You are here

Chinese Taipei - Shihmen

Key project features

Reservoir volume: 
309 Mm3 (original)
Installed capacity: 
90 MW
Date of commissioning: 

Typhoons are the main cause of sedimentation in the Shihmen reservoir. A reduction in reliable water supply in the region led to the implementation of a combination of sluicing and dredging schemes to achieve long-term sediment balance. 

Shihmen - better quality.jpg

View of the Shihmen reservoir

Shihmen dam is a rockfill embankment across the Dahan River, near Taoyuan City in northern Chinese Taipei. Construction began in 1955 and the dam was completed in 1964.

Owned by the Taiwan Water Resources Agency, it provides irrigation services in Taoyuan, flood control for the Taipei basin, hydropower (90 MW), and domestic water supply for more than three million inhabitants in northern Chinese Taipei. The annual water use is in the range of 800 to 1,000 Mm3. It is also a popular tourist spot in the region.

The project was designed by a US company, Tippetts-Abbett-MaCarthy-Stratton (TAMS). As shown in table 1, the original design consisted of a surface spillway, two 4.5 m diameter penstocks, a low-level permanent river outlet (PRO), an irrigation canal outlet, and a 2 Mm3 afterbay. About 10 years after the initial construction, two tunnel spillways were added to increase discharge capacity. Figure 1 depicts a plan view of the dam and outlet structures.

Hydrology and sediment

The 763.4 km2 watershed above the Shihmen reservoir has steep slopes, high-intensity rainfall during typhoons, and a weak geology, resulting in high sediment load. The lithologies are mostly sale, slate and sandstone, all heavily weathered and very erodible. Average annual precipitation is about 2,300 mm and peak discharge per unit area for different recurrence periods are shown in table 2. The peak 100-year inflow would be 763.4 km2 by 11.4 cm/km2, or about 8,700 m3/s.  

Sediment problems

Typhoon Gloria deposited 19.5 Mm3 of sediment when the dam was newly constructed, and to control sedimentation, 121 check dams with a total storage volume of 35.8 Mm3 were constructed by different agencies on the Dahan River and its tributaries. These dams trapped mainly coarse materials, but as they filled, the sediment load entering the reservoir increased.

Reservoir bathymetry is normally mapped each year at the end of the wet season, and 34 permanent cross-sections are prepared from this mapping to track sedimentation conditions. Longitudinal bed profiles for selected years are plotted in figure 2. Above section 24, the reservoir is constricted to a relatively narrow gorge. From section 24 to 29, the bed level has risen 15~20 m since 2003, and the delta has migrated downstream. Figure 2 also shows the grain size from sampling the deposited sediments. With the exception of the inlet area, most deposits are fine sand, silt and clay. The average (D50) diameter near the dam is 0.008 mm.

The 2015 reservoir volume of 204.7 Mm3 represents about a third of reduction in the original volume. Figure 3 shows the loss in reservoir volume over time, showing the loss of 19.5 Mm3 of storage due to typhoon Gloria when the reservoir was newly constructed, plus the effect of typhoon Aere. Together, these two typhoons accounted for about 45 per cent of total sedimentation over 52 years of operation. 

On 24-25 August 2004, Typhoon Aere affected northern Chinese Taipei, delivering 973 mm of precipitation over the watershed in two days. The resulting landslides and runoff deposited 27.88 Mm3 of sediment into the reservoir, reducing its volume by 11 per cent.

In addition to the loss of reservoir volume, the submerged laky of muddy water that formed in the deeper part of the reservoir in front of the intakes was too turbid to be purified, leading to a suspension of potable water supply for 18 days during a hot and humid summer. Figure 4 shows that on or before 1 September, the reservoir had formed a submerged muddy lake below El. 207 m; turbidity was less than 100 NTU above the interface versus 100,000 NTU below. From 1 September to 30 September, this interface lowered by only 7 m, at an average rate of about 1 cm/hr. This is about 1/180 of the fall velocity of a discrete 0.008 mm diameter particle, the average grain size of deposits near the dam, reflecting the hindered settling in the non-Newtonia mud-flow. Figure 5 shows the consistency of the mud that was in the penstock, as seen via an access hatch.

Public pressure from the interruption of potable water supply caused the government to pass legislation for renovation of Shihmen reservoir. The scope of renovation included construction of a new surface intake for potable water release during turbidity events and development of the schemes necessary sustain long-term reservoir capacity.

Estimate of average annual sediment inflow

To develop a sustainable sediment management strategy, an estimate of the average annual sediment inflow is essential. Two methods were used. First, sampling data were used to develop a sediment rating curve suspended load Qs (kg/sec) versus inflow Q (m3/s). This was applied to the historical inflow time series to estimate the average suspended load entering the reservoir from 1964 to 2010 as 3.392 Mt/yr. Assuming a 15 per cent bed load contribution the total sediment is thus about 4 Mt/yr. Taking an average unit weight of 1.13 t/m3, the equivalent volume in the reservoir would be about 3.53 Mm3/yr.

Sediment inflow can also be estimated from the following:

  • Average annual volume loss: 2.16 Mm3/yr
  • Average sediment discharge from PRO, turbine and spillway: 0.47 Mm3/yr
  • Average annual dredging and excavation: 0.32 Mm3/yr (since 1985)
  • Estimated additional annual sediment inflow could be expected due to loss of check dam trapping capacity: 0.21 Mm3

The summation of the above is 3.16 Mm3. An average annual sediment inflow of 3.53 Mm3 has been adopted for planning sediment management works. 

Most sediment inflow is associated with typhoon rainfalls. As seen in table 3, 89 per cent of total sediment inflow from 1963 to 2005 was produced by typhoon rains occurring in nine typhoon years. Given these data, a sedimentation management strategy was selected based on developing outlets capable of releasing inflowing sediment during and immediately following typhoon floods.

Sediment transport investigation

Hydraulic model study

To help understand sediment transport behavior in the reservoir and evaluate alternative outlet structures for releasing sediment, a 1/100 scale undistorted hydraulic model was constructed at the Water Resources Planning Institute of Water Resources Agency, Ministry of Economic Affairs. Typhoon Aere was used as the input condition in all tests. The model clearly demonstrated sediment transport by turbid density currents and the formation of a submerged muddy lake in front of the dam. Though the model cannot completely simulate the prototype phenomena due to scaling effects, it did provide a good qualitative physical comprehension of the transport phenomena within the reservoir and expected function of proposed outlet schemes.

Field sediment transport monitoring

To document in-situ turbidity current behavior a TDR (Time Domain Reflectometry) technique was developed and applied in Shihmen Reservoir to track density current movement and sediment concentration profiles. This method is based on suspending an array of sensors in the water column, and by measuring the round-trip travel time of a pulsed electro-magnetic wave the sediment concentration at each sensor location can be determined in real time. The measured result requires compensation for water temperature but is independent of particle size. Several monitoring stations were installed along the reservoir. One of these stations is shown in figure 6.

Monitoring data from typhoon Fungwong (figure 7) shows the time variation of suspended sediment concentration at the reservoir inlet, Section 24, and the irrigation canal intake. Reduction in peak concentration as the current flows along the length of the reservoir is clearly evident. In this typhoon it took about nine hours for the turbidity current to flow from the reservoir inlet to the dam, and the peak concentration decayed from about 28,000 ppm at the inlet, to 18,000 ppm at section 24, and then to 8,000 ppm at irrigation intake due to sediment deposition. The longitudinal deposit profiles previously shown in figure 2 reflect this deposition pattern. Figure 8 shows the vertical variations in sediment concentration at section 24 over time.

Sediment management strategies

There is no suitable alternative dam site to replace the function of Shihmen reservoir. Therefore, the only viable strategy is to sustain storage capacity in the existing reservoir. Based on sediment inflow and other characteristics of this reservoir, the following management strategies were selected to balance the long-term sediment inflow and outflow while preserving reservoir storage.

Watershed soil conservation

Typhoon-induced landslides are an important source of sediment, but studies showed these to mostly be due to natural process in non-accessible mountain regions. Therefore, soil conservation work to reduce sediment production from this source will be very limited. A reduction of sediment load by only 0.1 Mm3/yr has been allocated to watershed management.

Venting of turbid density currents and submerged muddy lake

The dam was designed ‘to store turbid water and to release overflow clear water’, a strategy which has maximised sediment trapping in the reservoir. Since this water supply reservoir cannot be lowered for flushing or drawdown sluicing, the adapted strategy has been to create new outlets for releasing turbid density currents as they flow along the reservoir bottom and also after they pond to create a submerged muddy lake at the dam.

Convert Existing Penstock to Vent Turbid Density Current

The reservoir was built with two 4.5 m diameter steel-lined penstocks feeding two Francis turbines. Sediment has now risen to a level slightly higher than the penstock intakes (see figure 2), meaning these intakes will be submerged by the muddy lake formed by typhoon-generated turbidity currents.

Figure 9 depicts the plan and longitudinal profile of the penstocks. To convert to sediment sluicing, penstock #1 was modified to power both turbines, while penstock #2 was converted to a 300 m3/s sediment sluice by removing a section and diverting into to a new 3.6 mf steel pipe, as shown in Figure 10. Within the new gate house, the 3.6m diameter steel pipe was bifurcated into two flow passages, each equipped with an upstream vertical gate, a downstream jet-flow gate and a flip-bucket terminal structure. Flow is discharged into the afterbay. Figure 11 shows a 3D view of this sediment release structure.

Construction was completed in 2012, and figure 12 shows the initial sediment sluicing operation in 2013 during Typhoon Soule. The contrast in discharge between the surface spillway and the silt sluiceway is obvious. Since 2013, turbid density currents have been vented through the sediment sluice during five medium to small scale typhoons, discharging about 1.575 Mt. As compared to the cost of dredging this same volume of sediment, the project benefit has already exceeded its construction cost.

Dawanping Silt Sluice Tunnel

To further enhance silt sluicing capability during typhoons, Dawanping was selected as the intake location for an additional sediment venting tunnel based on both numerical simulation and physical model testing. Figure 13 shows the overall layout of this project, which is not yet constructed.

The reservoir bed elevation at this intake site is about El. 195 m, which is 50 m below the normal water level of El. 245 m. To avoid constructing an intake cofferdam to this depth, two 10 m diameter ‘elephant-trunk’ steel pipes are planned to extend from the tunnel invert at El. 220 m down to the reservoir bottom, thereby withdrawing from the level of the turbidity current (see figure 14). A similar ‘elephant-trunk’ system has been successfully constructed and installed at another silt sluice tunnel project in southern Taiwan. This system is designed to discharge 1,600 m3/s by two parallel intakes and tunnels, regulating flow through from each passage by a radial gate.

Excavation and dredging

As turbid flow entering into a reservoir, coarser materials deposit near the reservoir inlet while finer sediments are deposited between the delta and the dam. Since flushing is not an option, dredging is the only means to remove these deposits. Dry excavation has been ongoing at upstream end of the reservoir since 1977 and dredging near the power intakes since 1985. However, the current dredging rate does not keep pace with sedimentation and additional dredging capacity must be added.

Dry excavation at reservoir inlet

Coarse sediment deposits at the reservoir inlet have been excavated over eight months of each year since 1977 using conventional earth moving equipment. Limited by road conditions, the plan is to remove at a rate of 400,000 m3/yr.

Amuping material handling facility

Core sampling of deposited material revealed that a change in grain size occurs in the vicinity of section 20 (figure 2). Upstream of this station the material is significantly coarser than downstream. About 40 Mm3­­­ of sediment is deposited between section 20 and the area of dry excavation. About half of this is sand which may be used in concrete aggregates. These materials can only be removed by dredging. Due to lack of storage space adjacent to the reservoir, the facilities shown in figures 15 to 18 were designed to transport the dredged slurry, to sort out useable sand, and dispose silt onto the Dahan River below the dam. This project has the following main components:

  • A 4 km long dual-purpose horseshoe tunnel (8 m wide x 7 m high) will convey dredged material to the sorting facility near the tunnel outlet via four 30 cm diameter slurry pipes installed near the crown of the tunnel. The lower half of the tunnel can discharge up to 600 m3/s during typhoon events.
  • A silt sorting facility is designed to segregate silt and sand. Sand will be used as aggregate while silt will be temporarily stored in a silt detention basin 600 m long and 62 m wide with storage capacity up to 200,000 m3.
  • The tunnel outlet discharges into the detention basin to flush accumulated sediment into the Dahan River downstream of the afterbay. Flushing water will come from pre-typhoon reservoir drawdown to empty a flood control pool, plus typhoon flood water.

Construction of this project began in June 2017 and is expected to be completed by 2021.

Dredging Near Power Intake

Sediment removal near the dam and power intake has been performed by a hydraulic dredge operating in water about 70 m deep with a planned removal rate of 500,000 m3/yr. The dredged fine silt (D50 0.008 mm) is transported by pipe to ponds located downstream of the afterbay for dewatering and subsequent trucking to disposal sites.

Sediment management strategies summary

Table 4 shows the planned long-term sediment discharge from the reservoir by each method needed to stabilise the volume of Shihmen reservoir. On average, about 55 per cent of the sediment inflow will be discharged through the PRO, turbidity current penstock sluiceway, Dawanping tunnel, and other outlets. The remaining 45 per cent will be trapped and subsequently removed by excavation of the delta, dredging in the upstream reach via the Amuping tunnel, and deep dredging near the dam.  


table 1.jpg

Table 1 - Shihmen reservoir outlet structures

Capture 2.jpg

Table 2 - estimated peak runoff from Dahan watershed above Shihmen dam for various return periods

Capture new table.jpg

Table 3 - major sedimentation events in Shihmen reservoir and associated typhoons through 2005


Table 4 - planned sediment discharge methods for Shihmen reservoir

Graphs and figures


Figure 1 - plan view of Shihmen reservoir outlet structures prior to modification for sediment release


Figure 2 - variation in longitudinal profiles for Shihmen reservoir over time due to sedimentation

Volumeloss_colours corrected.jpg

Figure 3 - Shihmen reservoir volume loss from 1963 to 2015


Figure 4 - turbidity in front of intake following Typhoon Aere, 2004


Figure 5 - consistency of residual mud that collected in the penstock during Typhoon Aere, as seen in an access hatch


Figure 6 - platform for TDR monitoring array at Shihmen reservoir illustrating method of sensor deployment


Figure 7 - turbidity current characteristics in Shihmen reservoir during Typhoon Fungwong, 2014


Figure 8 - vertical variation of sediment concentration at section 24 during Typhoon Fungwong, 2014


Figure 9 - general features of penstocks


Figure 10 - penstock modifications for sluicing of turbid density currents


Figure 11 - 3D view of penstock converted for silt sluicing


Figure 12 - simultaneous operation of spillway and penstock sluice venting turbid density current during Typhoon Soule


Figure 13 - plan view of Dawanping tunnel for venting turbid density currents


Figure 14 - Dawanping tunnel showing 'elephant trunk' type intake


Figure 15 - plan view of Amuping tunnel


Figure 16 - intake and cross-section of Amuping tunnel


Figure 17 - Amuping tunnel profile


Figure 18 - Amuping tunnel outlet and temporary storage basin

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 Shihmen reservoir.