Numerical simulations of heavily polluted fine-grainedsediment remobilization using 1D, 1D+, and 2D channelschematization

Jana Kaiglová & Jakub Langhammer & Petr Jiřinec &

Bohumír Janský & Dagmar Chalupová

Received: 31 July 2014 /Accepted: 2 February 2015# Springer International Publishing Switzerland 2015

Abstract This article used various hydrodynamic andsediment transport models to analyze the potential andthe limits of different channel schematizations. Themain aim was to select and evaluate the most suitablesimulation method for fine-grained sediment remobili-zation assessment. Three types of channel schematiza-tion were selected to study the flow potential forremobilizing fine-grained sediment in artificially modi-fied channels. Schematization with a 1D cross-sectionalhorizontal plan, a 1D+ approach, splitting the riverbedinto different functional zones, and full 2D mesh,adopted in MIKE by the DHI modeling suite, wasapplied to the study. For the case study, a 55-km stretchof the Bílina River, in the Czech Republic, CentralEurope, which has been heavily polluted by the chem-ical and coal mining industry since the mid-twentiethcentury, was selected. Long-term exposure to directemissions of toxic pollutants including heavy metalsand persistent organic pollutants (POPs) resulted indeposits of pollutants in fine-grained sediments in theriverbed. Simulations, based on three hydrodynamicmodel schematizations, proved that for events not ex-ceeding the extent of the riverbed profile, the 1D sche-matization can provide comparable results to a 2D

model. The 1D+ schematization can improve accuracywhile keeping the benefits of high-speed simulation andlow requirements of input DEM data, but the method’ssuitability is limited by the channel properties.

Keywords Cohesive sediment . Remobilization .

Floods .Modeling . Pollution . Bílina River

Introduction

Fine-grained particles originating from weathering andbiological processes play an important role in aquaticecosystems. Coarse sediment is mainly derived fromchannel erosion; fine-grained sediments (<0.1 mm)can originate from arable land wash or occasionallyfrom industrial or mining activities (Medek et al.2014). Haddadchi et al. (2013) stated that 70–85 % ofsediment transported as a suspended load originatesfrom surface erosion.

Due to selective transport, fine-grained sedimentsmostly contribute to the lower reaches of the channeland the riparian zone formation and thus are exposed toheavy pollution from industrial and extensive agricul-tural production. Toxic pollutants are attached to fineparticles and so can persist for a long time in the riparianecosystem (Droppo et al. 2009; Giere 2009; Grabowskiet al. 2011). Among other pollution issues, metals areinteresting while they are bound to particles with a lowprobability of degradation (Zeng et al. 2013). Accordingto Aksoy and Kavvas (2005), only a small portion ofsediment reaches the stream and can be transported

Environ Monit Assess (2015) 187:115 DOI 10.1007/s10661-015-4339-3

J. Kaiglová : J. Langhammer (*) :B. Janský :D. ChalupováDepartment of Physical Geography and Geoecology, Facultyof Science, Charles University in Prague, Albertov 6, Prague2 128 43, Czech Republice-mail: jakub.langhammer@natur.cuni.cz

J. Kaiglová : P. JiřinecDHI a.s., Prague, Czech Republic

during the short occasional period of a high-flow event.However, once toxic-bound fine-grained particles areremobilized, they can be transported for long distances(Komar 1988) and cause secondary pollution due todesorption of the pollutant in the water environment(Segura et al. 2006).

The threshold conditions for fine-particle remobiliza-tion can involve the sand-mud ratio (van Rijn 1993),organic material content (Grabowski et al. 2011), porewater content (Winterwerp and van Kersteren 2004),or even biotic factors, such as sediment burrowingand pellatization, and can significantly affect sedi-ment resistance and settling velocity (Giere 2009).

In addition to the appropriate sediment transportdescription, the problem of accurately calculatingflow parameters plays a crucial role whilepredetermining the hydrodynamic variables used asthe basis for the sediment transport model, such asflow velocity, water depth, and shear stress(Guerrero and Lamberti 2013). Hydrological andhydrodynamic modeling covers the entire rainfallprocess: the runoff process, including the causes ofhydrological events in the form of precipitation, therunoff dynamics of the water drained within theriver system, and dynamical morphological modelsof flow channels that describe the morphologicalconsequences of hydrological events. The parame-terization of the complex modeling system is highlysensitive and, due to a lack of monitoring data, oftenmust be estimated from experience and correctedduring the calibration process.

When dealing with sediment entrainment, the flowdynamics cannot be neglected. According to Yang et al.(2005); Simpson and Castelltort (2006); and Kiat et al.(2008), numerical flow and sediment transport shouldbe simulated simultaneously, as they significantly affecteach other. Coupling the water flow with sedimenttransport in streams meets with numerous conceptualproblems, e.g., including the simultaneous effects of anumber of channel properties, while the mechanisms ofsome of them, e.g., of bank entrainment, are still notsufficiently described (Simpson and Castelltort 2006).The effect of these properties differs with the appliedlevel of applied schematization. Hardy (2013) andPapanicolaou et al. (2008) provide a comprehensivereview of current state-of-the-art sediment transport pro-cess–based modeling on different levels of flow sche-matization and the problems, related to the appliedschematization levels.

The problem of insufficient data for sediment trans-port numerical modeling is one of the key issues andwas discussed in several studies comparing the results ofempirical or conceptual approaches with physicallybased numerical modeling. Comparative studies on sed-iment transport often test available sediment transporttheories and compare and discuss the results obtained byindividual approaches (Wu et al 2008; Cao et al. 2012).Van Griensven et al. (2013) compared results of sedi-ment routing obtained by the hydrological model(SWAT) with a 1D hydrodynamic model (SOBEK-RE) and found that a physically based model did notbring any significant improvement. Hostache et al.(2014) studied the Telemac-2D suspended sedimenttransport parameter sensitivity (Strickler’s coefficient,critical Shields parameter, and settling velocity) in dif-ferent physical conditions of a small river and pointedout the influence of the spatial variability of the modelparameter distribution on the results’ accuracy. Othercomparative studies discussed the importance of a hor-izontal plan schematization (Ghostine et al. 2012; Franket al. 2012), mostly advantaging the 2D models.

Our study contributes to the fluvial sediment trans-port modeling topics by introducing the influence of thehydrodynamic horizontal plan schematization on sedi-ment entrainment modeling method selection and re-sults. Compared to the preceding studies, focusingmostly on large rivers, our study applies the selectedmodeling techniques and schematization to smallstreams, which is quite a new direction in sedimenttransport modeling.

The goals of this study were as follows: (1) to test theapplicability of different hydrodynamic model schema-tizations and the accuracy of the simulation of fine-grained sediment remobilization by using MIKE 11and MIKE 21 modeling tools, (2) to identify dischargethreshold values for remobilizing sediment in the casestudy of the Bílina River, and (3) to discuss the efficien-cy, applicability, and limits of different model schema-tizations regarding the accuracy of the results and theoften limited availability of input data for setting up andcalibrating models.

Three types of channel schematizations were selectedto study the potential for remobilizing fine-grained sed-iment in an artificially modified channel: 1D schemati-zation, 1D+ (or pseudo-2D) approach, splitting the riv-erbed into different functional zones, and full 2D mesh.Using these schematizations, the potential forremobilizing polluted cohesive sediments with initial

115 Page 2 of 18 Environ Monit Assess (2015) 187:115

flood events was tested, and thresholds for initiatingpolluted sediments remobilization were identified.

The study area was a 55-km stretch of the BílinaR i v e r i n t h e no r t hwe s t e r n r e g i on o f t h eCzech Republic, Central Europe. This area serves as amodel example of an ecosystem damaged by long-termexcessive pollution from the heavy chemical industry:Old loads, including heavymetals and persistent organicpollutants (POPs), accumulated in fine-grained cohesivesediments are remobilized during flooding.

Material and methods

Sediment transport modeling

Flood events often modify the channel and riparianzone. These changes are accompanied by high rates ofsediment outwash and transport. Predicting flood-drivenmorphological changes has been a challenge for engi-neers for the past few decades. Sediment transport for-mulas are mainly based on flume experiments and re-quire extrapolation beyond the original data (e.g.,Komar 1988; van Rijn 1993). Thus, the applicabilityof various theoretical models in real cases has beendiscussed in numerous studies (e.g., Galappatti andVreugdenhill 1985; Mosselman 1992; Büttner et al.2006; Wu et al. 2008; Hardy 2013) regarding the nu-merical instabilities, lack of verification data, and theempiric base of the process description.

The sediment transport process can be divided intothree phases for simplification, even though the processis continuous: (1) initialization, comprising the erosionand remobilization process, (2) transport, and (3) accu-mulation. Sediment is transported within the river net-work, according to the flow variables, sediment proper-ties, and modes of transport.

In our study, only phase 1 (initialization of particlemovement) was considered and simulated, since it ex-presses critical conditions for triggering particle transportand accumulation, resulting in secondary contamination.

As for the initiation of particle movement, currentstudies usually focus on the relation between flow velocityand particle movement (e.g., Huang et al. 2006;Grabowski et al. 2011; Hardy 2013). The concept ofdisturbing forces (excess shear stress τe) affecting theparticles once the shear stress τ (N m−2) or shear velocityu (m s−1) exceeds the threshold value of particle resistance(gravity, cohesion) is frequently applied. This concept

requires a critical shear stress τc definition that integratesthe threshold cohesive, adhesive, frictional, and gravita-tional forces. The critical velocity uc or τc was first intro-duced by Shields (1936), while erosion power and coeffi-cient were derived from empirical flume observations.The intuitive empirical concepts (e.g., Hujlström,Postma, or Shields curves), Bangold’s (1966) stream pow-er, flow competence, or transport capacity model (Yanget al. 2005), and derived theoretical threshold values for τchave reduced meaning in cohesive sediments while omit-ting the inter-particle electro-chemical cohesive forces andmust be tested in case studies and confirmed with data(Grabowski et al. 2011). Cohesive clay particles do notbehave individually but tend to aggregate into flocks ofhigher resistance and settling velocity (Houwing and vanRijn 1998). The resistance to erosion is even strengthenedby the fine-particle bed decreased permeability describedby Wu et al. (2008), and entrainment thus is the result ofadverse lifting and settling forces.

According to Giere (2009), particle-bed substancesare slowly released into water columns. Thus, the studyis based on the hypothesis of toxic-bound particlestransported in the form of wash load for long distancesand deposited further downstream. Wu et al. (2008)rejected this statement and thought all particles contrib-ute to updating channel morphology.

Tested sediment transport models

The aim of this study was to find appropriate thresholdconditions for fine-grained sediment remobilizationbased on 1D channel schematization. To select the mostappropriate approach in terms of data and solution timerequirements, various approaches were tested. First, thesediment transport modules of MIKE by DHI were usedfor the 1D and 2D hydrodynamic descriptions. In thecase of cohesive sediments, the advection-dispersionformula (1) was used (MIKE 21 MT, MIKE 11ADCST). Erosion and deposition were describedwith the source/sink term (DHI 2012):

∂c∂t

þ u∂c∂x

þ v∂c∂y

¼ 1

h

∂∂x

hDx∂c∂x

� �þ 1

h

∂∂y

hDy∂c∂y

� �þ QLCL

1

h−S;

ð1Þwhere c is the depth average mass concentration, uand v are the depth average flow velocities, Dx andDy are the dispersion coefficients, h is the waterdepth, S is the deposition/erosion term, QL is the

Environ Monit Assess (2015) 187:115 Page 3 of 18 115

source discharge per unit horizontal area, and CL isthe source discharge concentration.

To investigate sediment remobilization, time-varyingshear stress maps were calculated from the 1D hydro-dynamic simulation result evaluation and discussed withthe other approaches. The threshold remobilization val-ue was found, according to the τc concepts and experi-ence with previous methods. Comparing these resultswith the 2D results should prove whether an appropriatesediment remobilization moment can be found withoutmodeling sediment transport modeling. Due to the prob-lematic calibration of the sediment transport model, asensitivity analysis of the main parameters of the sedi-ment transport equations was performed. The analysiswas based on the parameter variation across the valueinterval found in the literature, and the maximal expect-ed error was estimated.

To estimate the relevant time of remobilization sup-posing real environmental hazards, the objective remo-bilization criterion was adopted as the time at which theshear stress nearly reaches 1 N m−2, and consequently, asharp increase was used in the simulations based on theshear stress evaluation. The sediment transport simula-tions were evaluated according to the criteria defined asthe time step at which the bed level change equals−1 mm, and in the following time steps, the bed de-grades at the rate of at least 1 mm h−1.

Study area

The study area was the middle and lower reach of theBílina River (Fig. 1 and Table 1). The Bílina River isloca ted in the nor thwes te rn reg ion of theCzech Republic, where the river drains from an indus-trial area dominated by open-cast mining and the heavychemical industry. Because of long-term excessive pol-lution, the Bílina River is ranked among the most pol-luted rivers in the Czech Republic as well as in Europe(Přibylová et al. 2006; Langhammer 2010). Althoughexcessive industrial production has been restricted inmany areas since the 1990s, in the study zone, largeheavy chemical industrial sites are concentrated, withpetrochemical production, refineries, and agro-chemicalproducts.

The channel of the Bílina River was excessivelymodified in the past, while part of the river, crossingthe open-cast mining area, was converted to a pipeline.The present river channel in the assessed stream stretchhas been modified at 100 % of its length (Matoušková

and Dvořák 2011). The channel has been straightenedand dragged, and most of the segments have been con-verted to a prismatic shape with artificial banks. In theriver, numerous structures affect the flow dynamics.

The water and sediment pollution of the Bílina Riverhave a significant trans-boundary impact. The river is aleft-side tributary of the Elbe River at ÚstínadLabem, near the Czech-German border and signif-icantly contributes to the steep increase in pollu-tion levels of the Elbe River at its downstreamsegment (Langhammer 2007).

The catchment area of 935 km2 is characterizedby a fast runoff response to excessive rainfall. Theflood wave velocity ranged from 3.7 km h−1duringthe spring 2006 flooding with the return periodQ2–Q5 to 5.5 km h−1 recorded during the autumn2010 flooding with the estimated return period ofQ5–Q10. The recently observed flood events (since2002) with a return period up to Q10 were ana-lyzed to estimate a synthetic hydrograph with themost probable shape of the rising part (Fig. 2)consequently. Five flood waves of different mag-nitude and season were normalized at the time (x)axis to detect overlapping properties of the risingparts (before the culmination), chiefly the steep-ness (discharge increment per hour) and durationof definable phases. Three phases were observed atthe rising parts: (1) the slight initial rise, (2)followed by the main sharp rise, and (3) sloweddown again to gain the culmination. The steepnessand the duration of each phase were calculated forfive flood waves. The synthetic hydrograph waslater composed of the averaged properties of thosethree phases. This approach ensures the contribu-tion of all observed high-flow situations and non-predetermined final magnitude in any time step.The synthetic hydrograph was used to assess thesediment remobilization. The discharge value atthe remobilization time step was registered andevaluated in the form of the return period.Flooding events before 2002 were not used be-cause the course of the Bílina River has beensignificantly modified since this time (Jurajdaet al. 2010).

Sediment characteristics

The study involved fine-grained sediment layers andthe risk of remobilization of deposits that are heavily

115 Page 4 of 18 Environ Monit Assess (2015) 187:115

polluted with toxic and environmentally persistentmatter (Table 2). Data on the sediment samples werecollected during the field survey. Altogether, 24 sam-ples characterizing the nine most affected study siteswere collected and analyzed for granulometry andgeochemistry (Medek et al. 2014). The average grainsize of the bed differed dramatically and ranged fromsand to grave l . The channel modi f ica t ionspredetermined the fine-grained sediment to be hiddenat the berms or banks not exposed to the average flowconditions or in the calm backwaters. Thus, nine sites(B1–B9) with fine-grained sediment were selected,and the flow conditions were evaluated with numeri-cal hydrodynamic modeling (Fig. 1). The other siteswere integrated only to accurately describe the flowtransformation.

Due to the absence of other sediment character-istics, cohesivity was estimated from the grain sizeonly as a percentage of the silt and clay fractions.

According to Breusers (1986), only the finest sed-iments (9, 8, and 5) are expected to be partlyflocculated and cohesive. Nevertheless, in all sam-ples, a significant percentage of clay particles werefound; thus, the cohesivity throughout the studyarea was considered.

According to Amos et al. (2004), the empiricalrelationship between cohesive sediment critical shearstress τc and bulk density ρb results in τc=0.59–0.81for the Bílina sediment. For comparison purposes, theShields τc was calculated according to Wilcock(2004) (Table 2), even though it is proposed forcoarser fractions.

Although sediment with unimodal distribution ofgrain size can be described with the parameter D50

(grain size median), the behavior of the bimodal sedi-ment mixtures depends on the distance between the twopeaks on the granulometric curve (Wilcock 1993). Thethreshold for erosion in case of bimodal sediment

Fig. 1 Study area—modeled stretch of the Bílina River with selected areas of fine-grained sediment deposits

Table 1 Bílina River and catchment characteristics

Area Length Whole/studied reach

Annualprecipitation

Averagebottom slope

Qa Trmice (neardownstreamboundary)

Q10 Trmice Channelcapacitydischarge

C5 peakin TrmiceAugust 2009

Averagebed material

(km2) (km) (mm) ( %) (m3/s) (m3/s) QN (mg/l)

1071 81.9/55 634 0.17 6.5 37.2 10 418 Sand, gravel

C5 is the suspended sediment concentration recorded during flooding during 5/2009 with the return period year (data: Povodí Ohře 2010;CHMI 2013; Šípek et al. 2010)

Environ Monit Assess (2015) 187:115 Page 5 of 18 115

mixtures should be evaluated separately (Fig. 3), as theD50 parameter for those samples is not representative(e.g., Komar 1988; Wilcock 1993).

The clay and silt sediment embedded in the sidechannels, weir backwaters, and riparian water-courses, according to a recent survey (Medeket al. 2014), is heavily polluted by a wide rangeof toxic matter. In the sediment samples, elevatedconcentrations of persistent organic pollutants(PCB, DDT, HCH, PAHs, etc.), heavy metals,and others projecting into the downstream waterquality at the Elbe River were found (Table 3).The pollution in the sediment comes from theunregulated heavy industrial production during the

second half of the twentieth century that still con-tributes significantly to the Elbe River pollution.

Hydrodynamic model setup

To study sediment, three hydrodynamic transportmodels that solve the shallow water equations were setup with different levels of schematization: (1) the gen-eral model, using 1D channel schematization, coveringthe entire assessed stream stretch, (2) a 1D+ model,stemming from the 1D model while splitting the streamat suitable reaches into the functional parts, and (3) a 2Dmodel of two selected sites.

Fig. 2 Recently observed flood events at theMost gauging stationof the magnitude between Q5 and Q20 and resulting in a synthetichydrograph, culminating at magnitude Q10, used for the

remobilization assessment. Note: phase (1) is delineated bystraight gray line, phase (2) by dashed line, and phase (3) bydotted line

Table 2 Sediment basic characteristics

SITE Depth ofsediment layer

Length D50 CLAY SILT Character (AGU) Distribution Calculated τcShields curve

(m) (m) (μm) (%) (%) (N/m2)

B 1 0.05 500 32 18.08 48.9 Coarse silt Unimodal 0.05

B 2 0.20 500 95 8.10 28.5 Very fine sand Unimodal 0.03

B 3 0.05 940 80 10.17 31.2 Very fine sand Unimodal 0.03

B 4 0.15 430 9 34.39 47.7 Fine silt Bimodal 0.08

B 5 0.10 630 5 42.55 57.5 Very fine silt Unimodal 0.11

B 6 0.50 2100 103 10.53 23.9 Very fine sand Unimodal 0.03

B 7 0.20 240 36 18.89 47.0 Coarse silt Bimodal 0.05

B 8 0.15 500 3.4 57.51 42.5 Coarse clay Unimodal 0.13

B 9 0.15 560 3.4 57.51 42.5 Coarse clay Unimodal 0.13

The sediment deposition sites are sorted from upstream to downstream reaches (data: Medek et al. 2014)

115 Page 6 of 18 Environ Monit Assess (2015) 187:115

1D hydrodynamic model

The hydrodynamic model in general provides informa-tion about flow characteristics. In this case, it providesinformation on discharge, cross-sectional velocity, andwater depth for every computational point. Due to a lackof detailed channel bed morphology data, the one-dimensional schematization approach was chosen. Thehydrodynamic model of the Bílina River course was setup in MIKE 11 and consisted of 597 cross sectionsresulting from geodetical field survey (Povodí Ohře2010) that describe the channel characteristics in a spa-tial resolution of 100 m on average based on the localconditions. The database was constructed in order tomost accurately represent all the variations in the chan-nel longitudinal and cross-sectional patterns. The crosssections were transformed from previous studies ofBílina River flood risk mapping (Povodí Ohře 2010).The cross-sectional database was modified according tothe aims of the schematization. Cross-sections were splitin order to describe the channel marked by banks only.The channel capacity in the most of the reaches wassufficient for the Q10 value. Therefore, only the channelwas schematized, and the flood plain was neglectedwhile the expert assumption of the remobilization causaldischarge below Q10 was considered. The river network

consisted of three branches (main branch, left, andright divisions of flow in Trmice–lower Bílina).Weirs were schematized as structures where theenergy loss is calculated using the broad-crestedweir formula. Three bridges with limited capacitywere schematized as structures with culvert energyloss calculations. Another 90 bridges have a ca-pacity larger than Q10, and the energy loss causedby the flow contraction was schematized by usingthe compensatory roughness increase in the nearby(±20 m) cross sections. Such schematization resultsin flow energy loss, and thus, the water levelincreases above the structure, which was verifiedwith field observations.

The decisive calibration parameter for sedimenttransport studies is bed roughness (e.g., Guerrero andLamberti 2013). The model was calibrated by using datafrom two independent events with peak discharge belowor around Q10, within the channel capacity. The calibra-tion criterion was the time shift of the flood wave(Fig. 3) and the total fit.

Fine-grained sediment was expected at the bankstructures or in calm weir (bridge) backwaters. Thecross-sectional average flow description did not providesufficient information about the non-exposed parts ofthe channel. Thus, three improvement methods wereelaborated and tested in order to refine the flow descrip-tion along the cross section.

1D+ hydrodynamic model

The 1D+ model is based on splitting the riverbed profileinto zones where different fluvial processes are expect-ed. The suitable reaches of the Bílina River were dividedinto the main flow and side flow parts (Fig. 4) in order todistinguish the flow parameters at the berms not

Fig. 3 Example of granulometric curves of surveyed sediment layers of Bílina River. From left side: cohesive unimodal (B1), uni-modalnon-cohesive (B3), bimodal non-cohesive (B4). Data: Medek et al. 2014

Table 3 Exceedance of the target values of the most hazardouspersistent particle-bed substances in the monitoring profile Děčín(downstream from the confluence with the Bílina River)

Substance Cu Zn Cd Hg DDT PBC HCB

Exceedance of thetarget value

<2× <2× 2–4× >4× >4× >4× >4×

Adapted from Heise et al. (2008)

Environ Monit Assess (2015) 187:115 Page 7 of 18 115

exposed to the main stream power. The side riverbranches were interconnected in user-specified H pointswhere the water level was calculated and tended toequalize with the main channel. This approach reducedthe mean shear stress, further affecting the sedimentdeposits.

Shear stress maps based on 1D results

The interpolation routine of the cross-sectional distrib-uted 1D result data was used to refine the flow param-eters. The hourly mean flow characteristics, velocity anddepth, resulting from the 1D channel model were dis-tributed in the cross-sectional and longitudinal direc-tions. The flow velocity across the cross section wasestimated according to the conveyance distribution (2)(DHI 2012):

K ¼ ∫B0h5=3Mdx

m3

s

� �ð2Þ

where K is the conveyance, h is the local depth varyingacross the cross sections, B is the width of the watersurface, and M is Manning’s roughness coefficient.Once the cross section is distributed, the local velocity(average vertical value) can be defined according toEq. (3):

v x; tð Þ ¼ Q tð ÞKð ÞMh2=3

m

s

h ið3Þ

After approximation (2) and (3), the local velocityand depth were estimated and then linearly interpolatedto represent a pseudo-two-dimensional field. The series

of depth and velocity fields were used to manuallycalculate the shear stress (4):

τ ¼ ρghIRN

m2

� �Where IR ¼ v2

M 2h4=3

−½ � ð4Þ

where ρ is the water density, h is the depth, and IR is thedimensionless slope.

The resulting time-varying map of shear stress wasexplored to find areas with potential sediment depositthat were not exposed to the excess shear stress duringaverage flow conditions.

The areas with fine-grained sediment layers wereselected as places where the calculated shear stressduring normal flow conditions does not exceed0.2 N m−2. Those sites are mostly placed on the riverbanks not exposed to the normal flow conditions or inthe backwaters of the structure (Fig. 5).

2D hydrodynamic model setup

The most realistic schematization was obtained by de-scribing the channel properties with two-dimensionalcomputational mesh (e.g., Cook and Merwade 2009).Every mesh element contains information about eleva-tion, roughness, viscosity, grain size of the bed sedi-ments, and the thickness of the deposit layer. The scopeof this study did not allow the full solution based on the2D hydrodynamic description, as a long and narrowchannel would require a tremendous number of meshelements, and thus a dramatic increase in computationtime. The method was used in the experimental reach,merging two sites of sediment deposits B9 and B8. Theselection was based on the suitable channel

Fig. 4 Calibration and verification of the hydrodynamic model based on two flood events

115 Page 8 of 18 Environ Monit Assess (2015) 187:115

characteristic, including three structures of various ca-pacities and proximity to the upper boundary gaugingstation. Sediments were observed in the calm waters andthe bank zones; thus, the reach was representative.

A real 2D channel schematization was not availabledue to a lack of real bathymetry data. Therefore, thebathymetry was created by linear interpolating of crosssections with the river axis curvature and river bankwidth. The schematization was based on the orthogonalgrid (MIKE 21) with 1-m cell resolution. The smoothgrid consisting of approximately 450,000 cells enabledfast simulations while keeping the time step of 1 s. Theviscosity was calculated according to the Smagorinskyformula, dynamically varying with the flow velocity.The resistance represented by Manning’s M=40 wasreduced for the sediment, according to the estimatedNikuradse resistance length kt and set uniformly alongthe whole channel while the exact distribution could notbe known at the 1D bathymetry basis. To avoid numer-ical instabilities, flooding and drying depths of 0.05and 0.03 m, respectively, were used. The initialconditions were calculated in every cell based onthe mean annual flow conditions. The shear stressmap of the site B8-B9 resulting from recalculatingthe 1D interpolated velocity, depth, and roughnessmaps is shown in Fig. 6.

Sediment transport parameter sensitivity analysis

A sensitivity analysis of sediment transport parameterswas performed in the 2D model at experimental reachB9–B8. The 2D schematization was used to distinguishthe flow parameters in the cross-sectional direction anddefine the erodible layer (Fig. 7).

Due to the lack of data on sediment characteristics,the model parameters were tested within the expected

intervals, and the variation in the results was evaluatedto predict possible errors due to insufficient data(Table 4). Results of individual simulations were ana-lyzed at the five sites (Fig. 7) as the average values ofnumerous sub-sites.

Using the results of the sensitivity analysis, anexpected range of error due to insufficient datawithin 1.0–19.1 % for τ and 2.7–18.3 % for re-mobilization causal discharge was estimated. Themost sensitive parameters were kt, τc, and erosionpower, as it exponentially determines the speed oferosion and shifts the remobilization time. Thesettling velocity predetermines, according to Costaand Coelho (2011), the concentration profile andhas a positive effect on the erosion rate. Thus, theerosion rate slows down and the remobilizationtime results in shifts in later discharges and higherbed shear stress values. The lower erosion rateafter higher τc of erosion is caused by the de-creased excess shear stress value. Bed shear stressdecreases as the bed roughness decreases; thus, theshift in the remobilization time due to the de-creased value of the excess stress was obvious.The largest differences were observed in the ero-sion rate that, according to Houwing and van Rijn(1998), is a logarithmic function of excess shears t r e s s . The va lue s va r i ed f rom 1e − 5 t o8e−4 kg m−2 s−1. According to van Rijn (1993),the erosion rate varies significantly within severalorders of magnitude from 1e−5 to 5e−4 kg m−2 s−1.The range proposed by Huang et al. (2006) waseven larger, from 1.3e−7 to 5.8e−5 kg m−2 s−1.

The 2D schematization of the nearly prismatic chan-nel without dominant sinuosity was further evaluated asunnecessary and thus not used to evaluate the otherreaches.

Fig. 5 Principle of 1D+ schematization. Left is the main flow; right is the side where the sediment layers are expected

Environ Monit Assess (2015) 187:115 Page 9 of 18 115

Results

1D and 1D+ schematization

The simulation was based on the 1D channel schemati-zation for the fully dynamic, unsteady flow calculationand coupled with the cohesive sediment transportadvection-dispersion; the simulation was not successfuldue to the sediment properties and location. The mainnuisance was the non-uniformity in the sediment layerplacement, being located mostly near the banks, and thesignificant discrepancy between the grain properties ofthe cross-sectional average and the examined sediment.Thus, the cross-sectional averaged flow characteristicswere not applicable for the entrainment detection, due tothe overestimation of the local shear stress conditions.Even after considering the limits of sediment transportparameters (Table 4), the defined sediment layers erodedat normal flow (average discharge Qa) conditions atmost of the sites. At B9, B8, B7, and B1, the whole sitewas remobilized at Qa, while at B6 (94 %), B5 (64 %),B4 (39 %), B3 (54 %), and B2 (60 %), a substantial part

of the reach was eroded. It was caused by applyingcross-sectional (CS) average flow parameters for sedi-ment transport of non-uniformly stored sediments,which were only in some parts of the cross section thatwere not exposed to the main flow. The CS averageshear stress mostly varying within the interval of 3.6–5 N m−2 highly exceeded the threshold shear stress forcohesive sediment remobilization (Table 4). In the fullyprismatic channel, the CS average values for shear stressreached or exceeded 10 N m−2. Thus, the CS averagemethod had to be ignored in additional evaluation.

The shear stress obtained with the 1D+ approach,compared to the CS average value, decreased to 27 %at the B9 site, 30% at the B5 site, and 44% at the B8 site(Table 4). Nevertheless, the shear stress at the remobili-zation criteria time gained from the shear stress mapevaluation and the 2D sediment transport simulationexceeded 1 N m−2 at higher discharges; thus, the remo-bilization obtained with 1D+ should be evaluated inearlier time steps.

An example of a suitable locality is the reach B8-1(Fig. 8) with the well-developed step in the right bank.

Fig. 6 The shear stress map of the site B8–B9 resulting from recalculating the 1D interpolated velocity, depth, and roughness maps

Fig. 7 Definition of the erodible layer based on the normal flow condition evaluation accompanied by information obtained in a field survey(dark gray area). Rectanglesmark the focused sub-sites and black crosses the points for bed-level change and erosion parameter extraction

115 Page 10 of 18 Environ Monit Assess (2015) 187:115

The remobilization criteria were fulfilled at thecausal discharge of 4.2 m3s−1, corresponding tothe Q5. Nevertheless, remobilization occurred inone CS, only while in the overall left bank zoneof B8-1, and the remobilization criteria are ful-filled within the time interval from 43 to 60 h of

the synthetic hydrograph, corresponding to dis-charge values of 4.2–7.3 m3s−1 and local velocitiesreaching 0.2–0.4 m s−1. The method was not fur-ther used or developed because of the suitability inselected reaches: only where the channel is dividedinto significantly various parts.

Table 4 MIKE 21 MT sediment transport parameters, recommended value interval, and observed variance of results collected at 16 pointsselected within the erodible layer sub-sites (B9-x and B8-x) (Houwing and van Rijn 1998; van Rijn 1993; Grabowski et al. 2011; DHI 2012)

Parameter Tested range Influence on the results %Δ median±Δ St. dev (maximal expected error) Feedback reaction

Units Min Max Q (m3 s−1) atremobilization

τ (N m−2) at remobilization E (kg m−2 s−1) at remobilization

τc for erosion N m−2 0.6 0.8 3.7±6.8 (7.1)Positive

4.6±9.75 (14.35)Positive

45±57 (73.5)Negative

τc for deposition N m−2 0.05 0.1 2.4±0.6 (2.7)Positive

0.5±1 (1)Positive

1±3 (2.5)Positive

Settling velocity m s−1 0.0001 0.0005 2.5±12 (8.5)Negative

1.7±8.6 (6)Negative

90±40 (110)Positive

Power of erosion – 1 4.3 6±24.5 (18.3)Negative

2.1±18 (11,1)Negative

25±77 (63)Positive

Erosion coefficient – 0.00005 0.0001 1.2±11.9 (7.15)Negative

1.9±8.7 (6.25)Negative

88±40 (108)Positive

Bed Roughness kt m 0.05 0.07 2.4±4 (4.4)Negative

11.6±14.9 (19.1)Positive

28±30 (43)Positive

Dry bulk density kg m−3 600 800 2.4±13 (8.9)Positive

0.4±7.2 (4)Positive

1±23 (12.5)Negative

Fig. 8 Example of results for the 1D+ sediment transport simula-tion for the reach B8-1. On the left is the time series for shear stressin the reach B8-1 plotted for the main channel branch and the side

bank branch for the same cross section. On the right is thesediment layer evolution in time

Environ Monit Assess (2015) 187:115 Page 11 of 18 115

The CS average flow parameters were later distrib-uted along the y and x directions, and the shear stressmaps were processed. The process parameterization wasadjusted with the 2D sensitivity analysis results(Table 4) by defining the limiting value of the shearstress for the objective remobilization criteria. The eval-uation of the shear stress maps regarding the remobili-zation criteria resulted in discharge values ranging from3.6 m3s−1 at the uppermost B9 site to 24 m3s−1 at the B3site near the confluence with the left tributary Bouřlivec(Table 5). Although suitable for all sites, shear stressmaps were used for the remobilization criteria fulfill-ment analysis.

The causal discharges remain within the channelcapacity that, for the whole reach, suffices the Q10 flowconditions and varies with the longitudinal point ordiffusive flow increase.

The local current speed after the CS distribution(1D CS Distributed) drops below 0.5 m s−1, butmostly did not exceed 0.2 m s−1. The 1D CSaverage velocity is up to 0.8 m s−1, generally threeor four times higher than the 1D CS distributedvalues at the remobilization. Such variation ismultiplied by the water depth distribution, andthe ratio between shear stress distributed versusaverage reaches up to 0.7/12 in the nearly prismat-ic channel at the B3 site, while gaining in anaverage decrease to 20 % regarding the 1D CSdistributed values and 39 % compared to the CSaverage shear stress with 1D+ flow separation. The2D model results mostly correspond to the 1D CS

distributed values. The shear stress decreased 22 %with 2D schematization. The difference betweenthe flow velocity resulting from the CS averagevalue distribution and the real 2D simulation re-sults from the non-dynamics consideration in theconveyance-based approach.

2D model for the upper Bílina

The fully two-dimensional schematization resulted in aflow distribution and description in both directionsalong the cross section and the river axis. The remobi-lization criteria were fulfilled in the most of the bankstructure sites at discharge conditions ranging from 5.8to 8 m3 s−1, corresponding to the Q6–Q7, approximatelyin the interval between 36 and 74 h of the synthetichydrograph (Fig. 9). The shear stress values atremobilization ranged from 0.85 N m−2 to1.22 N m−2. The site upstream from the collapsedbridge (B9-3) differed due to increased exposure tothe transitional stream power caused by thecontracted flow. Thus, lower causal dischargevalues of 3.8 m3 s−1 were observed and evaluatedat approximately Q5, while the shear stress therewas observed ranging from 0.76 to 1.13 N m−2.Higher shear stress values were observed at thesites nearest the transitional flow, where the shearstress increase was steeper over time. The depthaverage velocity at remobilization ranged from0.27 to 0.45 m s−1; the most frequent value was0.38 m s−1.

Table 5 Flow conditions at the remobilization criteria derivedfrom the shear stress map evaluation in Bílina sites based on 1DCS distributed parameters, 1D CS average, 1D+ (extracted from

the side branch at the berm site if available), derived from the 2Dmodel if available

Site ID Chainage Local discharge Local current speed Local shear stress

Units km m3 −1 1D CS distributed 1D CS average 1D+ 2D 1D CS distributed 1D CS average 1D+ 2Dm s−1 N m−2

B1 0.3 16.0 0.2 0.3 – – 1.0 2.0 – –

B2 0.8 5.0 0.1 0.4 – – 0.6 3.0 – –

B3 10.1 24.0 0.2 0.8 – – 0.7 12.0 – –

B4 17.8 15.5 0.1 0.4 0.3 – 0.5 2.5 1.3 –

B5 29.8 20.0 0.1 0.5 0.3 – 0.5 5.0 1.5 –

B6 47.1 5.0 0.2 0.6 – – 0.8 11.0 – –

B7 51.1 4.5 0.5 0.6 – – 0.7 9.0 – –

B8 54.2 7.0 0.1 0.4 0.4 0.3 0.8 3.6 1.6 0.8

B9 54.8 3.6 0.1 0.4 0.3 0.2 0.7 3.6 1.0 0.9

115 Page 12 of 18 Environ Monit Assess (2015) 187:115

Comparison of results obtained with differentapproaches in the upper Bílina reach

The remobilization criteria fulfillment results from the2D sediment transport model of experimental reachesB8–B9 were compared with the results derived from the1D+ separate with sediment transport included andshear stress maps based on the 1D CS distributed valuesfor the same site in order to estimate the differences(Table 6).

The simple approach of the cross-sectional averageresult distribution did not differ dramatically from the2D true sediment transport calculation. The maximumobserved difference of 13 % in the shear stress valuemisinterpretation results in 9 % of the incorrect causaldischarge assessment, and for the B8-2 site, it wouldchange the final statistical interpretation from the floodfrom 5 years to approximately 5.3-year return period. Atthe B9-3 site, situated downstream from the collapsed

bridge with specific flow conditions, the evaluation ofthe shear stress map was not successful. Nevertheless,the whole reaches of B8 and B9, respectively, wereevaluated according to the lowest observed causal dis-charge values (B9-2; B8-2); thus, the resulting causaldischarge would be the same in both approaches. Theresults of the remobilization assessment based on the1D+ schematization with cohesive sediment transportdiffered mostly at the B8-2, with 40 % from 1D CSdistributed and 38 % difference from 2D regarding thecausal discharge evaluation, and B9-1, with 48 % dif-ference from 1D CS distributed and 28 % differencefrom 2D regarding the causal discharge evaluation, siteswhere the largest bed level differences (0.8 m) in thelongitudinal profile of the side functional zone wereobserved. The B9-3 site was evaluated similarly withthe 1D+ and 2D approaches, while the 1D CS distrib-uted approach brought the earlier remobilization assess-ment. In the B8-1 and B9-2 sites, the differences in the

Fig. 9 Bed-level change at the observed sites (B9–B8). The focused reach was subdivided into sites with possible sediment (e.g., B9-1).Each site is represented by two to three cells where the point time series of the bed level was extracted

Table 6 Comparison of results for the 1D CS distributed flow shear stress maps, 1D+ with sediment transport calculation, and 2D sedimenttransport calculation

Site ID Site description τ (N m−2) Q (m3 s−1)

1D CSdistributed

1D+ 2D 1D CSdistributed

1D+ 2D

B9-1 Downstream the weir 0.8 1.3 1.0 5.0 7.4 5.8

B9-2 Upstream the collapsed bridge 0.8 1.5 0.8 3.6 4.0 3.8

B9-3 Downstream the collapsed bridge 1.0 1.3 1.1 4.5 7.9 7.9

B8-1 Middle reach banks 1.0 1.1 1.0 7.0 8.1 8.0

B8-2 Lower reach banks 0.9 0.8 0.8 7.0 4.2 6.8

Environ Monit Assess (2015) 187:115 Page 13 of 18 115

causal discharge evaluation of the results for the 1D+approach differed by 11 and 15 %, respectively, whilecompared with the 1D CS distributed and 6 and 2 %,respectively, from the 2D causal discharge evaluation.The shear stress during remobilization was higher, at 1.3and 1.5 N m−2, than at the B9 site, and comparable, at0.8 and 1.1 N m−2, at the B8 site.

Discussion

In situations where descriptive data is scarce, an increas-ingly complex hydrodynamic model does not necessar-ily lead to better results (e.g., van Griensven et al. 2013).In special cases, where there are long, prismatic chan-nels represented and detailed topography data for thefloodplain is missing, a 1D model can sometimes per-form as well as a 2D model (Cook and Merwade 2009;Horritt and Bates 2002). Caviedes-Voullième et al.(2014) concludes that if the cross sections for a hydro-dynamic model are well located, the channel bed can bedescribed sufficiently, as in the case of detailed 2Ddescription. The bathymetric data available for thisstudy requires the undemanding 1D schematization offlow parameters as the standard praxis tool for floodmapping tasks (Cook and Merwade 2009) and sedimenttransport (Fang et al. 2008). The other reason forselecting 1D schematization was the presence of numer-ous structures that can be easily set up within MIKE 11modeling tool. The flow conditions were calculatedwithdifferent approaches in order to understand the possibleinaccuracies of each result. The pure application of 1Dchannel schematization was not successful, due to thenon-uniformly located sediment layers and the D50 be-ing much smaller than the average sandy-gravel channelmaterial. As expected, sediment started to erode imme-diately after overflowing as a result of high velocity andbed shear stress.

The separation of the flow into functional zoneswithin the channel (1D+) led to a substantial decreasein bed shear stress. Therefore, at sites where the channeldiffered from the pure prismatic (B4, B5, B8, B9), theremobilization assessment was successful. Other siteswith fully prismatic, artificial channel geometry did notsupport this type of schematization at all for mainly tworeasons: (1) no real functional zone diversion in reality,and (2) the link channel connecting two parallelbranches must be elevated to maintain the numerical

stability. The 1D+ schematization was not suitable forthe entire reach and thus had to be rejected.

The distribution of the flow parameters along andacross the channel by the conveyance module (1D CSdistributed) was reasonable, and fine-grained sedimentremobilization was evaluated at all sites since the bedshear stress was 0.5–1 N m−2.

The 2D simulation of the channel flow parametersrequired a 1-m cell resolution in order to get a gooddescription of the bed topography and channel capacity.That was caused by the channel width varying from10 m in the upper course to 20 m in the lower courseat the confluence with the Elbe. Thus, the computationaltime for the whole reach schematized within one 2Dmodel was unacceptable. The 2D schematization wasrejected because the results for the comparison of the 2Dresults and the 1D CS distributed and 1D+ results didnot differ dramatically.

Nevertheless, the results of the sensitivity analysisperformed with the 2D model served for the processparameterization within all tested approaches. Despitecomparative studies that were testing the importance ofthe horizontal plan schematization on the flood eventmodeling (Frank et al. 2012; Ghostine et al. 2012), andmostly advantaging the 2D models, we conclude thatthe increment of model complexity did not bring asignificant improvement. It is caused by the nature ofour study and the simplification of the sediment trans-port task to the stream entrainment competence.Another important reason is the reach characteristics.Thus, our findings are limited, especially for nearlyprismatic reaches and for the channel capacity flowconditions.

A sensitivity analysis was performed for all crucialparameters in order to realize the possible errors due tothe lack of data on sediment properties. The expectedmaximal range of error was 19 % in the τc evaluationand 18.3 % for the assessment of remobilization causaldischarge. Cohesive sediment erodibility cannot be ac-curately predicted, due to the strong inter-particle attrac-tions of cohesive sediment (Winterwerp and vanKersteren 2004; Grabowski et al. 2011). The criticalshear stress for cohesive sediments can be higher than1 N m−2 (van Rijn 1993) and even 2 N m−2 (Ariathuraiand Arulanandan 1978). Nevertheless, the values arevalid for calm backwaters without added sand sedimentsthat would increase erodibility. In the case of the Bílinasediment, which is often mixed with sand fractions, adecreased value of 0.8–0.9 N m−2 should be expected.

115 Page 14 of 18 Environ Monit Assess (2015) 187:115

The critical shear stress of bed material caused by higherdensity in lower layers increases with the depth of thesediment layer. Houwing and van Rijn (1998) presenteda three times increase in τc with a 1-cm increase in bedlayer depth. Nevertheless, the constant value can beused while focusing only on the threshold of the remo-bilization, arbitrarily set as a 1-mm degradation of thebed layer. The sensitivity analysis sediment transportparameters performed by Hostasche et al. 2014 pointout the settling velocity as the most sensitive parameterof the model. Nevertheless, they were focused on thesuspended sediment transport and pollutant desorptionrather than on the grain entrainment conditions, thefocus of our study. In our study, the settling velocityparameter influences the entrainment indirectly bypredetermining the concentration profile. The settlingvelocity of cohesive sediments is not only the functionof the particle weight, as it can be overcome by even asmall flow disturbance, such as turbulence or fluctuation(Grabowski et al. 2011), but also fine-grained sedimentschiefly in particle form. The shape of the grains couldnot be included due to the lack of data. McNeil and Lick(2004) reported the importance of spatial variability ofsediment distribution. They neglected the existence offine-grained sediments in transition flow zones. Thus,the Bílina fine-grained sediments were located only inchannel forms not affected by erosion during normalflow conditions.

To compare the results for all sites, the return periodparameter estimated from the duration curves of thenearest stations was adopted (Table 7). The downstreammonitoring station, the Trmice gauging station, was thebase for all assessments. The return period intervalswere further classified regarding the probability of

remobilization. The causal discharges (Table 5) dependon site-specific parameters and varied from 5 m3s−1,corresponding to the discharge not exceeded in Q210

days per year near the confluence with the Elbe (B2),to 24 ms−3, regarded as the discharge with the returnperiod of Q10 years at the most stable site at the conflu-ence with the Bouřlivec left tributary (B3). The mostexposed sites with the remobilization return periodsaround Q1 are located downstream of the Bílina River,near the confluence with the Elbe. The sediments arelocated in a concrete prismatic channel morphology andare exposed yearly to the transient flow, which entrainsthe sediment at velocities around 0.3 m s−1, that is,corresponding with the values found by Schwartz(2006) in the case of Elbe sediments. Moreover, thedownstream reach mostly tends to affect the water qual-ity in the Elbe River as the recipient of the flow; thus, theB1, B2, and B4 sites should be the focus of futureremediation plans and initiatives. The remote sites B5–B9 tend to be less exposed while the remobilizationreturn period is Q5–Q10, which nevertheless does notrepresent the overall environmental impact after sedi-ment is remobilized.

The results contribute to the toxic particle-bed sedi-ment in the Elbe basin assessment, often discussedqualitatively and quantitatively (the summary is givenby Heise et al. 2008; SedNet 2009). The particle-bedpollution of the Elbe originates at least partially fromupstream reaches. According to Heininger et al. (2003),the Czech Elbe upper course represents one of threemain pathways of German Elbe sediment pollution.The sediments tend to be toxically enriched severalmagnitudes more than the surrounding aquatic environ-ment. Thus, assessing fine-grained sediment remobili-zation in the headwaters of the Elbe basin plays a keyrole in understanding the water and sediment pollutionin the downstream watercourse.

Conclusions

This article analyzed the potential and limits of differentchannel schematizations for hydrodynamic modeling ofthe sediment transport in small streams. The appliedmodeling techniques were applied to the Bílina River,Czech Republic, Central Europe, representing a small-size stream with fine-grained sediments, contaminatedby heavy chemical industry.

Table 7 Return period for fine-grained sediment remobilization

Site ID Return period of thesediment remobilization

B1 Q1

B2 Q210days

B3 Q5–Q10

B4 <Q1

B5 <Q10

B6 >Q5

B7 Q5

B8 >Q5

B9 <Q5

Environ Monit Assess (2015) 187:115 Page 15 of 18 115

Along the 55-km-long assessed stream reach, the keyzones of the potential sediment remobilization weredetected. The critical spots in terms of remobilizingsediment were located the confluence with the ElbeRiver, where the remobilized sediments can affect thecontamination of the transboundary and downstreamElbe reaches at flow conditions even lower than a 1-year flood event.

The results achieved by different flow schematiza-tions indicated their varying applicability according thechannel characteristics and input data complexity. Theremobilization of fine-grained sediments was estimatedby using 1D, 1D+, and 2D types of flow horizontaldiscretization, while the 1D discretization was calculat-ed with CS average and CS distributed flow character-istics. The parametrization of the sediment transportmodel was tested by the sensitivity analysis, whichproved the key parameters to be the bed roughness ktand critical shear stress for erosion τc.

The results of 1D schematization with averagedcross-section parameters (1D CS average) were consid-ered as unrealistic, due to the non-uniformly locatedsediment layers and the D50 being significantly smallerthan the average sandy-gravel channel material. Thesediment started to erode immediately afteroverflowing, as a result of high velocity and bed shearstress. The results achieved by 1D schematization with adistribution of the flow parameters along and across thechannel by the conveyance module (1D CS distributed)were reasonable, while the fine-grained sediment remo-bilization was evaluated at all sites, since the bed shearstress was 0.5–1 N m−2.

The 1D+ schematization improved the flow de-scription by dividing the flow characteristics intoseparate zones, represented by independent values.The separation of the flow into the functionalzones within the channel led to a substantial de-crease of bed shear stress. Therefore, at siteswhere the channel differed from the pure prismatic(B4, B5, B8, B9), the remobilization simulationwas successful. However, for streams with prismat-ic segments, the functional zones cannot be iden-tified so such an approach is not applicable.

The 2D schematization is by far the most demandingin terms of input data quality as well as for the compu-tation time. The 1-m grid cell resolution, necessary forrealistic channel description resulted in unacceptablecomputational times while the accuracy of results wasnot significantly different to the 1D CS distributed or

1D+ schematizations. The application of the detailed 2Dschematization thus was considered as not necessary.

Our study proved that the channel properties, as wellas the quality and complexity of the input data, are thekey decisive factor for selection of appropriate modelschematization. Application of the simpler schematiza-tions does not necessarily lead to worse results.Conversely, it may result in a reduction of the compu-tational time for the entire reach by several magnitudeswhile retaining a similar level of accuracy. Nevertheless,the success of the simple 1D CS distributed or 1D+approach is limited by the appropriate channelproperties.

Acknowledgments This study was conducted as part of theELSA project 2/UPS-I/547.30 SedBiLa - Bedeutung der Bilinaals historische und aktuelle Schadstoffquelle für dasSedimentmanagement im Einzugsgebiet der Elbe, coordinatedby the International Commission for the Protection of the ElbeRiver (ICPER). The research presented in this paper was supportedby Czech Science Foundation project GAČR P209/12/0997. Thefirst author thanks DHI a.s. for providing software support.

References

Aksoy, H., & Kavvas, M. L. (2005). A review of hillslope andwatershed scale erosion and sediment transport models.Catena, 64(2–3), 247–271.

Amos, C. L., Bergamasco, A., Umgiesser, et al. (2004). Thestability of tidal flats in Venice Lagoon—the results of in-situ measurements using two benthic, annular flumes.Journal of Marine Systems, 51(1–4), 211–241.

Ariathurai, R., & Arulanandan, K. (1978). Erosion Rates ofCohesive Soils. Journal of the Hydraulics Division, 104(2),279–283.

Bagnold, R.A. (1966). An approach to the sediment transportproblem from general physics. U.S. Geological SurveyProfessional Paper 422-I.

Breusers, H. N. C. (1986). Lecture notes on sediment transportproblem. Delft: International course of hydraulic engineering.

Büttner, O., Otte-Witte, K., Krüger, F., Meon, G., & Rode, M.(2006). Numerical modelling of floodplain hydraulics andsuspended sediment transport and deposition at the eventscale in the middle river Elbe, Germany. ActaHydrochimica et Hydrobiologica, 34, 265–278.

Cao, Z., Li, Z., Pender, G., & Hu, P. (2012). Non-capacity orcapacity model for fluvial sediment transport.Proceedingsof the. ICE - Water Management, 165(4), 193–211.

Caviedes-Voullième, D., Morales-Hernandéz, M., López-Marijuan, I., & García-Navarro, P. (2014). Reconstructionof 2D river beds by appropriate interpolation of 1Dcross-sectional information for flood simulation. EnvironmentalModelling & Software, 61, 206–228.

CHMI (2013). Hydrological yearbook. CHMI, Prague.

115 Page 16 of 18 Environ Monit Assess (2015) 187:115

Cook, A., & Merwade, V. (2009). Effect of topographic data,geometric configuration and modeling approach on floodinundation mapping. Journal of Hydrology, 377(1–2), 131–142.

Costa, S., & Coelho, C. (2011). Suspended sediment concentrationimportance on cohesive sediment settling velocity. Journal ofIntegrated Coastal Zone Management, 11(2), 171–185.

DHI (2012). MIKE 21 & MIKE 3 flow model - Scientific docu-mentations. Horstholm.

Droppo, I. G., Liss, S. N., Williams, D., Nelson, T., Jaskot, C., &Trapp, B. (2009). Dynamic existence of waterborne patho-gens within river sediment compartments.Implications forwater quality regulatory affairs. Environmental Science &Technology, 43(6), 1737–1743.

Fang, H., Chen, M., & Chen, Q. (2008). One-dimensional numer-ical simulation of non-uniform sediment transport under un-steady flows. International Journal of Sediment Research,23(4), 316–328.

Frank, E., Sofia, G., & Fattorelli, S. (2012). Effect of topographicdata resolution and spatial model resolution on hydraulic andhydro-morphological models for flood risk assessment. In S.Mambretti (Ed.), Flood risk assessment and management.Southhampton: WIT Press.

Galappatti, R., & Vreugdenhill, C. B. (1985). A depth-integratedmodel for suspended transport. Journal of HydraulicResearch, 23(4), 359–377.

Ghostine, R., Vazquez, J., Terfous, A., Mose, R., & Ghenaim, A.(2012). Comparative study of 1D and 2D flow simulations atopen-channel junctions. Journal of Hydraulic Research,50(2), 164–170.

Giere, O. (2009). Meiobenthology. Heidelberg: Springer-VerlagBerlin.

Grabowski, R. C., Droppo, I. G., & Wharton, G. (2011).Erodibility of cohesive sediment: The importance of sedi-ment properties. Earth-Science Reviews, 105(3–4), 101–120.

Guerrero,M., & Lamberti, A. (2013). Bed-roughness investigationfor a 2-D model calibration: the San Martín case study atLower Paranà. International Journal of Sediment Research,28(4), 458–469.

Haddadchi, A., Ryder, D. S., Evrard, O., & Olley, J. (2013).Sediment fingerprinting in fluvial systems: review of tracers,sediment sources and mixing models. International Journalof Sediment Research, 28(4), 560–578.

Hardy, R. J. (2013). Process-Based Sediment Transport Modeling.In J. F. Shroder (Ed.), Treatise on Geomorphology, volume 2:Quantitative Modeling of geomorphology (pp. 147–159).London: Elsevier.

Heininger, P., Pelzer, J., Claus, E., & Pfitzner, S. (2003). Results ofLong-term Sediment Quality Studies on the River Elbe. ActaHydrochimica et Hydrobiologica, 31, 356–367.

Heise, S., Krüger, F., Förstner, U., Baborowski, M., Götz, R., &Stachel, B. (2008). Bewertung von Risiken durchFeststoffgebundene Schadstoffe im Elbeeinzugsgebiet. InZusammenfassungen der Studien. Hamburk: Hamburg PortAuthority.

Horritt, M. S., & Bates, P. D. (2002). Evaluation of 1D and 2Dnumerical models for predicting river flood inundation.Journal of Hydrology, 268(1–4), 87–99.

Hostache, R., Hissled, C., Matgen, P., Guignard, C., & Bates, P.(2014). Modelling suspended-sediment propagation and re-lated heavy metalcontamination in floodplains: a parameter

sensitivity analysis. Hydrology and Earth System Sciences,18, 3539–3551.

Houwing, E. J., & van Rijn, L. C. (1998). In situ erosion Flume(ISEF): Determination of bed shear stress and erosion of akaolinite bed. Journal of Sea Research, 39, 234–253.

Huang, J., Hilldale, R. C., & Greimann, B.P. (2006). Erosion andsedimentation manual: Chapter 4. In: Erosion and sedimen-tation manual. Denver: U.S. Department of the InteriorBureau of Reclamation.

Jurajda, P., Adámek, Z., Janáč, M., & Valová, Z. (2010).Longitudinal patterns in fish and macrozoobenthos assem-blages reflect degradation of water quality and physical hab-itat in the Bílina river basin. 2010(3), 123–136.

Kiat, C. C., Ghani, A. A., Abdullah, R., & Zakaria, N. A. (2008).Sediment transport modeling for Kulim River –A case study.Journal of Hydro-Environment Research, 2(1), 47–59.

Komar, P. (1988). Sediment transport by floods. In V. R. Baker &P. C. Parron (Eds.), Flood geomorphology (pp. 97–112).New York: John Wiley and Sons.

Langhammer, J. (2007). Modelling the Elbe River water qualitychanges. In P. Dostál, P., J. Langhammer (Eds.), Model. Nat.Environ. Soc. PřF UK, Praha (pp. 59–74).

Langhammer, J. (2010). Water quality changes in the Elbe Riverbasin, Czech Republic, in the context of the post-socialisteconomic transition. GeoJournal, 75, 185–198.

Matoušková, M., & Dvořák, M. (2011). Assessment of physicalhabitat modification in the Bílina River Basin. Limnetica,30(2), 293–306.

McNeil, J., & Lick, W. (2004). Erosion Rates and Bulk Propertiesof Sediments From the Kalamazoo River. Journal of GreatLakes Research, 30(3), 407–418.

Medek, J., Hájek, P., Král, S., Skořepa, J., Hönig, J., Kokšal, J.,Neuhöfer, M., Bednárek, J., Janský, B. Langhammer,J., Chalupová, D., Jiřinec, P, Indeguldová, E.,Kaiglová, J. (2014). SedBiLa - Bedeutung derBilinaalshistorische und aktuelleSchadstoffquellefürdas SedimentmanagementimEinzugsgebiet der Elbe.Project report. Povodi Labe, Hradec Kralove, 80 pp.Retrieved from http://elsa-elbe.de/assets/pdf/Fachstudie_Sedbila_DE.pdf.

Mosselman, E. (1992).Mathematical modelling of morphologicalprocesses in rivers with erodible cohesive banks, Comm. onHydr. and Geotec. Eng., 92 (3). Delft: Delft University ofTechnology.

Papanicolaou, A. N. T., Krallis, G., & Edinger, J. (2008). SedimentTransport Modeling Review — Current and future develop-ments. Journal of Hydraulic Engineering, 134(1), 1–14.

Povodí Ohřes.p. (2010). Studie záplavového území toku BílinaÚstí nad Labem - Liběšice (km 0.000 – 40.250) Libešice –Ervěnický koridor (aktualizace) (km 40.000 – 61.200).Povodí Ohře, s.p., Chomutov. (Flood extend assessment ofBílina river from Ústí nad Labem to Ervěnirký koridor - inCzech).

Přibylová, P., Klánová, J., & Holoubek, I. (2006). Screening ofshort- and medium-chain chlorinated paraffins in selectedriverine sediments and sludge from the Czech Republic.Environmental Pollution, 144(1), 248–254.

Schwartz, R. (2006). Geochemical characterization and erosionstability of fine-graineggryone field sediments of theMiddleElbe River. Acta Hydrochimica et Hydrobiologica, 34, 223–233.

Environ Monit Assess (2015) 187:115 Page 17 of 18 115

SedNet (2009).Observations on sediment management. SedNet e-news special. Retrieved from http://www.sednet.org/download/newsletter/E-News_special_on-RBMP_April2009.pdf.

Segura, R., Arancibia, V., Zúñiga, M. C., & Pastén, P. (2006).Distribution of copper, zinc, lead and cadmium concentrationsin stream sediments from the Mapocho River in Santiago,Chile. Journal of Geochemical Exploration, 91(1–3), 71–80.

Shields, A. (1936). Anwendung der Ahnlichkeitsmechanik und derTurbulenzforschung auf die Geschiebebewegung. Berlin:Mitteilung der preussischen Versuchsanstalt fur Wasserbauund Schiffbau, 26.

Simpson, G., & Castelltort, S. (2006). Coupled model of surfacewater flow, sediment transport and morphological evolution.Computers & Geosciences, 32, 1600–1614.

Šípek, V., Matoušková, M., & Dvořák, M. (2010). Comparativeanalysis of selected hydromorphological assessmentmethods. Environmental Monitoring and Assessment,169(1–4), 309–319.

Van Griensven, A., Popescu, I., Abdelhamid, M. R., Ndomba, P.M., Beevers, L., & Betrie, G. D. (2013). Comparison ofsediment transport computations using hydrodynamic versushydrologic models in the Simiyu River in Tanzania. Physicsand Chemistry of the Earth, 61–62, 12–21.

Van Rijn, L. (1993). Principles of sediment transport inrivers, estuaries and coastal seas. Amsterdam: AquaPublications.

Wilcock, P. (1993). Critical shear stress of natural sediments.Journal of Hydraulic Engineering, 119(4), 491–505.

Wilcock, P. (2004). Lecture notes – Sediment transport – Thesediment problem. Retrieved from http://calm.geo.berkeley.edu/geomorph/wilcock/WilcockSTLecture3.pdf.

Winterwerp, C., & van Kersteren, W. G. (2004). Introduction tothe physics of cohesive sediment dynamics in the marineenvironment. In T. van Loon (Ed.), Developments in sedi-mentology. Amsterdam: Elsevier.

Wu, B., van Maren, D. S., & Li, L. (2008). Predictability ofsediment transport in the Yellow River using selected trans-port formulas. International Journal of Sediment Research,23(4), 283–298.

Yang, C.T., Huang, J., & Greimann, B.P. (2005). User’s manualfbr GSTAR-ID (Generalized Sediment Transport model forAlluvial Rivers - One Dimension), version 1 .O. Denver: U.S.Bureau of Reclamation, Technical Service Center.

Zeng, S., Dong, X., & Chen, J. (2013). Toxicity assessment ofmetals in sediment from the lower reaches of the Haihe RiverBasin in China. International Journal of Sediment Research,28(2), 172–181.

115 Page 18 of 18 Environ Monit Assess (2015) 187:115