Water Resources Research

Water table dynamics and groundwater–surface water interaction during filling and draining of a large fluvial island due to dam-induced river stage fluctuations



[1] Dam-controlled river stage fluctuations alter groundwater–surface water interaction between persistent bars and islands and the rivers bounding them by rapidly changing hydraulic gradients and expanding hyporheic zones. A 300-m long and 80-m wide sand-gravel island with established vegetation located on the Colorado River (Austin, Texas, USA) is subjected to >1 m daily river stage variations due to upstream dam operations. Piezometer nests with probes monitored the evolution of the water table and groundwater flow paths through several cycles of dam-induced stage fluctuations. Results show that hydraulic head and the water table within the island closely track the river stage associated with dam release. Water table mounds and depressions which overlap in time were mapped through the course of one storage-release cycle over which >4,000 m3 of water moved in and out of the island. Dam operations have drastically altered groundwater–surface water connectivity between the Colorado River and the fluvial island aquifer by pumping substantial amounts of water in and out of the aquifer during dam release and storage cycles.

1. Introduction and Background

[2] Humans have altered and regulated rivers for improved navigation, water supply, flood control, and hydroelectric energy generation, and more than half of the world's large rivers are regulated by one or more dams [Nilsson et al., 2005]. Dam release and storage cycles continually change the natural hydraulic head gradients in the downstream reaches of a river modifying groundwater–surface water interactions and hyporheic processes. Since only 42 rivers longer than 200 km are free-flowing in the United States [Benke, 1990], and all watersheds greater than 2,000 km2 have dams [Graf, 1999], the effects of dam operations on groundwater–surface water interaction need to be further understood.

[3] Reversal of hydraulic head gradients in the riverbed and banks due to dam storage-release cycles affect the magnitude and direction of groundwater and hyporheic flow [Arntzen et al., 2006; Fritz and Arntzen, 2007], potentially causing gaining streams to become losing streams and vice versa. Furthermore, dam-induced stage fluctuations induce transient spatial pressure gradients at the sediment-water interface that may be much larger compared to those driving steady hyporheic exchange. Few field studies have investigated hyporheic exchange driven by dam operations. Recent studies examine the effects of dam-induced river stage fluctuations on hyporheic exchange in riverbeds and banks covering several kilometers, but with observations from usually one-dimensional transects at the meter scale [Arntzen et al., 2006; Curry et al., 1994; Hanrahan, 2008]; vertical head gradients are typically monitored along a few vertical profiles every several kilometers. However, the interactions between surface water and shallow groundwater are intrinsically three-dimensional and multiscale [Cardenas, 2008; Poole et al., 2006; Sophocleous, 2002; Storey et al., 2003], and the dynamics are impossible to completely comprehend with vertical profiles.

[4] Minimal research has been conducted on river islands subjected to regular and marked short-term changes in stage, and the effects of dam-controlled river stage fluctuations on groundwater–surface water interaction between persistent gravel bars and rivers are poorly understood. In fact, there is a scarcity of field studies mapping hyporheic flow paths in large fluvial islands even under simpler natural steady state conditions [e.g., Dent et al., 2007]. A step forward requires further scrutiny of the details of the near-stream processes that go beyond one-dimensional vertical transects in the streambed with limited measurement points. A recent study illustrated near-stream gradient reversals and associated thermal and chemical changes along a section of the riparian zone [Sawyer et al., 2009]. This note complements that study and examines the hydrogeology of Hornsby Bend Island (HBI), a large gravel bar with established riparian forest, in the Colorado River near Austin, Texas, that is subject to daily fluctuations in river stage due to upstream dam storage and release of water. Our goal is to characterize the dynamic water table in HBI, quantify associated fluid fluxes, and to assess any potential implications on stream-aquifer biogeochemistry.

[5] The study site, HBI, is located along the Colorado River in Texas 10 km southeast of downtown Austin (Figure 1a). HBI is downstream of a sequence of dams that are controlled for power generation and flood mitigation. Daily river stage fluctuations from water release for power generation exceed 1 m and may reach 2 m at the United States Geological Survey (USGS) Gaging Station (station 08158000) which is 2 km downstream of Longhorn Dam. The Colorado River is a regionally gaining river at the study area [Larkin and Sharp, 1992] but at a local scale switches from gaining to losing due to stage fluctuations [Sawyer et al., 2009].

Figure 1.

(a) Location map of the study site, Hornsby Bend Island (HBI), on the Colorado River relative to Austin, Texas, the Longhorn Dam (the nearest dam which is mostly flow-through), and the nearest upstream USGS gaging station. Inset map shows the state of Texas with the Colorado River outlined in blue and Travis County, whose center is Austin, in-filled with red. The Colorado River flows from northwest to southeast in the map. (b) Map of Hornsby Bend Island showing location of island piezometers (IP) and in-stream piezometers (ISP). The river flows from northeast to southwest in the map.

[6] HBI is a large 310-m long and 80-m wide sand-gravel bar in the middle of the Colorado River (Figure 1b). Its ground surface has an average elevation of 120.76 m above sea level with a relief of 2.49 m. HBI is 15 km downstream of Longhorn Dam and 13 km from the nearest permanently recording USGS gaging station. The island is composed of sand, gravel, and silt deposited by the Colorado River and has an average hydraulic conductivity of 5.25 × 10−4 m/s based on grain size data.

2. Methods

[7] Fourteen island piezometers (IPs) and four in-stream piezometers (ISPs) were installed via hand augering. Island piezometers were situated on HBI to create two northwest-southeast transects across the island approximately perpendicular to river flow and one long northeast-southwest transect approximately parallel to river flow (Figure 1b). Island piezometers were constructed of 5 cm (2 in) diameter PVC pipe and screened through the water table with the screen lengths covering the entire vertical range of the dynamic water tables. The pipes were screened throughout their entire length except for the aboveground portions. In-stream piezometers were positioned upstream and downstream of HBI, as well as on the northwest and southeast sides of the island (Figure 1b). In-stream piezometers were constructed with the same screened PVC pipes.

[8] Calibrated pressure probes were deployed in IPs and ISPs to log water table elevations. Three to six probes at a time were alternated through all the piezometers with each rotation lasting 3 days with logging at 15-min intervals and thus accurately capturing and covering daily fluctuations in the water table for close to one month in the summer. Although the measurements were not all collected simultaneously, a different day's response was assumed to be similar as other days. This assumption is justified by the regular and predictable stage variations over the study period (mid-August to mid-September 2008) as evidenced by stage and discharge observations at the nearest upstream USGS gaging station (Figure 2). In all the data sets presented here, none came from the few days when the stage/discharge variation was markedly different. In any case, any noticeable differences in the peak stage for a given day or subtle variations during the peak stage are damped out when the signal propagates downstream and reaches the study site.

Figure 2.

Time series data of discharge and gage height from the USGS gaging station (08158000) downstream of Longhorn Dam and upstream of HBI during the duration of the study.

[9] A differential geospatial positioning system was used to accurately determine the location and elevation of one piezometer (IP 13) from which all other coordinates were referenced. Each piezometer plus the island and riverbank outlines were surveyed with a total station accurate to within 1 cm for the distances used in the surveys. The water table data was interpolated and contoured using inverse distance weighting. Imaginary in-stream piezometers were added between the four ISPs to better define the water surface elevation along the island's perimeter. These imaginary ISPs were assigned elevations that were linearly interpolated based on distance between two real ISPs.

3. Results and Discussion

[10] Two kilometers downstream of Longhorn Dam at the USGS gaging station, daily dam operations led to daily discharge fluctuations from 4.2 to 4.8 m3/s during dam storage to 57 to 63 m3/s during dam release (Figure 2). River depth was at its minimum level at 10:00 h with a gage height of 0.5 to 0.6 m (Figure 2). The Colorado River at the station starts responding to dam releases between 10:00 and 10:15 h every day causing river depth to increase to a maximum gage height of 1.4 to 1.9 m between 21:30 and 23:00 h. This is equivalent to a stage change of 0.8 to 1.4 m daily.

[11] Thirteen km downstream of the gaging station, river stage at HBI was lowest between 13:00 and 14:00 h, and ranged from 117.26 to 117.54 m above sea level (Figure 3). The lag time for the response to dam release between the USGS site and the study site at HBI is roughly 3 to 4 h. At HBI, river stage was highest between 1:00 and 2:15 h, and ranged between 118.16 and 118.51 m above sea level. The amount of change in river stage at HBI ranged from 0.90 to 1.07 m. River stage fluctuations at HBI are expectedly attenuated compared to those at the upstream USGS gaging station.

Figure 3.

Comparison of time series of USGS gage height, stream elevation at an in-stream piezometer (IP), and water table elevations at three island piezometers (ISP). See Figure 1b for location of piezometers (see colored symbols for piezometers in Figure 1b; they correspond to the plot above).

[12] The groundwater flow paths at HBI respond quickly to the signal from dam operations. Water table elevation oscillations within HBI are further damped and lagged relative to river stage fluctuations in the Colorado River (Figure 3). When the river is at its highest (between 1:00 and 2:15 h), water is mostly entering HBI and flowing toward a water table depression in the center of the island near its tail (Figure 4 and Animation S1 in the auxiliary material). However, the water table is mostly flat near the island's tail with a broad, but subtle water table mound with relatively small head gradients. As the river stage begins to fall after 1:00 h, this mound becomes more pronounced. On the other hand, the water table depression near the tail vanishes and this is replaced with a ridge that is connected to the mound upstream of it. This suggests that water filling in the depression is not just river water infiltrating from the perimeter due to the higher river, but also has some water moving from the mound defining a flow path that traverses the core of the island from near its head to its tail. This is consistent with the persistent head difference between the upstream ISP 4 and the downstream ISP 1 which averaged 0.36 cm over a storage-release cycle. This downstream head gradient is responsible for the simultaneous existence of a mound and depression within the island since the fluctuations also propagate downstream across the island and not just in a direction normal to the banks. In addition to the water table fluctuations, the strong reversing head gradients along the island perimeter indicate pronounced exchange of water between HBI and the Colorado River.

Figure 4.

Time snapshots (every 3 h) of the water table at HBI. The ground surface at HBI has an average elevation of 120.76 m. Arrows show the direction and relative magnitude of groundwater flow. Note the pronounced variation in water table elevation with time and the depression in the water table that forms during high river stage and a mound during low river stage. Animation S1 illustrates the process in more detail.

[13] A simple calculation suggests that the volume of water pumped in and out of HBI over one dam release cycle is ∼4,770 m3/day. This estimate is based on the difference between the average water table elevation during low river stage and the average elevation during high stage (0.687 m) multiplied by the area of the island, which is roughly 18,046 m2, and corrected by an average porosity (38.5%). While a numerical groundwater flow model calibrated to the site is the best way to calculate fluxes, it is apparent that a substantial amount of water is being pumped into and out of HBI on a daily basis due to changes in river stage associated with dam storage and release.

[14] Our observations are possibly the first two-dimensional illustration of ‘bank storage’. Whereas a simple, one-dimensional model for bank storage would show a sinusoidal signal propagating into the aquifer, our observations indicate the co-existence of a water table mound and depression within the island. This is something that cannot be predicted from a one-dimensional model of bank storage or of a flood-wave propagating into an aquifer. Moreover, to our knowledge, despite ‘bank storage’ being a classical concept in stream-aquifer hydraulics with dozens of numerical and analytical modeling studies, there seems to be limited detailed near-stream (less than 10 m from the bank) measurements of the water table during a bank storage event and there seems to be no definitive or classical field data set. Based on our research of readily available literature, the most comprehensive field data are the observations by Squillace [1996], but that study focused more on chemical signals of water moving in and out of the bank. Prior to this, perhaps the most prominent experimental data investigating bank storage was from the phenomenon's original introduction 40 years earlier through Hele-Shaw model experiments by Todd [1955] which is the basis of the classic paper on the theory of bank storage by Cooper and Rorabaugh [1963]. A recent study illustrated dam-induced fluctuations along a bank transect using a transect of piezometers within close proximity of the river [Sawyer et al., 2009]. This current study is therefore potentially the first to illustrate two-dimensional water table dynamics across an entire fluvial island under dam-affected fluctuating river conditions. Moreover, we present a case where bank storage and hyporheic exchange are synonymous and indistinguishable.

[15] Given steady groundwater flow conditions, a sequence of redox conditions would normally exist at the interface between surface water and groundwater, serving as a natural biogeochemical filter [Hill, 2000]. Dam operations create short, transient hyporheic flow paths, forcing large amounts of water in and out the island which may not allow a typical redox ladder to persist. The transport of nutrients and energy along hyporheic flow paths is therefore altered, possibly degrading riparian and hyporheic processes. Fluvial islands and large bars experiencing pronounced water table fluctuations due to dam releases are likely subject to a complex but perhaps predictable biogeochemical and ecological regime. What is clear is that regulated stage fluctuations from upstream dams dictate hyporheic processes and groundwater–surface water interactions throughout HBI.


[16] We thank Kevin Anderson of the Austin Water Utility–Center for Environmental Research for access to the field site. This work was partly supported by student grants and fellowships to B.A.F. from the Geological Society of America, Nobel Energy, and the Geology Foundation at the University of Texas at Austin.