions. Questions have the answered such as how the waterquality is, where surface waters are polled, which sources cause stress tớ therivets, how effective specific actions or measures would be, Many existingmethodk have been proposed to answer these questions. They make use of riverswatergualty (WQ) observations in combination with models, WQ measurementsare limited to some re point long the river and at specifi sme moments“They are also affected by measurement and sampling errs. Therefore, it would

be incomplet, inaceurate and uneconomie ifthe WQ seated decision-makingwould only rely on the available measurements. By sohing mahematicacations detoibing mansport of pollstants snd their biochemical processes inthe aquatic environment, WQ modd can provide a more complete picture oftheWQ state, Theretons, the first main part of the thesit is applying models eosimulate the trảnqpof of pollutants inthe river system and analyse factors thatinfluence this transport

Tn the first phase of the research, the three sofware packages MIK 11,Infoorlks RS and InfoWorks ICM were used. In these sofeware packages, onlyMIKE II and InfolWorks RS allow to model the tine vuiadon in aver bedroughness coocint, The, therefore, were sdeeted for analysing thế seasonalaration oF river bed roughness impacts om water level, The time series ofroughness coefficient was calibrated tothe observed water level in MIKE 11, The

</div>Trang 9<div class="page_container" data-page="9">

The second application ofthe WO move the duy of the influence of modelstructure uncertaingy on river water quay assessment, First, the physicbiochemical processes were sereenel to obtain preliminary assessment on thecite processes and to determine the processes that require a more detdled‘comparison, Then, local sensivtyanalbs was erred oụt o specify the sensitive

pptamercrs and processes. Results show that the hyeeodmamic results, heat

transfer rate and ceseration simulations cause lange differences in moddiSimulagon oupots for water temperate and disolved oxygen (DO)concentrations. The ignorance of processes related to sediment trưa,phytoplankton and bacteria has a significant infuence on the higherconcentrations of organic matter and lower values of dissoled oxygen‘concentrations, The thice models show consensus on the main pollutant sources‘explaining ongunie matter and nitrate concentrations, but diagree on the main

factors explaining the DO conecntrdions

“The wse of software packages as MIKE. 11 and InfoWorks, which are based onfall hydrodynamic model, show dlificulies in quantifying model structurecertainty dục wo the gid Stucrres. Addisonaly, ro get convergence i results,a small time step is required. The simulation for long periods and big systemsrequires very long simulation time, Therefor, these models age inapplicable Forstudies in which many simulations or long simulation periods are needed. Onesolution isthe use of fast conceptual models. The main contibution of this thesisis extending the COnceptaal River Water Quality madel based on theInfoWorks RS (CORIWAQ-RS) processes and results, The research started from,the code that Keupets (216) developed for MIKE. 11, The model conceptuaizesrivers using cascades of reservoirs and lumps the advection diffusion physicabiochemieal processes. The hydeodynamie inputs are derived from the ours ofthe fall hydrodymamie models. We performed) comparative analysis on theCORIWAQRS and CORIWAQ-MIKEI models. Resuks indicate chat the

</div>Trang 10<div class="page_container" data-page="10">

Aloe »

‘concept medels perform equally wel asthe original MIKE 11 and IniAWorkeIRS models, but with much shorter simulation time (10 times). The successful‘esting ofthe conceptual mods opens a development avenue towards problemsling in che context of WQ contrl and management.

Next, CORIWAQ-RS was considered on the bass of neo «spss of applictions:uncertainty analysis of river WQ input and assessment ofeimate change impactsfon iver water quali. In addition to the incomplee knowledge on the WQpracestes, lack of pollution input data is known as one of the major uicerrinhysources in WQ modeling Model mmprovement, hence, should focus on obainingmore information on the inp/boundhuy conditions, Given that there ate so‘many pollusant sourees along 2 ver, model improvement actions should focusfon the most important souces to optimize the cost-benelit ratio of the ationsTnpat uncertsindes were modeled by the stochastic regression approach, iadding randomly time series of errors to estimated tine series by regression‘sations, Based on sensitivity indies related to the model output error variance,important poliudon sources were identified, The contibution of inputuncertainties tớ toal model output residuals was guandEod bythe total variance‘oF made! outpat errors đúc to lips,

The application of CORIW.AQ.RŠ for climate change impact analysis involved© quancfying climate change impacts on river lows and concentrations; (i)Setermining and evaluatinginpur factors that primarily contol the changesand Gi) evalsating the wncerajngy of climate change scenatios on the water

quality assessment, recpiation, potential evapotranspiaton and air empeature

in 30 climate change scenatios were sradedelly downscaled from 30 GeneralizedClimate Model runs, The influences ate evaluated for the observed pedod 2000-2010 and the target period 2050.2060. The hydrologieal made, regression andstochastic approaches were applied t transform the climate change signals formeteorclojedl variables to changes in runoff, nitrogen loads from catchments,DO concentitions and physico-biochemical rates. It is shown that climatechange ma lead co highly negative impacts for DO concentrations, The climatechange impacts ats, however, highly uncertain, especially for the high recorperiods

“The ease study sdected to implement al cbjecites of the doctoral research i theMoise Net sivereatchmeat ia Belgium,

</div>Trang 11<div class="page_container" data-page="11">

Beknopte samenvatting

ollie van riverwater stat bekend als đến van cúc voomaamaenilieuproblemen wereld, Om de dourzame onevikkeingsloslen te kunnenbenfken is het crucial om kwintitatieve informatie te anayeeren mbt. deriversaternaitt Ze spect cen crle tl in het itera strocmgebiedbeheerin hor algomecn en in het watefbealieiebchecr inet bữzondir. Op velelocuierwereld worden de waterhwaliteitsnormen no niet gehaald. Bovendien kan deKimuanernndeing de waelaaaltet van waterlopen negiúef belavloedenKHmaarertndring mượt immers voor een toemame in de Iueht enwsteremperstour, voor meer extreme wserloopdcbiren, zowel tocncmendepickavoeren als dalende hagwaterafvoeren, en gereltcerde wiigingen aan hetcologich ayvtcom. Tr is daatoor nocd aan sancti enLimautadapeaiestategicén. Om de ontuikkeling oF het omtwerp vận aulkestraps te ondersetnen wordt in modern waterbeheer gcbruik gomsskt vansimahdiemodelen. Deze moeten toelten de huidige waterkvalittsoestand tekwanlReeren en te evalicten, de oomalen na te gaan en acest effcgnteoploaingen voor te stellen, De moddlen worden camplemnnir aan debeschikbate waterhualiteismesingen en andere empirsche informatie gebroikt.Waeckvalitesmsetingen ain immers qplech slechss beschikbaar op cen beperktaannl locates lings de rivier of op slechts cen beperke aantal temomenten‘Veeder jn deze waarnemingen sterk ondethevig aan meet of aalysefoutea. Debeschibbare vaamemingen sjn dus onvolledig en onniuvkeurg. Het mưuincficént jn om watthwaltctsbcheer vitshitend te bascren op zulkeinformace. Vin wiskundige simulatemedellen kunnen de waaremingen fsischonderbouwd genterpolerd worden, Ook kunnen ze vit scenaro-anaysesetattapoloerd worden, vb inde ied = oa via soon’ inzake Klimaatopweming— of door simulate an bepslde wizigingen san de wterloop of aanpassingensan hee waterbeheet. Deze inter- en exttapolaies gebeuren door gebruik temaken van keanis over de fsische process en vervulingsbronnen die an debss Hggen van de watelewalielotoesannd van een waterloop. De modellen

</div>Trang 12<div class="page_container" data-page="12">

ii Benptesameneaning

lessen hieroe de tansporsprocessen op, zoal de advectie-dsperseverglikng onde voornaamste biochemische omzesingsprocessen. Er Zt cvenvel nog velamsekehel vervat ia lke medeleing Zeker in verging met-watediwamieitemoddlộng xoa hydelogisee en hydeaubsche modeling is

nog vec onderzock nog naar de nauskeurgheid van waterlwaltesmodliering,-owel wat de modlinvoce(verwulingsbronnen) betref als de proceskennis en

schematseing. Ex is nood aan verder ondetzock inzake de invloed van dermodestracuur (de set văn gelmplementecde procesvergelikiages) en hạtbepdlen van de meest optimale modelsruetuar voor sen speciBcke toepasting,Dit vag con verbeterdinzicht in de onzclerbeden betokken bij de modlleringzat cen goede afveging kan gemaake worden tusen de modelgedcalleerdheid,de bihotende rckend en de nauwkeurigheid, Vele seenatio-analyecs, oals hetsimuleren van de impact van de limaatverindering, andere extereinvlocdsfactoren en beptalde waterbcheersiatepicén vragen. lange-terminsimulaes (dmulade van tdreesen van meenderetentallenjaren) of een gfootaanal modelsimulaties. Hetelide gellt ook voor onderzock naar detmodelonackerheden. Verler is de kemis over de impact van de hienuor‘opgesomde tpen senaro's op de waterkwalitcstocstand lags waterlopen no,cendemmase. Dit doctorttsonderzock had als deel om voor cen deel aan dezenolen tegemoet te komen. De modeleing van de waterkwalteit langs de MoseNet in het Netebekhen fungeerde dambj als gevalstudi. De-waterkwalittsmealleing bepete vic tot de opgsoste ruuntofScin incase“organische vetontreinging,sksrofmutinten en de watertemperstiue

Meet speddick werden voor de gevalsudie đúc bestainde en frequent gebruiktesofoeatepabhetca bestodcerd: MIKE 11 lnfolWorks RS en InfoWorks ICM. Derodesracsaut van el wan deze madelen werd gesnalscerd en geimplementeerdin et COneepiual River WAer Quality (CORIWAQ) conepeeslmodllringsplatform dạt srler aan de Afdcing Hydraulics van de KU Leuvenwerd cmuikkeld, maar wea beperkt toe de MIKE. II-procesvergliagen.CORIWAQ caneptudlieenl wateslopen va een eascade vin reservoinnodelenem Xoợt sen mởHMỞjke aggregate door van dc advecie-dnpersie enbochemiche waterkwalitesprocessen per riviera (per reservoit). Alsmodelinvoer wordt tvoer van hydrodyaamische modellen en gegevens aferde verschillende vervulingsbronnen tangs

derelide invuer als gebruke in de MU

gedesillerdere waerhwaltcismodellen

de watedoop beschouwd. Dit isB11, InfoWorks RS en ICM

</div>Trang 13<div class="page_container" data-page="13">

Bedawpre samenating &Na vergeliking met de oorspronkelike, gedenlleerdere moderesulten enwaterkwalitesmetingen bleken de đúc conccptuele madellen een vergelikbarenauwheurgheid op te leveren De rekentia van de conceptucle modell was —iavogdjiing met de geleadlerdere moddlen — cen factor Út Kine, watperspectieven opent voor hel wat toepaednscn

Voor ok sin de đúc sets aan waterkwalcitsmodelverglikingen (MIKE. 11InfoWorks RS en ICM) werd via een gevoetgheidsanayse de htische processenen parameters geidentifeerd. Vour de mordeleultaten van watertemperamut ea‘opgsloste aaurstof bleken dit de hycradynamische modelresulaten (invoer voorsodclen) te in, alook de warmteoverdracht- en naruurhjkcberbeluchngsprocesen, De vervnitlozing vin sedimenttanspor, phytoplanktoncn bucterga bleck con bdaagejike impact te hebben op de hoge cuncenteaties aanonganische vermling en lige concentrates aan opgsloste zumestof, De diede watedealrer

meddseuchren bleken tor develde concusies te komen mbt. welkeverwuilingsbroanen vooral de concentrates san organisch maerial en nutiéntenbapulen. Ze bleken echưết verschillende concusics te geven over welkefaetoreade opgeloste susrstofouncentratesveklaen

Ben andere belangrike factor Đeck de sijhwatatie van de ruwheid van derivierbeding, door de sezoensvaritie in begrocing, op de waterpeen lang devatedoop. Vin de dre bestudeerde modellerngspakketten bien enkel MIKE: 11en TnfuWorks RS te om deze invloed rechiserecs te analyseren. Ke weed opbisis van beide modellen geconchideerd dat het niet inrskenen van descizoensvaratie in de rvierbedrusheid cen lange invloed kan hebben op dewatepelen in de zomer. Tidens pedodes met siverwassen is de invlocd danweer Mein.

De lage rekentd van de CORINVAQ moddlen hết toe om cơn geleaileerdeomsekghetloamlysc it te voetsa, wat voonl behmgijk bleck voor deemsekgheden op de modeimvoer. Voor de verschillende imechacingen aanvervuilingsbroanen a andere modelinvoer werden de verschillende tannamesgelaventascerd en de mogelike impact ervan op de invaer ggkuanlifeeenlDeve madelnvocronacketheden wenden sia een sochassche modeleringxmethode

becheten en gepropageerd dooheen het conceptueel_model. Viagevoclgheidsndices werd het seltieve belang van lk van de bestudcerde‘onzekethelsbronnen ingeschat. Dit it daamma te om de rota oraekeihdd viasen vatiantie decompositemethode op te len in de ndxieve b]dngen van elk

</div>Trang 14<div class="page_container" data-page="14">

rockomsthorizon 2050-2060. De esutatea tonen voor bepalde klimastscenaro'scen erg nogatcve impact, vootal voor de eonceneais aan opgeloste zuutstot enhoe erugkeerpeioden,

</div>Trang 15<div class="page_container" data-page="15">

Adsection Diffmdon

American Public Health AssocationBelgin Federstion forthe Water SectorBiological Oxygen Demand

‘Chemical Oxygen DemandClimate Change

‘Centeum voor Onderzock in Diergeneeskunde enAggochemie

‘COceptual River WAter Quality mode!CORIWA developed based 09 MIKE 11CORIWAQ developed based on TnfuWorks RSDepartment for Environment Pood & Rural AfiDanish Hydraulic Insite

Dissolved OxygenEuropean Commission

1 United ScareEnvironment Protection Agency

potential EvapeTranspirsionEuropean Statistics

Growing Degree DayHydeoDynamie

Global Water PamnerbipInhabitat EquivalentImpact Factor

</div>Trang 16<div class="page_container" data-page="16">

NIST SEMITECH,NNN

[Nasional Insitute Standards & TechnologySom of nate and iste

Nash-SotcifeEficency‘Total nnugen

River Water Quality Model

System forthe Bsahaton of Nutrient Transport toWarer

Sediment Oxygen Demand“Temperature

“Tweode Algemene WaterpasingBase Tempersture

</div>Trang 17<div class="page_container" data-page="17">

Vlaamse Milieu schappi

‘Water Eagineodng Time Seies PROcessing woolWorld Health Onginiation/ United NationChien’ Fund

</div>Trang 18<div class="page_container" data-page="18">

Reknopte samenvattingAcronyms

“Table of contentsList of FiguresList of Tables1 Introduction

11. Prablem satemene12. State of the arL3 Objecives

L4. Overview ofchaperc2 Molse Neetcatehment

21 Genetl chatacteistes22 Pallant sources

3 Emng models and methods

2s

</div>Trang 19<div class="page_container" data-page="19">

S11 Hydrological modes

ALA. Probability Distribution Model31.2. Generalized Hydrological Model32. Hydrodynamic models

33. River water quay models331 Advection and diffusion

332. Physieo biochemical transformation processes34 Mosel inputs

344 Model input data3422 Model input estination3⁄8. Climate change tol and seenatios36 POT extraction method

Development of a COnceptual River WAter Quality model

(CORIWAQ) with flexible model structure

41 Tnmnduelon42 Processes

421 Advection and diffusion

422. Physico-biochemicl ransformations43 Calibeation parametees

44 Implementation for MIKE 11 and InfoWorks RS441 MIKE tt

442 InfaVorks RS45 Performance evaluation46 Resnhs and discussions

</div>Trang 20<div class="page_container" data-page="20">

5‘Seasonal variation of iver bed roughness impacts on water level 83,BA Methodology Mã52 Ressls and diteusions a521 MIKE 11 calration roughness coetiient W5.22 InfoWorks RS time-varying roughness cosicient »S3 Conchasions 96Influence of model structure uncertainty on river water quality

assessment sỹ

6.1 Methodology ”62 Results and discussions wi621. Graphical seasitvty analysis 102622. Loca sensitivity analyse 0663 Conchsions neCORIWAQ input uncertainty analysis usTH. Methodology usTT. Results and disewsions 1207.2.1 Gap fling and stochastic modeling of input uncemimier 1207.22. Senddrin indices BsT23 Convergence 132Cangibsdon ofinpor uncerainy to total model residuals 13273. Concusions 133Climate change impacts on river water quality 1781 Methodology tr8.11 Propagation ofelinate chang to sver water quality input TẤT18.1.2. Impacts of climate change on extreme flows and determinant

‘concentrations 139

</div>Trang 21<div class="page_container" data-page="21">

8.13 Evaluation of the input factors controling the climatic changes0BALA Sensitivity of the hulmlogiol made! cdbrsdon to the climatetrang impact assessment 1H2. Results and discusions ae82.1 Propagation of climate change to ver water quality input 3822. Impacts of climate change on extreme flows and determinant

concentrations 1s

1423. Evaluation of the inpat factors controling the climatic changestí834. Sendiddg of the hydrological model calibration to the climate

‘change impact assessment 152Bồ. Conclosions 1559 Conclusions and perspectives 1879. Contributions ofthe doctoral research 1589.11 Quandesdvey evaluate and compare the reference river WQ.models and increase the insights in che river WQ divers 1589.12 Develop a river water quality concepual model based onInfos RS 1609.13. Uncertainty analysis and assessment oF elimate change impactsfon river water quality 1609.2 Puture research 162921 Development of CORIWAQ 163922. Improvement of dhe results 16s,

Bibliography 167

'Caieulam Vitae 183Publications by the author 18s

</div>Trang 22<div class="page_container" data-page="22">

List of Figures

Fig. L1 Impact mechanism of CC on river WQ, 1"Fig. 1.2 Structure ofthe thesis 15Fig. 21 Molse Nect catchment, 18Fig, 22 Livestock in Molse Neet catchment 18Fig. 23 Aquatic plants in Mose Nee sives in August 2015, 19Fig. 3.1 Medel arocture of The PDM rainfall rnot¥ model (Moore ets. 2007).24Fig. 52 Model srvetze ofthe VHM rainfall runoff mode! (Willems 2014)... 6ig, 33 Simulated and observed water temperature ar suöon 383000, Molse Netchết 45Fig. 34 Parameters used to sleet ney independent extreme Sows (Willems

2009, si

Tig. 41 Flow chất ofthe CORIWAQ operation for MIKE 1T and RS...0Fig. %2 Reserv determination in CORIWAQ foe MIKE 1 and RS. đoFig. 43, Rnfll ranoff from upstream sub-eatehment to Molse Neet đưếno 63Fig. 4 Calibration resus he location of sation 333000 forthe year 2008 for(4) water temperature, 6) DO, (9 NHN, (@) NOMN, () BODs,(D POP andLạ) PP-P concenteations in MIKE 11 66Fig. 4.5 Calration results a he lotion of sation 333000 for che year 2008 đọc

(6) water temperature, (6) DO, (6) NHAN, 4) NOWN, (6) BODs, () ON-N and

Fig 46 Box Whisker plots of eallrted fra) and ana (b) at reservoirs along,the Molke Nest sver for wet, mein and dy yeats corresponding to MIKE TT(ete) and RS (ht) modes 70Fig. 47 Seater plots of water temperature and DO concentrations in MIKE. (1and CORIWAQ MIKEII ot observed data at observation station 333000 alongthe Mose Nee ver after aplieasons of diferent erated parameter et...72N concentrations in RŠ co

</div>Trang 23<div class="page_container" data-page="23">

Fig. 48 Seater plots of tt temperature and DO coacentrsdons in RS andCORIWAQ RS or observed dan a observation station 333000 along the Molseeet ver after application of diferent calibrated parameter ets 2Fig. 49 Scattse plots of NHẠN and NON coneentrtions in MIKE 1 andCORIWAQ-MIKEI or observed dita at observation sation 333000 along the“Molse Nect ever after application of differen elated parkmeter se... 74Fig, 410 Seater ples of NHẠN and NON conecmrdlome in RS andCORIWAQ.RS or observed data st observation sition 333000 along the MoleNect ve aftr application of diferent calibrated parameter sets. 15Fig 5.1 Observed versus MIKE: 11 simulated water depths at Mecrhout, ater use‘of a constant Manaing coefiient of 0.35 4/1 seFig, 52 Calibrated Manning's e values per Sveday period with RMSE forcalibrated luc () and piecewise lines relation of the Manning's n versus day ofthe year () %9Fig. 53 Observed versus MIKE 11 simulated water depths at Meeshout,afersse of a time-varying Manning eveficient mFig Cumuladee probability plot for observed and simulated water depts Forset 2005, calibration () and summer 2007, ridadoa (), 90Fig. 55 Observed versus MIKE-11 simulated water depts, aftr use of a time

ent for sumer 2007 otvarying and fow-dependent Manaing coef

Fig 56 Rating curves for Manning cveffcens ranging rom 0.03 to 0.08 s/nswith an 005 interval (a) and relation of the niting curve parameters on theseasonal Mansing coefficient (, 92Fig. 57 Comparison of che estimated river discharges, after use of a Fixed andtime-varying Q = b relationship with the MIKE-L1 simulated discharges and che

precipitation time series fr summer 2005, 8

Fig. 58 Correlation between independent water depths in MIKE-It andInfoWorks RS; fo lw ) and peak (6) water depths in cabation peiod (2004205) aad for low (6) and peak (0) water depths invalidation pedod (2006-2007)'¢ 8.54 km along the Molse Net River sẽFig. 59 Local roughness, overall cross-section Manning's in InfoWorks RS andMIKE 11 at 8034 kin along the Molse Ner river on thước specific dụ 1February, 15 June and 3 Jul. 95Fig. 6.1 Proedae to implement model sưuetere wncersnty anal 101Fig. 62 Profiles of the masirmur and minimum water temperature (a), BODs (6),NILN (9, NOvN(@) and DO (6) concentrations slong Molse Net ier in theinal dmmuhdone tos

</div>Trang 24<div class="page_container" data-page="24">

Tàn of Figures iFig. 63 (ef) 10-percenle DO and 90-percentle BOD: concentation profilesbased on RS, ICM and MIKE 11 rouls in comparison with the Flemish standardA; right) coneenttions verse rerum period atthe observation station 333000.

bn aFig, 64 Comparison of CDY-rlationships for simulated DO concenttasons atstavion 333000 in MIKE 11, RS, ICM and fundamental interment standards‘wth areca period oFI month, 3 month and I year mFig. 7. Procede for the input unceruinty als. 119Fig. 7.2 ‘Trend in dschagge from factory Philips. iosFig. 73 Empirical and calibrated distibutions of (a) IKN concentration fromAjinomoto, &) KN concentration and (9 BOD, concentrations fom Philips. 126Fig. 74 Empiiel and clbnuel distibutions for DO concentration from

“Alinomoxo, Los

Fig. 15 Sensitivity indice 8, of poluant sources for (a) DO, (b) NH, (2)NOSN, (@) BOD: and (@) ON-N concentrations, for the ter concentrations atthe observed stition 333000 at the Molke Newt sver in the whole yea, wine,spring, summer and ase, tạiTy 76 S, for different sample sizes for ON-N concentrations from Ajnomotofactory to river ON-N concentations ngFig. T7 Contribution of input uncertainty to coal model residue, 133Fig. &1 Procedure co evaluate impact of uncertsinty ia CC scenarios on WQ andcontribution of cachinput from 14282 Box Whisker plots of (2) changes in LGP and (b) changes in ipcoefficient values fr specifi cops 148Fig. 83 Hox Whisker plots of the impact factors eaeulted for (a) monthlyprecpiaion and (b) monthly etal niteogen losses 1SFig. E4 Box Whisker plots of the changes in (a) water temperature and (5) DOSconcentration atthe boundaries of subseatchments i OC seemaio... LH”Fig. 85 Box: Whisker plots of the IP calelaed for temperate cosfñiciemf (3)`. 18Fig 8.6 Impact factors fr extreme (high flows and (9) low Bows, versus return

posi, 1

Fig, 87 Impact factors for exteme () high NHN, () NO“, (9 BODs,(€)ON and (low DO concentrations versus recur period 150Fig, 88 Impact factors calculated for extemely (a) high Bows and (0) low flowswith PDM mod parameters calibrated to Grote Nete Varendonk station, versusxen petiod 153

</div>Trang 25<div class="page_container" data-page="25">

gHẢ List of ges

Fig, 89 Impact fetes fr exeeme high (a) NHN, (ý NON, (© BOD. (0)CON N and (9 low DO coneenerasons at measurement station 333000 wil PDMmod parameters calibrated to Grote Nete Varendnk station, versus recom

</div>Trang 26<div class="page_container" data-page="26">

List of Tables

“Table 2:1 Manufactodes discharging to the Molse Neet river. amTable 22 Number of habiants discharging thơ wastewater to the Molse Nectwithout treatment and as point sources (UE-114363, UE-114372, UE-114373)“The ether inhabitants discharging thir wastewater untreated (eine sources with

VHASOL (distibuted sourcelower IE) ate đatibunel over the hydrological zone

UD-50)) 21Table 31 Physico-biochemieal processes and equations in InfoWarks R$/TCMand MIKE II 3B‘Table 32 WQ parameter values in TnfoWorks RS, ICM and MIKE 1 4“Table 3.3 Summary of available dan for diferent pllatantsoures, 2

Table 34 Inial ni) and fine-tuned valves (Ca) forthe nitrogen factions 0ŸAlfernt nitrogen components from sriclore, 7Table 35 Overview of 30 GCM suas considered 50Table 4i Minimum values of NSE of CORIWAQ-RS vs RS calculated withAltferent values of 12 WQ parameters forthe 7 WQ variables, osTable 5.1 Transition dep in Function of the tral unit roughness %6“Tale 61 Simulations conducted for lel sensitivity analysis, 108Table 62 Sendtivdy indices for the maximum and minimum temperatures andDO conecnteations (the white cells seer to RMSE and the grey els sơ $108Table 63 RMSE beeween the BODs, NHN and NON cancentations of thesiferent MIKE. 1 simulations and thore ofthe RS_1 simulation, 109Table 7.1 Coeficients of regression equations for the model input variables offactory Ajinomoto independent variables i the rows, dependent variables in thecolumns), The last row shows the etror dauibudons (arma dstibutionsindicating. the mean and standard deiadon as follows: Nữneim, standarddesi) fey

</div>Trang 27<div class="page_container" data-page="27">

“Table 72 Coefidents of regression equations for the model inpat vatabes offactory Philips Gndependent varibles in the rows, dependent variables in checolumns). The last row shows the error dieHbaions (somnsl distbusionsindisting the mean and standasd devittion sv follows: Nimesn, standardleviation) 123Table 73 Estimated user of inhabitants thar discharge tothe Molse Neet ier

135‘Table 74 Estimated wastewater discharge by domestic household 1

Table 75 DW vas forint variables of industrial polteies loads 126“Table 81 Standard deviason of the impact factors calculated for extreme values‘oF WQ vavables due o changes in 4 input fats mg“Table 82 Standard deviation of the impact factors ealeuated for extreme values‘of WQ vaiables due t changes in 4 input fetor with PDM mode parameterscalibrated tothe Grote Nete Varendonk sation 158

</div>Trang 28<div class="page_container" data-page="28">

1.1 Problem statement

Gan wate is an ineseasing concern to our society due ta its sương dieet andindtect influence on human's health The United Nations 2000) reported thatpeople with water-related diseases eontbue up to 50% of hospitalized patientsin the work. Warertlaed diseases cause aatly one of every five deaths under 5vests old (WHO/UNICEF 2005). Under impacts of eimate change (CO), thiseffect might be more severe duc to depletion of water quantity and degradationcof water gusty. Observations snd CC scenarios derived for different sconomicand societal conditions indtewe significant changes in air temperature and«xtreme precipitation (Ntegeka ct al, 2014). These ewo vaiales ae diving forces(of the physico-biochemieal processes in rivers. As a seul, in order to supportriver management plans that cope with CC, itis essential to quantitativelycvaltate the changes of river water quai (WQ) under diferent CC condiions,Such sedis ae, however, sill very limited (Michalak 2016)

Research on CC impacts on river WQ is implemented by fed experimentsand/or msmerieal modeling (Moss ts. 2003, 2008) "They both indicate negativeimpacts of CC on river WQ, Water temperature, bilogiel oxygen demand anditogen loads to rivers increase while dsslved oxygen in water eeduces (Wilby1993; Whitehead etal. 1997; Bouraoui eta. 2002 and ; Monteith etal. 2007).During the summer, the temperature increases, the flow decreases and theresidence time inctsses, which are good conditions for alge blooms, especiallyán low dissolved oxygen concentration and higher nutrient concentrations eiogen and phosphorus). In the winte, the higher temperature, onginie mater

</div>Trang 29<div class="page_container" data-page="29">

2 Chapter: ction

decy and soll mineralization cause more organic matter and nitrogen to beteunsparted to iver đường the storms. However, the ditecton and magnitude ofchanges in determinant concentrations for specific ease vnnlet can be venAiffersat, For instance, the simulation results for the Kennet river in the UKshowed increasing treads in both ammonium and atte conceateaions (Wilby eta, 2006). However, van Viet and Zvolman (2008) reported that in droughtvents, decreases in nitste concentration and increases ammoniumconcentation were observed at diffrent lacitions in the Meuse hen, The DOconcentrations atone station increase while thy decrease at anche station“To keep water clean, iis crucial co effectively contol and manage the WQ,

which requires quunsiaive knowledge on the water quanti’ and qua statsboth in sme and spac. The WQ stars ean be obtained from monitoring systems(Shrestha and Kazama 2007; Ghoiradc etal. 2016), However, sich systems areexpensive and usualy insufficient ro cover the high spatio-temporal variably ofWO vasables (Bary etal. 2012; Lessels and Bishop 2015). The

be incomplete, insccurite and sneconomie ifthe WQ related decision-makingsứ would‘would only rely on the avilable measurements, In that contest, mathematicaWQ modeling is a useful tool to provide such information, WQ modelling aoives decision makers insights into the cause-effect relationships and predicts thefare WQ states, and therefore, support the design of water resourcesmanagement stages (GWP 2013), By comparing simulation results for diffrentscenarios decision makers căn easily derive evidence-based decisions,

However the main problem in using mathematical models i thet accuracy. These of imple models exposes to many assumptions and simplcation, which cnreduce the accuracy of simulation results. On other hand, complex models thatintegrate 2 fall HydroDynamic (HD) madel with « detailed description of thephysico-biochemical processes allow to devive results with detailed temporal andspl variations, The WQ modeling software packages ate diferent in givensumptions, number and equations of physice biochemical processes that aretaken into account, These dlssmilrtes can led so discrepancies in thiesimulation eesuks. Therefore, it is crucial to quansiatvely compare WQmodeling pickiges to provide the usets with insights of each package andsupport them in selecting a ulable model for thế tly re.

“The typical ertor in HD river sues, which reduces the modelling accuracy, iđạt the roughness coetcint i usually assumed tobe constant, hence đưggtrHingthe seasonal effect of vegetation growth. ‘Tis assumption may lead toa bis in

</div>Trang 30<div class="page_container" data-page="30">

Chapter :meloedon 3

the estimation of diver Now velocities and water depth, Where high nutientloadings are present in dươn with low velocities, quae plants cản growbunlanly and postibly mitigate the detrimental effects of this palltion stressTndsed, a lower vslociy means longer residence time and higher sedimentationrates and thus a higher self pusfying capacity (Schulze al, 2003) Therefire iisecessiry to conser the senstvigy of the low rogime to seasonal variation ofriverbed roughness due to vegetation growth

Another challenge in WQ modelling and management is the seusig andeteogency of the WQ data available for model eaibation and validation,‘Various be of point and dfs pollutant sources ate spaily spread over the‘whale catchment and their concentatons and loads are highly vata in bothtime and space. The information abowt this variability i, however, often veryime pollatant sources even đo not have any measurement of thirdischarges and concentntions, Several authors reported that the pollutant inputuncertainties are one ofthe major sources of uncertainty in svee WO modeling,ine (Racivan et al. 20045 Freni and Mannins 2010;Willems 2012), For these reaeonr, it serial to carefully handle the missing damand to arses the influence the lack of input data bác on the simulation results,TẾ not the most important

Finally, among many WO models, one may pose the question which model issible 10 conduct CC impact assessment with nông high accuray. Asmentioned above, the detailed WQ models are able simile the temporsl andspatial vaiadons of sivee WQ variables. However, dey have long elculadonlies, which makes the use of these models impractical for many wypes ofapplications. The long simulation time poses diiehies on applications thatinvolve hage number of model run, itrtions and/or longterm simslations,such ae model uncerainty amalyjs, auœedbruioa, realtime conto,‘optimization, ee, Particulate assessment of CC on WQ, which involves 30ete or more long-term simulitons for several eenados is impractical witherailed WQ model. Therefore it is necessary to develop the conceptual model‘ht can derive simulation suk similar to the WQ modcls but in shore me

1.2. State of the art

‘ym (1980) defined a mathematical model to be “a tiple (S, Q, M) where Sis asystem, Ö is «question relating to, and Miss set of mathematical statements MF

Us 2s m) which can be used to answer Q". A river WO model isthe

</div>Trang 31<div class="page_container" data-page="31">

4 — Chaperirlmndhsdion

rmathematial model which depies the dvet (8) answer questions rey the fine‘pace volition of water quality variables alo the vor (Ợ). By solving mathematicaluations (M) detenbing transport of pollatants and their biochemical processesin the aquate envionment, WQ models can provide a move complete picture ofthe WQ sete, They allow one to beter identify posible stress positions andinporant polluant

actions. The slope of WQ profiles also helps the modslers sọ recognize thesensitive processes and to assess the tla oftheir parameter values.

curcss aswel sto evalte the effectiveness of remediation

Made per

“The Streter and Phelps equation (Smeder and Phelps 1923) for simulatingdissolved oxygen (DO) and biochemical oxygen demand (BOD) have formedthe basis of many WO models In terms of simulation compledy WQ modleca be elsif! into simple eg, TOMCAT, intermediate (eg. QUALZE) andcomplex (eg; Delt 3D) models (Cox 2003), Ahhossh th simple model requireslest input, ie considers only steady flow state and Limited phyieo-biocheprocestes, The intermediate models ate more complicated. Ye, some imporantprocesses, eg, back fous, ops in sverspstems and lateral snfall-unol, areot accounted for, The complex models do take these processes into account butae computationally expensive and may encounter the problem of parametstiendiubliy in calibration. The choiec of the mo appropriate model dependsfon the study si. However, WQ modeling software packages ate aot alwayslear about thất assumptions and application conditions, As aresul itis diffefor uses to te whether given model is stable or thir specific appictons.For example, differences in empiriel roughness equations for main channels andIhydaulc stracutes representation may cause important differences in the HDsimulations and, a a consequence lo in the WA resus (Warmink eta, 201)[Emulation modeling is known as a low-onler approximation of the detailedphysically-basod medels to rice their computational complexity (Castles eta. 2012). In this manne, the most relevant variables are taken ato accouat inthe emulator, The variables ace identified by date-based or structure-basedapproaches. In the đao ba approach, the variables ean be selected by +tistical measure of input-output relationship, eg pari mutual information

(Bowden etal, 2008) and minimum redundancy maximum relevance (Hej and

Ca 2009). In the structure-based approach, a model formulation is derived forcach possible combination of the variable replacement by constants The madel

</div>Trang 32<div class="page_container" data-page="32">

Chapter :meloedon 3

performances are evaluated by cdteda such as residual sum of squares, Akaike’information extsion and Bayesian information edtedon (eg. Cox et al 2006 andCro ot al 2100)

This reseatch mects the above mentioned needs and builds farther on the recentadvances by the development of 4 exible river WQ model CORIWAQ(COaesetual River WVAser Quay). CORIWAQ is a hybrid of eonceptual andphysically-based models to obtain more accurate simalaions than thế tationalJumped conceptual models and shorter computational time than dctedphysically-based reference models. Accordingly, HD information for CORIWAQJs obtained from detailed physial-bated medel ‘The biochemical transformationproceses se explicily simulated in the CORIWAQ model, similar to thephysically based model ater applying comecion factors. With the data from theeusiled models, the lumped model is implemented for the motions ofdeterminant concentrations. The advection and diffusion processes along diversegments are concepmulized using the teservoiewype approach. The phyicobiochemical processes along the dt segments ate presented by a set ofsqoadone with the incoming concentstions at the Fist cross-sections and‘uteoming concentrations at the ast cross-sections of the comresponding fiversegiens, With this approach, each river segment iseharteterized by one dieseries for each hydraulic characteristic and WQ variable. This approach has been‘widely used to transform prscpiedon to runoff (Pedersen etal. 980; Te Chowal, 1988; Weller 2005; Chetan and Sudheer 2006 and Nowra tal. 2009,groundwater recharge to discharge (Peters et al. 2003) and for ver otal hếtInydeales Wolf etal. 2015, Meect et a, 2016). However, these are only fewstudies on the application af lamped conccpsal models for river WO modeling

(eg. Whitchead al, 1997; Willems and Berlamont 2002; Radian etal. 2003

2004, Willems 2008). Two detaled physical-based models, implemented in theSoftware packages, MIKE 11 and InfoWorks RS (hereafter denoted showy asRS?) with different numerical schemes and different equations to simulatebiochemical ransformations, ate selected as reference modals, This reserch isafollow-up of the inital CORIWAQ developments bascd on MIKE. 11(CORIWAQMIKEI1) by Keupers and Willems 2017), CORTVAO-MIKEIT‘wat applied to simulate the inflocnce oF combined sewer system overfiows onriver WQ (Kewpers etal 2015) and to conduct global sensitivity analysis of WQparameters (Keupers and Willams 2015).

</div>Trang 33<div class="page_container" data-page="33">

6 — Chamerirlmmndhedionnti ini dling

oth dele and simplified concepraal WQ modeling involves several ypes ofuncertainties. These wnceraimiet are pally clsified imo model structureuuncertiny, parameter uncertainties, Ínput uncertanses and measurement ero.‘These pes of uncerinty collectively determine the total model ouput

uncertainty. Uncertainties ae wo WO modeling have long been recognizedbt only few studies quantified these uncertainties, For instance, eck (1978) and‘Tung and Yen (2006) named the types oF uneeroindy as wel as the means to«qaatity and apply uncertainties in general, but they did aot discuss speciôemodels ‘The measurement uncerainty was considered by Hatmel and Smith(2007 and parameter uncertainty was addressed by Manning and Vivas (2000),“There have been not many tuớiet on the model stratute and input uncertainties

while they ate known as so dominated uncertainty sources in river WQ‘modeling (Van der Petk 1997; Hikanson 2000; Van Griensven and Meixner 2006)

Titel nde gp si

“The total model ouput unectainy can be determined by different approachescorresponding to svalable dat for model validation rest (Refigaard etal, 2006;Van Griensven and Meisner 2006). When observation data are avalable, themodel parameters cin be achieved through cabraton. After split of the daa into2 periods, one for eaibeaton and one For evalation, the differences between themod simulation ress and observed data can be analysed for the evasionpeti, og by computing the model residual variance, This total uncertansy maybe decomposed in its main contibuting unecrointy sources by varianceAlecomporiton (Radvan and Willems 2008; Freni and Manaina 20108; Willers2012}, This ie done sfer quandĐjng and propagnting the inpst and parameteruncerdiniee in the model, computing their concribusions to the total modeltour variance, and considering the rest variance asthe result of model steuecureuncertainty apart from observation error),

Mads main

‘As cxplsined in the previous section, the model sưuetus wncertsinty can becomputed as the set uncertainy after subtracting the contributions of modelinput and parameter uncerindes from the total model output unceraigy. Whenobservation dita are available, an alternative approach isto cơndider severAfferent plausible model stractrcs and anaise che differences in rests ‘This

</div>Trang 34<div class="page_container" data-page="34">

Chapter :meloedon

vwas done by Van det Perk (1997) considering E đikrenr model structures toassess the model stucture uncertainty in simulating phosphorus concentrations.Instead of using multiple medels, Lindensshmiee et al. (2007) considered thexchange dita of sub-model. The data in the sub-modsle were linked by bnetrregression equations and stochastic etor tems weee added to these equations. Incase the dreet uncctding quantification is inapplicable to insufficient mptielmaterial the exper judgement bused (Hora 1993)

The model structure wncetsnty analysis bạc been intensively studied inInydrology, bu there has been bile effort on applying this analysis eo river WQscdling, Willems (2008) and (Feeni and Mannina 20108) quansifed the otalmcertaintes in utban WQ madels and attempted to decompose these in itmajor contributing uncertain sources. The vavianee decomposition approachwas ao implemented by Radwan ct a (WM) to quantify the sroctveunceriingy ofa river WO model

dapat sanh

"The input unecrsingy is primary eased by a lack of data Bstting approaches toInanle the missing data can be divided into ceo groups depending on whetherthe study area is measured/gauged or unmeasuted/ungauged. For the measuredWO variables, they ate clasitid into tational methods and "madesn” methods(ie. apphing masimem lkebhood estimation, Bayesian estimation and multiple“mpatsion) (Enders 2010).

lsewise deletion or by filing (eq using mean imputation and repressionJmputation hive et al. 2006; Kalteh and Hjonh 2009). These methods ate‘beneficial in computational east but may cause bss in the đng data etinaton+ traonal methods deal with missing data by

‘when the missing valuet are not completely random, For instance, the missingobserved data for iver flow during storm eveets can lead to underestimationwhen mean impuradon is used, As for the “moder” methods, the maximumHhelhood estimation ủ sermines the đng data with 4 given prabsbiltystation, bur this dveibation may be subject to high secondary uncertaintywhen the availabe sample size is small (NIST SEMATECH 2012) The Bayesianmethod estimates the poweior dstnbutions for inputs ftom 4 given prioedistabuton. Multiple imputation involve combining stochastic repression andBayesian estimation, Such “modeen methods aim 1 obtain unbiased estimatesfor the fling inpot dats distributions

method but i ean derive unbiased estimates with much less computational costthan the “modem” methods

Stochastic regression is « cational

</div>Trang 35<div class="page_container" data-page="35">

8 Chapter: ction

For ungauged basins, regionalization approaches ate often applied ia hydrologicalstudies In such studies, hydrological model parimeters for ungauged basins ateobtained from parameters of gaged basins spically using regression, spatialproximity or physic! simllad, In regression approach, the relationships berwocncachmene characteristics and model parsmeters ate obtained from large data sets(ex Young 2006). The spatial proximity approach consists of tansferingpaumeters of neighboring ettchments which are simiat to the ungungedcefchment in term of climate and catchment conditions (eq Paik st al 2007),When the mouel parameters ate undetermined for neighboring catchments thepiameters can be obtained trom the donor catchments which have inlarcatchment desriptors to the ungauged earchment (eg, Melaryre ee al 2005),‘To assess the model uncersingy related to the lace of medel input dts, the

worldwide used appreach is seasivty analysis (SA) alll etal 2008). SA isrecommended in the guidelines for extended impact assessment by the EuropeanCommission (EC 2002), Based on the testing are, SA classified ito loedl andgiobal methods. In the local methods, the expo variability is achieved by

changing the input factors around reference valies Pianos ct sl. 2016). Thesensi i then quantified asthe paral derivatives (Hill and Tiedeman 2007) 0zexplored by algebra “no box” SA (Norton 2015). The local approaches involve+ very low computational cost, However, they are only appliable for Enear orảtidwe models (Sle anở Annoni 2010) and i is impossble to compass theSensivigy of the diferent inputs. Additonal, these methods are notsteaightforsard when the iopas are varable ia time, Meanwhile, global SA earconsider the ouput vuiadon across the entire space of input factors farnonlinear modch. Commonly applied global SA methods are the ElementaryEfeet Tew (EET), ot the vanance-based methods (FAST and Sabo). The inpatsconsidered in such analyse are ot only the model input but also the mode! data(Hamm ee al 2006) and their resolutions (Baoni and ‘Tarantola 2014), The‘uncertainties considered in these researches are most or he mode! parameters

(og Noseot etal. 2011; Vanustrecht etl. 2014 Posters ct al, 2014), steady input

‘tables (Hamm etal. 2006) and their resolutions (Baroni and Tatanola 2014), Aimajor drauback is thatthe global methods ate computationally expensive withhuge number of model runs

Seana arian of ie el ress impacts on sư mi

(One of the assumpdons in steeofchcam WQ modgling and one of thecontributors to the model uncertiny is the assumption thất che river bed

</div>Trang 36<div class="page_container" data-page="36">

Chapter :meloedon Ũ

‘roughness does no văn overtime, Flow caries dhe secls of aquatic plants fomup-to downstream ofthe river When the flow velocity small esough the seedsprecipitate. The seeds grow into plants with the presence oF natrints and goodtemperature, The aquatic plants influence the hydrodynamics, geochemist,geomorphology and aquatic ecology (Koch 2001; Clarke 2002; Andesson eta“3006; Canporede eta, 2013) Their density, eight and distribution control theow resistance (Van Dik t a. 2013). The higher the densty and height, theInger bed resistance is. During the year, the density and height of aquatic plansare variable and highest im midsummer. Ths, the river bed resistance has aseasonal variation corresponding to the macrophyte growth periods. When theSư bed resister increases, the Dow velocity decreases, and water Level increasesGrowth of aquatic plants its known, common source of seasonal variability inMow resistance Watson 1987; Gurnell and Midgley 1994; Barnett and Shamsedin2000; King 2011). The effect of aquatic plants on discharge-water level‘relationships is potently large in small ives and steams, creating significant‘errors in How estimates and water level simulations1 túng of single mữngcurve to estimate discharges from water levele in different periods may led to‘over or underestimation of the discharges, For WQ studies, this may ead tobiased estimates of du flow velocities, ution and ether WO processes. Wheeehigh pardculate nutrient loadings ave pretent in ives with low velocities, aquatic

plants can grow abundantly and possibly mitigate the dtrimensal effects oF thcpollution sires. Indeed a lower velocity means longer residence time and higher

sedimentation rates and thus a higher self purifying eapaciy Schul etal. 2003)Mot of seveateh on vegetated channels and sivers are monly based on intensive‘experiments and measurement campaigns (Bakry et al. 1992; Green 2006; DeDoncker etal 2011; Pham e al. 2011). However, such derdled measurement<ampiga cannot be always conducted. In many cases, the density of water level‘gauging sadon seater low auch tha seldom mote than ne station i vallelong s river reach, This gives rie tothe need for + modeling framework thácvies the widely avaible water Ievel measurements to account forthe seasonalvariation of the sivr bed toughness. This neod has ao been addressed by Aicocal, 2009, 2010), who estimated discharge and chanael roughness simultaneously‘based on water level measurements at diferent setins along che ti,

</div>Trang 37<div class="page_container" data-page="37">

4Ò — CHaperlrimmodoedonClimate chung inact em rier WO

“The impact mechanlsm of CC on tiver WO is schematized in Fg. 11, The CCimpacts on weather conditions as precipitation, air temperature and potentialcrsporranspirstion (ET) influence the water balance in the extchment. Thi leadsto changes in the river ow, and consequent changes in the determinantconcentrations slong the river through dition effect, but ako due to changes inxeaedon times and rats, When runoff lows fom catchments change, thedeterminant loads from agriculture to the river aho change. Whether the riverWQ concentrations increase or decrease depends on the eave contnbution ofthe diferent influencing factors. When the retention time increases, the reactiontime is longer and determinant concentrations can increase (eg. slttateconcentration due to nitifeain from ammonium) or deeretle (eg, organicnitrogen concentration dục to bụi). Meanwhile, the teieration rateincreases, which brings more reacnel DO into the river water

To compute the infuence of changes in precipitation and air temperature onnitrogen losses from a watershed, cản runoff and ntsogen models need to becoupled, eg; Sol & Water Tool (SWAT) (Neisch tal 2011) and IntegratedCAtchments (INCA'N) (Whitehead etl 1998). Precipitationinduces nimogen deposition from the atmosphere, Precipitation and aietemperature indiecdh influence the niogen cycle inthe silvia soil moistureand sol temperature, At the same time, the changes in ai temperature inanethe cop growing peiod and the water and nurientx wptake, Nitrogen uptake isnoel by the growth stages with potential heat units, The potential hea unitor growing degree days (GDD, heat needed for ceeps until matty (Mile ea2001) foreach crop i a function oŸ masimum, minimum and base temperatureThe base temperature Temp.) is the miainuos aưnosphede tembenture forwhich the erops ean develop. The nittgen uptake depends on the avaiableniogen in the si, the maximum uprake rate and the sowing dae for each erep(Whieehead et al. 1998}. Crop phenology is characterized by the GDD andTenp,... When the aie temperature increases, the eat increases, the length oŸgrowing period (LGP, is shortened with unchanged sowing date, In combinationwith the inceas in peepietlen, th longer the unculvated time becomes andthe higher the nitrogen losses. The nitrogen is teansparted from catchments toNitrogen Mode for

the tơ through different components of mano, Hence, CC impacts theconcentration of nioge from the catchment through both the runoff Rows andthe nitrogen losses.

</div>Trang 38<div class="page_container" data-page="38">

Caper ltmsdeedon "[Al temperatue isthe ke factor conteoling water temperature. lê tư, the waterreesperaure constin the maximum oxygen dissolved into water, namely thedscolved oxygen saturation (DOS), When the water temperature inereases, theDOS concentration decreases. Additionally, changes in water temperature affectthe physico-biochemical WO processes. The higher the water temperate, themote oxygen is elerted into the water and the faster the reaction rates. Thesechanges im physico-biochemial sates combined with the changes jn Bow,“iogen loss loads, and DO in the inflowing water (ey, at the model boundaries)result in changes of the determinant concentrations long the river

| [„2E=- =. |

= Ps

Fig. Impact mechanism of CC on river WO.

‘To simulate the CC impaces on river WAQ, the SWAT model has beea frequentlyapplied (eg, Bouraoui ct al, 2002; Van Lew etal. 2012; Giavan eta, 2015 Spa.et a 2017). The equitions and parameters in the model ate developed fomexperiments For specific ease studies and can highly influence the accuray ofthemol simulion results, The flow hyelodynamies and physico-biochemicaprocesses along rivets are ofien ignored in the applications, Most of theresearches focus on analysing. changes in monthly/yealy narient loads oraverage DO concentrations, The extreme determinant concentrations and Hoend temperate relited contol factors ate seldom considered while they mayprovide imporsne information for CC remediation,

</div>Trang 39<div class="page_container" data-page="39">

CChapkr :lgmndhedon

13. Objectives

“The overall goal of thị doctoral dissertation is co improve understanding ofprocesses that influence the physco-biochemical characterises of vers and‘heir senidtty to ehanges inthe environment (polation sources,

the following thece main objecives:

(9 Quanttaively evaluate and compute diferent detailed river WO models andincrease the insight in the dược WQ divers

Implement WQ models in dee diferent and intematioally widelyapplied sofware packages (MIKE 11, InfoWorks RS, InfalWorks ICcvaluate the mode performances and intercompate thei esas

Investignte the contfbudons of diferences inthe WO process equationsand parameters to the WQ concentrations

Determine th role ofthe HD ress

[Evaluate how seasonal vegetation changes influence these HDD resultsIntercompate the simulation capacity ofthe different software packages(8) Develop aves W conceptual model based on kaWonks RS

Bvalste the conceptual model performance

ty amass and ascssment of CC impacts on river WQ

Define the most sensitiv input factors to WQ variable concentrationsEvaluate the cotsibution of the input uncertainties in the total modelAssess CC impacts on river WQ taking the uncertainy in the Saute CCDetermine the fctos that contol thế CCimpacts on ver WQ

</div>Trang 40<div class="page_container" data-page="40">

Chapter :meloedon "

1⁄4. Overview of chapters

The doctoral diseration is divided into 4 sections. Section 1 - Chapter 1inuoduee the content ofthe research and how the researches on siver WQ havelacen done, Section IT = Chapter 2 presents characteristics of the case sti ar,Chapter 3 describes the avilable soivare packages for river WQ modeling and‘methods, Section II - Chapter 4 shows the most fundamental development bythis doctoral esearch: the conceptual river WQ aiodel with Bexible modelstructs. Chapters 5to 6 present applications of the models andl Chapters? to 8focus on applying and testing the developed conceptual model. Section IV(Chapter 9 summarizes the main findings and future research recommendationsFig. L2 shows a schemate overview ofthe diferent chapters. The following text,briefly summaries the content oF each caper:

(Chapter 1 describes the importance and current use of river WQ models and themain problems related to the application of such models, Nest, the chapterprovides the state of att ia ver WQ modeling, simulation of dhe seasonalvariation of river bed rougines, estimation of dhe uncertain inluding modelsiracture and input uncertainties in iver WQ modelling, CC and its impactChapeer 2 introduces the study atca. Fist the general characteristics of thecatchment are presented. After that, che poluant sources that contbute to theriver polufon ate given, The general characteristics of the catchment inchidesinformation on topo-geography, meteorology, land use, soil gps, hydrology“hy, aquatic plants and corresponding avaiable data. The pollutant sourcesincluding 3 types, ie. idussil, domestic households and agicalere, are alsoprovided in this chapter.

Chap 3 presents che avalable modls for river WQ and the methods togenerate input fr these model. The thưc commercial sofewate packages namelyMIKE 11, InfoWorks RS and InfoWorks ICM ase selected for shis research The‘so humped conceptual mudels namely VHA and PDM are chosen to providerinfall-anof® and flow boundaries and the semi-conceptl/semi<mpitca“model SENTWA is presented to achieve nitrogen loads from agriculture, Thischapter also depicts methods to derive dhe inputs for which data is lacking orlise

</div>