You are here

Chinese Taipei - Tsengwen/Zengwen

Key project features

Reservoir volume: 
748 Mm3 (original)
Installed capacity: 
50 MW
Date of commissioning: 

Tsengwen reservoir, the largest in Taiwan, is mainly used for rice field irrigation. The need to maintain a minimum volume in the reservoir is threatened by typhoons, which are the major cause of reservoir sedimentation. A two-pronged approach to sediment management has been planned, consisting of venting density currents during the monsoon and typhoon seasons, and dredging.

banner_aerial view.jpg

Aerial view of Tsengwen reservoir
Tsengwen reservoir (also spelled Zengwen) was impounded by the construction of a 134 m-high rockfill dam across the Tsengwen River. Owned and operated by the Taiwan Water Resources Agency, water from this reservoir is discharged to the off-channel Wushantou reservoir for rice field irrigation in the Chainan plain. It also provides flood control, potable water supply and hydropower.

The dam creates a 17.4 km-long reservoir with a surface area of 1,700 hectares at an elevation 270 m. The average reservoir width is about 590 m.

The project was designed by Nippon Koei Co. Ltd of Japan, with technical consultation by the US Bureau of Reclamation. Construction began in 1967 and filling began in April 1973. Figure 1 shows a plan view of the dam and the original outlet structures.

Table 1 summarises the elevation and capacity of all outlets. The 177 m3/s capacity of the permanent river outlet (PRO) is small in comparison with the gated surface spillways, at 9,470 m3/s. Thus, the reservoir tends to store turbid and release clear water, leading to a high sediment trap efficiency and rapid storage loss.

Hydrology and sediment

Tsengwen dam has a catchment area of 481 km2. The highest elevation of the watershed is 2,160 m, while the lowest is 102 m at the dam site. The principal land use is forest, with 77 per cent coverage in the watershed.

Average precipitation in the watershed between 1975 and 2010 was about 2,980 mm/yr, with 83.5 per cent occurring during the May to September wet season, producing the seasonal water level variations shown in figure 2. There have been considerable variations in the reservoir level given the variability in annual inflow. Although the average annual inflow over 1975-2010 was 1,157 Mm3, it varied from just 432 Mm3 in 1980 to 2,248 Mm3 in 2005.

Turbine discharge has averaged 902 Mm3/year, 79.3 per cent of the average annual total outflow of 1,137 Mm3. Outflows through the PRO and spillway accounted for 4 per cent and 16.7 per cent, respectively.

The relation of the dam to its watershed is shown in figures 3 and 4. Figure 3 illustrates the schematic representation of the drainage area, where 34 percent of the watershed drains directly into the reservoir, and accounts for about 29 per cent of the runoff and 21.5 per cent of the sediment load. Figure 4 depicts a plan view of the watershed as well as locations of rainfall gauging stations.

The predominate lithologies in the watershed are sandstone and shale, and most soils are weathered sandy shale. The topography is generally very steep, with an average soil slope of 54 per cent and an average river slope of 0.015. The original river slope near the dam site is 0.005. These geologic and topographic conditions result in high rates of erosion, extensive landslide activity during typhoons, and high sediment transport into the reservoir. With an annual average inflow of ~6 Mt/yr, sediment yield is extremely high, at ~12,500 t/km2/yr.

Sediment problems

Typhoon-induced landslides are the major cause of reservoir sedimentation. Reduced reservoir volume has increased the probability of farmlands to lie fallow and sediment and submerged logs have also clogged the intake trash rack and hindered operations.

Figure 5 charts reservoir volume since 1973. By the end of 2016, the reservoir volume had declined from 748 to 462.7 Mm3, only 62 per cent of the original volume, for a loss of about 6.6 Mm3/yr (nearly 0.9 per cent per year). About 32 per cent of this volume loss resulted from typhoon Morakot in 2009.

Sediment has been regularly sampled at the locations shown in figure 6, and the results are summarised below:

  • Dam to Section A-2: Most materials are high-plastic clay (CH) and silty clay (CL). D50 is in the range of 0.002~0.006 mm.
  • Section A-3 to A-8: Most deposits are non-plastic silt (ML), with a grain size between silt and clay.
  • Section A-12 to A-18: Considerable vertical variation in deposited material, with alternating layers of clay, silt and sand (CL, CH, ML, SM, SC).
  • Section A-20: Deposited materials are mostly silty sand (SM) or sand (SP). D50 is in the range of 0.1~0.9mm.

The total 35 upstream check dams (including six on the main stem) trapped the coarser bed material, and reservoir deposits consist primarily of sand and fine particles (silt and clay). Figure 7 depicts the average grain size distribution in the reservoir; about 86 per cent and 72 per cent are smaller than 0.1 mm and 0.04 mm respectively.

A master plan for of Tsengwen reservoir sedimentation control has been formulated, which involves a combination of watershed conservation, sediment sluicing and dredging to maintain critical volume."

The sediment classification is also plotted on figure 8, which shows bed profiles for selected years. Deposition at the dam has reached el. 175.0 m, which is substantially higher than the intakes for the power plant and the PRO, except for a depression near the dam due to recent dredging. Sediment beds are extended nearly horizontally from the dam to about 4 km upstream, indicative of density current deposits. 

Morakot Typhoon event

From 8-10 August 2009, Typhoon Morakot affected Chinese Taipei, producing average precipitation depths of about 2,550 mm in the Tsengwen reservoir watershed, and causing 1,467 hectares of landslides. The reservoir lost 91.08 Mm3 of capacity as a result. Figures 9 and 10 depict scenes of the reservoir following the event. In addition to sediment, the upstream end also saw the accumulation of about 800,000 m3 of material over a distance of about 1 km, mostly consisting of logs and tree branches. Intake clogging by debris created a sizeable vortex above the intake, as shown in figure 11.

Typhoons play a key role in reservoir sedimentation in Chinese Taipei. At Tsengwen reservoir, nine years of significant typhoon impacts accounted for 191.4 Mm3 of storage loss, about 75 per cent of total silt accumulation. Table 2 displays the major sedimentation years and the associated typhoons.

Several methods were used to estimate average annual sediment inflow, expressed in terms of volume loss, in order to develop sediment removal strategies:

  • During project planning, the suspended sediment inflow was estimated based on a gauging station rating curve located below the future dam site, and bedload was calculated by the Schoklitsch equation. Annual sediment inflow was estimated to be 4.97 Mm3 and 0.64 Mm3 for suspended load and bedload, respectively, for a total volume loss of 5.61 Mm3/yr.
  • Based on measured reservoir volume loss averaging 6.6 Mm3/yr from 1973 to 2008, correcting for sediment trap efficiency and treating the sedimentation by Typhoon Morakot as a 200 year event, the annualised total sediment yield was computed as 5.62 Mm3.
  • The numerical programme PSED (Physiographic Soil Erosion-Deposition Model), developed in Chinese Taipei, was calibrated to the Typhoon Morakot sediment data. The calibrated model was used to calculate sediment inflow at different frequencies, resulting in an average annual storage loss of 6.51 Mm3 based on occurrence probabilities.

From the above results, an annual volume loss of 5.60 Mm3 was adopted for planning.

Sediment transport investigation

Field sediment transport monitoring

Since 2004, measurements of suspended sediment concentrations have been conducted at the reservoir inlet and at the power/PRO intake. In 2011, four stations using ultra-sonic sensors were also added in the middle portion of the reservoir and additional sampling stations were installed at the spillway and the #1 diversion tunnel outlet. Below is a brief summary of the results:

  • At Dapu check dam (just above the upstream limit of the reservoir): A concentration of 68,000 ppm was measured during Typhoon Morakot, prior to evacuation of monitoring personnel for safety reasons. The peak concentration was undoubtedly higher. Inflow concentrations during other typhoons were generally less than 30,000 ppm.
  • In the reservoir: Up to 45,000 ppm was detected during Typhoon Sinlaku (2008), but in most cases sediment concentrations were generally less than 10,000 ppm.
  • At the intake: A peak concentration of 370,000 ppm was measured during Typhoon Morakot.

In general, a low initial water level coupled with high inflow volume tends to create highly concentrated flows at the intake. Otherwise, the concentration is low. Table 3 summarises data for seven significant concentration events at the intake.

The sediment monitoring data indicates that in most rainfall events there is not a well-defined density current that transports sediment to the dam and the intake. Several factors may contribute to this phenomenon. First, as indicated in figure 3, about 34 per cent of the watershed drains directly into the reservoir, and these tributary inflows may dissipate the longitudinal movement of the density current. Second, the average reservoir width is nearly 600 m and the reservoir floor is flat, without a main channel through which the density current can flow. As a result, the current spreads out across the width of the reservoir, increasing the opportunity for sediment deposition and dilution with clear water, further weakening its forward motion. Finally, the reservoir floor about 4,000 m upstream from the dam is essentially horizontal, and the lack of longitudinal slope further impedes density current movement to the dam.

Hydraulic model study

A 1/100 scale model was constructed to gain a better understanding of sediment transport phenomena and to aid in future operations of outlet structures. Results to date indicate the following:

  • Downstream of section A-7 (figure 6), the movement of density current is substantially reduced due to flat bed slope.
  • For a substantial density current to travel to intake, at Dapu check dam a discharge of 3,000 m3/s plus a sediment concentration of 30,000 ppm would be required. 

Sediment management strategies

Maintaining a proper reservoir volume is vital to southern Chinese Taipei’s agriculture, potable water supply and industrial water use. To aid in establishing a reasonable goal for managing sedimentation of the reservoir, a study was conducted to define a critical minimum reservoir volume. Hydrologic analysis using historical rainfall and run-off data suggested that a live storage volume of at least 370 Mm3 would be needed.

A master plan for of Tsengwen reservoir sedimentation control has been formulated, which involves a combination of watershed conservation, sediment sluicing and dredging to maintain the abovementioned critical volume. A 10 per cent reduction in the inflowing sediment load is judged achievable through soil conservation, leaving the remaining 5.04 Mm3 (90 per cent of 5.6 Mm3) to be removed by operations at the dam and reservoir.

Turbid density current venting

Modification of PRO for sediment release

The initially constructed PRO conduit bifurcates into two parallel branches, which have been used mainly for flow release when the power plant is not in service. Each is equipped with a six-vane 1.95 m diameter Howell-Bunger valve, but cracks have appeared in the vanes, in part due to partial blockage of the valves at their leading edges by debris. With a decision to utilise the PRO for venting density currents and the submerged muddy lake, the Howell-Bunger valves were replaced by fully-open jet-flow gates. This will increase the discharge capacity of the PRO from 150 m3/s to 177 m3/s.

Construction of new outlet tunnel through left abutment

Following considerable study, a decision was made to construct a 995 m3/s tunnel through the left abutment, 50 m to the left of #1 diversion tunnel, for the venting of turbidity currents (as shown on figure 12). Figure 13 depicts the plan and longitudinal profile of the project, and figures 14 and 15 show the intake and outlet. The project has the following unique features:

  • The intake is a 10 m diameter ‘elephant-trunk’ shaped steel pipe which, as shown on figure 14, connects the existing reservoir bed level at El.170 m to the intake invert at El.190 m. This elephant-trunk intake allows the outlet tunnel to be constructed at a higher level, eliminating the need for a high coffer dam during intake construction, since the elephant-trunk can be pre-fabricated and lowered into position underwater. This arrangement also reduces the pressure at which the gates must operate.
  • Discharge will be regulated by a 6.8 ´ 6.8 m radial gate, and an upstream vertical slide gate 6.8 m (width) ´ 7.7 m (height) is provided for maintenance or emergency closure.
  • A cavern 18 m (width) ´ 90.0 m (length at the base) is used as a plunge pool to dissipate flow velocities that can reach 30 m/s in the tunnel. Cavern height varies from 15.1 m to 47.4 m.
  • The plunge pool is connected to dual 10 m wide tunnels prior to discharging onto Tsengwen River below the dam.
  • Epoxy coating 8 mm thick is applied to the entire flow section upstream of the radial gate and to bottom and lower part of the tunnel to protect against erosion by sediment laden flow and prevent cavitation.

Construction of this project began in 2013 and completion is expected by the end of 2017.

Estimated quantity of sediment release

To enhance density current movement to the new elephant trunk intake, a dredged channel of 15 m deep and 100 m wide is recommended to help guide the density current. The slope of this channel upstream and downstream of section A-15 would be 0.0050 and 0.0031 respectively. Based on this configuration, 32 typhoon events from 1987 to 2011 were simulated to estimate the concentration of suspended sediment that would arrive at the outlet. Considering only the density current discharge through the bottom outlets, figure 16 shows the computed sediment discharge ratio versus bottom outlet discharge capacity. For an outlet discharge of 1,170 m3/s, the sediment discharge ratio (outflow/total inflow) would be about 0.28, not taking into account sediment venting after formation of a submerged muddy lake. Thus, as a conservative estimate, the amount of sediment to be vented by these two bottom outlets will average 0.28 ´ 5.04 = 1.41 Mm3/yr.

Mechanical excavation

Dry excavation at reservoir inlet

The Dapu check dam was constructed at the inlet to the reservoir in 1989. Sediment trapped upstream of this check dam is removed by dry excavation but is limited to 50,000 m3/yr due to highway conditions. From the check dam to reservoir section 22, channel deposits are mostly sands and in this area the excavation of 100,000 m3/yr has been planned. However, this excavation can only be conducted when the reservoir is at low levels, and therefore the volume that can actually be excavated will vary depending on water levels each year.

Dredging near intake

Since 2012 dredging has been conducted near the intake and in front of the dam, extracting an average of 375,000 m3/yr. Dredged sediment is temporarily placed into the plunge pool downstream of the spillway and into the river channel below the intake to Wushantou reservoir. These sediments are scoured and washed downstream by spillway discharge during the typhoon season. River monitoring to date indicates that this ‘return sediment to river’ operation has not caused measureable adverse impacts on river channel downstream. Beginning in 2017, the dredging rate is set to increase to 0.5 Mm3/yr, and it is anticipated to increase further with time.

Additional note

With the anticipated average annual sediment inflow of 5.04 Mm3, it is clear that annual suspended sediment sluicing of 1.41 Mm3, land excavation of 0.15 Mm3 and dredging near the intake at 0.5 Mm3 will not be able to sustain reservoir capacity. Studies are being conducted on an additional sediment sluicing tunnel together with more extensive dredging. However, this analysis is still at a preliminary stage. Further action will be needed.


table 1.jpg

Table 1 - summary of Tsengwen reservoir original outlet structures


Table 2 - major sedimentation years and associated typhoons at Tsengwen reservoir


Table 3 - summary of high concentration events at the Tsengwen dam

Graphs and figures


Figure 1 - overview of dam and original discharge structures


Figure 2 - seasonal water level variation in Tsengwen reservoir


Figure 3 - schematic representation of drainage area, runoff and sediment loading above and within reservoir boundary


Figure 4 - Tsengwen reservoir in relation to its watershed


Figure 5 - decline in gross storage, Tsengwen reservoir 1973-2015


Figure 6 - plan view of Tsengwen reservoir and sediment sampling location


Figure 7 - grain size distribution of deposited sediments


Figure 8 - deposition profile and material classification in Tsengwen reservoir


Figure 9 - driftwood accumulation at upper end of reservoir following Typhoon Morakot at about section 18


Figure 11 - driftwood accumulation at upper end of reservoir following Typhoon Morakot at about section 21


Figure 11 - vortex formation due to driftwood clogging at intake following Typhoon Morakot


Figure 12 - location of new sluice tunnel for turbidity current release


Figure 13 - plan and profile of new low-level outlet


Figure 14 - intake and gate shaft of new low-level outlet


Figure 15 - plunge pool and new outlet for sediment release


Figure 16 - sediment discharge ratio versus bottom outlet capacity

This is part of a series of sediment management case studies collated by the International Hydropower Association with support from the South Asia Water Initiative (SAWI), trust funds to the World Bank. For more case studies, visit

Download the dataset for the Tsengwen reservoir