Evolution of hydraulic conductivity in the floodplain of a meandering river due to hyporheic transport of fine materials



[1] The hydraulic conductivity of riverbeds and shallow aquifers controls hyporheic flow and surface water-groundwater exchange, which is critical to aquatic systems. We show that the hydraulic conductivity of shallow alluvial aquifers subject to sinuosity-driven hyporheic flow is dynamic because of transport of fine materials. We analyze changes in hydraulic conductivity over one year at a meander-scale experimental river-aquifer system using water table monitoring, repeat in-situ hydraulic tests, grain size analysis, and flow modeling with particle tracking. Areas with relatively high initial hydraulic conductivity became more permeable after one year, while areas with lower initial hydraulic conductivity became less permeable. Particle tracking suggests fine materials were flushed from the more permeable portions of the aquifer and deposited in the less permeable areas down-gradient. Since hydraulic conductivity evolution has feedback on flow, static or synoptic hydraulic conductivity measurements in river-hyporheic-riparian systems may need to be revisited. Temporally changing hydraulic conductivity has the capacity to impact rates of ecological and biogeochemical processes.

1. Introduction

[2] The interaction between groundwater and surface water in the hyporheic zone significantly influences nutrient and thermal regimes in riparian environments, which creates an ecotone with unique ecological conditions [Brunke and Gonser, 1997]. Numerous processes have been shown to induce hyporheic flow, and much previous work has focused on processes related to the hyporheic flow in the vertical dimension. In particular, stream water flowing over bedforms such as ripples generates pressure gradients at the sediment-water interface that induce flushing of stream water in and out of the shallow subsurface along relatively short flow paths [Thibodeaux and Boyle, 1987]. On a larger scale, stream flow through a pool-riffle-pool sequence can also induce subsurface flow [Harvey and Bencala, 1993].

[3] Though previous work has focused on vertical hyporheic processes, there is a growing understanding that the hyporheic zone extends laterally to include stream meanders/point bars. Several recent studies have modeled hyporheic flow through meanders using stream surface elevation along a meander as a steady boundary condition [Boano et al., 2006; Cardenas, 2008, 2009a, 2009b]. These studies suggest that the drop in stream elevation around a meander creates a hydraulic gradient through the point bar that induces hyporheic flow. Modeling has also suggested that hydraulic gradients due to regional effects can significantly affect the extent of the intra-meander hyporheic zone depending on their direction and magnitude [Cardenas, 2009b]. Like the hyporheic zone beneath streambeds, it is thought that point bars provide conditions for chemical transformation of water, for instance due to reducing conditions [Boano et al., 2010].

[4] While modeling studies provide valuable insight into intra-meander processes, there remains a need to study natural systems to verify that the models are realistic in light of known and unknown natural complexity. In particular, hydraulic conductivity in point bars is commonly heterogeneous and influences spatial patterns of groundwater flow. It is unclear, however, if and how hydraulic conductivity in point bars evolves with time. Previous work has identified several processes that influence hydraulic conductivity in streambeds that may also be applicable to point bars and riparian areas. For instance, clogging caused by the transport and deposition of fine particles has been shown to induce both spatial and temporal variations in hydraulic conductivity. Settling particles can clog hyporheic interstices in the uppermost few centimeters of the streambed and reduce hydraulic conductivity near the bed surface, below which resuspension and deposition of fine particles may also occur [Brunke, 1999; Brunke and Gonser, 1997; Chen et al., 2008, 2009a, 2009b; Packman and MacKay, 2003]. Alternately, hyporheic flow, biological activity, and gas bubble nucleation in the subsurface is thought to increase vertical hydraulic conductivity near the bed surface, but not at depth [Song et al., 2007].

[5] Here, we examine whether these or other processes cause point bar aquifer hydraulic conductivity to change in natural systems. The results from this study help elucidate the complexities of intra-meander flow and transport, allowing for better understanding of its significance.

2. Study Site

[6] The field data for this study were collected from a meander-scale experimental stream-aquifer system at the Saint Anthony Falls Laboratory at the University of Minnesota. Called the Outdoor StreamLab (OSL), the system mimics a meandering low order stream and measures approximately 25 × 40 m (Figure 1). The OSL was designed and developed by the National Center for Earth-Surface Dynamics and is one of the largest facilities in the world designed to closely replicate a natural river-floodplain environment.

Figure 1.

View of the Outdoor StreamLab (OSL) at the Saint Anthony Falls Laboratory looking in the downstream direction. The OSL is a field scale stream-aquifer system designed to study riparian processes and measures approximately 25 × 40 × 2 m with a channel width of about 2 m.

[7] The OSL is constructed on a sloping dam spillway along the Mississippi River. Limestone makes up the base of the facility and stone walls contain it. Interlocking plastic sheet piles were installed inside of the walls to help mitigate groundwater leakage. The aquifer substrate is composed of sandy sediment excavated from a nearby construction and has a thickness of ∼1.5–2 m. The facility's stream has a controlled geometry with an average width of about 2 m. The stream completes one full meander and has riffles both upstream and downstream of the meander tip. Discharge is precisely controlled by an inlet valve on the upstream end of the facility and monitored with a gauging station.

[8] To access groundwater in the OSL, a nested piezometer network was installed in the summer of 2008. A total of 59 nests were installed, each with three piezometers: one screened through the water table, one screened from ∼30 cm through 60 cm below the water table, and one screened from about ∼100 through 130 cm below the water table. These are hereafter referred to as “water table”, “medium” and “deep” piezometers. The piezometers used were composed of PVC, with an inner diameter of 3.2 cm, outer diameter of 4.2 cm, and 0.1524 mm screen slot size. 41 out of the 59 nests were installed within the point bar along hypothesized hyporheic flow paths to give dense coverage. The top of each piezometer was surveyed to ∼1 mm accuracy. Stream surface elevation was also surveyed. A repeat survey of the piezometers was completed in the summer of 2009 to account for any shifting in the piezometers' locations due to settling or frost heaving.

3. Methods

3.1. Water Table Mapping

[9] Water table elevations were measured several times from all piezometers from August 2008 through spring 2010 using a Solinst electronic water level meter accurate to 1 mm and converted to elevation above sea level. These data, along with the surveyed stream surface elevations during normal discharge, were used to create potentiometric surface maps in the mapping and interpolation program Surfer. In Surfer, data were gridded using kriging interpolation with an isotropic linear variogram.

3.2. Hydraulic Conductivity Estimation and Grain Size Analysis

[10] Aquifer hydraulic conductivity (K) around medium and deep piezometers was determined using a pneumatic slug testing system following Cardenas and Zlotnik [2003]. Slug tests in 37 deep piezometers within the point bar were conducted in the summers of 2008 and 2009. In 2008, three tests were conducted at each well using different levels of pressurization to determine precision and repeatability. The repeat tests from 2008 showed very little variation; therefore, only one or two tests were conducted for the wells tested in 2009. The recovery curves were analyzed using the Bouwer and Rice method for unconfined aquifers as implemented in the AQTESOLV Pro aquifer testing software. Slug test K results were interpolated using kriging with an isotropic linear variogram. Percent change in K from 2008 to 2009 was also calculated and interpolated in the same way.

[11] Sediment push cores from the top meter were collected from the same approximate location in three places within the point bar in both 2008 and 2009, making it possible to track evolution of grain size within the aquifer. Gravimetric grain size analysis was conducted following standard sieving methods. Sieve sizes ranged from 0.044 to 31.4 mm.

3.3. Groundwater Modeling and Particle Tracking

[12] Steady-state finite-difference groundwater flow simulations were conducted using MODFLOW 2005 as implemented in Visual MODFLOW 4.3. Backward particle tracking was completed using MODPATH, a three dimensional post-processing package that works with MODFLOW. The numerical model consists of a 24 × 40 × 2 m domain representing the OSL. Rectangular elements were used, with a grid spacing of 0.25 m in the horizontal plane and a spacing of 0.5 m in the vertical plane, creating four layers. The top layer was treated as “unconfined”, while the bottom three layers were treated as “confined/unconfined”. The K field for the model was taken from the 2008 field observations. For the top two layers, the model used K measurements from the medium depth piezometers kriged using a linear variogram. The bottom two layers used K measurements from the deep piezometers interpolated in the same way.

[13] No-flow boundaries were used around the vertical edges of the volume, mimicking the sheet piles at the boundaries of the OSL. The bottom of the model was also a no-flow boundary; however, drains were imposed along the southern wall of the OSL to simulate postulated leaking. Constant head boundaries corresponding to the stream banks were imposed based on stream survey data.

[14] Backward-in-time particle tracking was employed to gain insight into the observed changes in K between 2008 and 2009. We hypothesized that filtering of fine particles from areas with high K and porosity to areas where they were lower induced K change. To determine if this hypothesis is consistent with flow direction and observed K change within the aquifer, we tracked particles terminating in areas where K decreased.

4. Results

4.1. Water Table

[15] In 2009, the observed water table has head equipotentials that are sub-parallel to the upstream meander limb. This indicates most water flow from the stream is directed towards the base of the point bar, which limits hyporheic flow (Figure 2a). The hydraulic gradient within the point bar is fairly uniform, with a typical value of about 0.01. The modeled water table is in good agreement with field observations verifying the presence of a leak near the base of the point bar (Figure 2b). The mean deviation of the model from the observed water table is 1.96 cm.

Figure 2.

Potentiometric surface maps of (a) observed and (b) modeled water table elevations in the point bar aquifer. Flow is primarily directed from the stream toward the base of the point bar, indicating the stream is partly losing. It is inferred that water is leaking from the base and walls of the OSL and influencing the direction of groundwater flow.

4.2. Patterns of Hydraulic Conductivity Change

[16] For 2008, average K for each deep well (n = 41) within the point bar ranged from 3.9 to 42.2 m/day with an overall arithmetic mean of 11.8 m/day and a standard deviation of 6.9 m/day. The mean range of the three tests for each individual well was 1.0 m/day. Normalized by the average value for each well, the average range was 9.6%. For medium depth wells (n = 38), average values ranged from 6.8 to 34.2 m/day with an overall average of 16.5 m/day and a standard deviation of 8.0 m/day. The mean range in results for each well was 1.9 m/day, or 12.75%.

[17] Due to the relatively low variation in K values from repeat tests at the same well in 2008, fewer repeat tests were conducted in 2009. In 2009, the range of average deep well values within the point bar (n = 37) was from 3.1 to 82.2 m/day with an overall average value of 16.2 m/day and a standard deviation of 14.0 m/day. Well averages for of the 9 medium wells tested ranged from 6.1 to 16.2 m/day with an overall average of 11.5 m/day.

[18] A number of spatial trends are present in the kriged K distributions for both 2008 and 2009 (Figures 3a and 3b). In both years, K is highest near the tip of the point bar, although it is slightly lower at the very tip. The area with the lowest K occurs within the base of the point bar slightly toward the downstream side. Despite these similarities, distinct differences are also present. Most notably, significant changes in K occurred between 2008 and 2009, with K increasing in some areas and decreasing in others (Figure 3c). The spatial distribution of K change is similar to that of 2008 K and 2009 K in general. Areas that had high K in 2008 tended to show increases in K, while areas with lower K tended to display reductions. A roughly similar trend exists for the medium depth data; however, not enough tests were conducted in 2009 to make a strong comparison.

Figure 3.

Hydraulic conductivity (K) distributions for (a) 2008 and (b) 2009, and (c) percent change in K between those years. Figure 3c also shows modeled backward-in-time particle tracks and the locations of grain size samples in Figure 4. Particles ending in areas where K decreased first passed through areas where K increased, which suggests fine materials are being mobilized then deposited along groundwater flowpaths.

4.3. Change in Grain Size

[19] All locations (shown in Figure 3c) consist of predominantly medium sand with minor fractions of coarse sand, fine sand, and silt and clay. Between 2008 and 2009, grain size distributions on the left (upstream) side and middle of the point bar experienced a slight decrease in the fraction of fines (0.1 to 1 mm) (Figure 4). In contrast, the proportion of fines on the right (downstream) side of the point bar increased by about 10%.

Figure 4.

Percent change in cumulative grain size distributions from 2008 to 2009 at three sample locations denoted in Figure 3c. At locations 1 and 2, where K increased, the proportion of grains between 0.1 and 1.0 mm decreased between 2 and 4 percent. At location 3, where K decreased or remained constant, the proportion of the same sizes increased. This is consistent with the mobilization and deposition of fine particles by groundwater causing K change.

4.4. Particle Tracks

[20] Backward-in-time particle tracking demonstrates that particles reaching the base of the point bar passed through areas experiencing large increases in K. In contrast, particles did not pass though the band across the tip of the point bar where K increased 40% or less. The particle tracks (streamtubes roughly) also broaden as they pass though the point bar, which suggests that groundwater discharge is greater near the bank, but decreases near the base of the point bar.

5. Discussion and Summary

[21] For the first time, we document meander-scale changes in hydraulic conductivity of an alluvial aquifer adjacent to a stream. The most plausible explanation for the K evolution is hyporheic transport of mobile fine materials in the aquifer driven by surface water-groundwater interactions. The observed patterns in K change, changes in grain size, and modeled particle tracks consistently point to this conclusion. The results also consistently suggest that fine materials are flushed from originally more permeable areas and then get trapped and accumulate in originally less permeable areas, potentially leading to some feedback. This study is novel in that the experiments were done under close-to-natural environmental conditions. This differs from most previous studies of subsurface particle transport, which are typically forced injection-extraction flows conducted at smaller scales.

[22] Other potential explanations for the K change are less plausible and easily refuted. Dissolution and precipitation reactions are unlikely since the OSL fill material consists primarily of quartz sand, which is chemically inert within the time-frame of the study and for the prevailing chemical conditions. Compaction is also not likely since repeat surveys of the OSL did not show substantial change in ground elevation. This would also lead to more or less uniform K decrease and cannot explain the patterns we observed. Temperature measurements of groundwater from both years had values well within 5°C (repeat slug tests were both conducted in the summer months), which corresponds to no more than a 10% difference in K if viscosity corrections are done. Biofilm growth cannot be completely ruled out, but would also lead to decreases in K, which does not account for the broad K increases in much of the OSL.

[23] The phenomenon we observed could have far-reaching consequences because it may be occurring in many alluvial systems subject to large-scale hyporheic flow. In particular, because this evolution took place in an artificial aquifer that was freshly deposited, the K evolution we observed may be a good analogue for changes occurring in new alluvial deposits from river avullsions, in point bar aquifers deposited during periods of rapid meander migration, or in alluvial aquifers soon after stream re-meandering efforts as part of river restoration.

[24] While it is clear that fine material transport in aquifers can influence hydraulic conductivity, it is unclear how it will do so over longer time and spatial scales. With a limited fine material supply, the K distribution observed in 2009 may persist, or alternately, the fine particles may be flushed from the point bar entirely and cause K to increase more broadly. On the other hand, if suspended sediment from the stream is able to enter the alluvial aquifer, the zone of K reduction may expand as particles reach areas with low K and are deposited. Any of these scenarios is plausible for natural alluvial aquifers depending on fine material availability and its mobility in the groundwater flow system. In some instances, periodicity in K changes may occur due to intermittent flooding. In this case, transport and deposition of fine material in hyporheic zones and shallow alluvial aquifers will be paced by the dynamics of the river.

[25] Previous studies have demonstrated K at or below the streambed changes due to particle deposition, clogging, and other processes. Now, it is clear that K should be considered a dynamic parameter even in areas several meters beyond the active layer of the streambed or bank where clogging and intermittent erosion and deposition occur. This phenomenon has the potential to affect all flow mediated ecological and biogeochemical reactions over time because K is a fundamental control on nutrient delivery and residence time in the subsurface.


[26] This research was funded by the American Chemical Society-Petroleum Research Fund (46655-G8), the STC program of the National Science Foundation via the National Center for Earth-surface Dynamics under the agreement EAR-0120914, a Geological Society of America student research grant, and the Geology Foundation at the University of Texas at Austin.