Journal of Oceanology and Limnology   2019, Vol. 37 issue(3): 982-997     PDF       
http://dx.doi.org/10.1007/s00343-019-8156-3
Institute of Oceanology, Chinese Academy of Sciences
0

Article Information

YU Xiaoxiao, LI Tiegang, GU Dongqi, FENG Aiping, LIU Shihao, LI Ping, XU Guoqiang, YAN Wenwen, ZHANG Zhiwei, ZHU Zhengtao
Sediment transport in the Luanhe River delta: grain size trend analysis
Journal of Oceanology and Limnology, 37(3): 982-997
http://dx.doi.org/10.1007/s00343-019-8156-3

Article History

Received May. 20, 2018
accepted in principle Jun. 12, 2018
accepted for publication Jul. 13, 2018
Sediment transport in the Luanhe River delta: grain size trend analysis
YU Xiaoxiao1,2,3, LI Tiegang1,2,3, GU Dongqi3, FENG Aiping3, LIU Shihao3, LI Ping3, XU Guoqiang3, YAN Wenwen3, ZHANG Zhiwei3, ZHU Zhengtao3,4     
1 Key Laboratory of Marine Geology and Environment, Institute of Oceanology, Chinese Academy of Sciences, Qingdao 266071, China;
2 University of Chinese Academy of Sciences, Beijing 100049, China;
3 The First Institute of Oceanography, Ministry of Natural Resources (MNR), Qingdao 266061, China;
4 State Key Laboratory of Estuarine and Coastal Research, East China Normal University, Shanghai 200062, China
Abstract: Sediment grain size in the deltaic environment of the Luanhe River (LR), Liaoning, China, contains sediment transport pathway information useful in elucidating the shoreline change and fluvialmarine interaction. In this study, we utilized numerical partitioning of the sedimentary components and geostatistical grain size trend analysis (GSTA) to define the sediment transport pattern in the Luanhe River delta (LRD) and interpolated the sediment transport pattern using content changes of end numbers (EM). EM1 (the mean grain size 7.12 Ф, fine silt), EM2 (2.37 Ф, fine sand), and EM3 (1.27 Ф, medium sand) components were identified by the numerical partitioning by GSTA. Kriging interpolation method was used to interpolate the parameters of the grain size for the regular grid, and the interpolation radius was 0.015 decimal degree. We chose 0.09 decimal degree as the characteristic distance for GSTA in the semivariogram model using the geostatistical method. The FB(-) case (finer, better sorted and more negatively skewed) was adopted in GSTA for its satisfaction in the Global Moran's I test. The result of the GSTA shows that the sediments in the south barriers (SBs) were transported to the southwest of the study area. The sediments in the north, in the SE direction of sediment transport trend from the river mouth, indicated that the sediments in the north of the study area were transported from the LR to the northern beaches, and to the south and east of the study area. The sediment transport trend that simplified by GSTA as the FB(-) case was approved by the content changes of sedimentary components (i.e. EM1, EM2, and EM3). In addition, the turbulent jet diffusion pattern indicated that the coarse sediments (EM3) were delivered by LR during the flood season, and the EM2 and EM1 were from wave and tide, respectively.
Keywords: Luanhe River delta    sediment transport    grain-size partitioning    geostatistical grain size trend analysis (GSTA)    
1 INTRODUCTION

Deltas are located in the intricate interfaces between marine and fluvial dynamics. Thus, the sediment transport, reworking and deposition processes in and near the deltaic regions are primarily driven by fluvial and marine hydrodynamic regimes (Bhattacharya and Giosan, 2003). There, the sediments, which are loaded from the downstream of the river, pass over the river mouth to replenish the distal side of the delta and finally cause deltaic progradation (Maillet et al., 2011). Coarse sediments (e.g. sand) transport is related closely to high-energy hydrodynamic regimes. For instance, fluvial force is the key driven factor for the transport of the bed-load particles, which is observable in the sediment transport processes of medium-scale river deltas (Dalrymple et al., 2003). Luanhe River (LR) is characterized by medium water discharge and high sediment flux with high sand concentration, due to the influence of monsoon climate. Thus, it provides an ideal study area for medium-scale deltaic sediment transport and deposition process (Feng and Zhang, 1998).

The Luanhe River Delta (LRD) is well known for its widespread barrier system (Wang et al., 2007). Liu (1989) and Duan et al. (2016) reported the spatial distribution characteristics of surficial sediments and divided the LRD into delta plain, delta front, and prodelta. Earlier studies (Xue et al., 2009; Chen et al., 2015) reported a dominant southwest-to-northeastoriented tidal current and seasonal variation of hydrodynamic regimes. Nevertheless, no focuses have been put on the characteristics of grain size decomposition and pattern of sediment transport in the LRD. Besides, some recent studies (Li and Feng, 2007; Chen et al., 2013; Kang et al., 2016) revealed that medium-scale rivers are seriously impacted by human activities because of their high sensitivities. Thus, the study of sediment transport processes in medium-scale river deltas not only owns academic contributions but also benefits for marine environmental sciences, marine economics, as well as marine managements and policies.

It has long been recognized that the grain size distributions of most hydraulic sediments are polymodal, and represent different transport and/or deposition processes (Tanner, 1964; Visher, 1969; Middleton, 1976; Ashley, 1978; Bagnold and Barndorff-Nielsen, 1980; Sun et al., 2002). Numerical partitioning of the sedimentary components provides an effective approach to separate specific sedimentary components (i.e. end members), which represent environmental information of sediment transport and/ or deposition processes (Sun et al., 2002). Such a method is even more efficacious and imperative in the coastal setting because the coast accepts several sources of sediment supply. With the identification of such components, we can easily analyze the sediment transport and deposition processes. This is because this method represents environmental information more directly by numbered end members (EMs) (Sun et al., 2008). Hence, numerical partitioning of the sedimentary components has been widely used to investigate sediment transport and deposition processes in loess (Sun et al., 2002), estuaries (Su et al., 2016), rivers (Sun et al., 2002), lakes (Sun et al., 2002).

Grain size trend analysis (GSTA) is an economic tool for sediment transport studies in open shallow marine environments (Poizot et al., 2008). Considerable studies have been widely practiced in different marine environments, e.g. deltas (Duc et al., 2007), estuaries (Li and Li, 2018), beaches (Liu et al., 2017), continental shelfs (Gao et al., 1994; Cheng et al., 2004), lagoons (Gao and Collins, 1992). GSTA is based on the net sediment transport trend that derived from three grain size parameters (i.e., mean size, sorting, and skewness) of surficial sediments (McLaren, 1981; Gao and Collins, 1992), which is also known as Gao and Collins method (1992). However, Asselman (1999) combined Gao and Collins (1992) method with the geostatistical approach, which is referred to as geostatistical GSTA method, endowed the Dcr with physical definition, and got better results.

Through using the GSTA method, the trends that we got are the sum trends of all the grain sizes of the bulk sample. However, one single bulk sample usually contains polymodal grain size distributions (McLaren et al., 2007), which is more obvious in the coastal areas as multiple sources of sediments are supplied. Hence, the sum trends only represent the dominant trend, and the other trends will be blotted. Therefore, we combined the geostatistical GSTA and numerical partitioning of the sedimentary components to define the pattern of sediment transport processes in the LRD and tried to clarify the contributions of different grain size to the sum trends. What is more, a detailed understanding of sediment transport processes will improve the accuracy of the end members result (Li and Li, 2018).

In this study, numerical partitioning of the sedimentary components was applied for the decomposition of grain size distributions in the LRD. The GSTA model together with the geostatistical approach was adopted to analyze the transport pattern of sediments. In addition, a combined approach based on the numerical partitioning of the sedimentary components and geostatistical GSTA was applied to elucidate the contributions of different grain sizes to the sum transport trend and facilitated geological explanation of grain size distributions.

2 REGIONAL SETTING 2.1 Luanhe River

The LRD (Fig. 1b), which is located adjacent to the Laoting and Changli County, was selected for this study due to the large sediment flux from the adjacent LR. The LR (Fig. 1a) meanders through the north of Hebei Province and originates on the northern piedmont of Bayanguer Mountain at elevations of above 1 600 m (Wang et al., 2007). After emerging from the west highlands, the braided channels of the LR flow southeast for approximately 1 200 km, traversing the Yanshan Range, Yanshan Foothills, LRCP (Luanhe River Coastal Plain) before terminating in a delta on the west coast of the Bohai Sea (Wang et al., 2007). The LR catchment (44 800 km2) is the largest on the Yanshan Range and is underlain by late Triassic to Cretaceous magmatite in the Yanshan Range, and Quaternary marine and non-marine deposits formed during a series of transgression in the LRCP (Li et al., 2008).

Fig.1 Location of the study area (a), the general introduction of the LRD and the location of surface sediment samples (b) The blue lines in Fig. 1a denote rivers flow into the Bohai Sea, the LR catchment is circled with black dotted line and, the solid black circle labeled QHD denote Qinhuangdao Oceanic Station. The white dotted lines in Fig. 2b denote the isobaths with water depth values (unit is meter) are labeled. Green dotted lines were labeled SBs (south barriers) and NBs (north barriers) in Fig. 1b are mouth barriers. The red solid circle is the position of flow velocity in the river mouth at 1980. The solid white line shows the position of the profile demonstrated in Fig. 10. The yellow line denotes coastline measured at 1985. The six subchannels (cyan dotted lines) in the Fig. 1b named as C-1, C-2, C-3, C-4, C-5, and C-6 were formed at 1915, 1939, 1958, 1952, 1959, and 1972, respectively.

The annual precipitation of the LR catchment reduces from 750 mm/a to 400 mm/a over the catchment in east-to-west trending (Feng and Zhang, 1998), and the average annual discharge is 4.56×109 m3. In mid-June to early September, the heavy rainfall, which caused by the monsoon climate, generates more than 70% of the water discharge. Moreover, this water discharge transports more than 90% of the sediment discharge of the LR in the wet season (Li and Feng, 2007) (Fig. 2a). Hence, the highly erodible lithology combined with the break-up flood pulse contributes large amounts of suspended sediments to the LR, resulting in a high annual sediment flux (20.1×106 t/a) to the coastal zone of the Bohai Sea (Li et al., 1983). It is noteworthy that both of the water discharge and sediment discharge significantly diverse between different years, respectively (Fig. 2b). The inter-annual variation of sediment discharge is even greater than that of the water discharge (Li and Feng, 2007).

Fig.2 Water and sediment discharges, tidal current and wave height The monthly average of water discharge (solid line) and sediment discharge (dotted line) (a), an annual average of water discharge (solid line) and sediment discharge (dotted line) (b) at Luanxian Station (from Li and Yin, 2010). The harmonized tidal current vectors within one tidal cycle for surface (c-1) and bottom (c-2) layer (from Xue et al., 2009). The wave direction-height distribution (d) at QHD (Qinhuangdao) Oceanic Station (Fig. 1a) between 1974 and 1984 (from Xue et al., 2009).
2.2 Morphology of LRD

The modern LRD is well known for its widespread barriers and sub-channels system in China (Wang et al., 2007) (Fig. 1b), and formed since 1915 when LR changed its course into Tianshuihe River (Fig. 1b). Since then, the river course diverted 7 times and formed 6 sub-channels (Fig. 1). However, only the major river channel is active during the dry season, and the others tend to be abandoned gradually. There are only a few of these abandoned sub-channels being active during flood season. Once these sub-channels were abandoned, no sediments would be transported into the delta. Then, the sediments in the river mouth would be transported landward and formed the fanshaped barriers (Wang et al., 2007). These barrier bars form the top part of the subaqueous lobe and occur as a shallow flat with a slope opposite to the flow of the LR (Liu, 1989). The delta front extends from the outer edge of the bars until 5 m isobaths in the south and 15 m isobaths in the north and is characterized by a steeply to gently slope from the south to north (Liu, 1989). However, the LRD is eroded seriously due to the sharply reduced sediment discharges as Qian (1994) reported. For instance, the north barriers (NBs) retreated more than 300 m during 1979–2007 (Li and Yin, 2010) (Fig. 1b). Therefore, the barriers are eroded seriously due to the heavy human activities (e.g. reservoir construction, sand mining, fishes farming). The shorelines retreat rate of the modern LRD since the late 20th century is estimated to be 15–30 m/y (Li and Yin, 2010).

2.3 Hydrodynamic conditions in the LRD

The influences of the river, wave, and tide are the major factors to classify or identify the deltas (Galloway, 1975). The nearshore of the LRD is characterized by a semidiurnal-mix semidiurnal tide averaging 0.73–0.91 m (Xue et al., 2009). Therefore, we classified the marine dynamic in the nearshore of the LRD as the microtidal environment. Moreover, Xue et al. (2009) surveyed the characters of tidal current in the LRD area around the 5m isobaths and 10m isobaths, respectively (Fig. 2c-1 and 2c-2). During both flood and ebb period, tidal current directions are confined to the SW (flood) and NE (ebb). The bottom tidal current velocity reached up to 50–75 cm/s during peak flood period and the net daily suspended sediment transportation is 32.4– 1 160.4 kg/m (Xue et al., 2009). The wave direction in the nearshore of LRD is dominated by the Asian monsoon climate. Therefore, the most frequent wave direction is southwest in late autumn to early winter. Then, it turns to be northeast in summer. The wave height in the nearshore of LRD is 0.5–1.5 m in fair weather (Feng and Zhang, 1998) (Fig. 2d). However, the maximum wind velocity generates wave height of 3.5 m in the SE direction by typhoon events (Xue et al., 2009).

3 MATERIAL AND METHOD 3.1 Samples grab and grain size analysis

The purpose of the sediment grab and grain size analysis was to investigate the spatial distribution and potential transport trend of the surficial sediments. Therefore, we can easily identify predominate factors that controlling the distribution of the surficial sediments. We collected 85 surficial sediment samples from the seafloor in the nearshore of LRD by a Van Veen sampler. Once the samples were collected, they were put into the sample sacks. Then, we carried the surficial sediment samples to the grain size laboratory of the First Institute of Oceanography, State Oceanic Administration for grain size measurement. Typical sample sizes of 2–5 g were retained for analysis. Then, we added 15 mL H2O2 (30%) and 5 mL HCL (10%) into each sample and kept stewing for 12 h to dissolve organic materials and carbonates. After that, each sample was centrifuged three times to wash the salts in the sediments. Then, each sediment sample was measured at least twice by Malvern Mastersizer 2000 laser particle size analyzer to improve the accuracy. This instrument has measurement ranges of 0.35–2 000 μm (0.25 Φ interval), thus giving 51 channels. Finally, we got a volume percentage of each portion of the 85 surficial sediment samples with mm units. Following standard practice, we converted sizes with mm units into Φ (phi) units using the equation Φ=-log2 (mm) for the purpose of computing the grain size parameters (Krumbein, 1938) i.e. mean grain size (Mz), sorting coefficient (So), skewness (Sk) and kurtosis (Ku) (McManus, 1988). The volume percent for each portion is centered at Φ/2.

3.2 Numerical partitioning of the sedimentary components

There are two essential tasks to do for numerical partitioning of the sedimentary components: 1) function formula definition; 2) estimations of function parameters (Sun et al., 2002). Sun et al. (2002) suggested adopting the Weibull distribution function for its high freedom than Possion, γ, F, c2 distribution functions. Hence, we chose the Weibull distribution function in this paper to determine the sedimentary components of the surficial sediments in the LRD. The formula of a general Weibull distribution function is as follows:

    (1)

Parameter a determines the shape, including the skewness and symmetry of the curve, while the parameter b controls the range of the curve. Parameter γ determines the position of the primary frequency of the curve, here referred to as the position of the modal grain size. Parameter x is a variable representing the grain size (Sun et al., 2002).

Once the function formula type is determined, it is easily to define the grain size function formula and its component based on the above strategy. It can be mathematically proven that the universal distribution functions can be calculated by the follows (Ashley, 1978):

    (2)

Parameter fi represents the function of component i with a total number n, and pi is the percentage in the sediment samples. There are n-1 coefficient and pi to be estimated due to the sum of percentages equaling unity. One to ten number of components, which are needed to best fit the data, can be generated in this approach (Paterson and Heslop, 2015). The parameters in the Eq.2 can be estimated by General Least Squares Fitting method (Chang et al., 2001) using the measured grain size data in 51 grain-size classes between 0.35–2 000 μm. Three statistical measurements, i.e. R2 (squared linear correlation between the measured data set and the date set constructed from the fitted modals), EM R2 (linear correlation between all fitted modals), and θ (the angular distance between the measured data set and the data set constructed from the fitted modals), are used to evaluate the fitness between the models and the measured data (Paterson and Heslop, 2015).

3.3 Geostatistical grain size trend analysis

Many previous studies (Pettijohn and Ridge, 1932; McLaren, 1981) have attempted to identify the grain size trend associated with the net sediment transport pathway. Both of the decreases (Pettijohn and Ridge, 1932) and increases (McCave, 1978; Nordstrom and McCluskey, 1985) between sediment grain-size distributions were recognized along the direction of transport. Most investigators preferred to use a combination of parameters (i.e., mean size, sorting, and skewness) to procedure sediment trend analysis as the uncertainties caused by the single parameter (McLaren, 1981; McLaren, 1985; Gao and Collins, 1992). Moreover, two methods (named 1-D and 2-D) were developed to proceed sediment transport trends.

McLaren (1981) developed an ideal model to define the sediment grain-size trend. In the following study, McLaren (1985) modified the previous conclusion and proposed an evolutionary hypothesis with two main scenarios. Taking d1 and d2 for examples, and d2 is the down-stream site of the transport direction between d1 and d2. Hence, sediment deposited at d2 is (1) finer, better sorted and more negatively skewed (noted FB(-)) than d1 or (2) coarser, better sorted and more positively skewed (noted CB(+)) than d1. This model has been practiced in many environments to improve coastal managements (Mclaren and Little, 1987; McLaren and Powys, 1993). However, some studies (Masselink, 1992; Gao et al., 2009) pointed out those practices using this model brought out interesting conclusions with controversial results in certain cases. Therefore, Van Lancker et al. (2004) suggested that this model would get better results in the significantly unitary environments (e.g. tidal inlet, tidal estuary) than in microtidal marine regions (Gao and Collins, 1992; Wu and Shen, 1999; Maillet et al., 2011). However, there are multiple sources in the LRD, and the dynamic regime is also complicated. Hence, the 1-D method was not used in this paper.

Gao and Collins (1991) proposed a 2-D method using a grid of sampling sites and derived trends by comparing the grain size parameters of each sample with their neighbors beyond the characteristic distance (noted Dcr) without directional limitation in the shallow marine environment. These trends are transformed into a "residual image model" by removing the "noise" from the original image containing both sediment transport and noise. Gao and Collins (1992) advocated FB(-) and CB(+) cases proposed by McLaren (1981), have a higher frequency of occurrence in the transport direction than other directions. This method is verified with good results in the studies of different open shallow marine environments (Mallet et al., 2000; Yang et al., 2004; Liu et al., 2017; Li and Li, 2018). Many of the study areas own multiple dynamics with different directions and sources of input (Li and Li, 2018), and the microtidal environment is one of them (Maillet et al., 2011).

Le Roux method (1994a) is also based on the 2-D approach. Considering a central station, four nearest sites are used to define transport vectors. Le Roux (1994a, Le Roux 1994b) assumed that the sediment grain-size transport vector direction for each sample is uniform with their directions of maximum gradient in grain size parameters. However, Asselman (1999) pointed out that the sediment grain-size transport trends are not always the same with the maximum gradient of it. Moreover, the maximum gradient direction of grain size parameters only represents the maximum direction of the hydrodynamic action model (Asselman, 1999). Hence, this method is also not used in the LRD.

Differences between distributions of surfical sediments grain size parameters and their derived statistical descriptors are spatially associated and contain process-related information (Zhang and Feng, 2011). Therefore, the geostatistical approach offers criteria for grain-size trend analysis in the spatial variation aspect. In considering this, Asselman (1999) advocated a spatial variation method to enhance the GSTA method proposed by Gao and Collins (1991). The optimization of sediment grid and choice of the sediment samples to consider are the major enhancements of the geostatistical GSTA method (Asselman, 1999; Flemming, 2007).

3.4 Regular grid generation

Gao (2009) reported that errors would occur if sample distances are not related to the sedimentary processes. When the sampling scale is too far, samples may be taken into different environments. Hence, the transport vectors between sampling stations are irrelevant. On the other hand, when the sampling scale is too close, "noises" proceeded by the GSTA may destroy the ordered patterns. Therefore, it is important to confine the characterized distance for GSTA (Gao and Collins, 1991, 1992; Asselman, 1999; Masselink et al., 2007; Poizot et al., 2008).

Matheron (1965) proposed the semi-variogram analysis into the regionalized variable theory to study the structural characteristic of a studied natural phenomenon. Then, Asselman (1999), Poizot et al. (2008) and Gao (2009) proposed to use this approach with the GSTA for two reasons. (1) It is useful to get a regular grid using Kriging interpolator. (2) The semi-variogram analysis can provide the best distance to define the trend vectors between sampling stations. Moreover, this method reduces the errors caused by the interval remodeling between irregular sampling stations.

However, the remodeling may also lead the grain size trend vectors to diverge from the original datasets without enough limitations, and hide the real results. To minimize these concerns, Poizot et al. (2008) proposed a protocol based on a five-step approach:

Step 1: generate a new regular grid from the original sampling scheme. Semi-variogram analysis was suggested to define the geostatistical model to interpolate the grain size parameters. Wilcoxon nonparameter test (in Matlab) was applied to evaluate the discrepancy between the original data set and interpolated values of different interpolation radiuses and grain size parameters.

Step 2: study the special variation of the grain size parameters on the new regular grid to define the geostatistical characteristic distance (Dg).

Step 3: choose the study trend (i.e., CB(+) and/or FB(-)).

Step 4: calculate the trend vector field following the guide of Gao and Collins (1991) approach via the interpolated dataset in step 1 and the Dg to replace Dcr.

Step 5: produce a Watson non-parameteric test to assess the autocorrelation of the different vector fields (i.e. CB(+) and FB(-) in this paper). The vector field of a particular trend case was then considered as reflecting a real sediment transport process. On the other hand, the vector field that failed in the Moran's I test indicates the direction of vector trend is random, therefore, would not reflect the proper sediment transport process. However, Watson non-parameteric test can't define and examine the spatial information. We proposed a Global Moran's I test to assess the spatial autocorrelation of the directions of the different vector fields in this study.

4 RESULT 4.1 Spatial distributions of sediment grain size parameters

The spatial distributions of the four grain size parameters (i.e. mean grain size, sorting, skewness, and kurtosis) were interpolated with a Kriging interpolation to produce a finer regular grid than the original irregular one. The Wilcoxon non-parameter test was performed to assess the differences between the original dataset and the kriging interpolated regular dataset (Table 1). Table 1 shows that the differences between the original and the interpolated dataset increases with the interpolation radius distance decreases. When the interpolation radius distance reaches up to 0.012 decimal degree, all the zero differences of three grain size parameters exceed the limitation. Therefore, we chose 0.015 decimal degree (DD) to interpolate the irregular dataset for the regular dataset, because the radius distances less than 0.015 decimal degree brought out unnecessary errors.

Table 1 Wilcoxon non-parameter test results performed for three grain size parameters (α=0.99)

The spatial distribution of mean grain size parameter in Fig. 3a illustrates that a medium sand zone is located outside the river mouth of the LR, and silt with seldom clay occupies the offshore with 10– 20 m isobaths. The sediments are moderately to well sorted in the sand zone and badly sorted in the silty zone (Fig. 3b). The spatial distribution of skewness is skewed negatively in the sandy zone and shows a very positively skewed trend in the seaward (Fig. 3c). The spatial distribution of kurtosis is a normal or near normal distribution in the sandy zone. However, most of the sediments in the silty zone present a sharp kurtosis (Fig. 3d). Moreover, all the four grain-size parameters present the similar spatial distribution trend.

Fig.3 The spatial distribution of sediment grain size parameters Mean grain size (a), sorting (b), skewness (c), kurtosis (d).
4.2 Spatial distributions of end members

Fitness quality of numerical partitioning of the sedimentary components in LRD (Fig. 4a, b) indicates that with the number of end members increases, the fitness quality (R2) increases. By comparing the R2 of different models, the minimum number of end members required to explain all the variances is three. Moreover, with the number of end members increases from three to five, the EM R2 increases from 0.038 to 0.105, which indicates that the correlation between end members increases. The fitness quality of three number is 89.24%, which is more than 85%. Only six sediment samples are not well represented by the three end members (R2 < 0.8), which means 92.94% of sediment samples have a correlation higher than 0.8. Hence, most of the sediment samples were well represented by those three end members (i.e., EM1, EM2, EM3).

Fig.4 itness quality of numerical partitioning of the sedimentary components at LRD Linear correlations (a) and angular deviation (b) of EMs. With the number of end members increases, the R2 value increases, the angular deviation decreases.

These components are illustrated by EM1, EM2, and EM3 from fine-grained to coarse-grained with medium grain size equals to 7.12 Φ, 2.37 Φ, 1.37 Φ, respectively (Fig. 5, Table 2). The EM1 component has a grain size range between -0.88 Φ and 11.36 Φ, with a maximum volume of 5.13%. The EM2 component has a grain size ranges between 0.62 Φ and 8.12 Φ, with a maximum volume of 16.1%. The EM3 component has a grain size range between 0.12 Φ and 6.37 Φ, with a maximum volume of 21.8%. EM1 to EM3 account for 32.38%, 32.37%, and 35.25% of the total deciphered variance. There are 69, 74, and 76 sediment samples which contain EM1, EM2, and EM3 components, respectively.

Table 2 Characteristics of three end numbers
Fig.5 Comparisons between three end members (a), four end members (b) and five end members (c) The EM1 component is almost the same as each other in the three figures. The percentage of EM2 end member increases with the particle range decreases from number three to five.

The spatial distribution of each end member abundance (Fig. 6) indicated that EM1 component (Fig. 6a) is mainly located at the seaward of the study area. The abundance is 40%–60%. Most of EM1 component is located in the 15–20 m isobaths. However, the EM1 component is also found in the west (5–15 m isobaths) of the study area (Fig. 6a). This observation of EM1component is similar to the work of Chen et al. (2015). EM2 component distributes in two zones: (1) along the barriers of LRD, and (2) between the 10–20 m isobaths. It is interesting that the abundance distribution of EM2 component (Fig. 6b) in the zone (>80%) along the barriers of LRD is much higher than the zone (30%– 60%) between 10–15 m isobaths. For EM3 component (Fig. 6c), nearshore of the river mouth between 5–12 m isobaths is the major spatial distribution area. Moreover, EM3 component abundance (50%–100%) is much higher than the other two EMs components.

Fig.6 Spatial distributions of end members a. EM1 component; b. EM2 component; c. EM3 component.
4.3 Geostatistical grain size trend analysis

Semi-variogram is a function of the distance between two sampling points along a particular direction (Maillet et al., 2011). It is imperative to have a global view of the semi-variogram values in all directions as the semi-variance changes in different directions at LRD. Therefore, we built the semivariogram map (Fig. 7), which is a preferential continuity direction produced by calculating semivariogram region by region, to check the anisotropy of the study area based on the interpolated dataset of mean grain size. It is well known that if the distance between samples increases, the semi-variogram value will rise until it reaches to the so called sill value. This is the maximum distance that samples are spatially correlated with each other. In this study, the sill is 0.09 decimal degree (Fig. 7d). This distance (0.09 decimal degree) is equal to six sample intervals (0.015 decimal degree), means that the mean grain size parameter is stationary over the study area. Further analysis indicates that the semi-variogram direction (55°) is the same as the tidal current direction (Fig. 2c). According to the objective of this paper, we focused on the GSTA of the whole study area. Hence, the 0.09 decimal degree distance was chosen to proceed the GSTA of the LRD.

Fig.7 Locations and semivariances of the irregular and regular grids Location of the original surfacial sediments (a) and the semi-variogram of the data from the original sample grid (b); location of the interpolated surfacial sediments (c) and the semi-variogram of the data from the interpolated sample grid (d). The geostatistical characteristic distance (Dg) was calculated with the interpolated data using GIS by the method suggested by Poizot et al. (2008).

According to the GSTA, two patterns of sediment transport trend vector field were performed for this study (CB(+) and FB(-)). Because each pattern is related to a specific environment, we plotted two sediment trends, denoted CB(+) (coarser, better sorted, more positive skewed) and FB(-) (finer, better sorted, more negative skewed). However, only the result based on FB(-) satisfied the Global Moran's I test (FB(-) case, 0.519; CB(+) case, 0.078) for its high spatial autocorrelation. Hence, only FB(-) was considered in this study (Fig. 8).

Fig.8 Sediment transport trend of LRD revealed by geostatistical GSTA (FB(-) case) The study area was divided into three zones (i.e. zone A, zone B and zone C) based on the vector trends. We also divided the zones into left, middle, and right parts for easy description of the sediment transport trends. Gao et al. (1994) suggested that vector direction has a significantly higher probability of occurrence in the direction of net transport than in any other directions. Furthermore, Poizot and Méar (2010) pointed out that vector length is a function of the confidence level associated with the transport in the corresponding vector direction.

Figure 8 shows that the sediment transport trends are seawards and/or landwards in the zone A, and parallel to the coast in the zone B. Moreover, zone C shows different trend direction with zone A and B. The vector directions of the zone C turn to NW (left) and SE (right). Further analysis indicates that the vector directions in the left zone A are N-to-S and NEto-SW. The high value of the vector length and similar directions indicate an isotropy of the transport trends in the left zone A. The vector directions in the right zone A are almost SW-to-NE and indicate a net sediment transport trend to the beach, which is located on the north of the LRD. The high value of the vector length indicates strong isotropy of the transport trends in the right zone A. Hence, the amount of sediments, which are replenished from the LRD to the beach, is considerable. Moreover, eight vectors are confined in the middle zone A, which the direction is NE-to-SW, indicates the river mouth feeding the LRD. However, the vector direction of zone B is isotropy, and the length of the vectors shows a decreasing trend in the left margin. What's more, the vector direction of the zone B is parallel to the tide current. The directions of vectors in zone C show an anisotropy. The vector direction in the left the zone C is SE-to-NW, which is a landward direction. The vector direction in the right zone C is NW-to-SE, which is parallel to the vector direction in the left zone C. However, the direction of vectors in the middle zone C is 45–90 degrees with the vector direction in the right part. Moreover, the length of the vectors in the left and middle zone C is longer than the right part. Hence, this indicates that the sediment transport trend to the SE is not strong.

5 DISCUSSION 5.1 Sediment transport pattern

The geostatistical GSTA determined the grain size trend vectors, which was deciphered in Fig. 8 to show clear trends of the sediment transport. Hence, we plotted the pattern of these trends on the map (Fig. 9). This is a more distinct way to simplify and highlight the sediment transport pattern.

Fig.9 Sediment transport pattern and the sedimentary components of typical samples

In the left zone A, the direction of sediment transport is SW from the SBs. This is coherent with the well-known erosion of barriers (Li and Yin, 2010). Therefore, we believe that the SBs is a sedimentary source in this study area. In the middle zone A, the sediment transport trend is SE direction from the river mouth. This sediment transport trend indicates the river feeding to the subaqueous delta (Li et al., 1985). The right zone B is another sedimentary source with sediment transport trend directions of NE (to left zone A), SW (to left zone B), SE (to right zone C), near S (to right zone C). In the right zone A, the sediment transport trends indicate that the sediments in the sedimentary source of right zone B are transported to the beaches, which is located at the north of the LRD. The sediment transport trend in the left zone B is more significant than the others are, and indicate a strong sediment transport pattern. In the right and middle zone C, the sediment transport trend directions are S and SE and indicate a sediment transport trend to the seawards. In the left zone C, the sediment transport trend direction is NW, which indicates a sediment feeding to the left zone B. The following section will provide interpretation of the sediment transport pattern by sedimentary components and geological interfaces.

Moreover, the length and direction of sediment transport trends in the LRD indicate that there are at least two sources in the LRD (Fig. 9). The first is the SBs (south barriers), which indicates a seaward direction to left zone B of the study area. The second source, which is located in the offshore of the river mouth, is in the right zone B of the study area. Further analysis indicates that the second source is the major source in the LRD. The sediments in the second source are transported to the right zone A (beaches), left zone B, right zone C of the study area. Beside these, there is a probable source where sediments in the left zone C are transported to the left zone B of the study area.

5.2 Interpretation of sediment transport pattern using end members

The SBs of LRD display a characteristic sediment signature (Fig. 9). The grain size distribution is medium (1–10 Φ), and is deciphered to EM2 (90.6%) and EM1 (9.4%) components, with a median grain size close to 3.3 Φ (Sample H25). However, the grain size distribution becomes wider (0–11 Φ) and is transformed to three end members (i.e., EM3, EM2, EM1) in the SW direction (Sample H10). Moreover, the percentage of EM2 decreases to 7.5% and the EM3 and EM1 components increase to 29.3% and 63.1%, respectively. Therefore, the sediment transport trend in the left zone A is caused mainly by the increase of EM1 and decrease of EM2. It is noteworthy that the percentage of EM3 component increases, and thus leads to a coarse trend. The increase is 29.3%, which is much smaller than the increase of fining trend decreases (70.7%). Hence, the FB(-) case reflects the sediment transport trend in the left zone A. For the sediment transport trend in the right zone A, taking Sample H14 and H91 for instance, decrease of EM3 component (5.5%) and increases of EM2 (1.5%) and EM1 (4.0%) components are the driving force of the sediment grain size trend in the right zone A.

The grain size distribution is narrow (0–4 Φ) in the right zone B (Sample H14) with only EM3 component, and the increase of EM1 component is almost three times than the increase of EM2 component in the NE direction. Further analysis indicates that the sediment transport trend of NE direction from right zone B is more significant than the others (e.g., SW, S, SE) (Table 3). The trend of SW direction is dominated by the increase of EM1 component (+63.1%) and a decrease of EM3 component (-70.7%). The changes in EM2 component percentage only contributed 7.5%. Hence, the sediment transport trend in the SW direction, which caused by the decrease of EM3 component and an increase of EM1 component, is stronger than the other directions. This conclusion also explains the result that the content of the EM3 component in the left zone B (Sample H10) is higher than the other area. The grain size distribution in the middle and left zone C (Sample H78) is wide (0.5– 11 Φ), and the sediment transport trend can be explained by the increases of EM1 and EM2 components and a decrease of EM3 component.

Table 3 Content changes of end members in typical samples from the right zone B (Sample 14) to the others
5.3 Geological interpretation of sediment transport pattern

As a mountainous river, LR carries a large number of coarse sediments into the Bohai Sea, in particular during the wet season (Feng and Zhang, 1998). Hence, the influence of the river cannot be neglected, especially during the flood period (Kang et al., 2016). The mean water depth of LR Mouth is 2 m, and the width of mouth section is about 400 m. The velocity of flow is about 1.0 m/s(Li et al., 1985) during the wet season, but increases to about 5.0 m/s (Li et al., 1985) during the flood period. We counted the Reynolds number of the LR Mouth (Pearce, 1966). The Reynolds number is 2.5×107 and much higher than 3 000, the lower limit of turbulent flow according to Pearce (1966). Hence, we plotted the turbulent jet diffusion mode of the LRD based on the turbulent flow pattern (Pearce, 1966) (Fig. 10).

Fig.10 The pattern of fully turbulent axial river mouth of the LRD The position of profile A-A' was marked in Fig. 1b.

According to the turbulent flow pattern, the length of the initial sector in LR is 2 080 m, and the width of the spraying jet, in the end, is 640 m (Fig. 10). The end of the initial sector reaches the 8-m isobaths and the scope is almost the same with the distribution of coarse sediments (Fig. 3a). Hence, we hold the point that the coarse sediments (EM3 component) are carried by the LR, and the fine sediments are washed away by strong wave and tide. As a result, only coarse sediments can be found. Besides, the terrain is steep between 0–5 m isobaths, whereas, the terrain slope becomes slight gently between 5–15 m isobaths. The coarse sediments can be transported to the 5–15 m isobaths during flooding period, because the strong spraying jet formed during flood period would deliver the coarse sediments further, in particular with the help of gravity or strong wave (Fig. 4c). The finer sediments (EM2 component) are carried by a wave and deposited surrounding the EM1 component sedimentary province (Fig. 4b). The fine sediments (EM1 component) are transported by tide from the south, deposited at the 10 m isobaths, and form a muddy argillaceous area (Fig. 4a). Hence, we believe that the EM3 component is carried by LR, deposits above 15 m isobaths, and form a sedimentary province in the northeast of the study area. Whereas, EM2 component is transported by a strong wave, and deposited around EM1 sedimentary province. EM1 component is transported by the tide. Sediment in the study area is delivered mainly by LR. The distribution of sediment is characterized with mountainous and seasonal river feature, which is closely related to wave and tide as reported by Liu (1989), Duan et al. (2016), Wang et al. (2007).

6 CONCLUSION

The spatial distribution of mean size parameter illustrated that a medium sand zone was located outside the river mouth in the LRD, and silt with seldom clay occupied the offshore with 10–20 m isobaths. Further analysis indicated that three grainsize end members (i.e. EM1, EM2 and EM3 components, in order of fine to coarse) were identified by numerical partitioning of the sedimentary components. The mean grain size of EM1 component was 7.12 Φ, which was equaled to fine silt. The mean grain size of EM2 and EM3 components were 2.37 Φ and 1.37 Φ, which belonged to fine sand and medium sand, respectively.

Kriging interpolation method was used to interpolate the grain size parameters for the regular grid. The interpolation radius was 0.015 decimal degree for its satisfaction in the Wilcoxon nonparameter test. We chose 0.09 decimal degree as the characteristic distance for GSTA by the semivariogram model using the geostatistical method. The FB(-) case was adopted in GSTA for its satisfaction in the Waston test. The pattern of sediment transport trend derived from the GSTA model indicated that the barriers and the right zone B were the main sedimentary sources of the study area. Further analysis indicated that the barrier sediments were transported to the SW. The right zone B sediments were transported to the north beaches, left zone B, middle, and right zone C.

A combined approach based on the numerical partitioning of the sedimentary components and geostatistical GSTA illustrated the sediment transport processes using content changes of end members. The influences of LR, wave, and tide on the sediment transport pattern were interpreted and discussed.

7 DATA AVAILABILITY STATEMENT

The datasets generated and analyzed during the current study are available from the corresponding author upon reasonable request.

8 ACKNOWLEDGEMENT

We thank the two journal reviewers for their constructive comments and suggestions.

References
Ashley G M. 1978. Interpretation of polymodal sediments. The Journal of Geology, 86(4): 411-421. DOI:10.1086/649710
Asselman N E M. 1999. Grain-size trends used to assess the effective discharge for floodplain sedimentation, River Waal, the Netherlands. Journal of Sedimentary Research, 69(1): 51-61. DOI:10.2110/jsr.69.51
Bagnold R A, Barndorff-Nielsen O. 1980. The pattern of natural size distributions. Sedimentology, 27(2): 199-207. DOI:10.1111/sed.1980.27.issue-2
Bhattacharya J P, Giosan L. 2003. Wave-influenced deltas:geomorphological implications for facies reconstruction. Sedimentology, 50(1): 187-210. DOI:10.1046/j.1365-3091.2003.00545.x
Chang C I, Du Q, Chiang S S, Heinz D C, Ginsberg I W. 2001.Unsupervised target subpixel detection in hyperspectral imagery. In: Proceedings Algorithms for Multispectral, Hyperspectral, and Ultraspectral Imagery VⅡ. SPE, Orlando, United States. 4381: 370-379.
Chen B, Yin P, Gao F, Liu J Q. 2015. Seasonal variation characteristics of hydrological environment factors and dynamic response relationship in Luanhe Estuary. Marine Environmental Science, 34(5): 729-735. (in Chinese with English abstract)
Chen G Q, Yi L, Chen S L, Huang H J, Liu Y X, Xu Y H, Cao J R. 2013. Partitioning of grain-size components of estuarine sediments and implications for sediment transport in southwestern Laizhou Bay, China. Chinese Journal of Oceanology and Limnology, 31(4): 895-906. DOI:10.1007/s00343-013-2304-y
Cheng P, Gao S, Bokuniewicz H. 2004. Net sediment transport patterns over the Bohai Strait based on grain size trend analysis. Estuarine, Coastal and Shelf Science, 60(2): 203-212. DOI:10.1016/j.ecss.2003.12.009
Dalrymple R W, Baker E K, Harris P T, Hughes M G. 2003.Sedi-mentology and stratigraphy of a tide-dominated, foreland-basin delta (Fly River, Papua New Guinea). In: Sidi H F, Nummedal D, Imbert P, Darman H, Posamentier H W eds. Tropical Deltas of Southeast Asia-Sedimentology, Stratigraphy, and Petroleum Geology. Society for Sedimentary Geology, Tulsa. p.147-173.
Duan X Y, Yin P, Liu J Q, Zhang D L, Cao K, Gao F, Chen X Y. 2016. Heavy metals and polycyclic aromatic hydrocarbons in surface sediments of Luan River estuary:distributions, sources and ecological risk assessments. China Environmental Science, 36(4): 1198-1206. (in Chinese with English abstract)
Duc D M, Nhuan M T, Van Ngoi C, Nghi T, Tien D M, van Weering T C E, van den Bergh G D. 2007. Sediment distribution and transport at the nearshore zone of the Red River delta, Northern Vietnam. Journal of Asian Earth Sciences, 29(4): 558-565. DOI:10.1016/j.jseaes.2006.03.007
Feng J L, Zhang W. 1998. The evolution of the modern Luanhe River delta, north China. Geomorphology, 25(3-4): 269-278. DOI:10.1016/S0169-555X(98)00066-X
Flemming B W. 2007. The influence of grain-size analysis methods and sediment mixing on curve shapes and textural parameters:implications for sediment trend analysis. Sedimentary Geology, 202(3): 425-435. DOI:10.1016/j.sedgeo.2007.03.018
Galloway W E. 1975. Process framework for describing the morphologic and stratigraphic evolution of deltaic depositional systems. In: Broussard M L ed. Deltas: Models for Exploration. Houston Geological Society, Houston, USA. p.87-98.
Gao S, Collins M B, Lanckneus J, De Moor G, Van Lancker V. 1994. Grain size trends associated with net sediment transport patterns:an example from the Belgian continental shelf. Marine Geology, 121(3-4): 171-185. DOI:10.1016/0025-3227(94)90029-9
Gao S, Collins M. 1992. Net sediment transport patterns inferred from grain-size trends, based upon definition of "transport vectors". Sedimentary Geology, 80(1-2): 47-60.
Gao S. 2009. Grain size trend analysis:principle and applicability. Acta Sedimentologica Sinica, 27(5): 826-836. (in Chinese with English abstract)
Gao S, Collins M, Mclaren P, Bowles D. 1991. A critique of the "mclaren method" for defining sediment transport paths; discussion and reply. Journal of Sedimentary Research, 61(1): 143-147. DOI:10.1306/D42676A9-2B26-11D7-8648000102C1865D
Kang X N, Yin P, Liu J Q. 2016. Variations in water and sediment discharges of mediam and small rivers and their response to human activities:a case study on the Luan River. Marine Geology & Quaternary Geology, 36(6): 1-6. (in Chinese with English abstract)
Krumbein W C. 1938. Size frequency distributions of sediments and the normal phi curve. Journal of Sedimentary Petrology, 8(3): 84-90.
Le Roux J P. 1994a. Net sediment transport patterns inferred from grain-size trends, based upon definition of "transport vectors"-comment. Sedimentary Geology, 90(1-2): 153-156. DOI:10.1016/0037-0738(94)90022-1
Le Roux J P. 1994b. An alternative approach to the identification of net sediment transport paths based on grain-size trends. Sedimentary Geology, 94(1-2): 97-107. DOI:10.1016/0037-0738(94)90149-X
Li C X, Chen G, Wang C G, Zhang Y M. 1984. On the Luanhe River alluvial fan-delta complex. Acta Petrolei Sinica, 5(4): 29-36. (in Chinese with English abstract)
Li C X, Chen G, Wang L. 1983. The abandoned deltas of the Luanhe River and the barrier-lagoon sedimentary systems. Acta Sedimentologica Sinica, 1(2): 60-72. (in Chinese with English abstract)
Li C X, Fan D D, Yang S Y, Cai J G. 2008. Characteristics and formation of the Late Quaternary incised-valley sequences in estuary and delta areas in China. Journal of Palaeogeography, 10(1): 87-97. (in Chinese with English abstract)
Li G, Yin Y. 2010. Recent geomorphological evolution of downstream channel and delta of Luanhe River. Geographical Research, 29(9): 1606-1615. (in Chinese with English abstract)
Li J Z, Feng P. 2007. Runoff variations in the Luanhe River Basin during 1956-2002. Journal of Geographical Sciences, 17(3): 339-350. DOI:10.1007/s11442-007-0339-8
Li T, Li T J. 2018. Sediment transport processes in the Pearl River Estuary as revealed by grain-size end-member modeling and sediment trend analysis. Geo-Marine Letters, 38(2): 167-178. DOI:10.1007/s00367-017-0518-2
Liu S H, Li P Y, Du J, Feng A P, Zhang Z W, Yu X X, Zhu Z T. 2017. Morphological variation and sediment dynamic in a cape-bay beach with a tidal channel around (Sanshan Daobeach, west coast of Laizhou Bay, China). Indian Journal of Geo-Marine Sciences, 46(5): 847-859.
Liu Z X. 1989. The influential factors and zoning of the modern Luanhe River Delta. Journal of Oceanography of Huanghai & Bohai Seas, 7(4): 55-64. (in Chinese with English abstract)
Maillet G M, Poizot E, Sabatier F, Vella C, Mear Y. 2011. Pattern of sediment transport in a microtidal river mouth using geostatistical sediment-trend analysis. Journal of Sedimentary Research, 81(1-2): 138-152.
Mallet C, Howa H L, Garlan T, Sottolichio A, Le Hir P. 2000. Residual transport model in correlation with sedimentary dynamics over an elongate tidal sandbar in the Gironde Estuary (Southwestern France). Journal of Sedimentary Research, 70(5): 1005-1016. DOI:10.1306/022900701005
Masselink G, Austin M J, O'Hare T J, Russell P E. 2007. Geometry and dynamics of wave ripples in the nearshore zone of a coarse sandy beach. Journal of Geophysical Research:Oceans, 112(C10): C10022. DOI:10.1029/2006JC003839
Masselink G. 1992. Longshore variation of grain size distribution along the coast of the Rhône Delta, Southern France:a test of the "McLaren Model". Journal of Coastal Research, 8(2): 286-291.
Matheron, G, 1965. Les variables régionalisées et leur estimation. Masson, Paris, France. 305 pp.
McCave I N. 1978. Sediments in the abyssal boundary layer. Oceanus, 21(1): 27-33.
McLaren P, Hill S H, Bowles D. 2007. Deriving transport pathways in a sediment trend analysis (STA). Sedimentary Geology, 202(3): 489-498. DOI:10.1016/j.sedgeo.2007.03.011
Mclaren P, Little D I. 1987. The effects of sediment transport on contaminant dispersal:an example from Milford Haven. Marine Pollution Bulletin, 18(11): 586-594. DOI:10.1016/0025-326X(87)90278-5
McLaren P. 1981. An interpretation of trends in grain size measures. Journal of Sedimentary Research, 51(2): 611-624.
McLaren P. 1985. Behaviour of diesel fuel on a high energy beach. Marine Pollution Bulletin, 16(5): 191-196. DOI:10.1016/0025-326X(85)90479-5
Mclaren P, Powys R I. 1993. Sediment pathways in a British Columbia fjord and their relationship with particleassociated contaminants. Journal of Coastal Research, 9(4): 1026-1043.
McManus J. 1988. Grain size determination and interpretation.In: Tuck M E ed. Techniques in Sedimentology. Blackwell Scientific Publications, Tucker. p.63-85.
Middleton G V. 1976. Hydraulic Interpretation of sand size distributions. The Journal of Geology, 84(4): 405-426. DOI:10.1086/628208
Nordstrom K F, McCluskey J M. 1985. The effects of houses and sand fences on the eolian sediment budget at Fire Island, New York. Journal of Coastal Research, 1(1): 39-46.
Paterson G A, Heslop D. 2015. New methods for unmixing sediment grain size data. Geochemistry, Geophysics, Geosystems, 16(12): 4494-4506. DOI:10.1002/2015GC006070
Pearce A F. 1966. Critical Reynolds Number for FullyDeveloped Turbulence in Circular Submerged Water Jets.National Mechanical Engineering Research Institute, Pretoria, South Africa.
Pettijohn F J, Ridge J D. 1932. A mineral variation series of beach sands from Cedar Point, Ohio. Journal of Sedimentary Petrology, 3(2): 92-94.
Poizot E, Méar Y, Biscara L. 2008. Sediment Trend Analysis through the variation of granulometric parameters:a review of theories and applications. Earth-Science Reviews, 86(1-4): 15-41. DOI:10.1016/j.earscirev.2007.07.004
Poizot E, Méar Y, Thomas, M ., Garnaud, S. 2006. The application of geostatistics in defining the characteristic distance for grain size trend analysis. Computers & Geosciences, 32(3): 360-370.
Poizot E, Méar Y. 2010. Using a gis to enhance grain size trend analysis. Environmental Modelling & Software, 25(4): 513-525.
Qian C L. 1994. Effects of the water conservancy projects in the Luanhe River Basin on Luanhe River Delta, Hebei Province. Acta Geographica Sinica, 49(2): 158-166.
Su Q, Peng C S, Yi L, Huang H J, Liu Y X, Xu X Y, Chen G Q, Yu H J. 2016. An improved method of sediment grain size trend analysis in the Xiaoqinghe Estuary, southwestern Laizhou Bay, China. Environmental Earth Sciences, 75(16): 1185. DOI:10.1007/s12665-016-5924-7
Sun D H, Bloemendal J, Rea D K, Vandenberghe J, Jiang F C, An Z S, Su R X. 2002. Grain-size distribution function of polymodal sediments in hydraulic and aeolian environments, and numerical partitioning of the sedimentary components. Sedimentary Geology, 152(3-4): 263-277. DOI:10.1016/S0037-0738(02)00082-9
Sun D H, Su R X, Bloemendal J, Lu H Y. 2008. Grain-size and accumulation rate records from Late Cenozoic aeolian sequences in northern China:implications for variations in the East Asian winter monsoon and westerly atmospheric circulation. Palaeogeography, Palaeoclimatology, Palaeoecology, 264(1-2): 39-53. DOI:10.1016/j.palaeo.2008.03.011
Tanner W F. 1964. Modification of sediment distributions. Journal of Sedimentary Research, 34(1): 156-164.
Van Lancker V, Lanckneus J, Hearn S, Hoekstra P, Levoy F, Miles J, Moerkerke G, Monfort O, Whitehouse R. 2004. Coastal and nearshore morphology, bedforms and sediment transport pathways at Teignmouth (UK). Continental Shelf Research, 24(11): 1171-1202. DOI:10.1016/j.csr.2004.03.003
Visher G S. 1969. Grain size distributions and depositional processes. Journal of Sedimentary Petrology, 39(3): 1074-1106.
Wang Y, Fu G H, Zhang Y Z. 2007. River-sea interactive sedimentation and plain morphological evolution. Quaternary Sciences, 27(5): 674-689. (in Chinese with English abstract)
Wu J X, Shen H T. 1999. Estuarine bottom sediment transport based on the 'McLaren Model':a case study of Huangmaohai Estuary, South China. Estuarine, Coastal and Shelf Science, 49(2): 265-279. DOI:10.1006/ecss.1999.0495
Xue Z, Feng A P, Yin P, Xia D X. 2009. Coastal erosion induced by human activities:a northwest Bohai Sea case study. Journal of Coastal Research, 25(3): 723-733.
Yang W, Mazzullo S J, Teal C S. 2004. Sediments, facies tracts, and variations in sedimentation rates of Holocene platform carbonate sediments and associated deposits, Northern Belize-Implications for "Representative" sedimentation rates. Journal of Sedimentary Research, 74(4): 498-512. DOI:10.1306/012004740498
Zhang C Y, Feng X L. 2011. Natural and human-induced effects on grain size of surface sediments along the Lianyungang muddy coast, China. Chinese Journal of Oceanology and Limnology, 29(2): 387-397. DOI:10.1007/s00343-011-0007-9