Advertisement

Journal of Soils and Sediments

, Volume 19, Issue 12, pp 4073–4085 | Cite as

Temporal changes in suspended sediment transport during the past five decades in a mountainous catchment, eastern China

  • Haiyan FangEmail author
Sediments, Sec 2 • Physical and Biogeochemical Processes • Research Article
  • 127 Downloads

Abstract

Purpose

Analysis of sediment transport is an effective approach for identifying sediment sources and for catchment management. However, a long-term analysis of sediment variability at multiple time scales is less available, especially in mountainous catchments. This study aims to determine sediment sources and to identify sediment transport dynamics, as well as the influencing factors, through analyzing long-term sediment fluxes at different time scales.

Materials and methods

In this paper, 32 years of sediment flux in an instrumented mountainous catchment in eastern Zhenjiang Province, China, was investigated at multiple time scales (i.e., monthly, seasonal, annual, and event). Sediment yields (SYs) during three time periods (i.e., 1964–1977, 1978–1989, and 2010–2015) were first classified by the Mann-Kendall and cumulative anomaly tests, and then sediment fluxes for each period were investigated and compared at multiple time scales.

Results and discussion

Annual SYs ranged from 29 to 308 t year−1 during the recording period and were significantly influenced by several high magnitude flood events. The mean annual SYs decreased from 153.82 t year−1 in 1964–1977, to 97.79 t year−1 in 1978–1989, and to 91.08 t year−1 in 2010–2015 due to improved soil conservation measures and increased reservoirs. At the seasonal scale, over 92% of the sediment was transported in spring and summer over the recording period. Heterogeneous sediment sources, partial areal distribution of rainfalls, and other factors led to complex suspended sediment concentration versus water discharge hysteresis loop patterns at the event and monthly scales.

Conclusions

The improved soil conservation measures and increased reservoirs over the recent decades decreased sediment availability, and the number and the magnitude of flood events from 1964 onward. However, the flood sediment fluxes in a few months were still high due to extreme precipitation events in recent years. The work can provide guidance for addressing sediment problems in this and/or other similar catchments.

Keywords

Hysteresis loop Mountainous catchment Multiple time scales Sediment dynamics 

1 Introduction

Information on sediment transport dynamics for a river catchment is vital to managing sediment-related issues such as river and reservoir siltation, and the transferring of nutrients and contaminant in river networks (Rickson 2014; Vercruysse et al. 2017). Suspended sediment is also linked to ecological degradation, flooding, and damage to infrastructure in an increasingly populated world (Taylor and Owens 2009; Kusimi et al. 2014). Sediment transport is highly variable and time dependent due to different erosion conditions, such as climate, topography, sediment availability, and catchment size (Ludwig and Probst 1998; Gay et al. 2014; Vercruysse et al. 2017). For example, sediment load is controlled by the hyperconcentrated flow on the Chinese Loess Plateau (Fang et al. 2008a, b), while alternatively snow-melt runoff in winter contributes a great deal of sediment in the glacierized Andean catchments (Mao and Carrillo 2017). Sediment load can also be affected by anthropogenic alterations. Human activities, such as soil and water conservation measures and reservoirs, can reduce sediment supply (Suif et al. 2016). On the contrary, human activities such as the intensification of land use, mining, and construction works can increase catchment sediment yields (SYs; Vercruysse et al. 2017).

Analysis of sediment transport is an effective approach for quantifying sediment load and identifying sediment sources for river and catchment managements. In recent years, many papers on temporal dynamics of sediment transport have been published in different mountainous regions such as in the Mediterranean (e.g., Lenzi and Marchi 2000; Rovira and Batalla 2006; Béjar et al. 2018), the UK (e.g., Walling and Webb 1982; Haifa 1984; Williams 1989; Worrall et al. 2013), the USA (e.g., Gomez et al. 1997; Gao et al. 2013), Israel (Alexandrov et al. 2007), on the Chinese Loess Plateau (e.g., Fang et al. 2008a, b), in the Beibu Gulf region in southern China (Li et al. 2017), and in Taiwan, China (Kao et al. 2005), with a variety of results. For example, Alexandrov et al. (2007) found that convectively enhanced high intensity rainstorms produced a clockwise (CW) hysteresis in the suspended sediment concentration (SSC)-water discharge (Q) relation, while low intensity frontal storms led to either counter clockwise (CCW) or no hysteresis in the SSC–Q relation. Lenzi and Marchi (2000) demonstrated that when the sediment source was in channels, CW hysteresis occurred, CCW loops occurred when sediment is from slopes, and different SSC–Q relations can coexist in the same flood when sediment sources changed. Similarly, Nistor and Church (2005) and Fan et al. (2012) identified sediment sources through analyzing sediment dynamics at multiple time scales. However, although much work on sediment transport fluxes and their temporal variations have been conducted, there are at least three chief limitations. First, studies on long-term sediment variability at multiple time scales are less available, with several exceptions. For example, nearly 30 years (1986–2014) of sediment fluxes in an Alpine basin in Rio Cordon, Italy, was investigated by Rainato et al. (2017), and a good relationship between peak Q and sediment load at flood event scale was found, with assessing the sediment contributions of single floods and seasonal sediment flux to annual SYs. Gao et al. (2018) studied the changes of flow-sediment relationships of 14 catchments on the Chinese Loess Plateau during 1956–2014 and found that land use change greatly influenced these flow-sediment relations at daily, seasonal, and yearly scales. Second, information of sediment dynamics is lacking in river mountainous catchments such as those in Zhejiang Province, China, because high forest coverage and low SSCs draw less attention from scientists and local managers. However, due to pervasive human activities (e.g., deforestation, road construction, and mining) in past decades, sediment transported from upper catchments in Zhejiang Province has raised some river beds by 0.5 to 1. 0 m, decreasing the inland river course by more than 1000 km (Yu 1994). Finally, the impacts of human activities on sediment dynamics are still required although extensive studies have been conducted on some large rivers, such as in the Nile River Basin (Frihy et al. 1998) and in the Changjiang River basin (Mei et al. 2015; Li et al. 2017; Dai et al. 2018) where suspended sediment load decreased greatly due to human activities.

Against this background, there is a clear need for analyzing sediment fluxes and the influencing factors at multiple time scales for catchment management. Therefore, a catchment in the mountainous regions of eastern Zhejiang Province was selected for this study to analyze sediment dynamics over a long-term time period (1964–1989 and 2010–2015). The specific aims of this study were to (i) identify sediment dynamics at different time scales and (ii) explain their changing trends during this long-term period.

2 Materials and methods

2.1 The Quxian catchment

The Qujiang River belongs to the Qiantang river system in eastern China (Fig. 1). The catchment is monitored at the Quxian hydrologic station and hereafter is called Quxian catchment. The catchment has an area of 5424 km2. The mean elevation is 360 m a.s.l, with a maximum of 1481 m a.s.l. The catchment has a subtropical monsoon climate and long-duration frontal rainfalls. Mean annual precipitation from 1964 to 2015 was 1794 mm, ranging from 1100 to 2495 mm. The rainfall amount in May or June accounts for approximately 33% of the total annual. The mean annual temperature is 17 °C. Steep topography characterizes the catchment, 70% of which is covered by mountains. The mean slope gradient of the catchment is 34%, ranging from 0 to 289%. The main soil types are humicacrisols, haplicalisols, and dystricregosols. Red sandstone dominates the hilly area. This kind of lithology is friable and easily eroded (Liao 2008).
Fig. 1

Maps showing the locations of (a) Zhejiang Province, (b) the Quxian catchment, and (c) the reservoirs, the meteorological and hydrologic stations within the study catchment

In the past decades, the landscape has experienced several changes since the establishment of the People’s Republic of China. From 1955 to 1958, some policies were installed to conserve the soil. Disorderly land reclamation was prohibited, and reforestation was advocated. However, during 1959–1961 and 1966–1976, large amounts of forests were destroyed, and steep lands were cultivated for grains. From 1980 onwards, local governments started to implement soil conservation measures again. These measures mainly include reforestation, contour cultivation, and planting pits. In 2000, a soil conservation plan was issued, and large-scale soil conservation measures were implemented. More agricultural lands on steep slopes, sparse forestlands, and some grasslands were implemented with conservation measures. Terraces were constructed in fields in combination with level ditches on sloping lands. Afforestation was conducted with fish-scale pits (a pit like a fish scale within which one tree is planted) in sparse forestlands. However, the lands with sparse grass and forest cover in regions over 500 m a.s.l were surrounded by fences for natural vegetation recovery. Land use maps indicated that until 2015, 72.50% of the catchment area was covered by forest lands, followed by 14.90% of dry lands. During 1985–2015, the area percentages of construction lands increased from 0.61 to 2.88% (Fig. S1 - ESM).

Reservoirs were built for the purpose of flood control in this region (Fig. 1c). Before 1966, a total of 12 reservoirs in three counties (i.e., Jiangshan, Kaihua, and Changshan) in the catchment were built, each having a capacity of less than one million cubic meters. From 1967 to 1977, 15 more reservoirs were constructed, including two large reservoirs. From 1980 to 1989, eight reservoirs were built. Up until 2005, there were a total of 35 reservoirs in the three counties, although data of constructed reservoirs in Chan County were not available before 1990 (Fig. S2 - ESM). Information regarding human activities was obtained from County Annals in the Quxian catchment.

2.2 Data source and treatment

The Quxian hydrologic station was built in the upper reaches of Qiantang River, and Q and SSC have been measured since 1964. At the hydrologic station, two or three water samples were collected for each sampling time. A horizontal sediment sampler was used to collect samples at the center of the river where the mean flow velocity was usually reached. The sampling intervals ranged from 5 min to 12 h, depending on the size of the flood. A total of 8149 samples were collected during the recording period (1964–1989, 2010–2015).

The samples were analyzed in the laboratory to determine their SSCs (kg m−3). Water level was observed at the staff gauge, and Q was obtained using previously established Q–water level curve. The curve has been calibrated with the discharges using current meters. The sampling method was kept unchanged during the past decades. The event and daily water yields (WYs; m3) and sediment yields (SYs; kg) were calculated using the measurements and time intervals. Monthly and yearly Q, SSC, WY, and SY were derived from daily data. In the catchment, rainfalls were recorded at five sites. Areal precipitation for the catchment was obtained by using Thiessen Method with the point station data.

All measurements of water stage, Q, and SSC at the Quxian hydrologic station, and rainfall data at six meteorological stations (i.e., Changfeng, Misai, Xujiacun, Shuangtadi, Xiakou, and Quxian) between 1964 and 1989 and 2000 and 2015 were compiled by the Qiantang River Water Resources Commission and printed for internal use. Because the SSC data in 1990–2009 were not available, sediment dynamics in this period were not analyzed in this paper. The TM images in 1985 and 2015 were available from Geospatial Data Cloud, Computer Network Information Center of Chinese Academy of Sciences (http://www.gscloud.cn), and several land use types were obtained using supervised classification (Fig. S1 - ESM).

2.3 Abrupt temporal change point test

2.3.1 Mann-Kendalltest

The Mann-Kendall (MK) test, proposed by Sneyers (1990), has shown robustness in temporal change detection (e.g., Yue and Pilon 2004; Ullah et al. 2018) and was employed to determine abrupt changes in annual precipitation, WY, and SY for this study using the following equations:
$$ {s}_{\mathrm{i}}=\sum \limits_{i=1}^n{t}_{\mathrm{i}} $$
(1)
$$ {t}_{\mathrm{i}}=\left\{\begin{array}{c}1,{x}_{\mathrm{i}}>{x}_{\mathrm{j}}\\ {}0,{x}_{\mathrm{i}}\le {x}_{\mathrm{j}}\end{array}\right.,\left(i=1,2,\dots, n;j=1,2,\dots .,i-1\right) $$
(2)
$$ E\left({s}_{\mathrm{i}}\right)=\frac{n\left(n-1\right)}{4} $$
(3)
$$ \mathrm{Var}\left({s}_{\mathrm{i}}\right)=\frac{n\left(n-1\right)\left(2n+5\right)}{72} $$
(4)
$$ \mathrm{UF}(s)=\frac{s_{\mathrm{i}}-E\left({s}_{\mathrm{i}}\right)}{\sqrt{\mathrm{Var}\left({s}_{\mathrm{i}}\right)}} $$
(5)
where xi is the ith variable, ti is the number of cases xi > xj, and E(si) and Var (si) are the mean and variance of the test variables, respectively. Two statistical curves UF and UB were obtained. The overlapping of these two curves can locate the beginning of a trend. If |UF(s)| ≤ UF(s)1 − α/2, a significant upward or downward trend exists. The confidence interval was taken as ± 1.96 at the 95% significance level.

2.3.2 Cumulative anomaly method

In combination with the MK test, a cumulative anomaly method was employed to identify abrupt temporal change points (e.g., Lozowsko et al. 1989; Jiang et al. 2017; Zhao et al. 2017). For a discrete series xi, the cumulative anomaly (Xt) can be calculated:
$$ {X}_{\mathrm{t}}=\sum \limits_{i=1}^t\left({x}_{\mathrm{i}}-{x}_{\mathrm{m}}\right),t=1,2,\dots, n $$
(6)
$$ {x}_{\mathrm{m}}=\frac{\sum \limits_{i=1}^n{x}_{\mathrm{i}}}{n} $$
(7)
where xm is the mean value of the series xi and n is the number of data points.

2.4 SSC–Q hysteresis loops and categorization

The hysteresis loop between SSC and Q can usually be used to investigate sediment sources (Williams 1989; Fang et al. 2008a, b). A CW hysteresis loop indicates a rapid delivery of sediment from channels, while CCW loops suggest the delay of sediment transport. The relations of SSC–Q at multiple time scales were categorized by using the method in Williams (1989).

3 Results

3.1 Variations of annual sediment load

During the recording period, the mean annual SY was 121 × 104 t, equivalent to an area-specific SY of 223 t km−2 year−1, varying from 29 × 104 t in 2013 to 308 × 104 t in 1983 (Fig. 2). Annual SY displayed a decreasing trend from 1964 to 2015. Both the MK and cumulative anomaly tests revealed a statistically significant change point of annual SY in 1977 (Fig. 3). The time period 2010–2015 was regarded as an individual time stage because the data for 1990–2009 were not available. Therefore, the entire time period was divided into three time stages: 1964–1977, 1978–1989, and 2010–2015. The mean annual SYs decreased from 154 × 104 t in 1964–1977, to 98 × 104 t in 1978–1989, and to 91 × 104 t in 2010–2015 (Table 1). The responding area-specific SYs changed from 284 to 168 t km−2 year−1.
Fig. 2

Temporal changes of annual precipitation (P), water yield (WY), sediment yield (SY), and suspended sediment concentration (SSC) during 1964–1989 and 2010–2015

Fig. 3

The change points of annual precipitation P (a, b), WY (c, d), and SY (e, f) calculated using the cumulative anomaly method (a, c, e) and the MK test (b, d, f). The dashed lines in the right figures indicate the confidence interval 1.96 (p = 0.05). UB and UF represent the series of progressive and retrograde values that were calculated by Eq. (5)

Table 1

Statistical characteristics of annual precipitation amount, annual sediment load and water yield for different time periods

 

1964–1977

1978–1989

2010–2015

Total

Mean

CV (%)

Mean

CV (%)

Mean

CV (%)

Mean

CV (%)

Precipitation (mm)

1842

17

1694

0.21

2060

0.22

1827

20

WY (× 108 m3)

65

29

55

33

72

31

62

31

SY (× 104 t)

154

41

98

72

91

41

121

55

SSC (kg m−3)

0.24

34

0.17

38

0.14

63

0.19

44

WY water yield, SY sediment yield, SSC suspended sediment concentration, CV coefficient of variation

Similarly, a statistically significant change point of annual WY was also detected in 1977 through those two tests, while no significant changes were detected for annual precipitation amounts. The yearly mean WYs first declined from 65 × 108 m3 in 1964–1977 to 55 × 108 m3 in 1978–1989, and then increased by 72 × 108 m3 in 2010–2015. Annual SSCs also varied greatly, ranging from 0.11 to 0.38 kg m−3 during the recording period. For the three time periods, annual SSCs decreased from 0.24 kg m−3 in 1964–1977, 0.17 kg m−3 in 1978–1989, to 0.14 kg m−3 in 2010–2015 (Table 1).

3.2 Monthly and seasonal sediment dynamics

Temporal variations in monthly SY and WY over the observed period are shown in Fig. S3 (ESM). The monthly SYs exhibited extraordinary fluctuations, ranging from zero to 137 × 104 t. Higher monthly SYs usually occurred in May or June, and lower values appeared in November, December, or January. Compared with the period in 1964–1977, less monthly SY peaks occurred in 2010–2015. In the recording periods, the monthly SYs were investigated as a function of monthly WYs, and a power-law regression line can be fitted for these two variables. The R2 of the regression curve was 0.80 in 1964–2015 (Fig. 4a). Specifically, the R2 values were 0.84 in 1964–1977, 0.79 in 1978–1989, and 0.80 in 2010–2015, and the intercepts (i.e., coefficient a) of the power-law regression curves varied from 0.06 in 2010–2015, 0.19 in 1978–1989, and to 0.54 in 1964–1977 (Fig. 4b). Student’s t test demonstrated that these three regression lines were significantly different at the 0.05 level.
Fig. 4

Relationships of monthly SY and monthly WY during the entire recording period (a) and during different time periods (b)

The monthly SSC values ranged from 0.03 kg m−3 in December to 0.30 kg m−3 in June for the entire recording period. In respect of the individual stages, the mean monthly SSC values ranged from 0.029 kg m−3 in January to 0.344 kg m−3 in June in 1964–1977, from 0.015 kg m−3 in January to 0.278 kg m−3 in June in 1978–1989, and from 0.020 kg m−3 in January to 0.254 kg m−3 in June in 2010–2015. The monthly SSC–Q hysteresis loops were complex with different loop directions and extents. Figure 5 also indicated that the monthly SSC and Q values were higher in spring (i.e., from March to May) and early summer (i.e., June and July) than those in autumn and winter. On average, the monthly SSCs were 0.073 kg m−3 in March, 0.086 kg m−3 in April, 0.081 kg m−3 in May, and 0.254 kg m−3 in June. However, during other months, the mean SSC was less than 0.050 kg m−3 or even close to zero (Fig. S4 - ESM). Due to a high correlation between monthly WY and SY values (R2 = 0.80; Fig. 4), over 92% of the sediment was transported in late spring and early summer (i.e., from April to June). However, due to lower monthly SSCs in 2010–2015, the monthly SYs were lower for almost all the months although the monthly WYs were usually larger than the counterparts in other two periods (Fig. S4 - ESM).
Fig. 5

Different examples of monthly hysteresis loop types. (a) Figure-eight I type in 1964. (b) CW plus counter-clockwise (CCW) loops in 1988. (c) CCW plus CCW loops in 2012. (d) clockwise (CW) plus clockwise (CW) loops in 2015

3.3 Sediment variability at the event scale

At the flood event scale, the instantaneous SSCs were plotted against Qs. A very wide scatter existed in the relationships between these two variables (Fig. 6a). A single Q or SSC value corresponded to different SSC or Q values. This means that sediment availability differed for each event. For instance, at a Q of 4200 m3 s−1, SSCs ranged from 0.2 to 3.9 kg m−3. Alternatively, an SSC of 0.92 kg m−3 was registered during flow conditions that ranged from 165 to 6961 m3 s−1. Due to the wide scatter of this SSC–Q relation, the coefficients of determination R2 were low. The R2 values were 0.51 in 1964–1977, 0.44 in 1978–1989, and 0.50 in 2010–2015. The slope of the linear fitting line in 1964–1977 was the largest (i.e., 2.53), and the slope in 2010–2015 (i.e., 1.65) was the smallest. Student’s t test indicated that these slopes were significantly different at the significant level of 0.05. Noticeably, the event SSCs in drought and flood years as defined by Mei et al. (2018) differed greatly. The event SSCs in drought years were generally higher than those in flood years. For example, five instantaneous SSCs in the extreme drought year 1979 were higher than 2.5 kg m−3, which were much higher than those in flood years during the time period 1978–1989. Similarly, the SSCs in extreme flood year 2015 were much lower than those in other years in 2010–2015.
Fig. 6

Relationships of event discharge (Q) and suspended sediment concentration (SSC) at the event flood scale for the three time stages during the recording period

As in other studies (e.g., Williams 1989; Mao and Carrillo 2017), the scattered SSC–Q relation showed a hysteretic behavior due to the lag time between peaks of water and sediment. Accordingly, the occurring sequences of the peaks of SSC and Q as well as the derived hysteresis loops were analyzed for the recording 201 floods. The floods within which sediment peaks (SSCp) appeared earlier than peak discharges (Qp; hereafter called type I floods) occupied approximately 73% of the total (Table 2). The floods within which SSCp had a good time agreement (hereafter called type II floods) or lagged behind Qp (hereafter called type III floods) occupied 14% and 13%, respectively. Hysteresis loops were carefully identified by plotting the SSC against Q for each event, and four types of hysteresis loops were found. They were CW, CCW, figure-eight, and complex loops (Fig. S5 - ESM). Over 50% of them were CW loops. Complex loops also occupied a major part (n = 70; 34.8%), followed by figure-eight (n = 27; 13.4%) and CCW (n = 3; 0.99%) loops.
Table 2

Discharge-sediment hysteresis loops and flood types of the recorded 201 flood events taking place during the recording period

Periods

Number of events

Hysteresis loops

Types

Clockwise

Counter-clockwise

Figure-eight

Complex

I

II

III

1964–1977

97

53

0

12

32

74

12

11

1978–1989

81

45

1

11

24

60

15

6

2010–2015

23

3

2

4

14

12

2

9

Total

201

101

3

27

70

146

29

26

I indicates that the flood within which the sediment peak (SSCp) precedes peak discharge (Qp), II indicates that the flood has a good time agreement between SSCp and Qp, and III indicates that the flood within which SSCp lags behind Qp

Flood characteristics, including SSCp, mean SSC (SSCm), SY, Qp, mean Q (Qm), WY, and flood duration, influence the flood type. All of these variables presented extraordinary fluctuations (Table S1 - ESM). Long duration floods frequently occurred. During the recording period, the mean duration of the floods was 124.8 h. The minimum and maximum durations of the floods lasted 1 day and 20 days, respectively. The Qm was 781.2 m3 s−1, ranging from 281.0 to 2290.4 m3 s−1. The Qp was above 7000 m3 s−1. A long duration and high Q yielded higher WY and SY values. For example, the flood from July 8–16, 1966, lasted 9 days and transported 94.1 tons of sediment, occupying over 60% of the total in 1966. A flood event in 1971 lasted 14 days and transported 143.0 million tons of sediment, which accounted for over 82% of the annual SY. For the three time periods, the mean event WY was the largest in 2010–2015 (i.e., 748.4 million m3). However, the SSCm of 0.1 kg m−3 in 2010–2015 was the lowest. For the recording periods, the mean event SY was 15.5 × 104 t, ranging from less than 0.1 × 104 to 143.0 × 104 t. Approximately 6.0 floods occurred each year during the recording period, and the annual mean numbers of flood events decreased from 6.9 floods in 1964–1977, 6.8 floods in 1978–1989, to 3.8 floods in 2010–2015.

4 Discussion

4.1 Flood events and SY

The event SSCs in the study catchment were not very high. The largest instantaneous SSC was less than 4.50 kg m−3 (Fig. 6). Although these SSC values varied greatly, the SSCs were much smaller than those in other regions. For example, the event SSC in the Chinese Loess Plateau is hundreds of kilograms per cubic meter (Xu 2002; Fang et al. 2008a, b; Zhang et al. 2016). However, the flood duration in Quxian catchment was quite long due to frontal storms that usually have a high intensity in summer (Feng 1990; Xu et al. 2013). For example, a rainstorm at Misai station started on May 28 and ended on June 26, 1964, with a maximum daily rainfall amount of 73.3 mm. For the recorded 201 events, the flood events lasting over 120 h (i.e., 5 days) and over 72 h (i.e., 3 days) at Quxian station occupied over 38% and 70% of the total, respectively (Table S1 - ESM). Due to similar climate condition, long durations of rainfalls and floods also occurred in the Changjiang River (Dai et al. 2016; Mei et al. 2018). A single flood event could thus transport a large amount of sediment, and greatly influenced the annual SY. During the recording period, the SY in the flood year 1983 (defined as the annual WY exceeding the 90th percentile threshold; Mei et al. 2018) was the highest, and as much as 304.4 million tons of sediment was transported out of catchment outlet during this year. The instantaneous Q larger than 1500 m3 s−1 continued from July 7 to 15 with water level higher than 5 m or even above 8 m. The five flood events during this year exported nearly 80% of the annual SY. This is consistent with studies in other regions (i.e., Smith et al. 2003; Fang et al. 2008a, b). For example, Turowski et al. (2009) observed that approximately 41% of the annual SY in 1982–2008 was transported out of the catchment by three flood events in the Erlenbach Basin, Switzerland. In the Chabagou catchment of the Chinese Loess Plateau, over 95% sediment of the total in 1966 was transported by four flood events (Fang et al. 2008a, b). Similarly, in the Loushui river catchment, Jianxi Province of China, the contributions of individual flood events to annual SY ranged from 36 to 91%, with an average of 78% (Sun et al. 2016).

4.2 Sediment source and SSC–Q relations

The water flow velocity on hillslopes is several orders of magnitude lower than that in channels. When a sediment source area is a channel, SSCp precedes Qp, or they occur simultaneously. On the other hand, when sediment source areas are on a catchment’s slope, or in the upper part of a catchment, SSCp lags behind Qp (Haifa 1984; Rovira and Batalla 2006). Thus, analysis of SSC–Q relations for individual events can be of assistance in identifying sediment source areas (Williams 1989). For the study catchment, the percentage of type I and II floods occupied 87% of the total events (Table 2). The sediment could come from slopes and be deposited into the channel network and/or be directly eroded from the channel bank. The available sediment in the channels was flushed away when intensive floods occurred.

The SSC–Q relations were generally shown as CW and CCW hysteresis loops (Table S1 - ESM). A single CW or CCW loop usually had a pair of SSC and Q peaks (Fig. S5b,d - ESM). The fast response of sediment for type I and/or II floods led to a higher SSC on its rising limb than that on the falling limb for a given Q, resulting in a CW hysteresis loop. For example, the SSC of the flood on April 3–7, 1982, was 3.2 kg m−3 on the rising limb, while it was only 0.046 kg m−3 on the recession limb, given a Q value of 395 m3 s−1 (Fig. S5b - ESM). Due to progressive exhaustion of sediment availability, CW hysteresis loops dominated the study area which is in agreement with the observations by Walling and Webb (1982), Asselman (1999), and Rovira and Batalla (2006). In contrast, only three floods had CCW hysteresis loops for the 201 floods. This finding was also similar to those in large catchments (Williams 1989). This can be explained by increased sediment supply from slopes or the upper catchment and a downstream increase in lag time between SSC and Q peaks during long rainfall durations (Klein 1984; De Girolamo et al. 2015). For example, the lag time of the example shown in Fig. S5d (ESM) was up to 8 h.

The hysteresis loops can thus be used to identify sediment source areas (Klein 1984; Williams 1989; Mao and Carrillo 2017). In the case of CW hysteresis loops, the sediment source is the channels or banks, whereas CCW hysteresis loops occur when the sediment source is in relatively faraway places (Oeurng et al. 2010). In this study, single CW loops occupied over 50% of the total, whereas CCW loops were less than 2%. Complex and/or figure-eight loops combined several single CW and/or CCW loops together. This phenomenon indicated that sediment in channels was initially activated during floods, and then the sediment from slopes or in the upper catchment came later.

Approximately 95% of the 201 floods occurred during late spring and summer (i.e., from April to August). This implies that the sediment was accumulated in other months and flushed away during this time period. Physical weathering is intense in the study area, and large amounts of sediment are produced due to the interaction of the freezing-thawing process (Liao 2008; Chen 2016). During autumn and spring, cultivation action on slopes also provides more available sediment. Some sediment may even fall into streams due to steep slopes. Therefore, sediment is relatively rich before flood comes. As a result, SSCs were high in drought years or in the initial stage of a flood year. This implies that in the future, soil conservation measures should be implemented both in channels and on slope surface areas.

4.3 Impact of soil conservation measures and reservoirs

The numbers of flood types and SSC–Q hysteresis loops changed over the three time stages, probably resulting from continuously improved soil conservation measures and reservoirs. The percentages of type I floods decreased from 76% in 1964–1977, 74% in 1978–1989, to 52% in 2010–2015. The percentages of single CW loops at flood event scale also decreased from some 54.5% in 1964–1977, 55.6% in 1978–1989, to 13% in 2010–2015. This implies that less sediment was available from channels over time. During the past decades, many agricultural fields along the rivers were replaced by construction lands (Fig. S1b - ESM). This land use change decreased soil erodibility and also prohibited sediment from directly entering into rivers. Furthermore, more soil conservation measures have been implemented since 1977. The built reservoirs in the catchment can also trap much sediment from upstream mountainous areas (Fig. 1c). In the Chinese Yellow River, the operation of large dams together with land cover change in the middle reaches, has caused stepwise decrease in sediment delivery to the sea (Wang et al. 2017). Therefore, some smaller floods were controlled by the implemented soil conservation measures and reservoirs. In contrast, larger floods were less influenced or not easily controlled and still occurred from May to July 2010–2015 (Fig. S4 - ESM). The shape and extent were the other two characteristics of an SSC–Q loop, and their quantifications have been given in literatures (e.g., Fang et al. 2011; Zuecco et al. 2016). Quantification analysis of these indices in the future can reflect more information of sediment flux. However, although this study did not analyze these indices, it can be estimated that the extents of the SSC–Q loops can become narrower with time due to lower SSC at a given Q in recent years (Fig. 6). In addition, the time duration for future flood monitoring can be shortened to a few months (Fig. S4 - ESM).

In the study area, precipitation can explain around 90% of the variation in annual WY for the entire recording period. This means that floods can be monitored and predicted through weather forecasting. However, only 22% of the variation of annual SY can be explained by annual precipitation (Fig. 7). This implies that sediment availability exerts a great influence on sediment transport in rivers, and more soil conservation measures should be implemented in the study area, although annual SY decreased with time in the past decades (Fig. 2). Studies (e.g., Asselman 2000; Dang et al. 2010) highlighted that the parameter a in the power function of SSC = aQb may give information on the soil erodibility, and the parameter b reflects the sensitivity of SSC to increase in discharge (Kusimi et al. 2014). The decreasing a values from 0.54 to 0.06 in 1964–2015 imply that land surface has become less erodible with time. In contrast, the increasing b value from 1.64 to 2.24 in 1964–2015 reflects that the SSC in rivers has become clearer over recent years (Fig. 4b). This inference can be verified by the relationship between annual precipitation and annual SY. Compared with other two time periods, the annual SY in 2010–2015 was the smallest although annual WY was the largest due to decreased SSC (Fig. S5 and S6). Similar phenomenon was also found in the neighboring Changjiang River catchment, where the Three Gorge Dam greatly reduced sediment discharge into the estuary, despite of almost unchanged water discharge during 2003–2013. The monthly SSCs and their changing trend in the Qujiang River were also similar to those in the Changjiang River (Dai et al. 2016). The changing sensitive nature of soil surface and sediment discharge to water discharge could be thus attributed to the improvement in soil conservation measures and the increased numbers of reservoirs in the catchment (Florsheim et al. 2011; Fig. S2 - ESM). This results in less sediment discharge into the river and the remobilization of deposited sediment in channel networks.
Fig. 7

Relations of annual precipitation, WY, and SY during the entire recording period (a, c), and the three time periods (b, d)

4.4 Explanation of SSC–Q relations

The catchment has an area of over 5000 km2, and there are multiple factors influencing sediment behavior, including sediment sources, rainstorm characteristics, spatial and temporal distributions of water and sediment, runoff amounts and rates, travel rates, and the distances of flood water in the main channel to the catchment outlet (Klein 1984; de Boer and Edmonton 1989; Seeger et al. 2004; De Girolamo et al. 2015; Pietroń et al. 2015; Vercruysse et al. 2017; Sadeghi et al. 2018). Due to fast response of sediment in channels, CW hysteresis loops dominated (Fig. S5b - ESM). However, this does not mean that sediment mainly came from channel networks. Due to the large catchment area, the sediment from slopes could first be deposited in channels and then was transported out of the catchment. This means that the SSC–Q hysteresis becomes more complex in the large catchment depending on storage-removal of sediments in rivers. Sediment source and SSC–Q relations could be altered by land use change. The conversion of cultivated fields to construction lands during 1985–2015 could directly reduce sediment delivery to rivers (Fig. S1 - ESM), resulting in less percentage of CW loops (Table 2). Furthermore, the asynchronicity between sediment production and flood areas in different subcatchments can also induce complex SSC–Q patterns. For example, the rainfall center of mass usually changes over time. As a result, the relative travel times of the flood waves and the sediment fluxes changed in different streams, inducing multiple SS–Q hysteresis loops (Fig. S5e,f - ESM). The figure-eight loop on April 30 to May 4, 2013, can be explained by different sediment source areas.

At the monthly scale, CW loops were first formed due to abundant previously deposited sediment in channels. On the contrary, when sediment was limited in channels, the eroded soils from slopes can supplement sediment again. As a result, monthly CW plus CCW loops formed. Areal and temporal distributions of precipitation and water flow can activate multiple peaks of monthly sediment and runoff flow, inducing complex loops. Monthly precipitations in 1972 and 1988 were shown as examples here (Fig. S6 - ESM). In 1972, the precipitation amounts ranged from 1570 mm at the Quxian station to 1854 mm at the Changfeng station. Multiple peaks of monthly precipitation occurred at the six meteorological stations, with inconsistent changed characteristics in this year. This can thus cause complex monthly SSC–Q loops. In contrast, the monthly patterns of precipitation in 1988 at these stations were simple and consistent, resulting in relatively simple loops (Fig. 5b). Therefore, sediment availability and rainfall patterns could be the main causes inducing complex monthly SSC–Q loops.

This study was to identify long-term sediment dynamics and to explain their changing trends in a catchment. These analyses at one site could be challenging to understand what is going in this large catchment and to discriminate the contributions of individual factors to sediment flux change. This information can be solved in the future study using spatially distributed soil erosion models such as the famous Soil and Water Assessment Tool SWAT. However, in this study, the detailed analyses and discussion of sediment fluxes at different time scales can still identify the changes of sediment flux over a long-time period, identify the general effect of the influencing factors, and give some implications for river and catchment managements.

5 Conclusions

The analysis of sediment dynamics at different time scales (monthly, seasonal, annual, and event) at Quxian catchment can provide insight into the characteristics of sediment fluxes in eastern China. In unraveling the temporal dynamics of sediment issue in the mountainous catchments, the present work contributes to an increase in the knowledge regarding the long-term (1964–2015) and multiple-scale sediment transport processes of river sediment transport in eastern Zhejiang Province, China.

Sediment transport varied greatly at different time scales, depending on water flow, sediment availability, and other factors. Abrupt temporal change points in the annual SY were first identified, and then three stages were classified. Due to improved soil conservation measures and reservoirs, mean annual SY and SSC continuously decreased over the recording period. For each individual time period, over 92% of the sediment was transported during the spring and summer periods due to high Q and long flood durations. Impacted by human activities (e.g., contour cultivation, reforestation, and constructed reservoirs), a large percentage of sediment was transported during few months (i.e., May to July) in 2010–2015. Water supply, sediment availability, and human activities exerted complex effects on monthly SSC–Q relations, inducing multiple hysteresis loops. Annual sediment transport was greatly influenced by several floods. From 1964 onwards, the numbers of annual flood events decreased. More CCW and complex loops occurred, due to changed sediment source areas. This study indicated that improved soil conservation measures and reservoirs were responsible for the decrease in the number and magnitude of flood events. However, the SYs in a few months (i.e., May to July) were still high due to increased precipitation amounts and extreme events in recent years. These analyses at one site could be challenging to understand what is going in the large catchment, and to discriminate their individual contribution. However, the findings of this study can contribute to providing guidance for addressing sediment problems for this and/or other similar catchments around the world by identifying potential sediment sources for further implementing soil conservation measures in rivers and on slopes, and by detecting flood behavior conditions in rivers in advance. In this manner, the sediment peaks in the few months could be mitigated or controlled.

Notes

Funding information

This work was financially supported by projects of the National Natural Science Foundation of China (grant number 41571271).

Supplementary material

11368_2019_2363_MOESM1_ESM.docx (1.6 mb)
ESM 1 (DOCX 1679 kb)

References

  1. Alexandrov Y, Laronne JB, Reid I (2007) Intra-event and inter-seasonal behavior of suspended sediment inflash floods of the semi-arid northern Negev, Israel. Geomorphology 85:85–97Google Scholar
  2. Asselman N (1999) Suspended sediment dynamics in a large drainage basin: the river Rhine. Hydrol Process 13:1437–1450Google Scholar
  3. Asselman NEM (2000) Fitting and interpretation of sediment rating curves. J Hydrol 234:228–248Google Scholar
  4. Béjar M, Vericat D, Batalla RJ, Bibbins CN (2018) Variation in flow and suspended sediment transport in a montane river affected by hydropeaking and instream mining. Geomorphology 310:69–83Google Scholar
  5. Chen RZ (2016) Soil profile weathering feature research of eroded weathering granite slope at different sections. Zhejiang University, Zhejiang, China (M.S. Thesis)Google Scholar
  6. Dai ZJ, Fagherazzi S, Mei XF, Gao JJ (2016) Decline in suspended sediment concentration delivered by the Changjiang (Yangtze) River into the East China Sea between 1956 and 2013. Geomorphology 268:123–132Google Scholar
  7. Dai ZJ, Mei XF, Darby SE, Lou YY, Li WH (2018) Fluvial sediment transfer in the Changjiang (Yangtze) river-estuary depositional system. J Hydrol 566:719–734Google Scholar
  8. Dang TH, Coynel A, orange D, Blanc G, Etcheber H, Le LA (2010) Long-term monitoring (1960-2008) of the river-sediment transport in the Red River Watershed (Vietnam): temporal variability and dam-reservoir impact. Sci Tot Environ 408:4654–4664Google Scholar
  9. de Boer DH, Edmonton IAC (1989) Spatial scale dependence of sediment dynamics in a semi-arid badland drainage basin. Catena 16:277–290Google Scholar
  10. De Girolamo AM, Pappagallo G, Lo Porto A (2015) Temporal variability of suspended sediment transport and rating curvesin a Mediterranean river basin: the Celone (SE Italy). Catena 128:135–143Google Scholar
  11. Fan XL, Shi CX, Zhou YY, Shao WW (2012) Sediment rating curves in the Ningxia-Inner Mongolia reaches of the upper Yellow River and their implications. Quat Int 282:152–162Google Scholar
  12. Fang HY, Cai QG, Chen H, Li QY (2008a) Temporal changes in suspended sediment transportin a gullied loess basin: the lower Chabagou Creek on the Loess Plateau in China. Earth Surf Process Landf 33:1977–1992Google Scholar
  13. Fang HY, Chen H, Cai QG, Huang X (2008b) Effect of spatial scale on suspended sediment concentrationin flood season in hilly loess region on the Loess Plateau in China. Environ Geol 54:1261–1269Google Scholar
  14. Fang HY, Li QY, Cai QG, Liao YS (2011) Spatial scale dependence of sediment dynamics in a gullied rolling loess region on the Loess Plateau in China. Environ Earth Sci 64:693–705Google Scholar
  15. Feng LH (1990) Hydrological law of the Qiantang River basin. Sci Geogr Sin 10:177–184Google Scholar
  16. Florsheim JL, Pellerin BA, Oh NH, Ohara N, Bachand PAM, Bachand SM, Bergamaschi BA, Hernes PJ, Kavvas ML (2011) From deposition to erosion: spatial and temporal variability of sediment sources, storage, and transport in a small agricultural watershed. Geomorphology 132:272–286Google Scholar
  17. Frihy OE, Dewidar KM, El Banna MM (1998) Natural and human impact on the northeastern Nile delta coast of Egypt. J Coast Res 14:1109–1118Google Scholar
  18. Gao P, Nearing MA, Commons M (2013) Suspended sediment transport at the instantaneous and event time scales in semiarid watersheds of southeastern Arizona, USA. Water Resour Res 49:6857–6870Google Scholar
  19. Gao GY, Fu BJ, Zhang JJ, Ma Y, Sivapalan M (2018) Multiscale temporal variability of flow-sediment relationships during the 1950s-2014 in the Loess Plateau, China. J Hydrol 563:609–619Google Scholar
  20. Gay A, Cerdan O, Delmas M, Desmet M (2014) Variability of suspended sediment yields within the Loire river basin (France). J Hydrol 519:1225–1237Google Scholar
  21. Gomez B, Phillips JD, Magilligan FJ, James LA (1997) Floodplain sedimentation and sensitivity: summer 1993 flood, upper Mississippi River Valley. Earth Surf Process Landf 22:923–936Google Scholar
  22. Haifa MK (1984) Anti clockwise hysteresis for suspended sediment concentration during individual storms: Holbeck Catchment; Yorkshire, England. Catena 11:251–257Google Scholar
  23. Jiang C, Zhang L, Tang ZP (2017) Multi-temporal scale changes of stream flow and sediment discharge in the headwaters of Yellow River and Yangtze River on the Tibetan Plateau, China. Ecol Eng 102:240–254Google Scholar
  24. Kao SJ, Chan SC, Kuo CH, Liu KK (2005) Transport-dominated sediment loading in Taiwanese rivers: a case study from the Ma-an stream. J Geol 113:217–225Google Scholar
  25. Klein M (1984) Anti-clockwise hysteresis in suspended sediment concentration during individual storms. Catena 11:251–257Google Scholar
  26. Kusimi JM, Amisigo BA, Banoeng-Yakubo BK (2014) Sediment yield of a forest basin in Ghana. Catena 123:225–235Google Scholar
  27. Lenzi MA, Marchi L (2000) Suspended sediment load during floods in a small stream of the Dolomites (northern Italy). Catena 39:267–282Google Scholar
  28. Li SS, Dai ZJ, Mei XF, Huang H, Wei W, Gao JJ (2017) Dramatic variations in water discharge and sediment load from Nanliu River (China) to the Beibu Gulf during 1960s-2013. Quat Int 440:12–23Google Scholar
  29. Liao CB (2008) Analysis on the processes of water and soil loess and measurement layout of soil and water conservation on red soil slopeland. Zhejiang University, Zhejiang, China (M.S. Thesis)Google Scholar
  30. Lozowsko EP, Charlton RB, Nguyen CD, Wilson JD (1989) The use of cumulative monthly mean temperature anomalies in the analysis of local interannual climate variability. J Clim 2(9):1059–1068Google Scholar
  31. Ludwig W, Probst JL (1998) River sediment discharge to the oceans: present-day controls and global budgets. Am J Sci 298(4):265–295Google Scholar
  32. Mao L, Carrillo R (2017) Temporal dynamics of suspended sediment transport in a glacierized Andean basin. Geomorphology 287:116–125Google Scholar
  33. Mei X, Dai Z, van Gelder PHAJM, Gao J (2015) Linking Three Gorges Dam and downstream hydrological regimes along the Yangtze River, China. Earth Space Sci 2:94–106Google Scholar
  34. Mei XF, Dai ZJ, Darby SE, Gao S, Wang J, Jiang WG (2018) Modulation of extreme flood levels by impoundment significantly offset by floodplain loss downstream of the Three Gorges Dam. Geophys Res Lett 45:1–9Google Scholar
  35. Nistor CJ, Church M (2005) Suspended sediment transport regime in a debris-flow gully on Vancouver Island, British Columbia. Hydrol Process 19:861–885Google Scholar
  36. Oeurng C, Sauvage S, Sánchez-Pérez JM (2010) Dynamics of suspended sediment transport and yield in a large agricultural catchment, Southwest France. Earth Surf Process Landf 35(11):1289–1301Google Scholar
  37. Pietroń J, Jarsjő J, Romanchenko AO, Chalov SR (2015) Model analyses of the contribution of in-channel processes to sediment concentration hysteresis loops. J Hydrol 527:576–589Google Scholar
  38. Rainato R, Ma L, García-Rama A, Picco L, Cesca M, Vianello A, Preciso E, ScusselGR LMA (2017) Three decades of monitoring in the Rio Cordon instrumented basin: sediment budget and temporal trend of sediment yield. Geomorphology 291:45–56Google Scholar
  39. Rickson RJ (2014) Can control of soil erosion mitigate water pollution by sediments? Sci Tot Environ 468-469:1187–1197Google Scholar
  40. Rovira A, Batalla RJ (2006) Temporal distribution of suspended sediment transport in a Mediterranean basin: the Lower Tordera (NE SPAIN). Geomorphology 79:58–71Google Scholar
  41. Sadeghi SH, Singh VP, Kiani-Harchegani M, Asadi H (2018) Analysis of sediment rating loops and particle size distributions to characterize sediment source at mid-sized plot scale. Catena 167:221–227Google Scholar
  42. Seeger M, Errea MP, Beguería S, Arnáez J, Martí C, García-Ruiz JM (2004) Catchment soil moisture and rainfall characteristics as determinant factors for discharge/suspended sediment hysteretic loops in a small headwater catchment in the Spanish Pyrenees. J Hydrol 288:299–311Google Scholar
  43. Smith BPG, Naden PS, Leeks GJL, Wass PD (2003) The influence of stormevents on fine sediment transport, erosion and deposition within a reach of the River Swale, Yorkshire, UK. Sci Tot Environ 314-316:451–474Google Scholar
  44. Sneyers R (1990) On the statistical analysis of series of observations. WMO technical note 143. World Meteorological Organization, GenevaGoogle Scholar
  45. Suif Z, Fleifle A, Yoshimura C, Saavedra O (2016) Spatio-temporal patterns of soil erosion and suspended sediment dynamics in the Mekong River Basin. Sci Tot Environ 568:933–945Google Scholar
  46. Sun LY, Yan M, Cai QG, Fang HY (2016) Suspended sediment dynamics at different time scales in the Loushui River, south-central China. Catena 136:152–161Google Scholar
  47. Taylor KG, Owens PN (2009) Sediments in urban river basins: a review of sediment-contaminant dynamics in an environmental system conditioned by human activities. J Soils Sediments 9:281–303Google Scholar
  48. Turowski JM, Yager EM, Badoux A, Rickenmann D, Molnar P (2009) The impact of exceptional events on erosion, bedload transport and channel stability in a steppool channel. Earth Surf Process Landf 34:1661–1673Google Scholar
  49. Ullah S, You QL, Ullah W, Ali A (2018) Observed changes in precipitation in China-Pakistan economic corridor during 1980-2016. Atmos Res 210:1–14Google Scholar
  50. Vercruysse K, Grabowski R, Rickson RJ (2017) Suspended sediment transport dynamics in rivers: multi-scale drivers of temporal variation. Earth-Sci Rev 166:38–52Google Scholar
  51. Walling DE, Webb BW (1982) Sediment availability and the prediction of storm-period sediment yields. In: Recent developments in the explanation and prediction of Erosion and sediment yield. IAHS Publication 137, IAHS Press, Wallingford, pp 327–337Google Scholar
  52. Wang HJ, Wu X, Bi NS, Li S, Yuan P, Wang AM, Syvitski PM, Saito Y, Yang ZS, Liu SM, Nittrouer J (2017) Impacts of the dam-orientated water-sediment regulation scheme on the lower reaches and delta of the Yellow River (Huanghe): a review. Glob Planet Chang 157:93–113Google Scholar
  53. Williams GP (1989) Sediment concentration versus water discharge during single hydrologic events in rivers. J Hydrol 111:89–106Google Scholar
  54. Worrall F, Burt TP, Howden NJK (2013) The flux of suspended sediment from UK 1974 TO 2010. J Hydrol 504:29–39Google Scholar
  55. Xu JX (2002) Implication of relationships among suspended sediment size, water discharge and suspended sediment concentration: the Yellow River basin, China. Catena 49:289–307Google Scholar
  56. Xu YP, Zhang XJ, Ran QH, Tian Y (2013) Impact of climate change on hydrology of upper reaches of Qiantang River Basin, East China. J Hydrol 483:51–60Google Scholar
  57. Yu ZY (1994) Soil of Zhejiang. Zhejiang Science and Technology Press, ZhejiangGoogle Scholar
  58. Yue S, Pilon P (2004) A comparison of the power of the t test, Mann–Kendall and bootstrap tests for trend detection. Hydrol Sci J 49(1):21–37Google Scholar
  59. Zhang LT, Li ZB, Wang SS (2016) Spatial scale effect on sediment dynamics in basin-wide floods within a typical agro-watershed: a case study in the hilly loess region of the Chinese loess plateau. Sci Tot Environ 572:476–486Google Scholar
  60. Zhao J, Xu JC, Li XC, Zhong Y, Han DH, Qiu H (2017) Characteristics analysis of spatial and temporal variation on extreme weather events in Anhui Province for recent 50 years. Nat Hazards 89:817–842Google Scholar
  61. Zuecco G, Penna D, Borga M, van Meerveld HJ (2016) A versatile index to characterize hysteresis between hydrological variables at the runoff event timescale. Hydrol Process 30:1449–1466Google Scholar

Copyright information

© Springer-Verlag GmbH Germany, part of Springer Nature 2019

Authors and Affiliations

  1. 1.Key Laboratory of Water Cycle and Related Land Surface Processes, Institute of Geographic Sciences and Natural Resources ResearchChinese Academy of SciencesBeijingChina
  2. 2.College of Resources and EnvironmentUniversity of Chinese Academy of SciencesBeijingChina

Personalised recommendations