Author’s Accepted Manuscript

Prediction of sedimentation in reservoirs by

combining catchment based model and stream
based model with limited data

Abebe Tadesse, Wenhong Dai

www.elsevier.com/locate/ijsrc

PII: S1001-6279(17)30150-6
DOI: https://doi.org/10.1016/j.ijsrc.2018.08.001
Reference: IJSRC186
To appear in: International Journal of Sediment Research
Received date: 9 August 2017
Revised date: 6 June 2018
Accepted date: 22 August 2018
Cite this article as: Abebe Tadesse and Wenhong Dai, Prediction of
sedimentation in reservoirs by combining catchment based model and stream
based model with limited data, International Journal of Sediment Research,
https://doi.org/10.1016/j.ijsrc.2018.08.001
This is a PDF file of an unedited manuscript that has been accepted for
publication. As a service to our customers we are providing this early version of
the manuscript. The manuscript will undergo copyediting, typesetting, and
review of the resulting galley proof before it is published in its final citable form.
Please note that during the production process errors may be discovered which
could affect the content, and all legal disclaimers that apply to the journal pertain.
 Prediction of sedimentation in reservoirs by combining
catchment based model and stream based model with
limited data

Abebe Tadessea, Wenhong Daib

aHohai University, Department of Water Conservancy and Hydropower Engineering,
E-mail address: abebulti@hhu.edu.cn
b Hohai University, Department of Water Conservancy and Hydropower Engineering, E-mail address :

wdai@hhu.edu.cn

1. Acknowledgment

This research is supported by the National Key Research and Development Program of China (Grant No.

2016YFC0402501), the National Natural Science Foundation of China (NSFC) (Grant No. 51479071), the

111 Project (Grant No. B12032 and B17015) and the Priority Academic Program Development of

Jiangsu Higher Education Institutions (Grant No. 3014-SYS1401). The authors also wish to express

gratitude to the Ministry of Water Resources, Irrigation and Energy of Ethiopia for the provision of the

data.

2. Corresponding author

Wenhong Dai, Water Conservancy and Hydropower Engineering in Hohai university, Email address:

wdai@hhu.edu.cn

∗Corresponding author

Email address: wdai@hhu.edu.cn (Wenhong Dai)

1
Abstract

Estimation of sedimentation in reservoirs helps in the management and design of the

reservoir’s useful capacity. This research was done on the Awash River basin at the Koka

Dam Reservoir in Ethiopia. The method applied was the loose integration of the Soil and

Water Assessment Tool (SWAT) model and Hydrologic Engineering Center-River Analysis

System (HEC-RAS) model for the estimation of the sediment load reaching the reservoir.

The SWAT model was used for the estimation of erosion at the catchment level, and the

HEC-RAS model was applied to estimate the sediment transport in the river channel. The

implemented method allows sedimentation in the floodplains and bed shear stress to be

considered in the sediment modeling, which cannot be considered in the SWAT model. In

addition, the river cross sectional properties and the hydrodynamic processes in the rivers

were considered in the modeling process. The data used in this study are a combination of i)

observed data collected by government agencies, ii) data available online in data

repositories, and iii) data extracted from remote sensing in the Shuttle Radar Topographic

Mission (SRTM) Digital Elevation Model (DEM). The calibration and validation of the

SWAT model was done by using Sequential Uncertainty Fitting (SUIF-2) calibration and

validation tools. The HEC-RAS model was calibrated by adjusting the roughness factor. The

output from the integrated approaches gives better estimates of flow and sediment near the

inlet to the reservoir, with coefficients of determination of 0.85 and 0.67, respectively, and

Nash Sutcliffe coefficients of model fit efficiency of 0.90 and 0.62, respectively, for daily

simulations.

Keywords: Reservoir sedimentation, Awash River, SWAT, HEC-RAS, Rating curve

2
1. Introduction

Sedimentation in reservoirs is a major problem in many parts of the world. The

construction of dams or reservoirs is important for human beings, since reservoirs

provide multipurpose uses which contribute significant value for the development of

society. The multiuse of the reservoirs includes flood prevention and provision of water

for different uses i.e. water for industry, agriculture, and domestic purposes. However,

rapid sedimentation resulting from catchment erosion from degraded land caused by

land use change (Ali et al., 2014; Garzanit et al., 2006) has a significant effect on the

sustainable use of reservoirs.

The study of the sedimentation process should give emphasis to the recognized sources

of the sediment. Those sources are erosion from the catchment and subsequent sediment

yield reaching the reservoir, which is affected by sediment transportation in the river

channel. The sediment from the catchment or sediment yield mainly focus on the soil

erosion process which depends on factors such as the variability of topography, hydrological

conditions, weather, land use, soil types, and vegetation which affect the subsequent runoff

and soil loss. The sediment transport in the river channels is controlled by the

hydrodynamics of the flow in the channel and the properties of the channel.

The sedimentation process can be studied through the application of different types

of models, which take different approaches in modeling the process. Simulation of the

runoff and soil erosion process from the catchments and routing runoff and sediment

through stream networks is the main task in most of models. The Soil and Water

Assessment Tool (SWAT) (Arnold et al., 2012; Neitsch et al., 2011), Agricultural Non-

Point Source Pollution model (AnnAGNPS) (Bingner et al., 2005), and Hydrological

3
Simulation Program-Fortran (HSPF) (Donigian et al.,1984) are some of the models

which are used for the study of the sediment yield process. On the other hand, river

hydrodynamic analysis models, e.g., the Hydrologic Engineering Center-River Analysis

system (HEC-RAS) (Brunner, 2016), MIKE-11 (Havnø et al., 1995), and SOBEK

(Deltares, 2018) are among the models that can be used to understand the flow and

sediment transport in river channels by solving the fully dynamic wave equations. These

models are able to simulate all cases including unsteady, non-uniform flow conditions.

A sedimentation process modeling study can be made through application of catchment-

based models, such as SWAT encompassing the stream in the basin. However, different

researchers have suggested that the model has drawbacks in considering the backwater

effects and sediment characteristics (e.g., bottom shear stress) effects in the model (

Betrie et al., 2011; Shrestha et al., 2013). Also sediment deposited in the floodplains and

sediment traps are not considered in process of catchment level sediment modeling

(Neitsch et al., 2011).

The processes of soil erosion and sedimentation include many sub-processes such as

the soil erosion process, hydrologic process, and hydraulics of flow. The all-

encompassing modeling of such processes, has a certain effect on the modeling

approach because of the interrelation and dependence of the processes on each other

(Shrestha et al., 2013). In order to consider the effects of each sub-process in the

modeling effort the separate approach for the soil erosion and sedimentation process at

the catchment level and the sediment transport in stream channels is one option, which

can make the representation of the modeling process reasonable.

In any modeling process, assumptions are made to simplify the complex real process

4
while keeping the overall effects of the simplification insignificant on the modeling

approach. However, when the process includes many sub-processes, i.e. many variables

(like erosion and sedimentation), the assumptions may have considerable effects on the

model output. The interdependence of the variables also may affect the results of the

separate model approach. The loose integration (Shrestha et al., 2013) of the catchment-

based and stream-based models is one option for modeling the sediment load process for

reservoirs. The method does not need additional coding but only linking the outputs of the

catchment model as input for the stream-based model. In addition, the approach addresses

the drawbacks in a catchment-based model, such as SWAT, and reduces the effects of the

assumptions on the interdependence of the variables between the models.

SWAT is a physical process based model to simulate continuous-time, semi-

distributed landscape processes at a catchment scale (Arnold et al., 2012; Neitsch et al.,

2005). The catchment is divided into hydrological response units (HRUs) based on soil

type, land use, and slope classes that allows a high level of spatial detail in the

simulation. The major model components include hydrology, weather, soil erosion,

nutrients, soil temperature, crop growth, pesticides, agricultural management, and

stream routing.

HEC-RAS (Brunner, 2016) is a hydraulic simulation model for calculating water

surface profiles for natural and manmade channels. It is a simulation model that has

one-dimensional and two-dimensional (1-D and 2-D, respectively), steady and unsteady

flow analysis options. It also has components for movable boundary sediment transport

computations and water quality analysis. Steady flow analysis allows calculation of

water-surface profiles for steady and gradually varied flow. HEC-RAS can model

5
subcritical, supercritical, and mixed flow regime water-surface profiles. It also can

model the flow and sediment movement around hydraulic structures such as bridges,

culverts, weirs, and spillways. Due to its extensive modeling capabilities, HEC-RAS is

widely used for channel and floodplain management and flood insurance studies to

evaluate floodway encroachments.

The combined SWAT and HEC-RAS model was applied for modeling sediment

transport into the Koka Dam Reservoir in Ethiopia, which includes a hydrological model,

SWAT (Arnold et al., 2012) and a hydraulic model HEC-RAS (Brunner et al., 2016) as the

major computational tools. In addition, a sediment rating curve was developed from long-

term sediment data in the catchment. The calibration and validation of the SWAT model for

sediment and flow was done using sequential uncertainty fitting (SUIF-2) in the SWAT-

calibration and uncertainty program (CUP) calibration and validation tool. The loose

combination of such hydrodynamics and catchment-based models enhances the opportunity

to consider the effects of river hydrodynamic processes and river cross section properties,

including the floodplains, on sediment transport. Hence, the situation helps to improve some

of the drawbacks in the SWAT model and gives insight to understand the sediment yield

process.

2. Materials and methods

2.1. Study Area

The Awash River is one of the rivers Awash in the eastern part of Africa, in the central

part of Ethiopia (Fig. 1). Koka Dam, which has a reservoir capacity of 1180 Mm3, was

constructed for hydropower generation in 1960 at an upstream reach of the Awash River,

which is located between latitudes of 8◦16’ and 9◦18’ Nand longitudes of 37◦57’ and 39◦17’

6
E. The river originates from the Ginich Highland area and flows downstream to the Afar

region; it ends in the lowland areas of Afar remaining in the sand at the Ethiopian

boundaries.

Fig. 1. Awash River upstream reach catchment above Kaka Dam study area

The river is recognized to be the most widely used river in Ethiopia. As the source of

the Awash River is from the highlands, the river is characterized by high flash floods,

which carry high amounts of sediment that significantly affect the capacity of the Koka

Reservoir. The sediment accumulated in the Koka Reservoir decreases the capacity and

creates a problem for power generation at the hydropower system. It can, thus, be

concluded that the water storage capacity of the reservoir is subject to severe

sedimentation problems (Ahemed et al., 2005).

2.2. The Approach for modeling

The combined use of the catchment-based SWAT (Arnold et al., 2012) model and the

stream-based HEC-RAS (Brunner, 2016) model was applied as the basic tool in this

research. It was understood that SWAT could model erosion and sedimentation from the

catchment including the streams in the basin. However, to address some of the

drawbacks of the SWAT in modeling process (Betrie et al., 2011) and to consider the

role of each sub-process in the event, the loose integration of the SWAT and HEC-RAS

models was preferred.

According to different researchers, it has been suggested that the system modeling

should be categorized into two parts (Vieira et al., 2018); the former is modeling the

flow and sediment at the sub-basin level in the upstream parts of the catchment using

7
the SWAT model. At this part of the catchment, the effects of stream processes are

smaller and a higher weight is placed on the soil erosion process and sediment yield

from the catchment. The latter part is stream modeling in the main river channel and

tributaries with the HEC-RAS model to estimate sediment transport and flow in the

river cross sections. The process in the latter steps places higher weight on the flow and

sediment transport in the river cross section; furthermore, the approach in the latter steps

includes the sediment and flow from the upstream parts of the catchment through the

outputs of the SWAT model. In the latter steps of modeling the flows in channels are

assumed to be more important in estimation of sediment transport in the downstream

sections of the catchment than the erosion process in the overall catchment. Furthermore,

the approach provides the opportunity to include sedimentation in the floodplain

through the stream cross sections input in the HEC-RAS model, which is not considered

in the SWAT model.

The system was subdivided to catchment-based and stream-based, to apply the

loosely integrated SWAT and HEC-RAS model as shown in Fig. 2. The SWAT model

was initially applied on the catchment by considering the outputs from the model to be

used as input to the HEC-RAS application in the stream channels.

Fig. 2. Modeling approach in the upstream reaches of the Awash River basin above

Koka Dam

2.3. Models descriptions

2.3.1. The SWAT model

The SWAT model (Neitsch et al., 2011; Williams et al., 2008) has become one of the

8
most widely used water quality watershed and river basin scale model worldwide,

demonstrating multiple decades of model development (Arnold et al., 2012; Gassman et al.,

2007; Williams et al., 2008). The model has been extensively applied for watershed

simulation modeling and has been used by water resources experts for watershed hydrology

related studies (Keshta et al., 2009; Santhi et al., 2001; Schuol & Abbaspour, 2006 ).

Flexibility in addressing a wide range of water resource problems, strong model support,

and the open access status of the source code are among the foremost reasons for its wide

application. It was developed to assist water resources managers in predicting and assessing

the impact of agricultural management practice on water, sediment, and agricultural

chemical yields in large ungagged watersheds or river basins. It is an operational or

conceptual model that operates on a daily time step.

The modeling application can be done in a large complex watershed with varying

soil, land use, and management conditions. The SWAT model uses discretization of the

catchment into a number of sub-divisions to describe the large scale spatial variability of

soil, land use, and management practices. The discretization has two stages; the first is

sub-dividing the large complex topographic catchment into sub-basins based on a

threshold area. Next the sub-catchments are further subdivided into one or more

homogeneous hydrologic response units (HRUs) considering unique combinations of

soil and land use. The computation of the output of water, sediment, nutrient, and

pesticides is affected by the modeling systems individually determined for each HRU

and further aggregated at the sub-basin level. Finally, different routing techniques are

used to associate the stream reaches and the catchment outlet in the channel network.

The hydrologic cycle of a sub-basin simulated with SWAT is based on the following

9
water balance equation:

∑ ( )

where SWt is the final water content on day t (mm), SW0 is the initial water content on

day i (mm), Rday is the amount of precipitation on day i (mm), Qsurf is the amount of

surface runoff on day i (mm), Ea is the amount of evapotranspiration on day i (mm),

Wseep is the amount of water entering the vadose zone from the soil profile on day i

(mm), and Qgw is the amount of groundwater return flow on day i (mm).

Sediment yield is estimated with the Modified Universal Soil Loss Equation (MUSLE)

(Williams, 1995) using the surface runoff, peak flow rate, soil erodibility, crop management,

erosion control practice factor, and slope length and steepness factors for each HRU.

where sed is the sediment yield on a given day (metric tons); Qsurf is surface runoff volume

(mm); qpeak is peak runoff rate (m3/s); areahru is area of HRU (ha); KUSLE is the soil

erodibility factor (0.013 metric ton* m2 *hr/ (m3 * metric ton * cm); CUSLE is the cover

and management factor; PUSLE is the support practice factor; LSUSLE is the topographic

factor; and CFRG is the coarse fragment factor.

In large sub-basins with the time of concentration greater than 1 day, all the flow

generated will not reach the main channel on that specific date, in such case SWAT

incorporates a surface runoff storage to lag a portion surface flow release to the main

channel. In addition, the sediment transported to the main channel in a given day

depends on the amount of sediment load generated in an HRU on that day, the sediment

stored or lagged from the previous day, the surface runoff lag coefficient, and the time

of concentration for the HRU. The sediment discharged to the main channel is

10
calculated based in the following equation (Arnold et al., 2012):

( ]

where sed is the amount of sediment discharged to the main channel on a given day

(metric tons); sed′ is the amount of sediment load generated in the HRU on a given day

(metric tons); sedstor,i−1 is the sediment stored or lagged in the HRU from the previous

day (metric tons); surflag is the surface runoff lag coefficient (hr); and tconc is the time of

concentration for the HRU (hr).

2.3.2. HEC-RAS Model

HEC-RAS is a public domain code developed by the U.S. Army Corp of Engineers

(Brunner, 2016). It does 1-D and 2-D, steady and unsteady flow calculations on a

network of natural or constructed open channels. Basic input data required by the model

include cross sectional geometry and alignment of the river, the channel network

connectivity, reach lengths, energy loss coefficients, stream junction information, and

hydraulic structure data. When changes in discharge, slope, shape, or roughness occur at

locations throughout a stream reach the representative cross sections at these locations

are important for proper modeling. Boundary conditions are necessary to define the

starting water depth at the stream system endpoints, i.e. upstream and downstream.

Water-surface profile computations begin moving upstream for subcritical flow or

downstream for supercritical flow. Discharge information is required at each cross

section in order to compute the water-surface profile. HEC-RAS can calculate water-

surface profiles for both steady and unsteady gradually varied flow. The steady flow

system is designed for application in floodplain management studies in rivers (Reed &

11
Maidment, 1995). Also, it has capabilities to assessing the change in water-surface

profiles due to channel improvements and levees.

The development of a Geographic Information System (GIS) interface called HEC-

GeoRAS at the University of Texas at Austin, by improving a previously issued Arc-View

extension (AV-RAS) (Maidment & Djokic, 2000) allows coping with quasi-2-D aspects of

flow through connecting the river geometry with a digital terrain model in the form of a

Triangulated Irregular Network (TIN). The distributed output provided by HEC-RAS for

each cross section is interpolated between cross sections resulting in a water depth and a

water velocity surface.

The 1-D continuous simulation in the HEC-RAS 5.0 model uses a sequence of steady

flows to represent the discharge hydrograph. It is designed to simulate and predict changes

in river profiles resulting from scour and/or deposition of sand, silt, and clay over moderate

time periods. The model inputs include flow, sediment transport, channel roughness, and

related changes in boundary geometry for simulation of an actual event. The HEC-RAS

model does the hydraulic computations, which include determination of water-surface

profiles and flow velocities at each cross section along the study reach. The water-surface

profile is calculated from downstream to upstream using the backward standard step method

to solve the one-dimensional energy equation (USACOE, 1990).

The HEC-RAS-5 software offers functions Laursen (1963), Engelund and Hansen (1967),

Toffaleti (1968), Meyer-Peter and Muller (1984), Yang (1984), Ackers and White (1993), and

Wilcock and Crowe (2003)for calculating sediment transport. The selection of the sediment

transport function is based on knowledge of the application conditions and practical

experience with the modeling process. A significant element of the sediment transport

12
calculation is the choice of a suitable function enabling good agreement to be obtained

between calculated values and field measurements of sediment data. The selected sediment

transport module for any model requires input of the hydraulic parameters. Therefore, HEC-

RAS computes the hydraulic parameters each time before the sediment computation or

updating the cross section because of the sediment transport progression. HEC-RAS-5

couples sediment transport computations either with quasi-unsteady or unsteady flow

hydraulics computations that are available in the model.

2.4. Model calibration and validation

2.4.1. SWAT model for the Awash Basin upstream of Koka Dam

The application of the SWAT model was done for the reach upstream of Koka Dam

in the Awash River basin. The model requires Digital Elevation Model (DEM), land

use, soil, and weather data for the simulation of the events in the modeling process.

Hence, DEM (Fig. 1), land use (Fig.3a), and soil data (Fig. 3b) with resolutions of 30 m

obtained from the Shuttle Radar Topographic Mission (SRTM) were obtained

(http://earthexplorer.usgs.gov and http://www.fao.org/soils-portal/soil-survey/soil-maps

and databases/harmonized-world-soil database-v12/en). The weather data was a

combination of limited local data and online data from http://waterbase.org a global

weather database for SWAT. The land use map was modified to suit the current

modeling work based on the actual land use information from the Ministry of Water

Resources, Irrigation and Energy of Ethiopia. Rapid land use change has occurred in the

basin and most of the land use in the area has been changed to agricultural land for crop

production purposes.

The catchment was subdivided into seven (Fig. 3a) sub-basins based on the gaging

13
stations. The sub-basins were delineated based on preferences (Neitsch et al., 2011) for

modeling of the flows and sediment at the gaging stations. This approach provides an

opportunity for the evaluation of the output of the model and observed data at the

gaging stations. The location of the outlets of the sub-basins also were selected in such a

way that the process can be separated into catchment level and in-stream level (Fig. 2),

to observe the effect of the combined approach on the modeling effort. In addition, the

SWAT model was applied for the whole catchment for comparison of the combined

approach and SWAT alone.

Fig. 3. a) Land use map of study the basin b) Soil map of the study basin

2.4.2. The combined SWAT and HEC-RAS model for the Awash Basin upstream of Koka Dam

The loose coupling of the aforementioned models was applied to the Awash River

upstream reaches to model the sediment and flow into the Koka Reservoir. The

extensive applicability of the models and the availability of the models for free makes

the study more plausible for practical application. The SWAT model was applied

individually for the simulation of the sediment and flow in the basin at the early stage.

The outputs of the model were extracted at gaging stations and served as input required

for the HEC-RAS model as shown in Fig. 4. The arrangement of the system in both

models was done considering the theoretical aspects discussed in the Introduction,

which places emphasis on the applications of the models in the catchment and stream

channels.

The input data for the combined approach includes flow and sediment data at the

14
upstream reaches (at the location where the streams are assumed to begin), cross section

data for the main river and tributaries and any hydraulic structure in the river network

which can affect the modeling process. Flow and sediment values at the upstream

locations of the main river and tributary channels were generated using the SWAT

model. The cross section, longitudinal profile, and any other additional information

(such as bridge dimensions, hydraulic structures, etc.) was prepared using the HEC-

GeoRAS (Ackerman, 2012) tools which work with ARCGIS. The data requirements for

the modeling depend on the case to be studied in the river reaches.

The HEC-GeoRAS tool is developed for the preparation of cross section, longitudinal

profile, bridge information, storage area, etc. (depending on the modeling plan) data for the

HEC-RAS model. The input data were extracted by the use of information from the DEM

or TIN and the actual topographic map of the catchment. The DEM used in this

modeling is the same as the DEM used for the SWAT modeling, the detailed procedures

for the preparation of the model input from the DEM or TIN is available in the HEC-

GeoRAS manual (Ackerman, 2012). The HEC-GeoRAS tool was prepared for flood

inundation modeling applications, but in this research, it was used for flow and sediment

modeling with the same approach that is applied for flood inundation modeling, by

providing some additional sediment data for the model to do sediment modeling.

Fig 4. HEC-RAS model schematization of the Awash River and its tributaries

upstream of Koka Dam (where XS means cross section)

In this research the quasi-sediment modeling using Ackers and White (1993) was

applied based on a detailed evaluation of the conditions for the application of the function

15
and the study basin. In addition, the HEC-RAS sediment modeling requires information

about riverbed materials, hence some samples of sediment from the riverbed were taken and

sieve analysis was done.

2.5. Uncertainty analysis, calibration, and validation

2.5.1. SWAT-CUP

The SUFI-2 (Abbaspour et al., 2007) routine in SWAT-CUP is a tool that can be

applied in calibration and uncertainty analysis in SWAT modeling. In SUIF-2,

uncertainty is described as the discrepancy between the observed and simulated

variables. Hence it important to include all measured values while the model output

uncertainty is quantified as the 95% prediction uncertainty (95PPU). The Latin

hypercube sampling method is applied to obtain the cumulative distribution of an output

variable. The main statistical parameters are the P-factor and R-factor, where the P-

factor is the percentage of observed data enveloped by the modeling result, the 95PPU.

The R-factor is the thickness of the 95PPU envelop, for details see Abbaspour (2015).

According to Abbaspour (2015) the target values for the P and R factors are not

fixed to a certain value but it was suggested larger values are better. A P-factor > 70%

and a R-factor around 1 for discharge, and a smaller P-factor and a larger R-factor could

be acceptable for sediment.

2.5.2. Model performance evaluation

The performance of the model can be evaluated using different model goodness-of-fit

measures that are widely used in hydrology (Moriasi et al., 2007). The coefficient of

determination (R2), the Nash Sutcliffe efficiency (NSE) (Nash & Sutcliffe, 1970) and the

percent bias (PBIAS) are commonly used methods related to the objective function of

16
calibration. During the calibration process, the calculation was based on the variables

within that period. When applied to the validation period, the initial variance was

calculated as the sum of squares of deviation from the mean Xm of the observed time

series of the calibration period. R2 was used to characterize the amount of variance in

the observed data described by the model results. The PBIAS value indicates the

average tendency of simulated outputs to be larger or smaller than observations (Gupta

et al., 1999). The NSE and R2 are calculated as follows:

∑ ( )
∑ ̅

∑ ( ̅ )( ̅)

√ ̅ ) ∑ ̅)
[ ∑ ( ( ]

where ̅ and ̅ are the mean of the measured and simulated time series, respectively, and

Xi,m, Yi,s are the observed and simulated variables, respectively.

The performance of the model can be explained based on the recommended ranges.

The R2 ranges > 0.7, 0.6-0.7, 0.5-0.6, and < 0.5 indicate very good, good, satisfactory,

and poor performance of the model, respectively. NSE values greater than 0.75 show

very good model efficiency, NSE between 0.75 and 0.65 shows good performance, NSE

values between 0.65 and 0.5 shows satisfactory performance, and value less than 0.5 are

considered unsatisfactory (Moriasi et al., 2007). In addition, PBIAS values < 10%,

10−25%, and > 25% indicate very good, satisfactory, and poor performance,

respectively (Moriasi et al., 2007; van Liew et al., 2005 ).

17
2.5.3. Roughness factor

The calibration of any hydrodynamic model can be done by varying the bed roughness

factor. Manning’s n bed roughness can be applied for the calibration of the flow model to

fit the simulated to the observed values in the calibration process. The Manning’s n value

for the main channel and flood plain are listed in Table 1. The selection of the bed

roughness factor can be done based on the channel characteristics and measurements at

the site. In this research, the values were determined through careful study of the

channel characteristics and recommendations from similar studies in different areas

(Phillips & Tadayon, 2006; USACOE, 2001; French, 1985). Based on the model results

further adjustment was done to fit the simulated to observed values, by staying within

the reasonable range. The final Manning’s n value for the main channel and the

floodplains were 0.035 and 0.050, respectively.

Table 1: Manning’s n bed roughness for the modeling application

Type of Minimum Average Maximum Fitted

channel value value value value

Floodplain 0.045 0.050 0.060 0.050

Main channel 0.025 0.035 0.045 0.035

2.6. Rating curve development

In the absence of the continuous sediment data, statistical methods, such as rating

curve development, from sample suspended sediment data can be applied in sediment

yield computation. The sediment rating curve was developed from the relation between

discharge and suspended sediment in the form of an exponential function as follows

18
(Ferguson, 1986; Talebi et al., 2015; Walling, 1977; Zhang et al., 2012; ).

where Qs - sediment load (tons/day), Qw - discharge (m3/s), and a and b are the equation

coefficients.

The current study tries to use the data available from the catchment areas, which

includes gaged flow data and limited suspended sediment concentration data. The

gaging data (flow and sediment) were obtained from the Ministry of Water Resources

Irrigation and Energy of Ethiopia for three different sites, Melika Kunture (1990- 2012),

Hombole (2004-2014), and Akaki (1992-2004) (see Fig. 2). The sediment data are not

continuous data, but sample data collected at different times at the aforementioned

stations. The suspended sediment data were used to develop rating curves for specific

locations.

The rating curve developed from the sample data using Eq. 6 is known to

underestimate the sediment loads (Ferguson, 1986; Walling & Webb, 1981 ) estimated

for a river reach. Different kinds of methods have been suggested to improve the

amount of underestimation yielded by Eq. 6. The bias correction factor (CF) has been

suggested to manage the underestimation. Some of the methods include the U.N. food

and Agricultural Organization (FAO) method (Jones et al., 1981), Quasi-Maximum

Likelihood Estimator (QMLE) (Ferguson, 1986), smearing factor Minimum Variance

Unbiased Estimator (MVUE) (Duan, 1983), among the others. However, in the case of

sedimentation in reservoirs, the rate of sedimentation in the reservoir is another source

of information, which can be used for the improvement of the rating curve developed

from the sample suspended sediment data. The long-term rate of sedimentation in a

19
reservoir is more reliable than the statistical methods because it represents the actual

physical process. The basic drawbacks of the statistical methods are they do not reflect

the physical processes in the basin because the approach is based on statistical

parameters rather than physical events.

The long-term sedimentation rate of the study reservoir was taken from research in

the area (Ahemed et al., 2005) as shown in Fig. 5. It is preferable to use this information

for the modification of the rating curve as it represents the sedimentation process more

than any statistical method. The historical data in Fig. 5 was also checked for recent

sedimentation in the Koka Reservoir before these data were used for calculation of the

modification of the improved rating curve.

Fig. 5. Koka Reservoir storage-elevation curves for different times (from Nile Basin

Capacity Building Network (NBCBN) (Ahemed et al., 2005)/ River Morphology

cluster 2005).

The loss of storage from sedimentation based on the storage capacity curves in Fig. 5

and other information available in the study area was about 13 M m3 annually. The

statistical methods (Eq. 6) also were used to estimate the average annual sediment yield

from the catchment near the inlet to the reservoir. Based on the values computed from

those two methods, the modification factor for the rating curve is 137%. In this

computation, the voids in the sediment deposited in the reservoir must be considered.

Hence, 30% voids were considered in the computation of the storage reduction factor.

The computed modification factor was applied to improve the values that are generated

from (Eq. 6) and those values were further used in the application the combined SWAT

20
and HEC-RAS models for the modeling of sediment yield from the catchment area.

3. Results and discussions

3.1. Results from SWAT modeling

SWAT modeling was done for the whole catchment to produce output at the selected

gaging stations, for the application of the loosely coupled model of SWAT and HEC-

RAS. The outputs from the SWAT model were selected based on the idea of catchment

based and stream based application of the models. The calibration of the SWAT model

was done at the Melika Kunture station from 1995-2004 for a monthly based simulation.

The results from the SWAT model outputs are shown in Fig. 6. The R2 and NSE values

were 0.80 and 0.72, respectively. Similarly, the validation of the model at the same station

was done from 2005-2012 based on monthly simulation and the R2 and NSE values were

0.76 and 0.68, receptively and the results are shown in Fig. 7.

Based on the recommendation of Moriasi et al. (2007) the performance of the model for

catchment was in very good of good ranges. Thus, it can be used for practical application in

the study basin. Sediment calibration and validation was done based on suspended sediment

samples, with two basic steps. The first step was rating curve development from the sample

sediment data, with the application of the modification factor from the reservoir

sedimentation information. The second step was generating continuous sediment values

from the relation of the rating curve and monthly flow in the specific reach. The calibration

and validation of the sediment load was done for similar times as the calibration and

validation of flow. Figures 8 and 9 shows the results of calibration and validation of the

sediment model, respectively. The R2 and NSE values were 0.69 and 0.65, respectively.

Similarly, the validation of the model at the Melika Kunture station was done based on

21
monthly simulation and the R2 and NSE values were 0.60 and 0.55, respectively.

Fig. 6. Calibration of flow at Melika Kunture station

The SUIF-2 in the SWAT-Cup model, allows including more output variables to be

considered during the calibration and validation process. In this research, the calibration and

validation of the flow and sediment was done at the same time. When combined variable

calibration is used, the parameters to be selected need serious consideration because still

further studies are required to identify the influence of each parameter in SWAT on the

selected output variables (Abbaspour et al., 2007). For practical application the calibration

of all variables, i.e. flow, sediment, and nutrients (if available), simultaneously is suggested,

since joint calibration represents the actual physical process better than the calibration of

each variable separately. According to the performance indicators of model fit (Moriasi

et al., 2007), the calibration and validation yielded good performance for practical

application in the study basin.

Fig. 7. Validation of flow at Melika Kunture station

In order to apply the calibrated and validated model for the proposed case, the model has

to be verified with observed data at another location in the basin. The verification result at

the Hombole station near the inlet of the dam are shown in Figs. 9 and 10. The flow and

sediment verification results are monthly simulations from 2004-2014 in the basin. The

verification results indicate good correlation of the observed and simulated flow and

sediment data.

Fig. 8. Calibration of Sediment load at Melika Kunture station

22
 The calibration, validation, and verification of the SWAT model was done and the

results were satisfactory for practical application. Nevertheless, in this research, these

results can be further improved by coupling these outputs from the upstream section

(the catchment), with the HEC-RAS model in the stream channels.

Fig. 9. Validation of Sediment load at Melika Kunture station

Fig. 10. Verification of the flow model at the Hombole station

Fig. 11. Verification of the sediment model at the Hombole station

3.2. Results from the coupled modeling

After SWAT model calibration and validation for the whole basin, the system modeling

was subdivided into catchment-based and stream-based modeling for further

application. The outputs of the SWAT model at the selected upstream locations on the

stream channels were prepared to use as input to the HEC-RAS model and then a 1-D

HEC-RAS model was applied in the river channels. The calibration of the coupled

model was done using the Manning’s n roughness listed in Table 1 for the main channel

and floodplains. In the calibration process, channel characteristics also were important

for estimation of the initial roughness values. The initial values were taken by combined

consideration of channel characteristics and practical experience with similar models,

later further adjustment was done on the initial values to fit the simulated output of the

observed values. The results of flow modeling are shown in Fig. 12 with R2 of 0.85 and

NSE of 0.9. These values indicate very good performance of the model to simulate the

23
observed flow. The result of the model is for Hombole station near the inlet of the

reservoir for the daily flows. The summary of the overall performance of the models and

the indicators are listed in Table 2. The performance indicators listed in Table 2 for the

coupled model is primarily are to show that the coupled model can be used to model the

flow and sediment on daily basis with good performance, whereas the SWAT model

alone performed poorly because its basic applicability is primarily for long-term studies

in any catchment.

Fig. 12. HEC-RAS model simulation daily flow output at Hombole station after

calibration

The sediment results are shown in Fig. 13 and the model output indicates a good

estimate of the sediment load results from the modeling process. It is known that

sedimentation process modeling is complicated and it is difficult to get a good estimate.

However, the sediment output results from the model give good estimates for practical

application, nevertheless, it was not as good as the flow result. The results of modeling

the sediment process depend on the quality of the sediment data and the length of the

periods which the data covered, here the period was too short, hence the short record can be

considered as a main source of reduced accuracy.

Fig.13. Daily sediment load at Hombole station from HEC-RAS model after
calibration

24
 Table 2. Model performance indicators

Model Name Variable Event (R2) NSE P R Status

SWAT(Monthly) Flow Calibration 0.80 0.72 0.56 0.88 Good

- - Validation 0.76 0.68 0.45 065 Good

- Sediment Calibration 0.69 0.65 0.2 0.35 Good

- - Validation 0.60 0.55 0.15 0.2 Satisfactory

SWAT (Daily for 2004) Flow Calibration 0.65 0.6 0.4 0.55 Satisfactory

Validation 0.55 0.45 0.3 0.4 Satisfactory

Sediment Calibration 0.45 0.4 0.1 0.25 Poor

validation 0.45 0.35 0.1 0.2 Poor

Loosely coupled

SWAT, HEC-RAS Flow Calibration 0.85 0.90 - - Very good

(Daily for 2004) Sediment Calibration 0.67 0.62 - - Good

The estimation of the sediment load can be further improved by using the specified rainy

seasons from June to October in the computation of sediment load. The rainfall distribution

in the catchment shows that most of the rainy season is from June to October. It is

understood that if the flow is low or insignificant, the sediment loads also are minimal; thus,

sediment load in dry season can be ignored while computing the sediment load for those

periods. However, the statistical methods like sediment rating curve development from

sample suspended sediment data may not consider these events and thus, generate some

values of sediment load in the low flow periods, which may be sources of error in the

25
computation of sediment loads in the current research.

4. Conclusions

The estimation of sedimentation in reservoirs is important for the design and

management of reservoirs. There are different kinds of methods that can be applied for

practical application in estimation of the sediment load to reservoirs. Methods like the

loose coupling of the HEC-RAS and SWAT models can be used for good estimation of

the sediment yields from the catchment and the sediment load in river channels. The

approach requires relatively minimal data to execute the modeling task. The methods

requiring minimal data are preferable considering the scarcity of sediment data at most

gaging stations.

In the loosely coupled SWAT and HEC-RAS models, it can be possible to provide

the cross section of the channel including the floodplains for the computation of the

sediment transport. This situation gives an opportunity to consider the sediment

processes in the floodplains. Therefore, the approach gives better consideration for all the

sources and sinks of sediment in the process.

The sediment computation in any modeling work should include both suspended load

and bedload in the stream channels. In this research, only suspended loads were

measured at certain gaging stations. The bedload was considered using the rate of

sedimentation in the reservoir, however, there must be further concise methods that

should consider both suspended load and bedload. Erosion from the riverbanks and beds

are other sources of sediment that need further consideration in sediment modeling

26
 applications.

6. References

Abbaspour, K.C. (2015). SWAT-CUP: SWAT calibration and uncertainty programs user

manual. Eawag; Swiss Federal institute of Aquatic Science and Technology, Dübendorf,

Switzerland.

Abbaspour, K., Yang, J., Maximov, I., Siber, R., Bogner, K., Mieleitner, J., Zobrist, J., &

Srinivasan, R. (2007). Modelling hydrology and water quality in the pre alpine/alpine Thur

watershed using SWAT. Journal of Hydrology, 333, 413–430.

Ackerman, C.T. (2012). HEC-GeoRAS GIS tool for support of HEC-RAS using ArcGIS 10 user’s

manual. U.S. Army Corps of Engineers, Hydrologic Engineering Center Report CPD-83, Davis,

CA.

Ackers, P., & White, W. (1993). Sediment transport in open channels: Ackers and White update.

Technical note. Proceedings of the Institution of Civil Engineers-Water, Maritime and

Energy, 101(4), 247 - 249.

Ahemed, M., Sherfi, M., Saghyroon, E., Muna, M., & Semunesh, G. (2005). Assessment of the

current state of the Nile Basin reservoir sedimentation problem. Report of the Nile Basin

Capacity Building Network (NBCBN), Cairo, Egypt.

Ali, Y., Crosato, A., Mohamed, Y., Abdella., S. H., & Wright, N. (2014). Sediment balances in

the Blue Nile River basin. International Journal of Sediment Research, 29, 316 – 328.

Arnold, J., Kiniry, J., Srinivasan, R., Neitsch, S., & Williams, J. (2012). Soil and water

27
 assessment tool, input/output documentation. Version 2012, Texas Water Resources Institute

TR - 439, College Station, TX

Betrie, G.D., van Griensven, A., Mohamed, Y.A., Popescu, I., Mynett, A.E., & Hummel, S.

(2011). Linking SWAT and SOBEK using open modeling interface (Openmi) for sediment

transport simulation in the Blue Nile river basin. Transactions of the ASABE, 54(5), 1749-

1757.

Bingner, R. L., Theurer, F. D., & Yuan, Y. (2005). AnnAGNPS technical processes

documentation, version 3.2. U.S. Department of Agriculture, Agricultural Research Service,

Beltsville, MD.

Brunner, G. W. (2016). HEC-RAS, river analysis system hydraulic reference manual. U.S. Army

Corps of Engineers, Hydrologic Engineering Center (HEC). Report CPD-69, Davis, CA.

Brunner, G. W., Warner, J. C., Wolfe, B. C., Piper, S. S., & Marston, L. (2016). HEC- RAS,

river analysis system application guide. U.S. Army Corps of Engineers Institute for Water

Resources, Hydrologic Engineering Center (HEC). Report CPD-70, Davis, CA.

Deltares (2018). 1D/2D modelling suite for integral water solutions. SOBEK – Hydrodynamics,

rainfall runoff, and real time control. Delft, The Netherlands.

Donigian, A., Imhoff, J., Bicknell, B., & Kittle, J. L. (1984). Application guide for

hydrological simulation program-FORTRAN (HSPF). Technical Report EPA-600/3-84-

065, Environmental Research Laboratory, U.S. Environmental Protection Agency,

Athens, GA.

Duan, N. (1983). Smearing estimate: A nonparametric retransformation method. Journal of the

American Statistical Association, 78, 605-610.

Engelund, F., & Hansen, E. (1967). A monograph on sediment transport in alluvial

28
 streams. Technical University of Denmark Ostervoldgade 10, Copenhagen.

Ferguson, R. (1986). River loads underestimated by rating curves. Water Resources Research,

22, 74-76.

French, R. (1985). Open-channel hydraulics. New York: McGraw-Hill.

Garzanit, E, Andó, S., Vezzoli, G., Meged, A. A. A., & El Kummar, A. (2006). Petrology of

Nile river sands (Ethiopia and Sudan): Sediment budgets and erosion patterns. Earth and

Planetary Science Letters, 252, 327-341.

Gassman, P., Reyes, M., Green, C., & Arnold, J. (2007). The soil and water assessment tool:

Historical development, applications, and future research directions. Transactions of the

ASABE, 50, 1211 - 1250.

Gupta, H., Sorooshian, S., & Yapo, P. (1999). Status of automatic calibration for hydrologic

models: Comparison with multilevel expert calibration. Journal of Hydrologic Engineering,

4, 135-143.

Havnø, K., Madsen, M. N., & Dørge, J. (1995). Mike 11 - A generalized river modeling package.

In: V.P. Singh (Ed.), Computer models of watershed hydrology. (pp. 733-782) Highlands

Ranch, CO.: Water Resources Publications.

Jones, K., Berney, O., Carr, D., & Barrett, E. (1981). Arid zone hydrology for agricultural

development. U.N. Food and Agricultural Organization Irrigation and Drainage Paper No. 37,

Rome.

Keshta, N., Elshorbagy, A., & Carey, S. (2009). A generic system dynamics model for

simulating and evaluating the hydrological performance of reconstructed watersheds.

Hydrology and Earth System Sciences, 13, 865-881.

Laursen, E. M. (1963). An analysis of relief bridge scour. Journal of the Hydraulics

29
 Division, 89(3), 93-118.

Maidment, D., & Djokic, D. (2000). Hydrologic and hydraulic modeling support with

geographic information systems. Redlands, CA: ESRI, Inc.

Meyer-Peter, E., & Muller, R. (1984). Formulas for bed load transport. Proceedings of 2nd

meeting of the International Association for Hydraulic Structures Research, Delft, (pp. 39-

64).

Moriasi, D. N., Arnold, J. G., Liew, M. W., Bingner, R., Harmel, R., & Veith, L. (2007).

Model evaluation guidelines for systematic quantification of accuracy in watershed

simulations, Transaction of the ASABE, 50(3), 885-990.

Nash, J. E., & Sutcliffe J. V. (1970). River flow forecasting through conceptual models: Part 1. A

discussion of principles. Journal of Hydrology, 10(3), 282-290.

Neitsch, S.L., Arnold, J.G., Kiniry, J.R., & Williams, J.R. (2005). Soil and water assessment

tool: theoretical documentation. Technical Report, U.S. Department of Agriculture,

Agricultural Research Service, Grassland Soil and Water Research Laboratory and

Blackland Research Center, Tempel, TX.

Neitsch, S. L., Arnold, J. G., Kiniry, J. R., & Williams, J. R. (2011). Soil and water assessment

tool theoretical documentation version 2009. Texas Water Resources Institute TR-406,

College Station, TX.

Phillips, J., & Tadayon, S. (2006). Selection of manning’s roughness coefficient for natural

and constructed vegetated and non-vegetated channels, and vegetation maintenance

plan guidelines for vegetated channels in central Arizona. U.S. Geological Survey

Scientific Investigations Report 2006-5108.

Reed, S. M., & Maidment, D. R. (1995). A GIS procedure for merging NEXRAD precipitation

30
 data and digital elevation models to determine rainfall-runoff modeling parameters. Center

for Research in Water Resources Report 95-3, University of Texas at Austin.

Santhi, C., Arnold, J., Williams, J., Hauck, L., & Dugas, W. (2001). Application of a

watershed model to evaluate management effects on point and nonpoint source pollution.

Transactions of the ASABE,44, 1559-1570.

Schuol, J., & Abbaspour, K. C. (2006). Calibration and uncertainty issues of a hydrological

model (SWAT) applied to West Africa. Advances in Geosciences, 9, 137-143.

Shrestha, N. K., Leta, O. T., de Fraine, B., van Griensven, A., & Bauwens, W. (2013). OpenMI-

based integrated sediment transport modelling of the river Zenne, Belgium. Environmental

Modelling & Software, 47, 193-206.

Talebi, A., Bahrami, A., Mardian, M., & Mahjoobi, J. (2015). Determination of optimized

sediment rating equation and its relationship with physical characteristics of watershed in

semiarid regions: A case study of Pol-Doab watershed, Iran. Desert, 20(2), 135-144.

Toffaleti, F. B. (1968). A procedure for computation of the total river sand discharge and

detailed distribution, bed to surface, Technical Report No. 5, Committee on Channel

Stabilization, U.S. Army Corps Engineers.

U.S. Army Corps of Engineers (USACOE) (1990). Hydraulics design of spillways.

Engineering Manual EM 1110-2-1603, Washington, DC.

U.S. Army Corps of Engineers (USACOE). (2001). The Columbia River system inside story.

Technical Report Bonneville Power Administration and U.S. Bureau of Reclamation and

U.S. Army Corps of Engineers.

van Liew, M., Arnold, J., & Bosch, D. (2005). Problems and potential of auto calibrating a

hydrologic model. Transactions of the ASABE, 48(3), 1025-1040.

31
Vieira, D., Wu, W., & Wang, S.S.Y. (2018). Modeling hydrodynamics and channel

morphology using CCHE 1D. available online at http://www.researchgate.net/profile/Dalmo_Vieira

/Publication/228707628_MODELING_HYDRODYNAMICS AND CHANNEL_MORPHOLOGY_

USING_CCHE1D/ links/ 555ce70208ac86c06b5.

Walling, D. (1977). Assessing the accuracy of suspended sediment rating curves for a small

basin. Water Resources Research, 13, 531-538.

Walling, D., & Webb, B. (1981). The reliability of suspended sediment load data. Erosion and

Sediment Transport Measurement, International Association for Hydrological Sciences

Publication no. 133, 177-194.

Wilcock, P. R., & Crowe, J. C. (2003). Surface-based transport model for mixed-size

sediment. Journal of Hydraulic Engineering, 129(2), 120-128.

Williams, J. R., Arnold, J. G., Kiniry, J. R., & Gassman, P. (2008). History of model

development at Temple, Texas. Hydrological Sciences Journal, 53, 948-960.

Williams, J. R. (1995). The EPIC Model, In: V.P. Singh (Ed.) Computer models of watershed

hydrology. (pp, 909-1000), Highlands Ranch, Co.: Water Resources Publications.

Yang, C. T. (1984). Unit stream power equation for gravel. Journal of Hydraulic

Engineering, 110(12), 1783-1797.

Zhang, W., Wei, X., Zheng, J., Zhu, Y., & Zhang, Y. (2012). Estimating suspended sediment

loads in the Pearl River Delta region using sediment rating curves. Continental Shelf

Research, 38, 35-46.

32
33
34
35
36
37
38
39
40
41
42
43
44