Jiménez, O., Rodríguez C., y Olsen, N., (2004), “Sedimentation in the Angostura Reservoir:
Studies and Experiencias”, Nine Internacional Symposium on River Sedimentation (ISRS),
Chiang, China, October
SEDIMENTATION IN THE ANGOSTURA RESERVOIR:
STUDIES AND EXPERIENCES
Oscar JIMENEZ and Carlos RODRIGUEZ
Instituto Costarricense de Electricidad, San José, Costa Rica. E-mail: email@example.com
Nils R. OLSEN
The Norwegian University of Science and Technology. . E-mail: firstname.lastname@example.org
Abstract: This paper describes the sedimentological studies for the Angostura Power Plant,
Costa Rica. The plant, in operation since October 2000, is located in the middle basin of the
Reventazon River, and has a reservoir of about 15x106 m3. The river up to the project site has a
drainage area of 1463 km2, with large slopes and high precipitation, exceeding 6000 mm in some
areas, which results in a heavy sediment production. Since the initial planning studies, very
detailed surveys were carried out to quantify the sediments entering the future reservoir. This
information helped to design a strategy for handling and routing the sediments including
operational procedures, definition of suitable bottom outlets, definition of flushing procedures
and appropriate monitoring of the reservoir.
One of the most important topics of these studies was the computation of the sediment deposition
pattern along the reservoir, which has a very complex geometry. For this, several procedures
were employed, from empirical methods to a fully 3D hydrodynamic model able to compute the
flow and suspended sediment fields. These computations were carried out back in 1994 and at
that time several shortcomings of the computation were evident. Currently, with the development
of wetting/drying algorithms, the use of unstructured grids, and more efficient numerical
techniques, many of these problems have been overcome. Nevertheless, the results were useful
in planning the future operation of the reservoir. Prototype measurements for the first years of
operation are reported.
Keywords: Angostura power plant, Reservoir sedimentation, Numerical model, Flushing
The 180 MW Angostura Power Plant uses water from the Reventazón, one of the main rivers
in the Caribbean watershed of Costa Rica (see Fig. 1) . The plant includes a rockfill, 38 m
high dam, from which up to 160 m3/s are derived through a 6.2 km tunnel and a 0.6 km
penstock to the powerhouse. The catchment area is 1463 km2.
Upstream of Angostura H.P., there are two existing powerplants (Río Macho, 120 MW and
Cachí 105 MW), as well as a new one under construction (La Joya 50 MW). With a combined
capacity of 450 MW these plants represents about 25% of the current installed capacity of the
country (1900 MW).
Angostura H.P. has been planned as a daily peaking powerplant. Therefore a reservoir of
about 17 million m3 of total volume, and 11 million m3 of useful volume has been provided.
Streactly speaking, for daily peaking purposes, only about 2.5 million m3 are required, but
because of the serious sediment problem, a larger reservoir was provided. The drawndown is
7 m, between levels 570 and 577 masl; the dead storage goes from level 570 to the river bed-
level 552 masl.
Fig. 1 Location of the the Reventazón Basin and the Angostura H.P
The Reventazon watershed is characterized by high slopes; the river runs from the 3000
meters high central mountains to the Angostura diversion at 577 m in about 60 km. In
addition, there are very high precipitation, with an average of about 3500 mm per year, but
reaching 6000 mm in some parts of the basin.
Although about 40% of the basing is covered by tropical forrest, the river carries a very high
sediment load, of about 3.5 million tons per year, or 2600 ton/km2/year. If all those sediments
were deposited in the Angostura reservoir its live storage would be loss in just few years. In
reality, only a porcentage of about 60% of the total sediments will be deposited during the
first years of operation, quantity that will diminished as more volume is lost to sedimentation
and trapping efficiency is reduced. Computation by various methods indicates that the
reservoir will be lost in less than 20 years, unless special actions are taken.
Upstream of Angostura, the Cachí Power Plant has been operating since the sixtees. This
plant has a 50 million m3 reservoir, that since 1973 has been the subject of annual flushings
due to the high sediment load. These flushing operations has been very succesfull (Morris &
Fan 1997) and in 40 years only about 10% of the reservoir volume has been lost due to
sedimentation. The yearly amount of flushed material is about 500 000 t, most of which is
carried downstream affecting the Angostura Reservoir, some 15 km dowstream.
Because of these difficult sediment problems, during the planning of the Angostura H.P.
especial care was required in order to assure the useful life of the reservoir. This paper aims to
describe the studies, planning estrategies, and initial operational experiencies of the Angostura
2. SEDIMENT PRODUCTION
Sediments in suspension and flushed sediments from the upstream Cachí reservoir account for
about 3.5 million tons per year. Also, there is a seizable amount of bed load, although its
quantification has been only roughly estimated at about 10% of the suspended load. Fig. 2
shows an estimation of the sediment load at the dam site, based on a long term correlation
between liquid and solid discharge, for the period 1967-2000.
Sediment Load (ton)
Mean Annual Discharge (m3/s)
Sediment Load Discharge
Fig. 2 Mean flow discharge and sediment load at dam site
As mentioned, the flushing operations at the upstream Cachí dam contributes with about 500
000 t per year of suspended sediments. From 1973 to 2002 that reservoir has been flushed 24
times. During those operations, the Cachí reservoir is emptied completely and free discharge
conditions are mantained for about 5 hours. The resulting sediment pulse reaches peak
concentrations of about 300 000 to 400 000 ppm inmediately dowstream of Cachí. The
sediment travels downstream and after 3.5 hours arrives at Angostura, with peak
concentration of about 150 000 ppm. Afterwards, the sediment load diminshes to about 15
000 to 25 000 ppm for one or two additional days (see Fig. 3). During the first 12 hours after
the flushing starts in Cachí, about 300 000 t pass through Angostura.
Station 9-03 Angostura
Cachí-Angostura Flushing, Oct 2002
15 20 25 30 35 40 45 50 55 60 65
Discharge Sediment Concentration
Fig. 3 Discharge and sediment concentration after Cachí flushing, 1992
3. RESERVOIR SEDIMENTATION
Fig. 4 shows a general plan of the Angostura Reservoir. As seen, a large part of the live
storage is located in a rather flat zone upstream of the reservoir, where naturally the river has
presented a braided pattern. This is the zone where the water velocity will decrease and where
most of the sediment deposits will take place. The computation of the sediment deposition
patterns is quite complex, given the uncertainties regarding the real sediment load, very
variable year to year, and because the reservoir geometry, with a three dimensional nature.
For this project four different methods where tried out, from the simplest, to more complex
Fig. 4 Angostura reservoir, upper view
The semi-empirical Brune-USBR method (Schwarz 1995)
The one-dimensional HEC-6 Corps of Engineers Method (Jiménez 1993)
A vertically bi-dimensional method called RESP (Lovoll 1995)
A 3-D method, the SSIIM code developed by Olsen (1994)
It is worthwhile to mention the 3D SSIIM method. This program solves the Navier Stokes
equations over a structured mesh, using the k-epsilon model to represent the turbulent
viscosities. The discretization of the convective terms is done by a control volume approach
using a second order upwind method. The SIMPLE method is used for pressure correction.
Once the flow field is computed, the convection-diffusion 3D sediment equations are solved
for each of the sediment fractions considered. From these the trapping efficiencias as well as
the sediment depositional pattern are obtained. For the Angostura Reservoir, a grid of
54x9x11 (about 5000 cells) was used. Since the computation should run time frames of 20 to
30 years into the future, several simplifying assumption where deemed necesary at that time
A constant sediment load was assumed year to year
The real water hydrogram was simplified to just one discharge of 350 m3/s, which
corresponds to the yearly flood. It was considered that that discharge represents well the flow
patterns during floods.
A constant water level was assumed, and therefore, the effect of drawndowns was not
Erosion of sediments was not considered in the computations.
The sediment granulometry was represented with three fractions.
Large time increments of 1.4 year were used for water discharge computations. In between,
sediment computations were done with a small time increment, using a special routine. This
routine scales the bottom deposition according to the computed sedimentation pattern by a
factor of about 1000 or larger. If part of the deposits fills the reservoir to a level higher than
the water level, the sediments are redistributed to neightboring cells by a heuristic procedure.
Fig. 5 shows the expected changes of trapping efficiencias and remaining live storage, while
Fig. 6 compares the live storage prediction according to the different methods. Fig. 7 shows
superficial velocity vectors at t=0 and t=7 years.
Fig. 5 Live Storage and trapping effic., model SSIIM Fig. 6 Live storage predictions without flushings
0 5 10 15 20 25
Years of operation
Trapping Efficiency (%)
Storage Efic. (%)
0 5 10 15 20 25 30 35 40
YEARS OF OPERATION
LIVE STORAGE (hm3)
HEC-6 (1-D) RESP (2-D) SSIIM (3-D)
Fig. 7 Superficial velocity, left: initial condition; right: after 7 years, model SSIIM
Regarding the 3-D model used, it is worthwile to point out its main limitation in relation with
more recent development. One was that a steady computation with large time step was used.
This caused a lot of sedimentation above the water line, and a not so very scientific algorithm
had to be used to move these sediments back into the water. Nowadays SSIIM allows time-
dependent computations with fairly short time step and several thousands time steps during
one simulation. Another shortcoming was that the water level was kept constant and that the
flushing process was not modelled. Wetting and drying algorithms with non-structured grid
are under development, where the water level is changed during the computations and the grid
moves horizontally by removing/adding cells in the horizontal direction. (Olsen 2003). Other
important topic, not yet satisfactorily solved, is the bottom boundary conditions for sediment.
At that time, an equilibrium equation for sediment concentration after Van Rijn (1987) was
4. RESERVOIR OPERATION
Three operational policies where proposed to minimize the impact of sedimentation in the
Operation at low levels during the rainy season.
Annual emptying and flushing of the reservoir.
Close monitoring of the reservoir bathimetry.
The first policy is aimed to minimize the sediment deposition in the reservoir live storage,
between levels 577 a 570 m; otherwise sediments will deposits in the ample and shallow
upstream part of the reservoir. Clearly, the flushings will not help to reverse the sediment
deposition on those areas.
Figure 8 shows the acumulated curve of sediment load versus discharge for a typical year. It
is seen that discharge lower that 90 m3/s produce less than 6% of the load; discharges between
110 and 160 m3/s, produce a 20% additional load, and finally, discharges larger than 160 m3/s
(design discharge of the plan) are responsably for 74% of the load. Therefore, the operational
levels according to Table 1 were suggested.
Table 1. Operational Levels
It was estimated that the above operation policy will diminish in 50% de sediment deposition
in the live storage, and that the useful life would be extended from 20 to more than 30 years.
Of course, the dead storage would decrease rapidly, and for that, the flushings would be
required, specially in synchrony with the upstream Cachí flushing.
Because of the Cachí flushings, and because the Angostura reservoir itself requires flushings,
two ample bottom gates were provided, 4 m wide and 5.5 m high, with sill 13 m below the
water intake sill, as ilustrated in Fig. 9. The Cachí flushing experience has shown that this
operation is particularly useful in small reservoir, where it is possible to empty and to
replenish it in a few days
without excesive production losses.
To estimate the effectiviness of the flushing at Angostura, the semi-empirical method of Fan
& Morris was employed (1992). This computation showed that in about 40 hours of flushing
it may be possible to discharge about 400 000 t of sediment, provided that water discharge of
about 150 m3/s are available.
Fig. 8 Curve of accumulated sediment load Fig. 9 Bottom outlet gates, intake at right
5. MONITORING SYSTEM
The last important point was the establishment of a system of control and monitoring with the
aim of measuring the reservoir loss, of identifying zones of deposition, and of measuring the
sediments removed during flushings.
Fan, J., & Morris, G.L., (1992), “Reservoir sedimentation II: reservoir desiltation and long-term storage capacity”, Jour. of
Hyd. Engrg., vol 118, no 3, March
% of sediments
Since Angostura is a rather large generating station in the Costa Rican Electricity Network,
the flushings are scheduled during weekends, when electric load is lower. The emptying of
the reservoir takes about a day until free flow conditions are stablished. Then, free discharge
conditions are mantained for about 30 to 40 hours. During this operation, continuous
sediments samplings are taken, as shown in Fig. 3. For instance, during the flushing of 2002,
the total sediment load discharged from Angostura attained about 1 068 000 t, while during
2001 it was only 662 000 t.
Other important measure is the survey of the reservoir with modern ecosounding equipment.
Up to now about 10 surveys has been carried out, from december 2000 to October 2003. Fig.
10 shows the actual evolution of the reservoir capacity. A loss of 4.4% in that period has
5 6 7 8 9 10 11 12 13 14 15 16 17
Total Volume (hm3)
May 02 Dic 00
Fig. 10 Volume versus elevation
The topography of the Angostura reservoir, with a shallow and wide upstream plain, is very
prone to sediment deposition. That is in contrast with the upstream Cachí reservoir, where
flushing has been very succesfull. In two years of operation, Angostura reservoir has lost
about 2.4% of its live storage and 2.0% of its dead storage. This loss is more or less as
expected during planning.
It is important to point out the importance of adequate planning of reservoirs with regards to
sedimentation, which in many cases requires more data gathering and modelling effort that
the hydraulic design of the plant itself. It is also important to consider much more carefully
the downstream environmental impact because flushing.
AB Hydroconsult, (1995), Sedimentation and Erosion Processes related to the H.E. Projects Angostura,
Guayabo and Siquirres, Part 1, AB Hydroconsult, Uppsala, Sweden
Fan, J., & Morris, G.L., (1992), “Reservoir sedimentation II: reservoir desiltation and long-term storage
capacity”, Jour. of Hyd. Engrg., vol 118, no 3, March.
Morris, G.L., & Fan, J., (1997), Reservoir Sedimentation Handbook, McGraw-Hill, New York.
Jiménez O., et. al., (1993), Informe preliminar sobre la Sedimentación en el Embalse del P.H. Angostura,
Instituto Costarricense de Electricidad (in Spanish).
Lovoll, A., (1994), Calculation of water and sediment flow in the Angostura Reservoir, Costa Rica,
University of Trondheim, Norway
Olsen, N.R., (1994), SSIIM-A three dimensional numerical model for simulation of water and sediment
flow, Hydrosoft-94, Porto Carras, Greece.
Olsen, N. R., (2003) "3D CFD Modeling of a Self-Forming Meandering Channel", ASCE Journal of
Hydraulic Engineering, No. 5, May
Rodríguez, A., y Jansson, M., (1992), Estudios Sedimentológicos en el Embalse de Cachí, Costa Rica,
ICE (in Spanish)
Rodríguez, C., (2001), Informe sobre sedimentos Proyecto Hidroeléctrico Reventazón, Costarrican
Institute of Electricity (in Spanish).
Van Rijn, L.C. (1987), Mathematical modeling of morphological processes in the case of suspended
sediment transport, PhD Thesis, Delft University of Technology.