Channel geometry adjustments in response to hyperconcentrated floods in a braided reach of the Lower Yellow River

Abstract

Hyperconcentrated floods with more than 200–300 kg/m³ sediment concentrations often occur in the Lower Yellow River (LYR) during flood seasons, which leads to unique fluvial processes in the braided reach of the LYR. The investigation of channel geometry adjustments in response to hyperconcentrated floods can not only help to gain a better understanding of associated fluvial processes, but also is significant for making flood control strategies in the braided reach. In this study, pre- and post-flood bankfull channel dimensions in the braided reach were calculated based on the observed cross-sectional profiles in 15 years with the occurrence of hyperconcentrated flood events. Adjustments in channel geometry at section- and reach-scales were investigated, with several factors influencing adjustments in reach-scale channel geometry being analyzed. It indicates that the mean sediment transport rate was a key factor influencing the adjustment index, although pre-flood channel geometry and sediment deposition can also affect the index to some extent. An empirical relationship was developed between the characteristic parameter representing the pre- and post-flood channel geometries and mean sediment transport rate in hyperconcentrated floods. Eleven datasets were used to calibrate the parameters in the empirical relation, with the datasets in 1973, 1988, 1995, and 2002 verifying the relation. The calculated post-flood characteristic parameter of channel geometry using the empirical relation agreed well with observed data, and the proposed method can be used to predict the reach-scale adjustment of channel geometry during hyperconcentrated floods in alluvial rivers.

Keywords

Channel geometry adjustment index hyperconcentrated floods braided reach Lower Yellow River

I. Introduction

Hyperconcentrated flows have been observed in various environments, such as rainstorm-induced landslides and dam collapses (Batalla et al., 1999; Kostaschuk et al., 2003; Pierson and Costa, 1987; Sohn et al., 1999; Wilcox et al., 2014) and volcanic and glacial settings (Breien et al., 2008; Lirer et al., 2001; Maizels, 1989; Smith, 1986). Hyperconcentrated flows also have been reported in large alluvial rivers with significant amounts of sediment naturally supplied from tributaries (Englund and Wan, 1984; Li et al., 2014; Wan, 1985; Wang and Ta, 2016; Xu, 2004; Zhang et al., 2015). For example, it is reported that the historical maximum sediment concentration was about 1600 kg/m³, measured during a hyperconcentrated flood in a tributary of the Middle Yellow River, which is the highest value worldwide (Shu and Fei, 2008; Van Maren et al., 2009a, 2009b). Hyperconcentrated floods in alluvial rivers can result in many problems, such as extreme high-water level, which threaten flood control engineering including levees and flow guide works (Wang et al., 2009).

The formation and mechanism of hyperconcentrated flows in alluvial rivers are rather complicated, which have attracted great concern of researchers in the contexts of river dynamics and geomorphology. According to the concentration and grain size of sediment, hyperconcentrated flows are generally classified into two types of pseudo-one-phase flow and solid and liquid two-phase flow in the Yellow River (Chien and Wan, 1998; Van Maren et al, 2009a). Existing studies indicate that a variety of unique fluvial processes including peak discharge increase and “tearing up of the river bottom” may occur in hyperconcentrated floods in the Yellow River, and rapid change in channel morphology is closely associated with these unique phenomena (Chien and Wan, 1998; Li et al., 2014; Van Maren et al., 2009a, 2009b; Wang et al., 2009; Winterwerp, 2006). Hyperconcentrated floods can lead to more remarkable changes in channel morphology, as compared with floods with low sediment concentrations (Qi and Li, 1996; Wang et al., 2009), but the complex morphological processes are still unclear. Therefore, it is necessary to investigate how the adjustments in channel morphology occur during hyperconcentrated floods.

Previous studies show that the cross-sectional geometry in the braided reach of the Lower Yellow River (LYR) can become narrower and deeper after a hyperconcentrated flood event (Chien and Wan, 1998; Qi et al., 2008; Wang et al., 2000). Some hyperconcentrated floods may also cause high sedimentation on the floodplains and severe erosion in the main channels, and some hydraulic engineers proposed that producing man-made hyperconcentrated floods may be a feasible way to transport sediment and reduce deposition in the main channel of the LYR (Fei, 1998; Qi et al., 2008). However, results of physical models and field investigations suggest that sediment transportation for a long distance should require a narrow and deep channel geometry, but the narrow-deep channel formed after hyperconcentrated floods can not remain stable in the braided reach of the LYR (Fei, 1998; Wang et al., 2000; Wang et al., 2009). Although many qualitative investigations have been conducted on morphology changes, the complex response of channel geometry to hyperconcentrated floods is still poorly understood. Some researchers attempted to investigate the influencing factors of channel geometry in hyperconcentrated floods by developing empirical relationships (Jiang et al., 1999; Sun et al., 2014; Zhang et al., 2002). For example, Jiang et al. (1999) developed an empirical relationship between cross-sectional area and discharge and sediment concentration. Zhang et al. (2002) found that the channel geometry after a hyperconcentrated flood was closely related to the sediment concentration, and the width-depth ratio decreased with an increase in sediment concentration. Sun et al. (2014) established a relationship between width–depth ratio and incoming sediment coefficient in the braided reach. However, these empirical relations are only applicable to a specified cross section, and the obtained results can not represent the channel adjustments in response to hyperconcentrated floods of a total river reach. A reach-averaged concept is more appropriate to describe the bankfull channel dimensions in a river because reach-scale values can provide more representative geometry and statistical characteristics (Harman et al., 2008; Wohl et al., 2004; Xia et al., 2014a).

The aims of this paper are: (a) to qualitatively analyze the channel geometry response to hyperconcentrated floods at section- and reach-scales in the braided reach of the LYR; (b) to determine the dominant factor influencing the adjustments in channel geometry caused by hyperconcentrated floods; and (c) to predict the characteristic parameter for channel geometry after hyperconcentrated floods. These results can contribute to provide more comprehensive understanding of morphodynamic behavior of hyperconcentrated floods in alluvial rivers, which is significant for designing river training works and making flood control strategies.

II. Study area and data collection

The Yellow River is located between 96–119° E longitude and 32–42° N latitude, as shown in Figure 1, and it has a drainage area of 795,000 km² and a length of 5,464 km (Miao et al., 2010). It originates from the Tibetan Plateau, carrying huge amounts of suspended load to the Bohai Sea. It is estimated that the sediment load carried by the Yellow River accounts for 6% of the total global river sediment to the oceans (Wang et al., 2007). The LYR is usually defined as the reach between Mengjin in Henan province and Lijin in Shandong province, with a length of about 731 km (Figure 1).

Figure 1.

Overview of the Lower Yellow River.

Hyperconcentrated floods often occur in the Yellow River, with sediment concentrations of greater than 200–300 kg/m³, leading to heavy sedimentation in the LYR. The LYR is confined by various river training works, resulting in the level of the riverbed rising above cities and farmland, so it is known as a “perched river” or “above-ground river” (Chu, 2014; Miao et al., 2016). The Sanmenxia Reservoir, located on the lower part of the Middle Yellow River, started to operate in 1961 mainly for flood control. However, severe sedimentation occurred in the Sanmenxia Reservoir, with the capacity of reservoir being almost filled up within several years. Another multipurpose project, the Xiaolangdi Reservoir, with a capacity of 12.6 billion m³, commenced impoundment in 1999. The operation of Xiaolangdi Reservoir remarkably altered the flow and sediment regime entering the LYR, leading to continuous channel degradation (Chen, 2012; Xia et al., 2014a).

1. Study reach

According to the geomorphological characteristics, the LYR can be divided into three different reaches of braided, transitional and meandering (Wu et al., 2005). The 275 km braided reach from Mengjin to Gaocun is selected as a study reach. The channel in the reach is relatively wide and shallow, with the bankfull width varying from 0.9 to 1.4 km and the bankfull depth of about 2.7 m (Xia et al., 2014a). The floodplains on both sides are very wide along the reach, with the widths between the levees at different sections varying from 5 to 20 km (Wu et al., 2005). The braided reach is rather unstable in that it is prone to rapid changes in channel geometry and frequent shifting of the main flow. Some groins along the main stream have been built in this reach to stabilize the main channel. However, most of the groins do not work well because the river channel shifted far away from them (Wu et al., 2005, 2008a). The Yellow River Conservancy Commission of China (YRCC) has set up three hydrometric stations including Huayuankou (HYK), Jiahetan (JHT), and Gaocun (GC) in the braided reach, and lateral profiles at 28 sedimentation sections are surveyed regularly twice a year (pre-flood in May and post-flood in October) in order to monitor the channel deformation of the reach.

2. Data collection

Hydrological data in hyperconcentrated floods were provided by the YRCC, including the mean and maximum discharges and sediment concentrations at the hydrometric sections. The pre- and post-flood cross-sectional profiles at 28 sedimentation sections and the amount of bed deformation were also provided by the YRCC. Hyperconcentrated floods occurred frequently in the LYR, with the sediment concentrations being relatively high in the 1970s. In 1977, the maximum sediment concentration at the Sanmenxia station even reached up to 911 kg/m³. In the 1980s and 1990s, the frequency of hyperconcentrated floods began to reduce because of the implementation of the water and soil conservation and the construction of various sediment trap dams in the Middle Yellow River (Wang et al., 2009). The hyperconcentrated floods in the LYR since the year of 2000 are from the turbidity currents releasing from the Xiaolangdi Reservoir, and these hyperconcentrated floods with very fine sediment particles are different from those occurring before the operation of the Xiaolangdi Reservoir.

There are no fixed criteria to determine whether a sediment-laden flow is regarded to be hyperconcentrated because the transition from non-hyperconcentrated to hyperconcentrated flows depends on both the sediment concentration and the grain size of sediment (Chien and Wan, 1998). The majority of the sediment transported in the LYR is suspended load, and the amount of bed load only account for less than 0.5% (Xia et al., 2014a). In this study, the floods are considered as hyperconcentrated floods, when the maximum daily mean suspended sediment concentration exceeds 300 kg/m³ at Sanmenxia in 1970−1999 or exceeds 150 kg/m³ at Xiaolangdi in 2000−2016. According to this criterion, 13 years with the occurrence of hyperconcentrated flood events in 1970−1999 and 2 years in 2000−2016 were identified, as shown in Table 1. It is noted that if there were several hyperconcentrated flood events occurring in the same flood season, the hydrological data in a hyperconcentrated flood with the highest sediment concentration were used for the statistical analysis because a flood with higher sediment concentrations had more remarkable effect on adjustments in channel morphology.

Table 1.

Flow and sediment regime and channel deposition during hyperconcentrated flood events.

Year	S_max (kg/m³)	Q_HYK (m³/s)	S_HYK (kg/m³)	Amount of sediment
Year	S_max (kg/m³)	Q_HYK (m³/s)	S_HYK (kg/m³)	deposition (10⁸ t)
1970	552	1890	225	5.289
1971	374	1904	143	1.978
1973	452	3385	195	3.075
1974	361	1311	97	1.481
1977	556	3598	197	5.255
1978	358	1889	118	1.339
1988	342	4538	102	2.615
1992	426	2992	190	3.766
1994	358	1958	77	1.906
1995	352	1184	74	1.048
1996	515	3287	89	5.252
1997	489	1200	186	3.17
1999	532	1735	122	2.139
2002	174*	592	109	0.103
2004	219*	2144	91	-0.142

S_max in 2002 and 2004 represents the maximum daily mean sediment concentration at Xiaolangdi, whereas S_max in other years represents the maximum daily mean sediment concentration at Sanmenxia; Q_HYK = mean discharge during a hyperconcentrated flood at HYK; S_HYK = mean sediment concentration during a hyperconcentrated flood at HYK.

III. Methods

1. Determination of section- and reach-scale bankfull channel geometry

The first step is to determine bankfull geometry at a section with a specified bankfull level and main-channel zone. Generally, the level of the bank top of an active floodplain is usually defined as the bankfull level at a section in terms of flood control, and the main passage between the two lips of the active floodplains on both sides is often defined as the zone of main-channel (Xia et al., 2014a, 2014b). In the case of the unclear bankfull indicator, earlier and later measurements of the cross-sectional profile are used as the reference information to determine the bankfull level at a section (Wu et al., 2008a). The width between these two lips is defined as the bankfull channel width (W_i), and the main-channel area under the bankfull level is defined as the bankfull cross-sectional area (A_i). The corresponding bankfull channel depth (H_i) is equal to the ratio of A_i to W_i. For example, Figure 2 shows the cross-sectional profile at Heishi measured before the 1999 flood season, located about 177 km downstream of the Xiaolangdi dam. According to the method mentioned above, the bankfull level at Heishi was determined to be equal to 85.81 m. The bankfull width was determined to be 2073 m, with a bankfull area of 2200 m² and a bankfull depth of 1.06 m. The pre- and post-flood bankfull channel dimensions at 28 sections during hyperconcentrated flood events can be identified by using the same method.

Figure 2.

Determination of the bankfull channel geometry at Heishi before the 1999 flood season.

The second step is to calculate the reach-scale bankfull channel geometry using an appropriate method. An approach proposed by Xia et al. (2014a) was adopted to calculate the reach-scale channel dimensions in the braided reach. This method assumes that the study reach with a channel length of L covers N cross sections, and the longitudinal distance between the ith and (i + 1)th sections is Δx_i. The corresponding reach-scale bankfull channel geometry ( $\bar{G}$ ) can be written as

\bar{G} = exp (\frac{1}{2 L} \sum_{i = 1}^{N - 1} (ln G_{i + 1} + ln G_{i}) \times Δ x_{i})

where $\bar{G}$ represents the dimensions of reach-scale bankfull width ( $\bar{W}$ ), area ( $\bar{A}$ ), and depth ( $\bar{H}$ ). Here G_i and $G_{i + 1}$ represent the bankfull dimensions at the ith and (i + 1)th sections, respectively.

2. Calculation of reach-scale adjustment index in channel geometry

The wetted perimeter is a common parameter in river dynamics, which can not only reflect the characteristics of channel geometry, but also plays a role in resistance to flow (Strelkoff and Clemmens, 2000). In open channel flows, it is defined as the surface of the channel bottom and sides in direct contact with the aqueous body. Hydraulic radius is another important parameter, which is often used by river engineers to assess the channel’s capacity to transport water. It is defined as the ratio of the wetted cross-sectional area to the corresponding perimeter (Cheng and Nguyen, 2011). The reach-scale bankfull hydraulic radius ( $\bar{R}$ ) can be expressed as

\bar{R} = \frac{\bar{A}}{\bar{χ}}

where $\bar{χ}$ represents the reach-scale bankfull wetted perimeter (m).

Both a reach-scale wetted perimeter and hydraulic radius can represent the characteristics of channel geometry, with the same dimension. Therefore, a dimensionless parameter M is defined as the ratio of $\bar{χ}$ to $\bar{R}$ , which is used to conduct the following analysis. It is difficult to determine the wetted perimeter in the whole reach accurately, so $\bar{χ}$ can approximately equal $\bar{W} + 2 \bar{H}$ in the braided reach in this study. Therefore, the characteristic parameter of channel geometry (M) can be written as

M = \frac{{(\bar{W} + 2 \bar{H})}^{2}}{\bar{A}}

In order to quantitatively analyze the variation in reach-scale bankfull channel geometry in hyperconcentrated floods, the adjustment index in channel geometry (η) is written as

η = \frac{M_{a}}{M_{b}}

where M_b and M_a represent the pre- and post-flood characteristic parameters of channel geometry.

IV. Results and discussion

1. Adjustments in section-scale channel geometry

A hyperconcentrated flood is termed overbank hyperconcentrated flood when the in-channel water level is higher than the bankfull level, otherwise it is termed non-overbank hyperconcentrated flood. In the selected 15 hyperconcentrated flood events, it was determined that hyperconcentrated floods in 1974, 1978, 1995, 1997, 1999, and 2002 are classified as non-overbank floods, and other floods belong to overbank floods.

Figure 3 shows the sketched adjustments in channel geometry during an overbank hyperconcentrated flood. At the beginning, the main-channel is characterized by a wide and shallow geometry, and sediment deposition occurs in the main channel (Figure 3(a)). During the rising limb of the flood, the water level rises rapidly with an increase in discharge, and then the water overtops the original bankfull level and inundates the floodplain (Figure 3(b)). The flow velocity on the floodplain is relatively small because of higher bed roughness and lower water depth. Therefore, the corresponding sediment transport capacity on the floodplain is relatively low, resulting in severe sedimentation. The flow velocity in the main channel increases with an increase in discharge, leading to an increase in the corresponding sediment transport capacity. Then the main channel undergoes intensive scour, with the bed level degrading remarkably (Figure 3(c)). During the receding limb of a hyperconcentrated flood, sediment deposition occurs on the floodplain marginal, with a new lip being formed, which decreases the bankfull width further (Figure 3(d)). However, it is noted that whether the main channel scours or deposits depends on the discharge (Q) and sediment concentration (S) during the receding limb. Previous studies show that the main channel would deposit when S/Q > 0.015, but would scour when S/Q < 0.015 (Wang et al., 2000). On the whole, the post-flood channel will become narrower and deeper, as compared with the pre-flood channel geometry.

Figure 3.

Sketched adjustments in channel geometry during an overbank hyperconcentrated flood.

A typical overbank hyperconcentrated flood occurred during the period from August 4 to August 12 in 1977 in the LYR, with the maximum daily mean sediment concentration of 556 kg/m³ at Sanmenxia. Figure 4(a) shows the variations in discharge and sediment concentration at HYK during this period. It shows that the maximum discharge was 10,800 m³/s, with the maximum sediment concentration of 437 kg/m³. Figure 4(b) shows the variations in water level and thalweg level in the main channel at HYK. It indicates that the thalweg level in the main channel was 87.68 m (Point a) at the beginning, and it degraded to 85.88 m (Point b) due to an intensive scour in the main channel. During the receding limb, very quick siltation occurred in the main channel, which caused the thalweg level to aggrade to 86.92 m (Point c). Figure 5(a) shows the pre- and post-flood cross-sectional profiles at HYK in 1977. It can be seen that the bankfull channel width reduced from 1180 m to 441 m, with the floodplain level aggrading by more than 1 m due to severe sedimentation.

Figure 4.

The variations in flow and sediment regime and bed level during a hyperconcentrated flood in 1977: (a) discharge and sediment concentration; (b) water level and thalweg level.

Figure 5.

Channel adjustments in different flood events of: (a) an overbank flood in 1977; (b) a non-overbank flood in 1995.

Channel geometry adjustments in non-overbank hyperconcentrated floods are different from those in overbank hyperconcentrated floods. At first, a huge amount of sediment carried by the hyperconcentrated flood deposits in the main channel, which gradually reduces the channel width. There is an increase in flow velocity in the main channel with an increase in discharge and a decrease in channel area. Subsequently, the main channel undergoes scouring from deposition due to the increased sediment transport capacity. The bankfull depth increases greatly and the main channel becomes narrower and deeper, while the floodplain remains unchanged. A typical non-overbank hyperconcentrated flood occurred from July 19 to 28 in 1995 in the LYR, with the maximum daily mean sediment concentration of 352 kg/m³ at Sanmenxia. Figure 5(b) shows the pre- and post-flood cross-sectional profiles at HYK in 1995. The bankfull width decreased from 1357 m to 671 m, and the deepest bed level also incised by 1.3 m. Figure 5 indicates that there were some differences in channel geometry adjustments between an overbank and a non-overbank hyperconcentrated flood: (a) a new lip of floodplain was formed in an overbank hyperconcentrated flood, but the lip of floodplain in a non-overbank flood did not change; (b) the adjustments in channel geometry during an overbank hyperconcentrated flood were more severe than those during a non-overbank flood.

The channel response to a specified hyperconcentrated flood is different at various cross-sections. Table 2 shows the pre- and post-flood bankfull widths and depths at 28 cross-sections in the braided reach in 1996. It can be seen that the pre-flood bankfull widths were greater than the post-flood bankfull widths at 20 cross-sections, and the maximum difference reached up to 1906 m. The post-flood bankfull depths were greater than the pre-flood bankfull depths at 17 cross-sections, with a maximum difference of 1.05 m. Therefore, bankfull channel dimensions change greatly along the braided reach of the LYR, and the variation in channel geometry at a specified section cannot represent the characteristics of the whole reach, and it is necessary to analyze the adjustments in reach-scale channel geometry after hyperconcentrated floods.

Table 2.

Pre- and post-flood bankfull widths and depths at 28 cross-sections in 1996.

Sections	Pre-flood		Post-flood
Sections	bankfull width (m)	bankfull depth (m)	bankfull width (m)	bankfull depth (m)
CS1	743	3.31	743	3.13
CS2	1410	2.51	1385	1.47
CS3	619	3.33	601	2.9
CS4	1254	2.15	1207	1.77
CS5	1150	2.04	974	2.15
CS6	3405	0.81	1498	1.1
CS7	1144	2.46	1097	1.98
CS8	1057	1.4	752	1.93
CS9	1222	1.62	1204	1.9
CS10	961	1.84	963	1.7
CS11	827	1.39	1030	1.77
CS12	1745	1.06	1054	2.11
CS13	594	2.19	627	2.12
CS14	3164	1.13	2861	1.12
CS15	2448	1.25	2335	1.17
CS16	1823	1.51	1415	1.65
CS17	992	1.41	789	1.54
CS18	1352	1.26	589	1.88
CS19	1496	1.36	1365	2.03
CS20	802	1.52	980	2.03
CS21	1702	1.58	1044	1.36
CS22	666	1.88	663	2.32
CS23	2607	0.87	2662	1.13
CS24	3304	1.93	2652	1.99
CS25	1399	0.92	931	1.61
CS26	579	1.91	592	2.74
CS27	676	1.53	690	1.78
CS28	719	1.6	607	2.12

2. Adjustments in reach-scale channel geometry

The dimensions of reach-scale channel geometry in the braided reach were calculated based on the observed pre- and post-flood cross-sectional profiles during the selected 15 years with the occurrence of hyperconcentrated floods. Table 3 shows the calculated pre- and post-flood reach-scale bankfull widths ( ${\bar{W}}_{b}$ , ${\bar{W}}_{a}$ ), bankfull depths ( ${\bar{H}}_{b}$ , ${\bar{H}}_{a}$ ), bankfull areas ( ${\bar{A}}_{b}$ , ${\bar{A}}_{a}$ ), and characteristic parameter representing the pre- and post-flood channel geometries (M_b, M_a) in the braided reach, respectively. It can be seen that the pre- and post-flood reach-scale bankfull widths in the 1970s and 1980s were larger than those in the 1990s because the braided reach of the LYR underwent continuous aggradation owing to the frequent occurrence of hyperconcentrated floods, while the bankfull depth increased slightly in the 1970s–1990s. In the 2000s, both the reach-scale bankfull width and depth increased remarkably owing to severe degradation caused by the operation of the Xiaolangdi Reservoir. It also shows that most of the pre-flood bankfull widths were greater than the corresponding post-flood bankfull widths, and most of the pre-flood bankfull depths were less than the post-flood values. It means that the whole braided reach became narrower and deeper after hyperconcentrated floods. In addition, all the post-flood characteristic parameters of channel geometry (M_a) were less than the corresponding pre-flood parameters (M_b), with the values of η being less than 1.0. Table 3 also shows that the values of η in non-overbank hyperconcentrated floods were generally greater than those in overbank hyperconcentrated floods, and overbank hyperconcentrated floods have more effect on adjustments in channel geometry than non-overbank hyperconcentrated floods. It is because more intensive exchanges of flow and sediment occurred between the main channel and floodplain during overbank hyperconcentrated floods. The exchanges resulted in sedimentation on the floodplain, with clear water going back to the main channel, which led to severe scour in the main channel. It is noted that the narrow and deep channel formed after hyperconcentrated floods would be unstable, and the channel in braided reach may become wide and shallow again owing to the collapse of the floodplain banks caused by flows with low concentrations during non-flood seasons (Fei, 1998; Zhang et al., 2002).

Table 3.

Reach-scale bankfull channel dimensions before and after hyperconcentrated floods.

Year	${\bar{W}}_{b}$	${\bar{W}}_{a}$	${\bar{H}}_{b}$	${\bar{H}}_{a}$	${\bar{A}}_{b}$	${\bar{A}}_{a}$	M_b	M_a	η
1970	1923	1575	1.28	1.26	2460	1988	1507	1251	0.830
1971	1660	1164	1.12	1.03	1867	1198	1481	1134	0.766
1973	1468	1277	1.03	1.66	1504	2118	1436	774	0.539
1974*	1430	1380	1.28	1.36	1825	1883	1124	1015	0.903
1977	1979	1635	1.39	1.46	2752	2381	1427	1128	0.790
1978*	1928	1787	1.54	1.60	2975	2856	1253	1122	0.895
1988	2400	2044	1.37	1.54	3291	3162	1754	1325	0.755
1992	1783	1255	1.34	1.69	2389	2117	1335	748	0.560
1994	1313	1158	1.65	1.49	2164	1725	800	782	0.977
1995*	1268	1262	1.32	1.38	1669	1742	967	918	0.949
1996	1260	1082	1.57	1.77	1983	1918	805	614	0.763
1997*	938	938	1.57	1.72	1477	1613	600	549	0.915
1999*	969	943	1.53	1.56	1486	1469	636	609	0.958
2002*	1009	1058	2.05	2.25	2067	2378	497	475	0.956
2004	1147	1144	2.61	2.70	2999	3085	443	428	0.968

Year* represents a year with the occurrence of non-overbank flood events.

3. Factors influencing adjustment index in channel geometry

The response of channel geometry to hyperconcentrated floods are complex, which is affected by many factors. In natural alluvial rivers, the factors affecting the adjustment in channel geometry can be summarized as the flow and sediment regime and channel boundary conditions, and the first factor is of primary significance (Julien and Wargadalam, 1995). The reach-averaged concept is more appropriate to describe the bankfull channel dimensions in the braided reach because reach-scale values can provide more statistical characteristics. In this study, adjustment index in channel geometry (η) was used to represent the variation in reach-scale bankfull channel geometry in the braided reach during hyperconcentrated floods. Several influencing factors were tested in order to determine a dominant factor influencing the channel geometry adjustment in hyperconcentrated floods.

3.1 Amount of sediment deposition

Hyperconcentrated floods can cause intensive scour in the main channel and high sedimentation on the floodplain, and the amount of sedimentation is generally larger than the value of scour (Wang et al., 2000; Wang et al., 2009). Channel geometry adjusts mainly through the sediment exchange between hyperconcentrated floods and bed and bank materials, and the total amount of sediment deposition can represent the effect of this interaction to some extent. Figure 6(a) shows the relationship between the adjustment index in channel geometry (η) and the amount of sediment deposition (V_d) in hyperconcentrated floods in the braided reach. It shows that η decreased with an increase in V_d, with the square of correlation coefficient of 0.28 based on the linear regression. It shows that the reach-scale channel geometry would become narrower and deeper in case of severe sediment deposition, because severe sediment deposition is caused by more intensive sediment exchanges in the river channel. However, the low correlation degree indicates that the amount of sediment deposition is not closely related to adjustment index in channel geometry in the braided reach.

Figure 6.

Relationships between channel geometry adjustment index and influencing factors of: (a) sediment deposition; (b) pre-flood characteristic parameter of channel geometry; (c) sediment concentration; (d) mean sediment transport rate.

3.2. Previous channel geometry

The preceding channel geometry can represent the characteristic of the channel before a hyperconcentrated flood, which may also influence the channel adjustment. Zhang et al. (2002) found that the variation in width-depth ratio would be more severe if the previous channel geometry is wider and shallower. In this study, the pre-flood characteristic channel geometry parameter (M_b) was used to represent the preceding channel geometry. The relationship was tested between the M_b and η in the braided reach, as shown in Figure 6(b). It can be seen that the value of η decreases with an increase in the pre-flood characteristic parameter of channel geometry, with the square of correlation coefficient of 0.43 based on the linear regression. It indicates that the reach-scale channel adjustment would be more remarkable with the wider and shallower channel geometry before a hyperconcentrated flood. This is because hyperconcentrated floods are more likely to overtop the bankfull level if the preceding channel geometry is wider and shallower. When the flood level approaches the bankfull level, the sediment carried by hyperconcentrated floods will deposit on the marginal floodplain, which greatly reduces the bankfull width in the main channel. The corresponding sediment transport capacity will increase more significantly than non-overbank floods in the main channel, which results in more intensive scour in the main channel and more sedimentation on the floodplain. Therefore, the pre-flood channel geometry can affect the adjustment index to some extent, but is not the key influencing factor.

3.3 Flow and sediment regime

Flow and sediment regime is generally considered as the main influencing factor of channel geometry (Wu et al., 2008a, 2008b; Xia et al., 2016). Previous studies show that the channel at the specific section would become narrower and deeper after a hyperconcentrated flood with higher sediment concentrations (Jiang et al., 1999; Zhang et al., 2002). Figure 6(c) shows the relationship between the mean sediment concentration (S) at HYK and η in the braided reach. It is obvious that the value of η decreases with an increase in sediment concentration, with the square of correlation coefficient of 0.32 based on the linear regression. The results indicate that a higher sediment concentration can cause the reach-scale channel geometry to become narrower and deeper, which is similar to the results obtained at section scale in previous studies (Chien and Wan, 1998; Qi et al., 2008; Wang et al., 2000). However, the correlation degree between S and η is relatively low, and the sediment concentration is not the primary parameter to influence the adjustment in channel geometry caused by hyperconcentrated floods. Schumm (1977) proposed that sediment transport rate can be used to predict the development of channel geometry, and the channel became narrower and deeper with a decrease in sediment transport rate. However, this conclusion was drawn based on the investigation into low sediment concentration floods. Zhang et al. (2002) investigated the adjustments in channel geometry in the LYR, and found that the cross-sectional geometry became narrower and deeper with an increase in sediment transport rate during hyperconcentrated floods. In order to determine the relationship between sediment transport rate and adjustments in reach-scale channel geometry, the mean sediment transport rate (Q_s) at HYK was calculated by the function of $Q_{S} = Q \times S / 1000$ . Figure 6(d) shows the relationship between Q_s and η in the whole braided reach. The linear regression shows that the value of η decreases with an increase in the mean sediment transport rate, with a square of correlation coefficient of 0.67. Therefore, the theory obtained from rivers with relatively low suspended sediment concentrations can not be used to explain the behavior of fluvial hyperconcentrated floods. The mean sediment transport rate is the dominant factor influencing the adjustments in channel geometry during hyperconcentrated floods, and the larger the mean sediment transport rate is, the narrower and deeper the channel becomes.

4. Prediction of post-flood channel geometry

The above investigation into various factors influencing the channel geometry adjustment indicates that η varies primarily with the mean sediment transport rate, and the pre-flood channel geometry can partly affect the adjustment index during hyperconcentrated floods in the braided reach. In order to predict the post-flood characteristic parameter of channel geometry, the mean sediment transport rate (Q_s) and pre-flood characteristic geometry parameter (M_b) were used to conduct a multiple nonlinear regression analysis. The best-fit equation for M_a can be expressed by

M_{a} = k {(Q_{s})}^{α} {(M_{b})}^{β}

where k, α, and β are parameters that need to be calibrated by the measurements in the braided reach. The data of M_a, Q_s, and M_b in 15 years with the occurrence of hyperconcentrated flood events were divided into two parts. One group consisting of 11 datasets was used to calibrate the parameters, and another group including the datasets in 1973, 1988, 1995, and 2002 was used to verify the predictive accuracy of equation (5). A multiple nonlinear regression analysis was conducted using SPSS software, and the results indicated that the calibrated parameters (k, α, and β) were equal to 3.434, −0.153, and 0.920, respectively. It indicates that M_a has a negative correlation with Q_s and a positive correlation with M_b. Figure 7(a) shows the relationship between M_a and a comprehensive factor including Q_s and M_b, with the square of correlation coefficient of 0.88. Figure 7(b) shows a comparison between the observed M_a and the calculated M_a using equation (5) in the braided reach. The calculated M_a using equation (5) were, respectively, 32%, 5%, and 15% greater than the observed values in 1973, 1995, and 2002, and the calculated M_a was 5% less than the observed one. Therefore, the calculated post-flood characteristic parameters of channel geometry using equation (5) generally agree well with the observed data.

Figure 7.

Calibration and verification of equation (5): (a) relationship between M_a and the comprehensive factor including Q_s and M_b; (b) comparison between the calculated and observed values of M_a in the braided reach.

Compared with the previous studies (Jiang et al., 1999; Sun et al., 2014; Zhang et al., 2002), the effects of flow and sediment regime and preceding channel geometry on adjustments in channel geometry are clarified quantitatively, and the post-flood characteristic parameter of channel geometry after hyperconcentrated floods can be predicted using the proposed method in alluvial rivers. Hyperconcentrated floods can cause intensive scour in the main channel, which leads to severe changes in thalweg migration and river regime. The abrupt and significant changes in channel geometry may threaten the safety of river training works. Therefore, it is necessary to monitor the discharge and sediment concentration during hyperconcentrated floods and to predict the adjustments in channel geometry. Besides, the river training works and flood control projects should be improved to prevent the potential risk caused by hyperconcentrated floods in the braided reach of the LYR, especially in the local region with a larger adjustment index in channel geometry. However, the empirical relation can not fully represent the complex response of channel geometry to hyperconcentrated floods. Whether the hyperconcentrated floods are overbank or non-overbank can also have an impact on the post-flood channel geometry, and the adjustments in overbank hyperconcentrated floods were generally greater than those in non-overbank hyperconcentrated floods. In addition, the flows with low concentrations during non-flood seasons can also restore the narrow and deep channel formed by hyperconcentrated floods.

V. Conclusions

Hyperconcentrated floods are hydrodynamically complex and usually cause severe changes in morphology, which are still poorly understood. This study introduces a parameter of reach-scale adjustment index in channel geometry to investigate the complex fluvial processes of hyperconcentrated floods. Adjustments in channel geometry at section- and reach-scales in the braided reach of the LYR were investigated in the selected 15 years. Several factors influencing the reach-scale adjustment index in channel geometry were identified, including the amount of sediment deposition, preceding channel geometry, and incoming flow and sediment regime. The conclusions obtained from this study include as follows:

Hyperconcentrated floods can cause the cross-sectional geometry to become narrower and deeper, with severe sedimentation on the floodplain. Overbank hyperconcentrated floods have a more remarkable effect on adjustments in channel geometry than non-overbank hyperconcentrated floods. However, the shaped narrow and deep channel is unstable, and the channel may become wide and shallow again owing to bank erosion after low concentration flows during non-flood seasons.

The reach-scale adjustment indexes in channel geometry (η) were all less than 1.0 in hyperconcentrated floods. Various factors influencing channel geometry adjustments were investigated quantitatively, and it was found that the mean sediment transport rate was the key parameter to determine the channel adjustments in hyperconcentrated floods.

An empirical relation including the variables of the mean sediment transport rate (Q_s) and pre-flood characteristic parameter of channel geometry (M_b) was developed to predict the post-flood characteristic parameter of channel geometry (M_a) in the braided reach of the LYR. The empirical relation can also be used to quantitively analyze the channel adjustments at reach scale in other alluvial rivers, which can help to provide more understanding of the morphology changes in hyperconcentrated floods.

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The study reported herein was supported partly by the Program of the National Key Research and Development Plan (Grant No. 2017YFC0405501), and it was mainly supported by the National Natural Science Foundation of China (Grant Nos. 51725902, 51579186 and 51379156).

ORCID iD

Junqiang Xia

References

Batalla

Jong

Ergenzinger

. (1999) Field observations on hyperconcentrated flows in mountain torrents. Earth Surface Processes and Landforms 24(3): 247–253.

Breien

De Blasio

Elverhøi

. (2008) Erosion and morphology of a debris flow caused by a glacial lake outburst flood, western Norway. Landslides 5(3): 271–280.

Chen

(2012) Reservoir sedimentation and transformation of morphology in the lower Yellow River during 10 year’s initial operation of the Xiaolangdi reservoir. Journal of Hydrodynamics 24(6): 914–924.

Cheng

Nguyen

(2011) Hydraulic radius for evaluating resistance induced by simulated emergent vegetation in open-channel flows. Journal of Hydraulic Engineering 137(9): 995–1004.

Chien

Wan

(1998) Mechanics of Sediment Transport. Reston, VA: ASCE Press.

Chu

(2014) The dramatic changes and anthropogenic causes of erosion and deposition in the lower Yellow (Huanghe) River since 1952. Geomorphology 216: 171–179.

Englund

Wan

(1984) Instability of hyperconcentrated flow. Journal of Hydraulic Engineering 110(3): 219–233.

Fei

(1998) Mechanics and application of sediment transport at long distance in hyperconcentrated floods. Journal of Sediment Research 3: 55–61. (in Chinese).

Harman

Stewardson

Derose

(2008) Variability and uncertainty in reach bankfull hydraulic geometry. Journal of Hydrology 351(1): 13–25.

10.

Jiang

Zhang

Zhao

. (1999) Study of bed building law of hyperconcentrated flood and river facies relation. Yellow River 21: 12–15. (in Chinese).

11.

Julien

Wargadalam

(1995) Alluvial channel geometry: theory and applications. Journal of Hydraulic Engineering 121(4): 312–325.

12.

Kostaschuk

James

Rishi

(2003) Suspended sediment transport during tropical cyclone floods in Fiji. Hydrological Processes 17(6): 1149–1164.

13.

van Maren

Wang

. (2014) Peak discharge increase in hyperconcentrated floods. Advances in Water Resources 67(4): 65–77.

14.

Lirer

Vinci

Alberico

. (2001) Occurrence of inter-eruption debris flow and hyperconcentrated flood-flow deposits on Vesuvio volcano. Sediment Geology 139(2): 151–167.

15.

Maizels

(1989) Sedimentology, paleoflow dynamics and flood history of jökulhlaup deposits: paleohydrology of Holocene sediment sequences in southern Iceland sandur deposits. Journal of Sedimentary Research 59:(2) 204–223.

16.

Miao

Kong

. (2016) Functional degradation of the water–sediment regulation scheme in the lower Yellow River: spatial and temporal analyses. Science of the Total Environment 551–552: 16–22.

17.

Miao

Borthwick

AGL

(2010) Recent changes of water discharge and sediment load in the Yellow River basin, China. Progress in Physical Geography 34(4): 541–561.

18.

Pierson

Costa

(1987) A rheologic classification of sub-aerial sediment-water flows. In: Costa

Wieczorek

(eds) Debris Flows/Avalanches: Process, Recognition, and Mitigation. Geological Society of America Reviews in Engineering Geology. Boulder, CO: Geological Society of America, 1–12.

19.

(1996) Evolutional characteristics of hyper-concentrated flow in braided channel of the lower Yellow River. International Journal of Sediment Research 11(3): 49–57.

20.

Sun

. (2008) Efficient sediment transport theory of the hyperconcentrated flow in the lower Yellow River. Journal of Sediment Research 4: 74–82. (in Chinese).

21.

Schumm

(1977) The Fluvial System. New York: John Wiley and Sons.

22.

Shu

Fei

(2008) Sediment transport capacity of hyperconcentrated flow. Science in China Series G: Physics, Mechanics & Astronomy 51(8): 961–975.

23.

Smith

(1986) Coarse-grained non-marine volcaniclastic sediment: terminology and depositional process. Geological Society of America Bulletin 97(1): 1–10.

24.

Sohn

Rhee

Kim

(1999) Debris flow and hyperconcentrated flood-flow deposits in an alluvial fan, northwestern part of the Cretaceous Yongdong Basin, Central Korea. The Journal of Geology 107(1): 111–132.

25.

Strelkoff

Clemmens

(2000) Approximating wetted perimeter in power-law cross section. Journal of Irrigation and Drainage Engineering 126(2): 98–109.

26.

Sun

Liu

Zhang

. (2014) Alluvial river adjusting response to the flood-sediment process: for the wandering reach of Yellow River as an example. Advances in Water Science 25(5): 668–676. (in Chinese).

27.

Van Maren

Winterwerp

Wang

. (2009a) Suspended sediment dynamics and morphodynamics in the Yellow River, China. Sedimentology 56(3): 785–806.

28.

Van Maren

Winterwerp

. (2009b) Modelling hyperconcentrated flow in the Yellow River. Earth Surface Processes and Landforms 34(4): 596–612.

29.

Wan

(1985) Bed material movement in hyperconcentrated flow. Journal of Hydraulic Engineering 111(6): 987–1002.

30.

Wang

Yang

Saito

. (2007) Stepwise decreases of the Huanghe (Yellow River) sediment load (1950–2005): impacts of climate change and human activities. Global and Planetary Change 57(3): 331–354.

31.

Wang

Chen

Zhou

(2000) Analysis on characters and mechanism of fluvial processes of high sediment-laden flow in wandering channel. Journal of Sediment Research 1: 1–6. (in Chinese).

32.

Wang

(2016) Hyper-concentrated flow response to aeolian and fluvial interactions from a desert watershed upstream of the Yellow River. Catena 147: 258–268.

33.

Wang

Melching

(2009) Fluvial hydraulics of hyperconcentrated floods in Chinese rivers. Earth Surface Processes and Landforms 34(7): 981–993.

34.

Wilcox

O’Connor

Major

(2014) Rapid reservoir erosion, hyperconcentrated flow, and downstream deposition triggered by breaching of 38 m tall Condit Dam, White Salmon River, Washington. Journal of Geophysical Research: Earth Surface 119(6): 1376–1394.

35.

Winterwerp

(2006) Stratification effects by fine suspended sediment at low, medium and very high concentrations. Journal of Geophysical Research 111: C05012.

36.

Wohl

Kuzma

Brown

(2004) Reach-scale channel geometry of a mountain river. Earth Surface Process and Landforms 29(8): 969–981.

37.

Wang

. (2005) Case study: river training and its effects on fluvial processes in the Lower Yellow River, China. Journal of Hydraulic Engineering 131(2): 85–96.

38.

Wang

Xia

. (2008a) Response of bankfull discharge to discharge and sediment load in the lower Yellow River. Geomorphology 100(3): 366–376.

39.

Xia

. (2008b) Effect of altered flow regime on bankfull area of the lower Yellow River, China. Earth Surface Processes and Landforms 33(10): 1585–1601.

40.

Xia

Deng

. (2016) Dynamic channel adjustments in the Jingjiang Reach of the Middle Yangtze River. Scientific Reports 6: 22802.

41.

Xia

. (2014a) Response of reach-scale bankfull channel geometry in the Lower Yellow River to the altered flow and sediment regime. Geomorphology 213: 255–265.

42.

Xia

Zong

Zhang

. (2014b) Prediction of recent bank retreat processes at typical sections in the Jingjiang Reach after the TGP operation. Science China Technological Sciences 57(8): 1490–1499.

43.

(2004) Hyperconcentrated flows in the slope-channel systems in gullied hilly areas on the Loess Plateau, China. Geografiska Annaler 86(4): 349–366.

44.

Zhang

(2002) Impact of flood events from different source areas on channel transects adjustment at the wandering-braided reach of the Lower Yellow River. Journal of Sediment Research 6: 1–6. (in Chinese).

45.

Zhang

Wang

. (2015) An experimental study of fluvial processes at asymmetrical river confluences with hyperconcentrated tributary flows. Geomorphology 230: 26–36.