Sediment Flushing Operation Mode During Sediment Peak Processes Aiming Towards the Sustainability of Three Gorges Reservoir

Abstract

Asynchrony between the movement of water and sediment in a reservoir will affect long-term maintenance of the reservoir’s capacity to a certain extent. Based on water and sediment data on the Three Gorges Reservoir (TGR) measured over the years and a river network model, optimization of the dispatching mode of the reservoir’s sand peak process was studied, and the corresponding water and sediment dispatching indicators were provided. The results show that (1) sand peak discharge dispatching of the TGR can be divided roughly into three stages, namely the flood detention period, the sediment transport period, and the sediment discharge period. (2) According to the process of the flood peak and the sand peak, a division method for each period is proposed. (3) A corresponding scheduling index is proposed according to the characteristics of the sand peak process and the needs of flood control scheduling. This research can provide operational indicators for the operation and management of the sediment load in the TGR and also provide technical support for sustainable reservoirs similar to TGR.

1. Introduction

The Three Gorges Reservoir (TGR) is a key backbone project for managing and developing the Yangtze River. It has significant comprehensive benefits such as flood control, power generation, navigation, and water resource utilization. The sediment problem is one of the key technical challenges in the construction and operation of the TGR [1,2,3,4]. Sedimentation in the TGR affects the reservoir’s lifespan and inundation and the evolution of the channel and port area at the reservoir’s end, drawing considerable attention [5,6]. In the demonstration and preliminary design stage, to ensure an effective long-term storage capacity, the operation mode of “storing clear water and releasing muddy water” was developed [7,8]. During the flood season (from mid-June to September), when the sediment input is high, the reservoir is maintained at a low water level (145 m) to preserve its flood storage capacity, manage potential floods, and maintain a large water surface gradient to support sediment discharge. In the case of large floods, flood storage operations are carried out; after the flood season, when the sediment load decreases in October, its water storage begins to reach 175 m [9]. Similarly, in the Yangtze River Basin in China, the Mississippi River Basin in the United States, the Wylye River Basin in the United Kingdom, and the Petzenkirchen River Basin in Austria, there are obvious asynchronous propagation phenomena of flood peaks and sand peaks. Due to an imbalance between water and sediment, the asynchronous propagation of flood peaks and sand peaks can lead to problems such as “scouring in the flood season” and “silting in the flood season” [10,11,12].

Since the operation of the TGR, due to upstream rainfall, sediment retention, soil and water conservation, and other factors, the sediment inflow to the reservoir has significantly decreased [13,14,15]. The sediment transport of the main upstream tributaries, Jinsha River and Jialing River, has significantly decreased, while the interception of various reservoirs; soil and water conservation in the basin; and the combined effect of river sand mining are the main reasons for changes in water and sediment [16]. In the TGR, the flood season water levels set in the preliminary design are adjusted and optimized based on changes in water and sediment conditions and the multi-objective needs of operation. Under manageable flood risk conditions, the reservoir uses its flood control capacity to keep the discharge below 56,700 m³/s and conducts medium and small flood operations as appropriate. This not only improves the utilization of flood resources and enhances power generation and navigation benefits but also reduces the flood control pressure in the middle and lower reaches of the Yangtze River.

However, with the implementation of medium and small flood operations, reservoir water levels rise during the flood season, and the duration above the flood limit increases, resulting in lower sediment discharge ratios and relatively greater sediment accumulation in the reservoir [17,18,19]. Thus, it is necessary to study how to reduce the risk of siltation caused by these operations [19]. As the TGR is a typical river-type reservoir, the sand and flood peaks are asynchronous, with the sand peak usually lagging [20]. An increase in the water level in front of the dam reduces the flow velocity and sediment-carrying capacity, slows down the propagation speed of the sand peak, and continuously reduces the peak value along the way, ultimately resulting in the sand peak in the reservoir area lagging behind the appearance of the flood peak for a longer period of time along the way [21]. Using this difference, the TGR, when managing small and medium floods, uses the lag between the flood and sand peaks to schedule sediment operations—cutting the flood peak to reduce downstream flood pressure and increasing the discharge when the sand peak approaches to release sediment [22]. Since 2012, the TGR has carried out flood season sand peak sediment discharge scheduling experiments and drawdown period tail sediment reduction scheduling experiments, which have had a good effect on reducing effective storage capacity sedimentation [1]. This has been a beneficial attempt to develop the “storing clear water and releasing muddy water” model.

Currently, sediment operations in the TGR rely on real-time sediment monitoring and forecasting. When the sand peak reaches approximately 60 km from the dam, the reservoir discharge is increased. As the sand peak propagates from the reservoir’s end to the dam, its transport speed and attenuation depend on the flood peak process and reservoir operation. Therefore, it is necessary to manage the entire sand peak transport process and further optimize its operation into a full-process management strategy. Based on the asynchronous propagation of flood and sand peaks and considering the sediment transport characteristics and flood control needs, this study classified the operation stages; calculated and analyzed the impact of different operations at each stage on the sediment transport and erosion–deposition changes; and proposed an operational index for sand peak management in the TGR. It provides technical support for efficient sediment regulation and sustainable utilization of the TGR and other similar reservoirs.

2. Materials and Methods

2.1. The Study Area

The TGR is a key backbone project for the governance and development of the Yangtze River. Its sediment problem is one of the key technical issues that runs through the demonstration, design, construction, and operation stages for the TGR and has always been highly valued. Extensive, in-depth, and continuous hydrological and sediment prototype observations and research have been conducted at different stages of the construction and operation of the TGR, targeting key sediment issues and regions, providing rich basic data for this study. This study focuses on the TGR area, starting from Zhutuo and ending in front of the dam, as shown in Figure 1. The inflow, sediment transport, characteristic water levels in the reservoir, and other information on the TGR during the sand peak discharge scheduling test are presented in

Table 1. Statistics on the effect of the sand peak discharge scheduling test.

Year	Time of Sand Peak Entering the Reservoir	Amount of Sand Entering Reservoir/104 t	Time of Sand Peak Emerging from Reservoir	Time	Amount of Sand Emerging Reservoir/104 t	Water Level Before Dam/m
2012	14 July to 29 July	5530	45,100	20 July to 12 August	1967	158.77
2013	11 July to 18 July	5740	32,000	19 July to 26 July	1760	150.00
2018	11 July to 17 July	7440	51,400	15 July to 25 July	2144	152.94
2020	13 August to 24 August	12,650	52,200	17 August to 24 September	3390	160.81

.

2.2. Data

The data for this study are the daily averaged flow discharge (Q) and the sediment concentration (SC) measured at hydrological stations within the study area. The data collected in this article are from the Hydrology Bureau of the Yangtze River Water Conservancy Commission. By setting up the corresponding flow- and sediment-measuring equipment in the control hydrological station, the water level (Z), Q, and SC data for the control section are monitored and controlled (

Table 2. Monitoring data.

Name	Data Type	Time Series	River (Lake)	Source
Zhutuo	Q, SC, Z	1900~2022	Yangtze Rive	Bureau of Hydrology, Changjiang Water Resources Commission
Beibei			Jiangling River
Wulong			Wu River
Cuntan			Yangtze Rive
Qingxichang
Wanxian
Huanglingmiao		2002~2022

). Notably, hydrological tests of the control stations usually follow the principle of intensive observation during the flood season. For example, during the period when the incoming water and sediment change significantly in the flood season, intensive observation is required at the hourly scale. In general, water and sediment monitoring data will be compiled into daily-scale data by means of compilation. Among these, the data measured at Zhutuo (ZT), located at the entrance the study area, represent the input flow and sediment from upstream of the reservoir. There are three hydrological stations located in the reservoir area, namely Cuntan (CT), Qingxichang (QXC), and Wanxian (WX), used to monitor the spread of the flow and sediment in the reservoir. The characteristics of the flow and sediment discharge from the TGR are represented by the data measured at Huanglingmiao (HLM) station downstream of the dam. Those for the two major tributaries, Jialing River and Wu River, flowing into the TGR are represented by the data measured at Beibei (BB) and Wulong (WL) stations, respectively. The distribution of each hydrological station is shown in Figure 1.

2.3. Methods

2.3.1. Definition of Sand Peak Characteristics

The steepness or flatness of the sand peak is an important indicator of sediment concentration processes [23]. The commonly used kurtosis coefficient is expressed as $\eta =S_P/S_A$, where $S_P$ is the peak value of SC and $S_A$ is the average SC.

To account for variations in sand peak durations across floods, the waveform coefficient of sand peak propagation was calculated by introducing the duration parameter $k=S_P/T$, where $T$ was the sand peak duration. During sand peak propagation, attenuation occurs due to deposition, while flattening results from longitudinal dispersion [24,25,26]. The flattening rate is represented by the change in the waveform coefficients between upstream and downstream control stations: $p=k_{\mathrm{u}\mathrm{p}}/k_{\mathrm{d}\mathrm{o}\mathrm{w}\mathrm{n}}$.

2.3.2. The Numerical Model for Sand Peak Scheduling

The main and tributary structures in the study area present a tree-like river network; therefore, this study adopts a river network model [27] to simulate the sand peak scheduling. This model has good simulation accuracy for the reciprocating flow and suspended sediment transport in tidal river sections. In this study area, except for the significant differences in bedload sediment movement and sediment flocculation compared to those in tidal river sections, the other sediment movements are basically similar to those in the tidal river sections. Therefore, special treatments were used for bedload movement and sediment flocculation in this study, while the methods used in [27] were applied to the other sediment movements.

The calculation range of the water and sediment transport model in the TGR area is from Zhutuo to the front of the dam. The length of the main stream is about 760 km, and 13 tributaries, including the Jialing River, Wujiang River, and Xiaojiang River, are considered. A total of 411 sections were set up in the main stream and tributaries of the Yangtze River, with a section spacing ranging from 300 to 1600 m.

• The bedload sediment transport rate

This study adopts a formula for the bedload sediment transport rate in the upper reaches of the Yangtze River with high accuracy [28]:

$${\displaystyle \frac{V_d}{\sqrt{gd}}}~{\displaystyle \frac{g_b}{d\sqrt{gd}}}$$

Among these, $V_d={\displaystyle \frac{m+1}{m}}{\left({\displaystyle \frac{H}{d}}\right)}^{-{\displaystyle \frac{1}{m}}U}$, $m=4.7{\left({\displaystyle \frac{H}{d_{50}}}\right)}^{0.06}$.

In the formula, $g_b$ is the bedload sediment transport rate, $g$ is the gravitational acceleration, $d$ is the sediment particle size, $H$ is the water depth, $U$ is the flow velocity, and $d_{50}$ is the median particle size.

• The impact of sediment flocculation

At present, most research on sediment flocculation has been focused on estuary areas, where the main factor affecting flocculation is salinity. For this study area with a large water depth, the coefficient $F$ was used to adjust the sediment settling velocity [29]:

$$F=0.0013\times d^{-1.9}$$

To consider the factor of the flow velocity further, this study made corresponding improvements to bring it closer in line with the actual application in the TGR:

$$F=\left\{\begin{array}{cc}{\displaystyle \frac{U}{0.3}}\times 0.0013\times d^{-1.9}& U<0.3\mathrm{m}/\mathrm{s}\\ {\displaystyle \frac{0.6-U}{0.3}}\times 0.0013\times d^{-1.9}& 0.3\mathrm{m}/\mathrm{s}\le U<0.6\mathrm{m}/\mathrm{s}\\ 1& 0.6\mathrm{m}/\mathrm{s}\le U\end{array}\right.$$

• Model validation

Due to the extreme floods that occurred in the Yangtze River Basin in 2020 [30], this study selected data measured from January 2019 to March 2021 to validate the model. The calculation period covered the flow and sediment processes in 2020, and the results are shown in Figure 2. It shows that the calculated SC at the CT and BD stations is basically consistent with the measured values, indicating that the parameters used in this model are reasonable and can be used to simulate the sand peak process in the TGR. The maximum error in the sand peak value is 20%, and the phase difference in the peak time is less than 12 h.

3. Results

3.1. Propagation Characteristics of the Sand Peak in TGR

3.1.1. Asynchronous Propagation Characteristics of the Flood Peak and the Sand Peak

Due to the effects of different mechanisms of sediment production and loss in the upper reaches, different sources of water and sediment, bank collapse, landslides, and other accidental events, 45% to 50% of the flood peaks and sand peaks in the TGR propagate asynchronously [10,31]. In recent years, with the successive implementation of water storage in cascade power stations along the lower reaches of the Jinsha River, the proportion of sand peaks lagging behind flood peaks has increased significantly, from 22% in 2003–2012 to 32%, as shown in Figure 3a. Compared with those before the impoundment of the TGR, there was no significant change in the propagation characteristics of the flood peak and the sand peak in the Cuntan–Wanxian section from 2003 to 2012. From 2013 to 2022, the proportion of the sand peak lag in the Cuntan–Wanxian reach increased, the sand peak value was large, and the sand peak duration was short. At the HLM station in the lower reaches of the TGR area, due to sediment deposition and sand peak flattening, the sand peak value becomes smaller, and the sand peak duration becomes longer.

After the operation of the TGR, due to the rise in the reservoir’s water levels and the slowdown of the flow velocity, the transport time of the sand peak increased significantly compared to that under pre-impoundment conditions. According to data measured during the flood seasons from 2003 to 2022, the average propagation time from CT station to the dam front is 6.9 days, which is about 4 days longer than the natural propagation time of 3 days. The propagation time of the sand peak is inversely proportional to the flood peak entering the reservoir: the larger the flood peak, the shorter the propagation time; the smaller the flood peak, the longer the propagation time. When the flood peak into the reservoir is less than 3000 m³/s, the sand peak’s propagation time in the reservoir area exceeds 9 days. The flood peak propagation time is inversely proportional to the reservoir water level—the higher the water level, the shorter the propagation time, and vice versa. Generally, when the water level is below 155 m, the flood peak propagation time from CT to the dam is 18–30 h, with an average of 22 h. When the water level is 155–165 m, the propagation time is about 18 h. When the water level is 165–175 m, the time is about 12 h. Therefore, due to the distinct propagation characteristics of the flood and sand peaks in the reservoir area, the sand peak significantly lags behind the flood peak. The proportion of sand peaks lagging the flood peak increases from 32% at the CT inflow station to 72% at the HLM outflow station, as shown in Figure 3c.

3.1.2. Peak-Type Characteristics of the Sand Peak in the Reservoir

Due to the typical floods in the Yangtze River Basin in 2012, 2013, 2018, and 2020, the sand peak samples of the control stations of the TGR into and out of the reservoir in the four years were selected for analysis. Among them, the Cuntan station into the reservoir has 37 groups of sand peak samples, and the Huanglingmiao station out of the reservoir has 15 groups of sand peak samples. The sand peak sample periods of the two control stations are shown in

Table 3. Sample periods of sand peaks at Cuntan station and Huanglingmiao station in 2012, 2013, 2018, and 2020.

Year	CT Station	HLM Station
2012	22 July to 27 July	25 July to 5 August
2013	10 July to 18 July	17 July to 30 July
2018	10 July to 21 July	16 July to 27 July
2020	12 August to 29 August	17 August to 16 September

.

A total of 37 sand peak samples from CT station since the TGR began operation were analyzed. The kurtosis coefficient $\eta$ ranged from 1.35 to 3.00, with an average of 1.79. The typical sand peak process at CT station lasted 6–7 days, with pre-peak times of 1–4 days (average: 2.4 days) and post-peak times of 2–6 days (average: 4.3 days), indicating that most sand peaks were asymmetric. The average pre- to post-peak time ratio was 1:1.8. Figure 4 shows a schematic diagram of the sand peak shapes.

Fifteen sand peak samples from HLM station were selected for statistical analysis. The kurtosis coefficient $\eta$ ranged from 1.37 to 9.84, with an average of 3.03, indicating that the sand peaks here were more flattened compared to those at CT station. The sand peak process at HLM station generally lasted 10–11 days, with pre-peak times of 3–5 days (average: 4.1 days) and post-peak times of 5–10 days (average: 6.9 days). The average pre- to post-peak time ratio was 1:1.7. Compared with those at CT station, both the pre- and post-peak durations increased. However, the ratios remained relatively close, suggesting a possible longitudinal dispersion mechanism during sand peak propagation in the reservoir area.

According to the statistical distribution of the sand peaks during each flood event, at CT station, the sediment transport during the central 50% of the sand peak (1.2 days before and 2.2 days after the peak, totaling 3.4 days) accounted for 83% of the total sediment transported during the flood.

At HLM station, due to peak flattening, the average sediment transport from 2.0 days before to 3.5 days after the peak (5.5 days) accounted for 77% of the total. Within a tighter window (1.0 days before to 1.7 days after the peak, totaling 2.7 days), 56% of the total sediment was transported.

3.1.3. Attenuation and Flattening Evolution of the Sand Peak Process in the Reservoir

In the 2012 flood season, both flood and sand peaks appeared at CT station on 24 July. The peak SC was 2.60 kg/m³. The sand process exceeding 1.0 kg/m³ lasted from 22 to 27 July (5 days), with an average SC of 1.27 kg/m³ and a peak–flat coefficient η of 2.0. At HLM station, the peak appeared on 28 July with an SC of 0.414 kg/m³, and values above 0.3 kg/m³ lasted from 25 July to 5 August (11 days), with an average SC of 0.35 kg/m³. The flattening rate of the sand peak was 7.9%, as shown in Figure 5a.

Based on data measured from 2003 to 2020, sand peak flattening in the TGR was analyzed. Thirteen single-peak flood events showed that the sand peak shape remained largely intact at the dam front (

Table 4. Thirteen single-peak flood events during 2003-2018.

Number	Year	Time
1	2003	30 August to 6 September
2	2005	18 July to 26 July
3	2005	9 August to 22 August
4	2007	25 August to 2 August
5	2008	6 August to 18 August
6	2009	30 July to 10 August
7	2009	15 August to 23 August
8	2010	17 July to 24 July
9	2010	20 August to 28 August
10	2012	26 June to 6 July
11	2012	22 July to 28 July
12	2013	10 July to 18 July
13	2018	10 July to 19 July

). Across different sections, the attenuation was relatively small from CT to QXC (81.6% on average) and from QXC to WX (80.5% average). Greater attenuation occurred from WX to the dam site, ranging from 20% to 40% and averaging at 34.2%. Overall, the average sand peak flattening rate from CT to HLM was 22.3%.

3.2. Ideas and Methods for Sand Peak Process Dispatching

3.2.1. The Division of the Sand Peak Process Dispatching Stages

After the operation of the TGR, the rise in the reservoir’s water levels led to a reduction in the flow velocity and a notable increase in the sand peak transport time compared to those under pre-impoundment conditions. A clear inverse relationship emerged between the flood peak propagation time and the water level at the dam—higher water levels corresponded to shorter propagation times. Given that the sand peak typically lags 6–9 days behind the flood peak, the sand peak dispatching strategy is based on the following principles:

(1)

During significant upstream flood and sand peaks, inflows are stored to raise the reservoir to a target water level.

(2)

Outflows are regulated to maintain an inlet–outlet balance, enabling the sand peak to migrate toward the dam front (near Badong).

(3)

Based on the sediment prediction results, once the sand peak reaches the dam front, the reservoir’s discharge should be moderately increased to exceed the inflow, maximizing sediment release. In essence, high sediment concentrations should be discharged as much as possible during their peak presence at the dam.

According to this strategy, achieving both a high sediment content and an adequate water volume at the dam front is essential—yet these two factors are interdependent and often conflicting. While early-stage water storage slows sediment propagation and diminishes sediment concentrations near the dam, the sediment discharge efficiency ultimately depends on the product of the discharge volume and the sediment concentration. However, if the increase in discharge during the later stage compensates for earlier reductions in sediment concentrations due to impoundment, the sediment discharge performance can still remain effective.

Accordingly, the sand peak process dispatching was divided into three stages based on the transmission characteristics of the flood and sand peaks in the reservoir: the flood detention period, the sediment transport period, and the sediment discharge period at the dam front (Figure 6). (1) The flood detention period: This stage primarily involved flood peak attenuation and storage when both the flood peak and the sand peak entered the reservoir. (2) The sediment transport period in the reservoir area: The main objective during this stage was to ensure that the sand peak was transported as efficiently as possible to the dam front while minimizing its attenuation along the way. (3) The sediment discharge period: In this final stage, as the sand peak approached the front of dam, the reservoir outflow was increased to enhance the sediment discharge. During the sand peak dispatching process, effective control of reservoir water levels and discharge—including its magnitude and timing—was essential. A reasonable distribution of water among the three stages—flood detention, sediment transport, and dam-front discharge—was critical to accelerating the propagation of the sand peak, minimizing its attenuation along the dam, and improving the sediment removal at the dam front.

3.2.2. An Example of Sand Peak Process Scheduling Phase Division

• The Year 2013

During the 2013 flood season, TGR experienced two flood peaks with complex asynchrony between the sand and flood peaks. At 02:00 on 13 July, the flood peak at CT station reached 39,600 m³/s, while the sand peak occurred six hours later at 08:00, peaking at 6.42 kg/m³. A second upstream flood peak occurred on 21 July at 49,000 m³/s, one day after the sand peak arrived at the dam front. The sand peak took 8 days to propagate through the reservoir. Based on the transmission features of the sand and flood peaks, the dispatching process was divided as follows: (1) the flood detention period: 11 July, 02:00 to 15 July, 06:00; (2) the sediment discharge period: 18 July, 10:00 to 29 July, 01:00, marking the main sand peak duration at the dam front; and (3) the sediment transport period: 15 July, 06:00 to 18 July, 10:00. Figure 7 shows the phase division of the sand peak dispatching in 2013.

• The Year 2018

Figure 8 shows the phase division of the sand peak dispatching in 2018. On 14 July 2018, the CT station recorded a flood peak of 59,300 m³/s, whereas a sand peak of 6.69 kg/m³ occurred one day earlier at 08:00 on 13 July. The sand peak’s propagation time was 6 days. Due to this shorter duration and the longer flood peak process, the sediment transport stage was negligible. Therefore, only two stages were defined:

(1)

The flood detention period: 4 July, 08:00 to 16 July, 08:00;

(2)

The sediment discharge period: 16 July, 08:00 to 26 July, 08:00.

3.3. The Influence of Different Scheduling Methods on Reservoir Erosion and Deposition During the Sand Peak Process

Based on the inflow water and sediment data measured during the 2013 and 2018 flood season, the sand peak process scheduling for the TGR was analyzed to assess how different dispatching modes at various stages influenced the sediment transport and riverbed scouring and deposition in the reservoir area.

3.3.1. The Scheduling Calculation Scheme

(1)

Scheduling in 2013

According to the sand discharge dispatching stages of the TGR during the flood season in 2013 (Figure 9), the reservoir dispatching schemes for each stage are determined. Among these, scheme 1 adopts full discharge during the flood control and peak reduction and reservoir sand pulling periods and adopts a maximum flow of 37,500 m³/s of the remaining water volume during the dam-front sand discharge period. Scheme 2 reduces the discharge flow in flood control and peak reduction and increases the discharge flow in reservoir sand pulling, and the discharge flow during the dam-front sand discharge period is consistent with that under scheme 1. Scheme 3 reduces the discharge flow during the reservoir sand pulling period and increases the discharge flow during the dam-front sand discharge period compared to those under scheme 2. The control parameters for each calculation scheme are shown in

Table 5. The calculation scheme for sand peak dispatching in the reservoir area in 2013.

Scheme	Discharge (m3/s)			The Highest Water Level After the First Flood Peak (m)	The Highest Water Level After the Second Flood Peak (m)
	Flood Detention Period	Sediment Transport Period	Sediment Discharge Period
1	30,000	30,000	37,500	149.51	150.38
2	24,000	37,500	37,500	153.04	150.55
3	24,000	24,000	42,000	154.20	153.76

, and Figure 10 is a diagram of the dispatching process for scheme 3.

During the flood control and peak reduction period, scheme 1 uses 30,000 m³/s for controlled discharge. At the end of the peak reduction period, the sediment content in front of the dam is 0.25 kg/m³. Schemes 2 and 3 both use 24,000 m³/s for controlled discharge. At the end of the peak reduction period, the sediment content in front of the dam is 0.22 kg/m³; see

Table 6. Calculation results for sediment discharge in different dam site schemes in 2013.

Scheme	Flood Detention Period (×104 t)	Sediment Transport Period (×104 t)	Sediment Discharge Period (×104 t)	Total Sediment Load (×104 t)	Sand Peak (kg/m3)	Sand Load Ratio (%)
1	260	532	1994	3317	0.88	49
2	223	518	1969	3245	0.85	48
3	223	284	1947	3018	0.78	45

and Figure 10.

During the sand pulling period in the reservoir area, scheme 1 uses 30,000 m³/s for controlled discharge. After the sand pulling period, the sediment content in front of the dam rises to 0.51 kg/m³. Scheme 2 has the largest discharge flow (37,500 m³/s). After the sand pulling period, the sediment content in front of the dam rises to 0.52 kg/m³. However, due to the high discharge flow, the sediment discharge at the dam site during the sand pulling period in the reservoir area increased by 1% when compared with that in scheme 1. Scheme 3 reduces the sediment discharge at the dam site during the sand pulling period in the reservoir area by 50% when compared with that under scheme 1.

During the sediment discharge period in front of the dam, the discharge flows in scheme 1 and scheme 2 were 37,500 m³/s. A sand peak emerged on 20 July, with peak values of 0.88 kg/m³ and 0.85 kg/m³, respectively. The discharge flow in scheme 3 was 42,000 m³/s. However, due to the small discharge flow during the flood control and peak reduction period and the reservoir sand pulling period, the reservoir’s operating water level was high, and the flow velocity was low. The time for the sand peak to arrive in front of the dam was delayed by about 1 d compared with those in scheme 1 and scheme 2, and the peak value of the sand peak was 0.78 kg/m³. During the sediment discharge period in front of the dam, the sediment discharge at the dam site was reduced by 2% when compared with that in scheme 1.

The sediment upstream of the dam is affected by the scheduling of the sediment discharge period in front of the dam, which alters the sediment content process upstream. It takes a certain period for the sediment to evolve to the front of the dam. Therefore, at the end of the sediment discharge period in front of the dam, there is a late impact stage for sediment discharge at the dam site.

The sediment discharge at the dam site in scheme 2 during the entire stage was reduced by 2% when compared with that in scheme 1, and the sediment discharge at the scheme 3 dam site was reduced by 9% when compared with that in scheme 1. The sediment discharge ratio during the sand peak process was calculated, and that in scheme 2 was reduced by 1 percentage point when compared with that in scheme 1. Furthermore, the sediment discharge ratio in scheme 3 was reduced by 4 percentage points when compared with that in scheme 1.

The obvious increases in deposition in scheme 2 and scheme 3 are mainly located in the section from Qingxichang to Wanxian, whereas the deposition in the section from Fengjie to the front of the dam is reduced. Compared with that scheme 1, the increase in siltation in scheme 3 is located mainly in the reservoir section of S287+2~S221.

(2)

Scheduling in 2018

According to the scheduling phase division in Figure 11, three dispatching schemes were developed. In scheme 1, a uniform discharge rate of 39,500 m³/s was maintained during both the flood detention and sediment discharge periods. The water level in front of the dam dropped to 145 m by the end of the sediment discharge period. The second scheme emphasized increasing the discharge during the flood detention period and implementing peak cutting to reduce the rate of attenuation in sand peak propagation in the reservoir area, with the remaining water allocated to the sediment discharge period. The third scheme focused on substantial peak cutting and water storage during the flood detention period, followed by the maximum discharge during the sediment discharge period to enhance the sediment removal efficiency. The scheduling and discharge control strategies for each scheme are detailed in

Table 7. The calculation scheme for sand peak process dispatching of the TGR in 2018.

Scheme	Discharge Flow (m3/s)		Maximum Water Level (m)
	Flood Detention Period	Sediment Discharge Period
1	39,500	39,500	156.95
2	50,000	35,000	153.26
3	34,000	47,000	162.51

, and the scheduling process for scheme 3 is illustrated in Figure 11.

3.3.2. Analysis of the Calculation Results

During both the flood detention and sediment discharge periods, higher discharge flows increased the flow velocity and reduced the operating water levels in the reservoir area. These changes effectively decreased the attenuation of sand peak propagation and enhanced the sediment transport to the dam front. In scheme 1, a discharge rate of 50,000 m³/s was used during the flood detention period, resulting in a significant increase in the sediment concentration (SC) at the dam front from 0.19 kg/m³ to 0.54 kg/m³. Scheme 3 improved the sediment transport further, increasing the sediment load by 63% (Figure 12;

Table 8. Calculation results for sediment transport at dam sites with different schemes in 2018.

Scheme	Flood Detention Period (104 t)	Sediment Discharge Period (104 t)	Late Effects (104 t)	Total Sediment Load (104 t)	Sand Peak in Front of the Dam (kg/m3)	Sediment Discharge Ratio
1	1243	1953	597	3792	1.09	35%
2	1460	2228	559	4247	1.33	39%
3	897	1714	623	3234	0.85	30%

).

During the sediment discharge period at the dam front, the high discharge flow in the early stage, combined with a lower reservoir water level and higher velocity, contributed to reduced sand peak attenuation and a higher SC. Despite the lower discharge rates in the later stage, the sediment load remained substantial. In scheme 2, the discharge rate decreased from 50,000 m³/s to 35,000 m³/s, yet the sediment load was still 30% higher than that in scheme 3.

Considering the entire sand transport process, increasing the discharge during the flood detention and sediment transport periods in the reservoir area effectively reduced the attenuation in the sand peak propagation. During the sediment discharge period at the dam front, the remaining water volume was efficiently used to enhance the sediment discharge. Scheme 2, which prioritized increased discharge during the flood detention period, achieved a 31% increase in the sediment discharge ratio at the dam site compared to that under scheme 1, which focused on increased discharge during the sediment discharge period. The improvement reached 9% during the sediment discharge period alone.

Upstream sediment transport was also influenced by the discharge operations. As sediment propagated to the dam front, a delayed response phase was observed in the sediment transport dynamics following the end of the sediment discharge period.

From a spatial perspective, the increase in discharge during the flood detention and sediment transport periods reduced the deposition of sediment between the CT and Fengjie sections, resulting in greater sediment transport to the dam front, as shown in Figure 13.

4. Discussion

4.1. Factors Influencing Sand Peak Process Dispatching

4.1.1. Sediment Load Conditions

The primary objective of sand peak process dispatching is facilitating multiple sediment discharge events. The volume of sediment discharged during floods serves as a key performance indicator. Dispatching during sand peak events must be performed in real time, beginning as sediment enters the reservoir. The quantity of sediment discharged is directly influenced by the inflow conditions, particularly under significantly reduced inflow scenarios. Hence, the sediment volumes entering and exiting the reservoir during floods are critical indicators that determine the initiation of sand peak dispatching. If the inflow is insufficient, the discharge volume will also be inadequate, rendering dispatching ineffective and wasting available flood resources. Conversely, setting the dispatch threshold too high makes the operation unfeasible.

Table 9. A statistical table of the sand peak process characteristics of the inflow and outflow of the TGR.

Index	CT Station					HLM Station					Sediment Discharge Ratio
	Peak Time	Sand Peak (kg/m3)	Time Duration (d)	Sediment Load (104 t)	Average Flow (m3/s)	Peak Time	Sand Peak (kg/m3)	Time Duration (d)	Sediment Load (104 t)	Average Flow (m3/s)
1	28 July 2007	4.72	8	4034	27,467	3 August 2007	1.4	10	2370	38,336	59%
2	10 August 2008	3.89	12	4353	25,785	17 August 2008	0.58	14	1427	31,493	33%
3	3 August 2009	2.17	12	4558	33,538	9 August 2009	0.787	13	1759	33,779	39%
4	18 August 2009	1.17	8	1612	26,611	26 August 2009	0.315	9	478	27,740	30%
5	19 July 2010	2.53	7	4238	40,288	23 July 2010	0.475	8	1072	38,444	25%
6	22 August 2010	2.19	8	3569	32,022	31 August 2010	0.222	14	440	26,193	12%
7	24 July 2012	2.60	8	3656	46,717	28 July 2012	0.414	11	1333	41,067	36%
8	13 July 2013	6.42	8	5828	28,778	21 July 2013	1.19	10	2340	32,569	40%
9	13 July 2018	6.69	11	7360	41,117	19 July 2018	1.21	12	2167	34,592	29%
10	14 August 2020	4.86	5	3940	43,040	20 August 2020	0.689	6	1022	46,950	26%
11	20 August 2020	3.95	12	8685	44,276	26 August 2020	0.936	12	2378	34,800	27%

presents statistical data on the major sand peak processes observed since the beginning of the operation of the TGR at 145 m during flood seasons. A total of 12 sample groups were analyzed, with Groups 11 and 12 representing the compound sand peak events in 2020. Except for Groups 4 (18 August 2009) and 6 (22 August 2010), all events recorded sediment loads between 10 and 25 million tons, with an average of 15 million tons. Notably, in 2012, 2013, 2018, and 2020, peak dispatching events were executed, and the corresponding sediment load in each instance exceeded 10 million tons. Following the construction of the XLD and XJB hydropower stations, the sediment inflow into the TGR declined significantly, with most of the sediment now concentrated during typical flood periods. Considering that a typical sand peak event lasts approximately 11 days, the dispatch target for the TGR was defined as an event lasting for at least 11 days with a minimum sediment load of 10 million tons. During the demonstration phase, the TGR’s sediment discharge ratio was 30%, and the average sand retention period was 7 days; therefore, the sediment load at the CT station should not be less than 30 million tons.

Based on Table 9, the proposed sediment inflow and outflow conditions for initiating sand peak dispatching indicate that apart from Groups 4 and 6, all sand peak events met the minimum sediment discharge threshold of 10 million tons.

If the incoming sand peak is narrow and sharp—with a high peak but insufficient total sediment—it may be unsuitable for dispatching. For instance, on 25 August 2019, the average daily sediment concentration at the inlet was 2.14 kg/m³, and the peak reached 2.31 kg/m³. However, the sand peak duration was brief, with concentrations above 1.0 kg/m³ sustained for only two days. Consequently, the total sediment volume transported was only 12.5 million tons, and the maximum sediment concentration at the dam front was below 0.1 kg/m³. These conditions did not constitute a significant sand peak event, and dispatching was deemed infeasible.

4.1.2. Sand Conditions

The sand peak propagation from the CT station to the dam front typically lasted about 8.6 days, with a peak coefficient ranging from 1.4 to 3.0, averaging 1.79. Based on these parameters, the storage-phase sediment load should be no less than 30 million tons for dispatch initiation. The corresponding average sediment concentration during this phase is estimated at 1.4 kg/m³, with an average peak daily value of 2.5 kg/m³. Considering the lower threshold, the minimum peak daily average is 2.0 kg/m³.

The sediment discharge phase at the HLM station lasted approximately 11.5 days, with a peak coefficient ranging from 1.66 to 4.77, averaging 3.03. Taking a minimum discharge threshold of 10 million tons, the average peak daily sediment concentration was 0.3 kg/m³, with a corresponding sand peak average of 0.9 kg/m³. At the lower limit, this value was 0.5 kg/m³. Therefore, in addition to the inflow and outflow volumes, a minimum inlet peak of 2.0 kg/m³ and an outlet peak of 0.5 kg/m³ are necessary to initiate sand peak dispatching. As shown in Table 3, all groups except Groups 4 and 6 met these criteria.

4.1.3. Discharge Conditions

This study indicated that higher inflow rates led to shorter sand peak propagation times, reduced attenuation within the reservoir area, and higher sand discharge rates. Simultaneously, a lower water level at the dam front also contributed to faster propagation and increased sediment discharge.

Based on measured data from the reservoir, the average sediment transport ratio during sand peak events was derived through curve fitting, yielding the formula

$${\displaystyle \frac{S_o}{S_i}}=1.30e^{{\displaystyle \frac{-0.0094V}{Q}}},$$

where $S_i$ and $S_o$ represent the average sediment concentrations entering and leaving the reservoir, respectively; $V$ denotes the reservoir’s storage capacity; and $Q$ is the average discharge during the transport process (Figure 14). Using this equation, the minimum sediment discharge requirements for various water levels at the dam front were estimated. Specifically, when the water level at the dam front was 145 m, the minimum inflow required to initiate sand peak dispatching was calculated to be 25,000 m³/s. This discharge threshold ensured sufficient sand transport from the point of inflow to the point of outflow across the reservoir. For the other water levels, the corresponding minimum inflow values can be referenced in

Table 10. The minimum inflow conditions for sand peak process dispatching under different water levels in front of the dam.

Index	Mean Water Level in Front of the Dam (m)	Minimum Inlet Flow Value (m3/s)
1	145	25,000
2	150	30,000
3	155	35,000
4	160	40,000
5	165	45,000

.

When scheduling sand peak dispatching at the dam front, increasing the discharge flow during the arrival of the sand peak improved the sediment release. Given the considerable water depth at the dam front, changes in the sediment concentration (SC) were minimally influenced by water depth and were instead primarily controlled by the discharge rate. Therefore, higher discharge flows were more effective in enhancing sediment flushing.

Sand peak transport through the reservoir can be divided into three stages: flood blocking and peak clipping; reservoir sand conveyance; and sediment release at the dam front. During the flood detention and sediment transport phases, increasing discharge rates and lowering reservoir water levels helped accelerate the flow velocity and reduce sand peak attenuation. In the sediment release phase, residual water was utilized to flush the sediment, resulting in a better discharge performance.

As shown in Table 3, aside from Groups 4 and 6—which had relatively low sediment discharge volumes—all other dispatch events featured discharge flows exceeding 30,000 m³/s. This indicates that effective sediment flushing typically requires the discharge rate to be maintained above this threshold. Considering the sand peak type, the average pre-peak and post-peak durations at the HLM station were 4.1 and 6.9 days, respectively. Notably, 77% of the sediment transport occurred within the central 50% of the sand peak duration. According to dispatching principles, increasing the discharge before the sand peak reaches the dam significantly improves the sediment removal. Hence, during the critical five-day window—spanning from two days before to three days after peak arrival—the discharge flow should be maintained at no less than 35,000 m³/s. However, actual discharge decisions must also accommodate multiple operational constraints, including flood control, navigation, and flood resource management.

4.1.4. The Water Level Conditions in Front of the Dam

In the process of flood control in flood season, the water level in front of the dam is affected by the inflow and outflow flow regulation and storage, which is a changing process. According to the study on the inflow flow conditions required for the operation of the sand peak process, to maintain the transport of the sand peak in the reservoir area during the period from the inflow to the outflow of the sand peak, different water levels in front of the dam should have different lower limits. Therefore, when the inflow flow is low, different inflow flow conditions also need different lower limits on the water level in front of the dam, and the corresponding values can be calculated according to Table 9.

4.2. The Dispatching Strategy for the Sand Peak Process in the TGR

The primary objective of sand peak dispatching in the TGR is to enhance the sediment concentration (SC) during high-flow discharges and to maximize the sediment removal, thereby minimizing the overall siltation within the reservoir. The cascade reservoirs of the Yangtze River and some of the cascade reservoirs in the upper reaches of the Yangtze River are also operated jointly with water and sediment to maximize the benefits of flood control, power generation, sediment discharge, navigation, water supply, and ecology. Following the construction of upstream reservoirs on both the mainstem and the tributaries of the Yangtze River, the sediment influx into the TGR has significantly declined, with the majority of the sediment load now concentrated during the flood season. Given the seasonal characteristics of the sediment inflow, sediment discharge operations during sand peak events in flood season have proven beneficial in mitigating sediment deposition in the TGR.

Sand peak dispatching should be initiated based on sediment forecasting, with optimization and simulation of the dispatch scheme implemented as needed. The recommended operational modes for sand peak dispatching are as follows:

(1)

Triggering Conditions Based on Real-Time Monitoring:

When real-time monitoring indicates that the SC of the sand peak at the CT station reaches or exceeds 0.5 kg/m³, sediment forecasting should be initiated to guide reservoir discharge operations during the sand peak event.

(2)

Forecasting Thresholds for Initiating Dispatching:

Based on forecast data, the discharge from the CT station should meet the minimum thresholds indicated in Table 4, with conditions including a sand peak SC of 2.0 kg/m³, an average SC of no less than 1.4 kg/m³, a duration exceeding 7 days, and an incoming sediment load of no less than 30 million tons.

For the HLM station, the sand peak should reach 0.5 kg/m³, the average SC should be no less than 0.3 kg/m³, and sediment-rich discharge should last over 11 days with a total sediment transport of at least 10 million tons.

(3)

Scheduling the Dispatch Process:

Once the decision to begin sand peak dispatching is made, the process is divided into three phases: flood detention, sediment transport, and sediment discharge.

• During the flood detention period, floodwaters are stored in compliance with flood control guidelines, leading to a gradual rise in the reservoir’s water level.
• During the sediment transport period, sediment movement through the reservoir is prioritized, with the minimum discharge required adjusted based on the corresponding reservoir water level, as shown in Table 4.
• During the sediment discharge period, the primary objective is to increase the outflow to flush high-SC water from the dam front. This intensified discharge should begin two days before the sand peak reaches the dam and must not fall below 35,000 m³/s. The elevated discharge should be sustained for at least three days following the sand peak’s exit from the reservoir. Completion of the dispatch process should be determined based on the sediment concentration and reservoir water level conditions.

(4)

The Reservoir Level and the Outlet Utilization Strategy:

During the discharge phase of the sand peak process, the reservoir’s water level should be kept as low as possible. Emphasis should be placed on increasing the discharge flow during the flood detention and sediment transport phases to reduce sediment attenuation within the reservoir. In the sediment discharge phase, when the SC at the dam front remains high, sediment flushing should prioritize the use of sediment outlets. If additional discharge is required, drift holes and deep flood outlets may be employed to preferentially release turbid bottom water, which enhances the sediment transport efficiency and optimizes the sediment distribution near the dam front.

In general, according to the model’s calculation results, when the boundary conditions for water and sediment change in the future, they can be calculated and analyzed under different scenarios, which provides technical support for water and sediment transport in the reservoir area under different reservoir operation modes in the future.

5. Conclusions

In this study, measured data were analyzed to study the asynchronous characteristics of the sand peak and the flood peak of the TGR, the peak type of the sand peak process, and the evolution of attenuation and flattening so as to provide technical support for the formulation of effective scheduling schemes in the TGR. Furthermore, it was determined that different water and sediment conditions from recent years should be combined for optimized water and sediment dispatching. The main conclusions are as follows:

(1)

The sand peak discharge scheduling for the TGR is divided into three stages: the flood detention period, the sediment transport period, and the sediment discharge period. The flood detention period mainly involves flood peak reduction and storage when the flood peak and the sand peak enter the reservoir. The sediment transport period mainly maintains smooth propagation of the sand peak to the river section in front of the dam, whereas the sediment discharge period increases the discharge when the sand peak is about to reach the dam for greater sand discharge.

(2)

Starting conditions and control indicators for sand load scheduling during the sand peak process are proposed. The goal of sand load scheduling during the sand peak process of the TGR is to discharge sand for 11 days, and the sand load is no less than 10 million tons. The average sediment content is 0.3 kg/m³, and the average daily maximum value of the sand peak is 0.9 kg/m³. If the lower limit is considered, the daily maximum value of the sand peak is 0.5 kg/m³. In the time period from 2 d before the sand peak reaches the dam to 3 d after the sand peak (a total of 5 d), the water discharge should not be less than 35,000 m³/s.

(3)

The dispatching methods in the TGR during the three periods of sand peak discharge are presented. During the flood detention period, the reservoir’s water level rises as the reservoir is intercepted and stored. During the sediment transport period, when the average water level in front of the dam changes within the 145–165 m range, the corresponding minimum inflow is 25,000–45,000 m³/s. During the sediment discharge period, the intensified discharge should begin 2 d before the sand peak reaches the dam and must not fall below 35,000 m³/s. The elevated discharge should be sustained for at least 2 d following the exit of the sand peak from the reservoir.

(4)

The sand peak discharge scheduling method proposed in this study provides a reference for applications in actual project scheduling, which can provide an important scientific basis for the sustainable utilization of the TGR and other similar reservoirs.