ONE-DIMENSIONAL SEDIMENT TRANSPORT MODEL FOR THE

TAIACUPEBA RESERVOIR BASIN

Samuel Barsanelli Costa 1, Paolo Alfredini 2, Carlos Lloret Ramos 3

ABSTRACT – One-dimensional modeling has been successfully adopted since the 1980s in the
engineering practice to estimate long-term sediment transport. It is also an important tool for
understanding the sediment dynamics in watersheds and for managing sediments in reservoirs.
This paper aims to present a one-dimensional HEC-RAS sediment transport modeling for the
Taiacupeba reservoir basin, located in Sao Paulo, Brazil, focused to estimate long-term total loads
into the reservoir. The results were skewed low because of drought-like conditions, as data
collection period overlapped with the 2013-2015 drought in Southeastern Brazil. Such condition
contributed to the critical analysis of the results highlighting the use of one-dimensional sediment
modeling as an effective tool for validating sediment rating curves. The Laursen-Copeland and
Toffaleti transport capacity functions have been shown to be applicable to this river system, as
well as the Active Layer and Exner 7 bed sorting and armoring methods from HEC-RAS. Both the
sediment rating curves and Manning coefficients were considered preliminary and additional
field sampling at higher discharges are encouraged to enhance model accuracy and better
estimate loads.

Keywords – Fluvial hydraulics, sediment transport, one-dimensional model.

1) Institute for Technological Research – IPT. Postal address: 532 Professor Almeida Prado Avenue, Building 59, Butanta 05508-901, Sao Paulo/SP,
Brazil. Telephone number: +55 11 3767 4852. E-mail: sbcosta@ipt.br. (corresponding author).

2) Department of Hydraulic and Environmental Engineering, Polytechnic School of the University of Sao Paulo. Postal address: 120 Professor Lucio
Martins Rodrigues Avenue, Butanta 05508-020, Sao Paulo/SP, Brazil. Telephone number: +55 11 3091-5396. Email: alfredin@usp.br.

3) Hydraulics Technological Centre, Sao Paulo State Department of Water and Electric Energy. Postal address: 120 Professor Lucio Martins
Rodrigues Avenue, Butanta 05508-020, Sao Paulo/SP, Brazil. Telephone number: +55 11 3039-3201. Email: clramos@sp.gov.br.
XIII Brazilian Meeting of Sediment Engineering 1
I Particles in the Americas
1 - INTRODUCTION

Fluvial sediment transport is a natural phenomenon, which occurs in the form of suspended
loads and bed loads, being responsible for forming channels, river plains, and coastal areas. Rivers
are the main conductor of weathered materials from watersheds, and changes in channel
geometry or the introduction of obstacles to the flow, such as dams, affect the dynamic
equilibrium of the sediment regime.
Reservoir-related sediment studies in Brazil became popular after the joint effort of the
National Electric Energy Agency and the National Water Agency on publishing Regulation no.
003/2010 (ANEEL; ANA, 2010), which introduced sediment monitoring rules for hydroelectric
reservoir operators. The yearly monitoring program includes collecting four mandatory
suspended-sediment samples which shall be used to establish and update sediment rating curves.
However, its operational details do not cover the evaluation of rating curves representativeness,
such as estimating long-term loads and checking them against observed data.
Hydraulic modelling is one of the most used tools for assessing long-term changes in channel
morphology and sediment loads, and its derived data can be later used to evaluate reservoir
management alternatives (CASTILLO; CARRILLO; ALVAREZ, 2015). One-dimensional modeling has
been successfully adopted since the 1980s in engineering practices for this matter (SPASOJEVIC;
HOLLY JR., 2008), and the Hydrologic Engineering Center's River Analysis System – HEC-RAS one-
dimensional computing software (US ARMY CORPS OF ENGINEERS, 2016) stands out as a free and
widely used modelling suite (GIBSON; PAK; FLEMING, 2010).
This paper aims to present a one-dimensional sediment transport model for the Taiacupeba
reservoir basin, implemented on HEC-RAS, focused to determine the sediment transport regime of
its main tributaries and to estimate long-term total loads, which accounts for both suspended and
bead loads. Furthermore, the representativeness of sediment rating curves is discussed and
suggestions to the sediment sampling program proposed by Regulation no. 003/2010 are
presented.

2 - STUDY AREA

The Taiacupeba reservoir is located in the Upper Tiete River Basin, East from Sao Paulo City,
Brazil (Figure 1). The dam was built in 1976 on the Taiacupeba Acu river plain, a tributary of the
the Tiete River, originally designed for flood control to retain water from a drainage area of
224 km2. However, with the population growth of Sao Paulo Metropolitan Area in the 1980s, the
Taiacupeba reservoir was interconnected to a series of reservoirs through an artificial transfer
channel and its water begun to be treated and distributed to supply 5 million people from the East
Sao Paulo Metro Area.
XIII Brazilian Meeting of Sediment Engineering 2
I Particles in the Americas
 Figure 1 - Study area. Figure 2 - HEC-RAS river system schematic.

The upper Taiacupeba basin is occupied by native vegetation, while vegetable crops are
predominant in the middle and lower basin and a few urban developments spread along the
Taiacupeba Mirim river. Clayey texture soils are predominant in the area. Recent bathymetric
surveys in the reservoir were compared to its project design depth-area-volume relationship,
indicating a volume loss of 5x106 m3 (or 5 hm3) due to sedimentation, which corresponds to
approximately 5 % of storage capacity (COSTA; SOUZA; DEMARCO, 2017).
Rainfall is the main source of erosion and sediment transport in the basin, where a humid
subtropical climate dominates with an annual precipitation of about 1,500 mm. However, a
historical drought occurred in Southeastern Brazil during data collection, October of 2013 through
September of 2015. The drought was caused by an anomalous high pressure system that blocked
the flow of moisture from the Amazon to the South Atlantic Convergence Zone (MARENGO et al.,
2015), which is responsible for most of the rain in the area.

3 - MATERIAL AND METHODS

3.1 - Sediment data collection

There were no streamgages in operation in the study area, so discharge was measured
during sampling events only. Discharge was measured with a flowmeter by the Area-Velocity
method (BUCHANAN; SOMERS, 1976), at all sampling sites (Figure 1). The Equal-Width-Increment
method (EDWARDS; GLYSSON, 1999) and a US DH-48 hand sampler were used to collect the
XIII Brazilian Meeting of Sediment Engineering 3
I Particles in the Americas
suspended sediment samples. Bed material was collected using a Van Veen grab sampler. The
particle-size distribution of the suspended-sediment and bed material was assessed by sieving for
sand and gravel (≥ 0.0625 mm) and through sedimentation analysis for fine fractions, both
methods according to Carvalho (2008). Sampling events were determined according to the
weather forecast and real-time conditions broadcasted by meteorological radars, in order to cover
a wide range of flow conditions.
Sediment rating curves were set as upstream boundary conditions for the one-dimensional
sediment transport model. The sediment rating curves were developed as power functions in R,
using the “nlstools” package (BATY et al., 2015), by least squares non-linear regression method, in
the form of , where and are the regression coefficients, [mg/L] is the
3
suspended-sediment concentration (dependent variable) and [m /s] is the discharge
(explanatory variable). Randomness and normality of the residuals were evaluated graphically, the
association between these two variables was tested by Kendall’s Tau coefficient (KENDALL, 1975)
and the efficiency of the regression model was assessed by the R2 index, as proposed by Nash and
Sutcliffe (1970).

3.2 - Cross-sectional survey

All sampling sites represented the upstream limit of the modeled river system, while the
downstream boundary was set just before the reservoir backwater zone (Figure 2). Cross-sections
were surveyed using a laser rangefinder and a Differential Global Positioning System (DGPS)
receiver. The length of the each river was determined by aerial imagery at a scale of 1:10,000.
Mean slopes were estimated by leveling the first and last cross-section of each river stretch
through static Global Navigation Satellite System (GNSS) surveying using a L1 receiver with
integrated OmniSTAR® antenna for sub-metric correction. Observations (minimum of 600 at 1 Hz)
were later post-processed with base-station data located 50 km from the area.

3.3 - Model calibration and validation

The water surface profile calibration was based on adjusting the Manning coefficient by
comparing observed and modeled maximum water depths at each surveyed cross-section, for
different measured discharges at the upstream boundary and known reservoir pool levels at the
downstream boundary. Later, a sensitivity analysis was performed between different
combinations of the sediment transport capacity functions from Ackers-White (1973), Engelund-
Hansen (1967), Laursen (1958) modified by Copeland (1989), Toffaleti (1968) and Yang (1973) and
bed sorting and armoring methods Active Layer (HIRANO, 1971) and Exner 7 (COPELAND, 1993) to
select the set of parameters which lead to the most stable morphological evolution of the bed at
the end of the simulation (balance between erosion an deposition).

XIII Brazilian Meeting of Sediment Engineering 4

I Particles in the Americas
 In order to simulate flow in the tributaries over 41 years (between 1976 and 2017), daily
streamflow records of streamgages that operated between 1960 and 1976 were replicated
consecutively throughout the simulation. These streamgages were installed for the design of the
dam and their location was coincident to sites 1, 2, and 3. For the previously mentioned sites,
sediment rating curves were used to generate daily suspended loads, which were subdivided by
grain class according to the gradation analysis of the suspended-sediment. For site 4, sediment
rating curves did not make sense once flow and sediment are controlled by an upstream reservoir
in the artificial transfer channel. Mean values from the entire and dataset collected were
used instead.

4 - RESULTS

All raw and derived data are preserved in an online digital repository, available at
https://figshare.com/projects/Taiacupeba_Sediment_Model/25723, available without restrictions
as long as this paper is appropriately mentioned.

4.1 - Sediment regime

A total of 102 events were sampled to characterize the sediment transport regime,
averaging 20 samples per site (Table 1). During the data collection period the natural tributaries
contributed approximately 25 % of the incoming flow, whereas sediment load were predominantly
originated at its own drainage area. Sampled events covered a limited range of streamflow
conditions due to drought-like conditions.

Table 1 - Descriptive statistics of discharge and suspended-sediment concentration data

Site 1 2 3 4 5
2
Drainage area (km ) 68.3 35.7 97.8 - -
23 25 20 20 14
Minimum 0.254 0.064 0.351 1.746 0.323
3 Mean 0.763 0.295 0.932 8.452 1.128
(m /s)
Maximum 4.958 1.218 1.826 12.899 1.517
Stand. Dev. 0.982 0.214 0.420 2.727 0.354
Minimum 4.0 0.4 0.3 1.2 0.8
Mean 23.3 23.2 9.1 6.8 2.1
(mg/L)
Maximum 227.2 333.3 43.3 30.6 3.1
Stand. Dev. 45.9 65.1 10.0 7.0 0.6

Sediment rating curves (Table 2, Figure 3 - Figure 5) showed a good fit, with coefficients
being statistically significant at a 95 % confidence level, high positive correlation between and
(Kendall’s Tau higher than 0.65) and high model efficiency in predicting the dependent

XIII Brazilian Meeting of Sediment Engineering 5

I Particles in the Americas
variable (R2 greater than 0.85). The regression residuals did not adhere to the normal distribution,
however they were random and centered at zero.
Bed material for site 1 showed a d50 of 0.564 mm (coarse sand), 0.181 mm for site 2 (fine
sand), 0.329 mm for site 3 (medium sand) and 0.389 mm for site 4 (medium sand). Fine particles
(< 0.0625mm) were predominant in suspended-sediment samples (Figure 6).

Table 2 - Summary of the nonlinear regression analysis

Site 1 2 3
Coefficient
Estimate 26.350 1.344 208.783 2.371 6.687 2.921
Standard Error 1.038 0.026 2.701 0.047 1.137 0.360
-16 -16 -16 -16 -05 -07
p-value < 2x10 < 2x10 < 2x10 < 2x10 1.44x10 2.02x10
Confidence interval 2.5 % 24.191 1.289 203.195 2.272 4.298 2.164
Confidence interval 97.5 % 28.508 1.398 214.371 2.469 9.075 3.678
Kendall’s Tau 0.655 0.731 0.818
0.995 0.997 0.852

Figure 3 - Fitted sediment rating curve for site 1. Figure 4 - Fitted sediment rating curve for site 2.

Figure 5 - Fitted sediment rating curve for site 3. Figure 6 - Cumulative frequency distribution of
suspended-sediment (SS) and bed material samples.

XIII Brazilian Meeting of Sediment Engineering 6

I Particles in the Americas
4.2 - Sediment transport model

A total of 17 cross sections were surveyed and interpolated every 100 m to set the fluvial
system geometry in the model (Figure 2). Estimated mean slope for Taiacupeba Mirim river was
0.0013 m/m, 0.0016 m/m for Balainho river, 0.0009 m/m for Taiacupeba Guacu river and 0.0018
m/m for the transfer channel.
Simulations performed through Ackers-White, Engelund-Hansen and Yang sediment
transport capacity functions resulted in a processing error and unrealistic adjustment of the bed in
one or more cross-sections. It is important to note that the hydraulic characteristics of the studied
channels falls outside of the data range originally used to generate these functions.
The set of parameters that resulted in the most stable bed-change evolution were
Toffaleti/Exner 7 for Taiacupeba Mirim river (Figure 7), with coefficients ranging from 0.02 to
0.035, and Laursen-Copeland/Active Layer for the others (Figure 8 - Figure 10), where
coefficients ranged from 0.02 and 0.14 for Balainho river, 0.023 and 0.05 for Taiacupeba Guacu
river, and from 0.17 and 0.29 for the transfer channel. The highest values of were observed close
to the reservoir backwater zone, meanders or confluences, where higher resistance to flow is
expected.
The total sediment load produced during 41 years of simulation was 430,738.92 m3,
equivalent to 26.3 ton/day. Modeled bedload represented 4 % of the total load from Taiacupeba
Mirim river and 1 % from the other rivers. Rouse numbers along all channels were predominantly
below 2.5, indicating that suspension indeed dominated the overall transport. In terms of particle-
size distribution, modeled total load was composed of 3 % of sand (≥ 0.0625 mm), 13 % of clay
(< 0.004 mm), and 84 % of silt (≥ 0.004 mm and < 0.0625 mm).

Figure 7 - Profile plot of the sediment transport Figure 8 - Profile plot of the sediment transport
simulation for Taiacupeba Mirim river. simulation for Balainho river.

XIII Brazilian Meeting of Sediment Engineering 7

I Particles in the Americas
 Figure 9 - Profile plot of the sediment transport Figure 10 - Profile plot of the sediment transport
simulation for Taiacupeba Guacu river. simulation for the transfer channel.

5 - DISCUSSION

The drought-like conditions that overlapped with the period of data collection restricted the
range of streamflows at sampled events. The absence of sediment samples at high flow events
diminished the accuracy of the model (0.4 hm3 of modeled total load against an expectation of
5 hm3). As the model inputs accounted only for fine suspended-sediments at low concentrations,
there was no replacement of bed material to the system during simulation, inducing the armoring
of the bed hence reducing the overall transport.
However, monitoring under such conditions highlights the importance of assessing the
representativeness of the sediment data collected. Despite the extensive sampling carried out for
this research, with an average of 20 samples per site, and the statistical significance of the rating
curves, its use as an input of a dynamic hydraulic model evidenced that the depicted sediment
regime is representative for a dry period. In the present study, the one-dimensional model would
benefit from more representative sampling and a longer data collection period so samples could
be collected at higher discharges and model results would greatly benefit reservoir management.
It should be noted that setting a prior number of samples would not assure model accuracy,
and therefore the sampling program proposed by Regulation no. 003/2010 should be reviewed to
include the necessary tools to assess the sediment rating curves validity, such as the modelling
approach presented in this paper. Besides, the results suggest that it is pivotal to plan sampling
based on events (as opposed to calendar-based sampling) covering a wide range of flow
conditions (particularly high discharge events), and to indirectly validate sediment rating curves
through dynamic hydraulic models.
In addition, continuous monitoring of suspended-sediments with surrogate technologies
such as turbidimeters (RASMUSSEN, 2011), are encouraged for futures works and should also be
considered by regulators. If resources are scarce, at least one streamgage should be installed at
site 1, 2, or 3 and its continuous data should be regionalized to the others.
XIII Brazilian Meeting of Sediment Engineering 8
I Particles in the Americas
6 - CONCLUSIONS

The one-dimensional hydraulic model fulfilled the long-term sediment transport assessment
for the Taiacupeba basin, meeting the objectives proposed by the research. Both Laursen-
Copeland and Toffaleti transport capacity functions were applicable to this river system, as well as
the Active Layer and Exner 7 bed sorting and armoring methods from HEC-RAS. Sediment rating
curves and Manning coefficients should be considered preliminary due to the drought-like
conditions encountered during data collection and extra field sampling is encouraged in order to
provide more accuracy in the total load estimation.
Sampling sediment during a period of historical drought in Southeastern Brazil was beneficial
to the development of the study. It revealed that the proposed modelling approach is useful to
assess rating curves representativeness, hence suggesting topics to be considered when reviewing
Brazilian regulation on sediment monitoring programs.

ACKNOWLEDGMENTS

The authors acknowledge the East Water Resources Division of Sao Paulo Water Company –
Sabesp, for supporting data collection and providing technical information for this research, and
Joel Groten from the United States Geological Survey – USGS Minnesota Water Science Center, for
the technical review of the paper.

BIBLIOGRAPHY

ACKERS, P.; WHITE, W. R. Sediment Transport: New Approach and Analysis. Journal of the
Hydraulics Division, v. 99, n. HY11, p. 2040–2060, 1973.
AGENCIA NACIONAL DE ENERGIA ELETRICA; AGENCIA NACIONAL DE AGUAS. Resolucao Conjunta
n. 3, de 10 de agosto de 2010. Diario Oficial da Uniao, Brasilia, 2010.
BATY, F. et al. A Toolbox for Nonlinear Regression in R : The Package nlstools. Journal of Statistical
Software, v. 66, n. 5, p. 1–21, 2015.
BUCHANAN, T. J.; SOMERS, W. P. Discharge measurements at gaging stations. Washington, 1976.
cap. A8. 65p. (Techniques of Water-Resources Investigations of the U.S. Geological Survey, Book 3,
Application of Hydraulics).
CARVALHO, N. O. Hidrossedimentologia Pratica. 2.ed. Rio de Janeiro: Interciencia, 2008.
CASTILLO, L. G.; CARRILLO, J. M.; ALVAREZ, M. A. Complementary Methods for Determining the
Sedimentation and Flushing in a Reservoir. Journal of Hydraulic Engineering, v. 141, n. 11, p. 1–10,
2015.
COPELAND, R. R. Numerical Modeling of Hydraulic Sorting and Armoring in Alluvial Rivers. PhD
XIII Brazilian Meeting of Sediment Engineering 9
I Particles in the Americas
Thesis, University of Iowa, Iowa City, 1993.
COPELAND, R. R.; THOMAS, W. A. Corte Madera Creek Sedimentation Study. Sacramento: USACE,
1989. (Technical Report HL 89-6).
COSTA, S. B.; SOUZA, L. A. P. de; DEMARCO, L. F. W. Programa de monitoramento de controle dos
processos erosivos e de assoreamento do reservatorio de Taiacupeba, SP – Fase de alteamento
da barragem. Sao Paulo: IPT, 2017. 33p. (Relatorio Tecnico 150.607-205).
EDWARDS, T. K.; GLYSSON, G. D. Field Methods for Measurement of Fluvial Sediment. Reston:
1999. 89p. (Techniques of Water-Resources Investigations of the U.S. Geological Survey: book 3,
chap. C2).
ENGELUND, F.; HANSEN, E. A Monograph on Sediment Transport in Alluvial Streams.
Copenhagen: Teknisk Forlag, 1967.
GIBSON, S. A.; PAK, J. H.; FLEMING, M. J. Modeling Watershed and Riverine Sediment Processes
with HEC-HMS and HEC-RAS. In: WATERSHED MANAGEMENT CONFERENCE 2010.
Proceedings....Madison, Wisconsin: ASCE, 2010. v. 394. p. 1340–1349.
HIRANO, M. River Bed Degradation with Armoring. Proceedings of Japan Society of Civil
Engineers, n. 195, p. 55–65, 1971.
KENDALL, M. G. Rank Correlation Methods. 4.ed. London: Charles Griffin, 1975.
LAURSEN, E. M. The total sediment load of streams. Journal of the Hydraulics Division, v. 84, n.
HY1, p. 1530-1 até p.1530-36, 1958.
MARENGO, J. A. et al. A seca e a crise hidrica de 2014-2015 em Sao Paulo. Revista USP, n. 106, p.
31, 2015.
NASH, J. E.; SUTCLIFFE, J. V. River Flow Forecasting through conceptual models. Part I - A
discussion of principles. Journal of Hydrology, v. 10, p. 282–290, 1970.
RASMUSSEN, P. P. et al. Guidelines and Procedures for Computing Time-Series Suspended-
Sediment Concentrations and Loads from In-Stream Turbidity-Sensor and Streamflow Data.
Reston: 2011. 52p. (U.S. Geological Survey Techniques and Methods, book 3, chap. C4).
SPASOJEVIC, M.; HOLLY JR., F. M. Two- and Three-Dimensional Numerical Simulation of Mobile-
Bed Hydrodynamics and Sedimentation. In: GARCÍA, M. H. (Ed.). Sedimentation Engineering:
Processes, Measurements, Modeling and Practice. Reston: ASCE, 2008. cap. 15. p. 683–761.
TOFFALETI, F. B. A procedure for computation of the total river sand discharge and detailed
distribution, bed to surface. Vicksburg: USACE Committee on Channel Stabilization, 1968.
(Technical Report No. 5).
US ARMY CORPS OF ENGINEERS. HEC-RAS: River Analysis System, version 5.0.3. Alexandria, VA:
Institute for Water Resources, Hydrologic Engineering Center, 2016.
YANG, C. T. Incipient motion and sediment transport. Journal of Hydraulic Division, v. 99, n. 10, p.
1679–1704, 1973.
XIII Brazilian Meeting of Sediment Engineering 10
I Particles in the Americas