The heterogeneity of vertical hydraulic conductivity (Kv) is a key attribute of streambed for researchers investigating surface water–groundwater interaction. However, few three-dimensional (3-D) Kv models with high spatial resolutions have been achieved. In this study, in-situ permeameter tests were conducted to obtain Kv values. A 3-D model with 443 Kv values was built comprising 10 lines, 10 rows, and five layers. Statistical analysis was done to reveal the spatial characteristics of Kv. The influence of bedform on Kv values was restricted to the near-surface streambed. Kv increased with the increasing distance from the south river bank for the upmost layer, but it was not the case for other layers and the combined Kv values of five layers; the spatial pattern at transects across the channel did not differ significantly. The Kv values of each layer pertained to different populations; the sediments of individual layers were formed under different sedimentation environments. The coupling of erosion/deposition process and transport of fine materials primarily contributed to a reduction of the mean and median of Kv values and an increase of heterogeneity of Kv values with depth. Thus, a collection of Kv values obtained from different layers should be considered when characterizing the heterogeneity of streambed.
INTRODUCTION
The heterogeneity of streambed vertical hydraulic conductivity (Kv) is a major controlling factor in water and solute fluxes through the streambed–aquifer interface (Cardenas & Zlotnik 2003; Salehin et al. 2004; Ryan & Boufadel 2007). Overexploitation of groundwater in northern China, especially high-Kv areas, resulted in base-flow reduction and stream infiltration, and the degradation of stream ecosystem (Chen & Shu 2002). Therefore, the concern about induced stream depletion urgently requires a more detailed picture illustrating the heterogeneity of streambed. Especially, the spatial variability of vertical hydraulic conductivity and bedform is an important aspect of streambed heterogeneity.
Spatial heterogeneity of streambed hydraulic conductivity has been widely studied by many researchers. Chen (2005) and Genereux et al. (2008) reported the significant spatial variation of streambed Kv in their individual study reaches. The spatial variability of hydraulic conductivity under different topographic features has also been frequently surveyed (Genereux et al. 2008; Sebok et al. 2015). To quantitatively characterize the heterogeneity of streambed, the probability distribution was applied to describe the spatial variability of streambed hydraulic conductivity (Chen 2005; Genereux et al. 2008; Cheng et al. 2011). As is well known, the influence of multiple layered sediments on hyporheic exchange is significant (Landon et al. 2001). However, studies about the variability of streambed hydraulic conductivity with depth have only been mentioned by a limited number of researchers (Ryan & Boufadel 2007). Song et al. (2007) investigated the variations of streambed Kv with two connected depths in three rivers of Nebraska. Chen (2011) even produced Kv vertical profiles of the channel sediments to a depth of 15 m below. However, those two studies only dealt with the one-dimensional issue regarding spatial variability of streambed Kv. Leek et al. (2009) compared the spatial patterns of horizontal hydraulic conductivity (Kh) at 0.3–0.45, 0.6–0.75, 0.9–1.05, and 1.2–1.35 m depth intervals in two selected sites in Touchet River. Ryan & Boufadel (2007) introduced the probability distribution of Kh values of two depth intervals of 7.5–10 cm and 10–12.5 cm in an 80 m reach of Indian Creek in Philadelphia. However, the variability of spatial pattern in streambed K with varied depths was beyond a full understanding.
Uncertainties have existed in Kv estimation, which are partially derived from the layout of measurement locations. Leek et al. (2009) and Chen et al. (2014) underlined the huge influence of the distribution of test locations on probability distribution of Kv. Kennedy et al. (2008) and Genereux et al. (2008) also emphasized the role of sampling density in estimation of streambed attributes. Landon et al. (2001) even found that the method used may matter less than making sufficient measurements to characterize spatial variability adequately. To diminish the uncertainties in estimation of the Kv as much as possible, more three-dimensional (3-D) measurements carried out in a streambed are essential. The objectives of this study are to build a 3-D Kv model and to reveal the spatial characteristics within and between individual layers. The Kv values of stream channel were obtained from permeameter tests. To achieve high-resolution characterization of Kv, a total of 443 measurements in a cubic space of 8.5 m in length, 11 m in width, 0.5 m in depth were logged. Statistical analysis was used to reveal the spatial heterogeneity of streambed Kv.
STUDY AREA AND TEST SITE
Map showing the study area (a) and schematic of the layered streambed where permeameter tests were conducted (b). The term ‘hyporheic zone’ is used to refer to the zone beneath and adjacent to a river or stream in which groundwater and surface water mix (Environment Agency 2005).
Map showing the study area (a) and schematic of the layered streambed where permeameter tests were conducted (b). The term ‘hyporheic zone’ is used to refer to the zone beneath and adjacent to a river or stream in which groundwater and surface water mix (Environment Agency 2005).
Field tests were performed in a reach with a straight longitudinal plan-form. The field investigation was conducted on 5–7 June 2012. Stream discharge and water level were relatively constant in the previous several months. Spatially continuous silt/clay layers were not observed in the shallow streambed. Grain-size analysis shows that channel sediments tested consisted largely of sand and gravel, with a very small amount of silt and clay. The silt and clay (particle diameter %3C0.075 mm) accounted only for about 2.6% of the sediments by weight. More than 70% of the weight percentages of the sediment fell into the category of sand with particle diameter between 0.075 and 2 mm. During the investigation period, the averaged width of the submerged channel was 21 m and the averaged water depth measured at test sites was 0.71 m. Amounts of invertebrates and small bubbles produced by microbes were observed in sediments of streambed. Water temperature was measured using temperature sensors when field tests were being performed and ranged from 19.87 to 28.50 °C with an average temperature of 25.23 °C. Although temperature may have effects on the estimation of Kv values, the influence of diurnal oscillation of water temperature was neglected in this paper. The groundwater was dominantly fed by stream water within the tested channel during the period of field testing (Zhu et al. 2013). A total of 100 test locations comprising 10 columns and 10 rows at a horizontal grid coordinate were determined covering the half part of the channel. In terms of spatial correlation range in Kv (just under 0.5 m) and complicated bed form, the grid spacing was determined to be 0.5 or 1 m along the longitudinal channel and reached a maximum of 2 m except for 0.5 or 1 m along the traverse section. In consideration of the small thickness (just under 1 m) of this hyporheic zone (Zhu et al. 2013), at each grid point, permeameter tests were carried out at five depths: 0.1, 0.2, 0.3, 0.4, and 0.5 m below the streambed. Consequently, a 3-D Kv model consisting of a total of 443 values was built up (Figure 1(b)).
METHODS
In-situ permeameter test
To estimate Kv of the streambed, in-situ permeameter test using the falling head method was used in this study. This method has many advantages, such as plug and play, low cost, short duration, and determination of hydraulic conductivity along any direction compared to other methods, such as grain-size analysis (Landon et al. 2001), slug and bail tests (Leek et al. 2009), and pumping test (Kelly & Murdoch 2003). The principle of the permeameter test has been introduced by a number of papers (Chen 2000, 2005; Genereux et al. 2008; Cheng et al. 2011). The main procedure includes three steps: step one, inserting a standpipe into the unconsolidated sediments; step two, pouring river water to fill the pipe; step three, timing the elapsed decline of hydraulic head using a stopwatch as soon as water inside the pipe falls because of hydraulic head differences. Note that it is assumed that hydraulic head at the bottom of the sediment column is approximately equal to ambient water level that is considered to be constant during testing. The tube of permeameter is 150 cm in length and 4.5 cm in inner diameter. The wall of the tube is about 3.5 mm thick. Thus, the disturbance of the tube on surrounding sediments can be neglected when the tube is being pressed into the streambed. Water heads during each permeameter testing were recorded at least six times and the test was relatively quick (<20 min). Kv can be calculated using any pair of water head readings at a given time interval. To achieve Kv values of streambed at different depths, permeameter tests were conducted at the depth intervals of 0–10 cm, 0–20 cm, 0–30 cm, 0–40 cm, and 0–50 cm, respectively. After the permeameter testing at the depth interval of 0–10 cm was done, the tube was then pressed vertically into the sediments to a depth of 20 cm. The permeameter test was again conducted at the depth interval of 0–20 cm. The same procedure was repeated for the other three depth intervals of 0–30, 0–40, and 0–50 cm.

On the basis of sedimentary structure and composition of the sediment columns tested, 1.2 can relatively be an appropriate value for m in the study area (Mutiti & Levy 2010; Lu et al. 2012). A non-linear regression method was applied to eliminate measurement noise and enhance the precision in the computation of Kv by taking into account all the head readings simultaneously (Chen 2005).
Calculation of Kv for five sediment layers
Profile of the permeameter test at varied depths of the streambed sediments.
Statistical analysis
One-sample Kolmogorov–Smirnov (K-S) test was used to determine whether Kv values of each sediment layer were distributed normally or lognormally. The analysis was performed using SPSS v18.0 that returned a p value. If p < 0.05, it rejects the null hypothesis that the Kv values did not belong to the normal or lognormal distribution. If p > 0.05, it accepts the null hypothesis that Kv values belonged to the normal or lognormal distribution.
The non-parametric Kruskal–Wallis test and Wilcoxon rank-sum test were conducted at a significance level of 0.05. Kruskal–Wallis test is used to compare two independent samples that may have different sample sizes and to determine whether they originate from the same population; whereas Wilcoxon rank-sum test is applied to the determination of the similarity for matched pairs that are relevant. They have the same hypothesis. The null hypothesis is that the Kv values for the two layers are drawn from the same population. The hypothesis will be rejected if p < 0.05, or else if p > 0.05, the null hypothesis will be accepted.
During permeameter tests, water depth at each test location was measured to determine the correlation with streambed Kv. Correlation coefficient produced using SPSS v18.0 was considered as a representative factor to reflect how closely streambed Kv was associated with water depth, implicitly showing the influence of bedform on Kv of streambed.
RESULTS AND DISCUSSION
Statistics of Kv
Box plots of Kv values from individual layers (layers 1–5) and combined layers (layers 1–5) during the investigation period. □ indicates the mean, - the 1st or 99th percentile, × the minimum or maximum, and ○ the outliers. Note that ▪ stands for coefficient of variation of each sediment depth interval.
Box plots of Kv values from individual layers (layers 1–5) and combined layers (layers 1–5) during the investigation period. □ indicates the mean, - the 1st or 99th percentile, × the minimum or maximum, and ○ the outliers. Note that ▪ stands for coefficient of variation of each sediment depth interval.
Spatial variability of Kv in horizontal dimension
Interpolated map of Kv values from individual layers (layers 1–5). The arrow points downstream. The black lines in layer 1 are grid lines along the cross section.
Interpolated map of Kv values from individual layers (layers 1–5). The arrow points downstream. The black lines in layer 1 are grid lines along the cross section.
Bivariate plots showing the relationship between water depth and Kv values from individual layers (layers 1–5). R indicates the coefficient of correlation and 1 Pearson correlation coefficient; ** and * represent the significance level of 0.01 and 0.05, respectively.
Bivariate plots showing the relationship between water depth and Kv values from individual layers (layers 1–5). R indicates the coefficient of correlation and 1 Pearson correlation coefficient; ** and * represent the significance level of 0.01 and 0.05, respectively.
Boxplots of combined Kv values of layers 1–5 with the distance from the south bank.
Boxplots of combined Kv values of layers 1–5 with the distance from the south bank.
Broken line chart of Kv values for individual lines from the upstream (8.5’–8.5’) to the downstream (0’–0’) (as shown in Figure 4) across the channel for the first layer.
Broken line chart of Kv values for individual lines from the upstream (8.5’–8.5’) to the downstream (0’–0’) (as shown in Figure 4) across the channel for the first layer.
Spatial variability of Kv in vertical dimension
Coefficients of variation were given to determine the variability of Kv for each depth interval (Figure 3). The mean values for layers 1 and 2 were almost equal to the median values, but those for layers 3–5 exceeded the corresponding median values; the mean for layer 5 approached near the 75th percentile indicating large variability in Kv and the statistical effect of the few largest values. There was an increasing trend in heterogeneity of Kv with depth (Figure 3). Ryan & Boufadel (2007) investigated the heterogeneity of Kh from two layers and also found the increasing trend with depth.
P values from one-sample Kolmogorov–Smirnov test of Kv and lnKv from individual layers (layers 1–5)
. | Layer 1 . | Layer 2 . | Layer 3 . | Layer 4 . | Layer 5 . |
---|---|---|---|---|---|
Kv | 0.7* | 0.9* | 0.1* | 0.2* | 0.003 |
lnKv | 0 | 0.003 | 0.001 | 0.025 | 0.5* |
. | Layer 1 . | Layer 2 . | Layer 3 . | Layer 4 . | Layer 5 . |
---|---|---|---|---|---|
Kv | 0.7* | 0.9* | 0.1* | 0.2* | 0.003 |
lnKv | 0 | 0.003 | 0.001 | 0.025 | 0.5* |
*For the significance level of 0.05.
P values from the Kruskal–Wallis test and Wilcoxon rank-sum test of Kv values between any pair of layers
. | . | 1 vs. 2 . | 1 vs. 3 . | 1 vs. 4 . | 1 vs. 5 . | 2 vs. 3 . | 2 vs. 4 . | 2 vs. 5 . | 3 vs. 4 . | 3 vs. 5 . | 4 vs. 5 . |
---|---|---|---|---|---|---|---|---|---|---|---|
P-value | Kruskal–Wallis | 0.01 | <0.001 | <0.001 | <0.001 | 0.02 | <0.001 | <0.001 | 0.2* | 0.001 | 0.01 |
Wilcoxon rank-sum | 0.004 | <0.001 | <0.001 | <0.001 | 0.001 | <0.001 | <0.001 | 0.006 | <0.001 | 0.002 |
. | . | 1 vs. 2 . | 1 vs. 3 . | 1 vs. 4 . | 1 vs. 5 . | 2 vs. 3 . | 2 vs. 4 . | 2 vs. 5 . | 3 vs. 4 . | 3 vs. 5 . | 4 vs. 5 . |
---|---|---|---|---|---|---|---|---|---|---|---|
P-value | Kruskal–Wallis | 0.01 | <0.001 | <0.001 | <0.001 | 0.02 | <0.001 | <0.001 | 0.2* | 0.001 | 0.01 |
Wilcoxon rank-sum | 0.004 | <0.001 | <0.001 | <0.001 | 0.001 | <0.001 | <0.001 | 0.006 | <0.001 | 0.002 |
*The similarity is statistically significant (P > 0.05). The number i (i = 1,…,5) represents the ith layer.
Probability plots of Kv and lnKv from individual layers (layers 1–5).
Empirical cumulative distribution of Kv values from individual layers (layers 1–5).
Empirical cumulative distribution of Kv values from individual layers (layers 1–5).
SUMMARY AND CONCLUSIONS
The study site was located in a reach of the Dawen River, China. In-situ falling head standpipe tests were carried out in channel for the achievement of 3-D Kv model comprising 10 lines × 10 rows × 5 layers. The correlation analysis of Kv values for individual layers with water depth was performed. The statistical analysis of Kv values obtained was made to reveal the spatial distribution of streambed Kv under horizontal and vertical dimensions, simultaneously. The following conclusions were drawn:
The sediments primarily consisted of sand/gravel, with a small amount of silt/clay. No colmation occurred at the surface of the streambed. As a tributary of the Yellow River, the Dawen River has distinctly different streambed sediment Kv values up to two orders of magnitude greater than the lower reach of the Yellow River. This study reach had a high degree of spatial variability.
The correlation between Kv values and water depth was significantly positive for layer 1 whereas that was not the case for layers 2–5. Thus, the influence of bedform on Kv values was restricted to the near-surface streambed (0–10 cm depth interval) although stream discharge and water level remained smooth in the last few months. For layer 1, spatial patterns in Kv at transects across the channel remained almost unchanged within a distance of about 10 m along the channel; Kv had an increasing trend with the distance from the river bank. As well, the variation of Kv was larger at the midstream than the bank sides.
Mean and median Kv values decreased with depth in the shallow streambed; on the contrary, heterogeneity of Kv tended to increase with depth. The intrusion of fine materials contributed to larger heterogeneity in Kv for the lower layer. The histograms, P-P plots, and K-S test all suggest that the Kv values of each layer belonged to the normal distribution for layers 1–4 and those for layer 5 tended to be distributed lognormally. When the Kv values of layers 1–4 were combined as a single data set, the assembled data set did not belong to the normal/lognormal distribution. However, the combined data set from which the values less than 1.0 m/d were removed belonged to the normal distribution. Therefore, fine-grained materials carried by hyporheic exchange strongly affected the spatial distribution of Kv with depth. In addition, similarity analysis suggests that the Kv values of individual layers came from different populations; the sediments of each layer were formed under different sedimentation environments. This phenomenon was primarily attributed to the coupling of erosion/deposition process and transport of fine materials. It is noted that, to better capture spatial heterogeneity of streambeds and hyporheic exchange, a collection of Kv values should depend on measurement points at different depths.
ACKNOWLEDGEMENTS
This research was supported by the National Natural Science Foundation of China (41201029), Specialized Research Fund for the Doctoral Program of Higher Education of China (20120094120019), Fundamental Research Funds for the Central Universities of China (2012B00314), and China Postdoctoral Science Foundation (2013M540410).