Effect of experimental wood addition on hyporheic exchange and thermal dynamics in a losing meadow stream

Authors


Corresponding author: A. H. Sawyer, Department of Earth and Environmental Sciences, University of Kentucky, Lexington, KY 40506, USA. (audrey.sawyer@uky.edu)

Abstract

[1] Stream restoration structures such as large wood can enhance shallow river-groundwater exchange, or hyporheic exchange, and alter temperature dynamics in restored reaches. We added and then removed channel-spanning logs in a second-order mountain meadow stream to test short-term impacts on hyporheic exchange, streambed temperatures, and surface water temperatures. Based on vertical seepage measurements and numerical simulations of hyporheic fluid and heat flow, large wood addition increased hyporheic exchange and altered streambed temperatures. In this losing stream, meter-scale hyporheic exchange cells formed beneath large wood. Upwelling pore water downstream of logs stabilized diel temperature cycles across <8% of the streambed, creating localized but potentially valuable thermal refuge. Exchange rates were <0.1% of channel discharge—too small to impact the range of diel temperature signals in surface water. However, the lag between downstream and upstream diel temperature signals was slightly greater with large wood, which may indicate that surface storage zones rather than hyporheic storage zones increased thermal retardation. Losing conditions limited the spatial extent and rates of hyporheic exchange near large wood. Impacts of large wood reintroduction on hyporheic exchange depend on ambient groundwater discharge or recharge, streambed permeability, channel Froude number, large wood blockage ratio, and large wood spacing. In many streams, large wood reintroduction may increase hyporheic habitat volume and complexity but may not increase exchange rates enough to alter surface water temperature or chemistry. Surface storage zones such as eddies and pools can still influence heat and solute retention in the channel.

1. Introduction

[2] Large wood is a natural morphologic feature in streams that impacts channel roughness and flow conveyance [Curran and Wohl, 2003; Manga and Kirchner, 2000; Shields and Gippel, 1995], traps sediment [Montgomery et al., 1996], controls pool spacing [Montgomery et al., 1995], and drives river-groundwater (hyporheic) exchange [Lautz et al., 2006; Mutz et al., 2007; Sawyer et al., 2011]. However, modern riparian land use and channel clearing practices have decreased large wood abundance in streams [Montgomery and Piégay, 2003]. Large wood reintroduction is a common tool for restoring diversity in hydrologic environments and in-stream habitat [Shields et al., 2006].

[3] One potential benefit of large wood reintroduction is an increase in hyporheic exchange [Boulton, 2007], which connects streams with their surrounding aquifers. In stream restoration efforts, hyporheic exchange has traditionally received limited attention [Hester and Gooseff, 2010; Ward et al., 2011], despite its potential influence on the transport of heat, organic matter, nutrients, and contaminants within the stream corridor. Hyporheic exchange exposes surface water to microbially active sediment, providing opportunities for nutrient retention and contaminant degradation. The hyporheic zone is also habitat for invertebrates and fish embryos, which are sensitive to temperature, dissolved oxygen, and other biophysical parameters controlled by fluid flow [Brunke and Gonser, 1997]. Hyporheic restoration may therefore improve water quality and habitat in both the channel and streambed.

[4] The few published studies that have monitored hyporheic responses to large wood reintroduction or removal report variable findings. Lautz and Fanelli [2008] demonstrated that a log dam restoration structure contributed to heterogeneity in vertical head gradients, temperature, and pore water geochemistry. In contrast, Kasahara and Hill [2006] observed little impact of a large wood-constructed step on oxygen concentrations within the hyporheic zone, presumably due to siltation. Wondzell et al. [2009] observed a short-term decline in hyporheic exchange following experimental wood removal in a low-gradient stream. Scealy et al. [2007] monitored hyporheic fauna in response to wood reintroduction in pools and riffles and observed the greatest increases in abundance and diversity near riffle locations.

[5] These mixed outcomes suggest that the success of large wood reintroduction for restoring hyporheic processes may depend on site-specific variables such as stream morphology, streambed permeability, and ambient groundwater discharge or recharge rates. In particular, ambient groundwater flow reduces hyporheic exchange driven by bed form-current interactions [Boano et al., 2008; Cardenas and Wilson, 2006] and should similarly reduce hyporheic exchange due to current interactions with large wood. Gaining and losing conditions may also influence ecological responses to restored hyporheic flow paths near large wood. For example, restored hyporheic flows may particularly benefit losing reaches for two reasons. First, losing reaches typically have less hyporheic exchange (and presumably less hyporheic habitat complexity) than neutral reaches of similar morphology and sediment permeability due to prevailing downward head gradients [Cardenas, 2009; Lautz et al., 2006]. Second, losing reaches are less thermally buffered against air temperatures and radiative heating because groundwater and hyporheic contributions to the channel are locally minimal or absent. Restoring hyporheic exchange in losing reaches could improve both habitat complexity and thermal buffering.

[6] We added and removed channel-spanning logs in a losing mountain meadow stream to assess the impact on hyporheic exchange and thermal dynamics in the streambed and surface water. Large wood addition formed meter-scale hyporheic zones and stabilized streambed temperatures within small upwelling zones but did not significantly alter surface water temperatures exiting the reach due to low exchange rates. Ambient groundwater recharge limited the spatial extent and rates of hyporheic exchange near large wood. We first present measured and numerically simulated effects of large wood addition and removal on hyporheic exchange and streambed temperatures. We then use numerical models to further test how losing and gaining conditions impact hyporheic responses to large wood addition.

2. Study Site

[7] Experiments were conducted in San Antonio Creek, a second-order meadow stream in Valles Caldera National Preserve (VCNP) in the Jemez Mountains of New Mexico (Figure 1). VCNP was created by an act of US Congress in 2000 for conservation, grazing, hunting, and recreation. Mean annual temperature and rainfall in the region are 9°C and 476 mm, respectively [Bowen, 1996]. Forty percent of annual precipitation falls primarily as snow between October and April. Another 50% of annual precipitation falls during the monsoon season in July and August [Bowen, 1996]. This investigation was conducted after the hydrograph peak associated with spring snowmelt but prior to monsoon rains. Discharge ranged from 96 to 127 L s−1, and water depth was ∼20 cm (Figure 2).

Figure 1.

(a) Location of experimental reach on San Antonio Creek in Valles Caldera National Preserve, New Mexico. (b) Map of experimental reach. Locations of channel-spanning logs are numbered from upstream to downstream. High-resolution hyporheic temperature measurements were conducted in the subreach near the second, third, and fourth logs. Monitoring locations for reach-scale surface water temperature are labeled SW1 and SW2. (c) Panoramic photograph of reach.

Figure 2.

Water surface elevation profile before log addition (IC) and after (L12). Arrows indicate log positions. Experimental subreach is shaded gray. Logs altered local water surface elevation most visibly near the first, eighth, and tenth logs, marked with asterisks.

[8] The experimental reach is 2606 m above sea level and lies at the head of an unconfined valley (Figure 1). The reach was selected for its relatively simple morphology, permeable streambed sediment, and lack of preexisting large wood. Woody riparian species are absent in the meadow valley, due in part to temperature and historically frequent surface fires [Coop and Givnish, 2008]. Riparian vegetation is dominantly grasses and forbs. Bed slope is 0.006 m m−1 (Figure 2), and channel width ranges from 1–3 m. The bed is armored with cobbles and gravel. Underlying deposits consist of very poorly sorted, unconsolidated volcaniclastic and colluvial sediment ranging from silt to boulders.

[9] The 175-m experimental reach is losing throughout and includes two straight runs separated by a meander (Figure 1). The reach also contains subtle pool-riffle structures (Figures 1 and 2). Logs were added and removed throughout the reach to assess large-scale changes in surface water temperature. Hyporheic dynamics were monitored at high spatial and temporal resolution within a representative “subreach” near three logs in the upstream run (Figure 1).

3. Methods

3.1. Experimental Log Configurations

[10] Channel-spanning logs were added and removed over five sets of experiments: initial control without large wood (IC), three experiments with successively greater log separation distances of 12, 24, and 48 m (L12, L24, and L48), and a final control without large wood (FC). The goal of the final control was to resolve any effect of declining water table and discharge over the season.

[11] Logs were cut to stream width and securely mounted with wire on pairs of steel stakes, which were driven approximately 75 cm into the streambed. Logs were positioned against the bed, but a vertical gap of several centimeters remained between each log and the relatively deep, central portion of the bed. All logs were fully submerged. Average log diameters ranged from 10 to 19 cm, and blockage ratios ranged from 0.46 to 0.88 (Table 1). Log addition and removal did not significantly erode or disturb the banks or armored bed.

Table 1. Log Dimensions
LogLength (cm)Diameter (cm)Blockage Ratio
  • a

    Logs in experimental subreach. Blockage ratio is log diameter normalized by channel flow depth.

1104110.46
2a114160.64
3a104180.82
4a105110.65
5163100.67
6100170.61
7169150.88
8127150.83
9110190.79
10144100.67
11128120.60
12170140.70
13153160.55
14120130.54

[12] The five sets of experiments (IC, L12, L24, L48, and FC) spanned approximately 6 weeks. Equilibration time for hyporheic exchange due to log addition or removal was ∼1 day or less, based on changes in diel temperature cycles within the hyporheic zone. With the exception of initial control (IC), surface water temperatures were monitored at the upstream and downstream ends of the reach within the main channel (SW1 and SW2 in Figure 1). To characterize hyporheic flow patterns around several representative logs within an experimental subreach, we measured vertical head gradients, surveyed water surface elevations, and monitored streambed temperatures.

3.2. Measurement of Hyporheic Responses to Log Addition and Removal

[13] Nine in-stream piezometers were installed within the subreach to measure vertical head gradient in the streambed (Figure 3). Several additional piezometers were installed in the larger reach to confirm that vertical head gradients and losing conditions were similar throughout. In-stream piezometers were constructed from 1.7-cm outside-diameter (o.d.) steel pipe and had 3.0-cm screens, which were positioned 30 cm below the top of sediment. Piezometer diameter was less than one-tenth of log diameter.

Figure 3.

Photographs of subreach (a) without logs and (b) near the third log. Streamflow directions are indicated with white arrows. Examples of thermistors (T) and in-stream piezometers (IP) are labeled.

[14] Head at in-stream piezometers was measured relative to the surrounding stream's water surface elevation using a chalked wire and was averaged from three readings. Vertical head gradients were used to calculate vertical specific discharge rates (q) from estimates of hydraulic conductivity (K) and Darcy's law. Water surface elevation was surveyed along the experimental reach using a Sokkia Total Station. Streambed elevations, bank locations, and piezometers were also surveyed.

[15] Sixteen vertical arrays of temperature sensors were installed within the subreach (Figure 3). Vertical thermistor arrays were constructed by mounting four HOBO TMFC0-HD sensors on steel rods, which were inserted in solid, water-filled, 2.13-cm o.d. steel pipes located along the channel centerline. Temperature sensors were installed to depths of 5, 10, 20, and 30 cm below the sediment-water interface in most cases. Five arrays alternately recorded temperatures at greater depths of 60 and 80 cm. Surface water temperatures were also recorded at the upstream and downstream ends of the reach (SW1 and SW2) and at several locations within the subreach. Temperatures were logged every 5 min using four-channel HOBO U12 outdoor data loggers with specified accuracy of 0.25°C and resolution of 0.03°C. Streambed specific heat and thermal conductivity were measured in situ using a Decagon Devices KD2 Pro probe.

[16] Vertical seepage velocities were estimated from the two shallowest temperature signals at every location according to the method of Hatch et al. [2006] implemented in a MATLAB program called “Ex-Stream” [Swanson and Cardenas, 2010]. The approach is based on an analytical solution that relates vertical seepage velocity to the ratio of amplitudes of diel temperature cycles at two depths within the streambed. The solution assumes one-dimensional advective-conductive transport of a diel thermal signal from the streambed into homogeneous sediment under steady porous flow conditions. Accuracy in estimated vertical flux is greatest when the horizontal flux component is minimal [Lautz, 2010].

3.3. Hydraulic Conductivity Estimates

[17] Based on Darcy's law, vertical seepage velocities from heat tracing (v) and head gradients from in-stream piezometers (Δh/Δz) were used to calculate streambed hydraulic conductivity (K) at five collocated piezometers and thermistors:

display math

where q is specific discharge. Porosity (n) was assumed equal to 0.3

[18] To further constrain hydraulic conductivity, pneumatic slug tests were conducted at four locations according to the method of Cardenas and Zlotnik [2003] and analyzed according to Bouwer and Rice [1976]. The slug test apparatus consisted of a 3.1-cm o.d. steel piezometer with a 30 cm-long screen. The top of the screen was positioned at least 30 cm below the sediment-water interface.

[19] Estimated K are typical of silty or clean sand [Freeze and Cherry, 1979] and decrease gradually with distance downstream in the subreach (Table 2). Estimates from slug tests are approximately 2–5 times greater than estimates from VHG and temperature measurements (equation (1)). Where in-stream piezometers were not collocated with thermistors, K values were manually interpolated across the subreach so that vertical fluid fluxes could be calculated from VHG (equation (1)). In our interpolation, we attempted to honor all K values from slug tests and collocated piezometers and thermistors. The maximum interpolation distance was 3 m.

Table 2. Hydraulic Conductivity Estimates From Piezometer-Thermistor Data (Kpt) and Slug Tests (Kslug)
x (m)Kpt (m d−1)Kslug (m d−1)
17.723.516
23.721.37.9
30.322.04.4
32.120.47 
39.130.30 

3.4. Numerical Simulation of Hyporheic Fluid and Heat Exchange

[20] Vertical heat tracing and head gradient methods provide a one-dimensional estimate of hyporheic exchange. In order to assess two-dimensional exchange patterns near the third log, we numerically modeled fluid and heat flow within the streambed using the approach of Sawyer et al. [2012]. Two-dimensional porous flow in sediment was solved using the steady state groundwater flow equation:

display math

where K is hydraulic conductivity, and h is head. Darcy velocity or groundwater flux, q, equals the term in brackets. Hydraulic conductivity was 4.0 m d−1 based on field estimates (Table 2) and was assumed homogeneous and isotropic. Heat flow within the sediment was linked with fluid flow using the advection-conduction equation:

display math

where T is temperature, t is time, ρc is bulk volumetric heat capacity, q is Darcy velocity from pore water flow simulations, and κ is bulk thermal conductivity. Subscript w denotes a property of the pore water. Thermal and hydraulic properties of the sediment were chosen based on field measurements: bulk thermal conductivity was 1.6 W/K-m, bulk volumetric heat capacity was 3.0 × 106 J/m3-K, and fluid volumetric heat capacity was 4.2 × 106 J/m3-K. Numerical simulations did not include the effects of thermal dispersion, which were small [Sawyer et al., 2012].

[21] Pressure along the top boundary (sediment-water interface) was predicted from blockage ratio (B, the ratio of log diameter to channel depth), gap ratio (G, the distance between the log and bed normalized by channel depth), and channel Froude number (Fr, with channel depth as the length scale) [Sawyer et al., 2011]. The thermal condition along the top boundary was mixed: most of the sediment-water interface was assigned a sinusoidal diel temperature signal with amplitude of 5°C, similar to measured surface water temperatures, but regions of strong upwelling were treated as advective-flux boundaries, after Sawyer et al. [2012]. Strong upwelling was defined by an upwelling velocity with thermal Peclet number (Pe) greater than 2, where:

display math

Td is the diel timescale (1 day). Pe represents the balance between advective and conductive heat transport. Values greater than unity indicate that advection dominates heat transport. By basing the boundary condition on a Pe threshold, we allowed hyporheic flow dynamics to influence temperatures in upwelling zones and only neglected conduction along the streambed where it was small. At the model base, we assigned a downward fluid flux of −3 cm d−1, similar to estimated rates from in-stream piezometers and thermistors in Trials IC and FC. The basal boundary condition for heat flow was convective-flux. Model sides were spatially periodic in pressure and temperature. Domain width was chosen as 12 m to represent the log spacing in L12. Domain depth was 3 m, which minimized the influence on fluid and heat flow near the log. We report simulated streambed temperature dynamics as A*, the amplitude ratio of the diel temperature cycle in the sediment to that in the surface water.

[22] Hyporheic exchange was also simulated for various rates of ambient groundwater upwelling and downwelling (qgw) to explore how gaining or losing conditions impact exchange. Boundary conditions at the model sides and top were unchanged, while the specified fluid flux at the base was adjusted. The domain depth was increased to 5 m to minimize the impact on hyporheic exchange near the log, particularly for scenarios with weak ambient groundwater flux and deep hyporheic exchange. Average hyporheic exchange across the streambed (qhyp) was calculated for each simulation as the average specific discharge, q, across the portion of the bed with hyporheic exchange (not across the entire bed, which would include regions without hyporheic exchange). Hyporheic zone area (A) was also estimated, and a characteristic hyporheic residence time (τ) was calculated as the area divided by the integral of specific discharge across portions of the bed with hyporheic exchange.

4. Results

4.1. Measured Surface Water Temperature Response to Log Additions

[23] Differences in mean surface water temperatures at the upstream and downstream ends of the experimental reach were negligible and within the error of temperature sensors for L12, L24, L48, and FC (downstream temperature was not recorded for IC). Mean temperature, averaged over a 2 day record, decreased with distance downstream by 0.002°C and 0.005°C for L12 and L24 and increased by 0.030°C and 0.024°C for L48 and FC. Amplitudes of diel temperature cycles at upstream and downstream locations were also equal to within sensor accuracy: downstream amplitudes were 0.086, 0.097, 0.083, and 0.121% greater than upstream amplitudes for L12, L24, L48, and FC, respectively. Diel temperature cycles at the downstream location lagged the upstream location by 40, 29, 37, and 23 min, respectively. These temperature lags may correspond with advective transport and retardation of the diel temperature signal in the downstream direction. Advective travel times in the experimental reach were on the order of 10 min, based on measured stream velocities. Logs may have formed stagnant zones that increased retardation—the temperature lag was greatest for L12, where the density of logs was greatest.

4.2. Measured Hyporheic Response to Log Additions

[24] Without large wood (IC and FC), vertical fluid flow across the bed was downward throughout the subreach (Figure 4). Average downwelling rates were −3.0 cm d−1 at the beginning of the season (IC) and −1.8 cm d−1 at the end of the season (FC). With initial log addition (L12), downwelling rates increased immediately upstream from the second and third logs (Figure 4). Immediately downstream from the second, third, and fourth logs, regions that were previously downwelling became upwelling. In L24 and L48, the second and fourth logs were removed. Uniform downwelling resumed where the second log had been located. In contrast, local upwelling rates temporarily increased where the fourth log had been located. The zone of downwelling broadened near the third log.

Figure 4.

Vertical specific discharge estimates (q) along the subreach for five experiments from heat tracing (white bars) and vertical head gradients, or VHG (black bars). Fluxes from heat tracing were estimated using the shallowest sensor pairs (5 and 10 cm depth). Piezometer screens for VHG calculation were at 30 cm depth. Positions of second, third, and fourth log (as numbered in Figure 1b) are indicated with black dots. The subreach is naturally losing (IC and FC), but logs create zones of localized upwelling (L12, L24, and L48).

4.3. Measured Streambed Temperature Response to Log Additions

[25] Diel streambed temperature cycles varied strongly in response to log addition and removal. Without logs, diel temperature cycles were essentially one-dimensional—amplitudes decayed with depth but varied little with distance along the stream (Figure 5). At four thermistor arrays (T7 through T10), average amplitudes were 57, 39, 19, and 8.6% of river amplitudes at depths of 5, 10, 20, and 30 cm, respectively. With logs present, amplitudes were greater immediately upstream from each log (Figure 5, T7 and T8) and lesser downstream (T9 and T10). At T8 where downwelling near the third log was greatest, amplitudes were 91, 75, 36, and 15% of river amplitudes at depths of 5, 10, 20, and 30 cm. At T9 where upwelling was greatest, amplitudes were 34, 20, 12, and 6.6% of river amplitudes at the same depths. These temperature patterns near the third log remained relatively constant with manipulation of neighboring logs in L12, L24, and L48. The streambed thermal response to log addition or removal occurred over a few hours (Figure 6).

Figure 5.

Cross-sections of diel streambed temperature amplitudes normalized by surface water temperature amplitudes (A*) before and after removal of the third log (experiments L48 and FC). Streamflow in surface water (not shown) is left to right. Log is drawn to scale. Thermistor positions are indicated with circles, and vertical arrays are labeled as T6-T12.

Figure 6.

Time series of streambed temperatures recorded immediately upstream and downstream of the third log (T8 and T9) before and after the time of log removal, indicated with vertical line (between experiments L48 and FC). Positions of T8 and T9 are indicated in Figure 5.

4.4. Simulated Response of Hyporheic Exchange and Streambed Temperatures

[26] Simulations of hyporheic fluid flow predict similar patterns upstream and downstream of the log (Figure 7). However, predicted vertical flux magnitudes from 2-D simulations are greater than estimates from 1-D field measurements, particularly in the upwelling zone downstream of the log (Figure 7c). One potential source of model error is the prescribed pressure boundary at the sediment-water interface predicted from channel Froude number, log blockage ratio, and gap ratio. Surveyed water surface elevation appears to agree well with predicted pressures though (Figure 7a). Another source of error is the choice of streambed permeability and assumption of homogeneous, isotropic media. Additional error in field-based vertical flux estimates can arise from the application of one-dimensional heat tracing approaches to multidimensional hyporheic flows [Lautz, 2010].

Figure 7.

(a) Prescribed head along the bed (upper boundary condition for hyporheic flow simulation) and surveyed water surface elevation. Prescribed head depends on log diameter (18 cm), height above bed (∼4 cm), channel flow depth (25 cm), and channel Froude number (0.26) [Sawyer et al., 2011]. (b) Simulated specific discharge across the bed, evaluated at the sediment-water interface. (c) Simulated specific discharge across the bed, evaluated based on head gradients between the stream and 30 cm depth (black line), for direct comparison with specific discharge estimates from L24 based on in-stream piezometers (blue bars) and thermistors (red bars). (d) Simulated amplitude of diel pore water temperature signal normalized by amplitude of surface water temperature (A*) near the third log. Simulated amplitudes are in fair agreement with measurements (Figure 5). White lines are simulated streamlines.

[27] Despite discrepancies in the flux magnitude, simulated thermal patterns in the hyporheic zone are similar to field observations (Figures 5 and 7d). Amplitudes are large immediately upstream of the log where downwelling occurs. Downstream of the log, upwelling water stabilizes streambed temperatures. The simulated zone of stable temperature is narrower than that observed in the field, suggesting that simulations predict a more focused upwelling region.

[28] Ambient groundwater recharge spatially restricts hyporheic exchange. The hyporheic zone, which comprises all flow paths that originate from and return to the channel, extends only a meter from the log in each direction. Hyporheic zone area (volume per channel width) is 1.54 m2. The simulated average hyporheic flux over the length of streambed where exchange occurs is 2.6 × 10−6 m s−1 (22 cm d−1), compared with an ambient rate of groundwater recharge of −3.5 × 10−7 m s−1 (−3 cm d−1). The characteristic hyporheic residence time is 2.7 × 105 s (3.2 days).

4.5. Sensitivity of Hyporheic Exchange to Ambient Groundwater Flow

[29] Hyporheic exchange area decreases with ambient groundwater flux (Figures 8 and 9a). Losing conditions (Figures 8a–8d) skew the hyporheic zone toward the downstream side of the log and extinguish longer hyporheic flow paths that would originate far upstream of the log. Gaining conditions (Figures 8f–8h) skew the hyporheic zone toward the upstream side of the log and extinguish longer hyporheic flow paths that would exit far downstream of the log. Small rates of groundwater recharge or discharge substantially reduce hyporheic zone area due to the extinction of broad areas of slow hyporheic flow farther from the log. Additional increases in ambient recharge or discharge cause moderate declines in hyporheic zone area associated with the extinction of faster potential flow paths close to the log. Further increases in the rate of recharge or discharge eventually overwhelm all exchange (Figure 9a).

Figure 8.

(left) Streambed fluxes for various rates of ambient groundwater discharge or recharge (lower right bar plot indicates relative strength and direction of qgw). White line delineates the hyporheic zone. Flow in the surface water (not shown) is left to right. Log is shown for scale. (right) Streambed diel temperature response. White line indicates where thermal Peclet number equals 1—inside this region, advective heat transport dominates conduction. Parameters in Figure 8c represent the third log (see Figure 7). The stream is losing for Figures 8a–8d, neutral for Figure 8e, and gaining for Figures 8f–8h.

Figure 9.

(a) Hyporheic exchange area versus ambient groundwater flux for gaining and losing conditions. (b) Average hyporheic flux versus ambient groundwater flux. Hyporheic flux is only averaged over the portion of streambed where hyporheic exchange occurs—not over the entire bed. (c) Characteristic residence time versus ambient groundwater flux. τ is calculated as the hyporheic zone area divided by the integrated hyporheic flux over the portion of the bed where exchange occurs. Dotted and dashed lines represent values of ambient groundwater flux required to extinguish hyporheic exchange for gaining and losing conditions, respectively.

[30] The average hyporheic exchange rate initially increases with ambient groundwater flux and then declines (Figure 9b). The initial increase is due to the extinction of broad regions of slow exchange—only a small zone of relatively fast hyporheic exchange remains. Additional increases in ambient groundwater flux overwhelm hyporheic exchange, causing a reduction in average exchange rates. Notably, if hyporheic flux were averaged over the entire streambed, the average exchange rate would simply decline monotonically with increasing ambient groundwater flux. The characteristic hyporheic residence time declines rapidly with small increases in ambient groundwater flux, due primarily to the decline in hyporheic zone area and extinction of long, slow flow paths far from the log (Figure 9c). Additional increases in ambient groundwater flux cause a modest reduction in the characteristic residence time, until ambient groundwater flux eventually overwhelms exchange.

[31] Simulations indicate that gaining conditions restrict hyporheic exchange more than losing conditions: hyporheic area and average exchange rate are smaller for a given magnitude of groundwater discharge than recharge, and the characteristic residence time is longer (Figures 8 and 9). A groundwater discharge rate of 2.8 × 10−5 m s−1 (2.4 m d−1) or a recharge rate of −6.5 × 10−5 m s−1 (−5.6 m d−1) would eliminate all hyporheic exchange near the third log in this study. The greater sensitivity to gaining conditions is due to the asymmetric distribution of fluid flux across the bed, which results from the asymmetric pressure distribution (Figure 7). The predicted zone of upwelling is focused in a narrow zone downstream of the log. As a result, maximum hyporheic upwelling rates are greater than maximum downwelling rates under neutral conditions. Therefore, a greater ambient rate of groundwater recharge is required to eliminate upwelling. Near the third log in the experimental reach, however, the measured upwelling distribution appears broader than the simulated distribution (compare Figures 5 and 7). If simulations generally over-predict flow focusing downstream of logs, hyporheic responses to gaining and losing conditions may be more similar than models indicate.

[32] Despite sensitivity of hyporheic exchange to ambient groundwater flux, diel streambed temperature dynamics near the log are relatively insensitive (Figure 8). San Antonio Creek has a moderate streambed permeability, so fluid fluxes and thermal Peclet values are only significant immediately adjacent to the log. As a result, temperature signals are only sensitive to hyporheic dynamics immediately adjacent to the log, but this region is also where hyporheic dynamics are most insensitive to ambient groundwater flux. The permeability of the sediment essentially moderates the impact of ambient groundwater fluxes on diel streambed temperature dynamics.

5. Discussion

5.1. Implications for Streambed Environments

[33] Large wood addition increases hyporheic exchange and thermal heterogeneity in the streambed, which may increase benthic and hyporheic habitat diversity and the availability of thermal refuges. However, effects may extend only meters from large wood, particularly in losing or gaining reaches. In this losing stream, hyporheic exchange was presumably negligible prior to log addition, due to pervasive downwelling. Introduction of channel-spanning logs created localized upwelling zones. If each log converted two linear meters of streambed to a region of hyporheic influence (Figure 7), logs cumulatively transformed surface water-groundwater interactions across 17% of the streambed in L12. New hyporheic exchange cells increased thermal heterogeneity in the streambed, but thermally buffered zones near each log were less than one linear meter long (Figures 5 and 7). Large wood addition likely converted less than 8% of the streambed to thermal refuge in L12. Upon log removal, downwelling resumed and thermal dynamics became longitudinally uniform again.

5.2. Implications for In-Stream Environments

[34] Large wood addition may buffer surface water temperatures if hyporheic exchange rates increase by a significant fraction of the channel discharge rate, a condition most likely met in rare conditions where channel discharge is small, streambed permeability is large, and ambient groundwater discharge or recharge is low. In this study, the volumetric hyporheic exchange rate due to an individual log was ∼0.01 L s−1, which indicates that total exchange was only ∼0.14 L s−1 in L12, or ∼0.1% of stream discharge. As a result, log addition did not significantly change diel temperature ranges in the surface water. However, the timing of diel temperature cycles did change with log addition and removal. The stream's diel temperature cycle peaked later at the bottom of the reach, and this lag generally increased with log addition. Either large wood retarded downstream heat transport, or unrelated variables such as shade from meadow grasses influenced the timing of diel temperature cycles. If large wood retarded heat transport, the mechanism was likely surface transient storage (eddies) rather than hyporheic exchange. In flume and numerical studies, Sawyer et al. [2012] also demonstrated a limited potential for large wood to impact diel surface water temperature signals. They showed exchange rates near logs are typically small relative to stream discharge, and the fastest return flows have diel temperature cycles that resemble the surface water. In another study of hyporheic flow beneath a weir, Hester et al. [2009] demonstrated that solar radiation overwhelmed the impact of hyporheic heat exchange on surface water temperatures.

[35] Analogous solute transport experiments suggest that large wood enhances solute retention, but pools and eddies rather than hyporheic zones are responsible. In three independent tracer injection studies, channels with woody obstructions had greater median travel times associated with transient storage (Fmed) and, in most cases, proportionally greater transient storage areas (As/A) [Ensign and Doyle, 2005; Jin et al., 2009; Stofleth et al., 2008]. Ensign and Doyle [2005] attributed changes in solute retention to surface storage because visual estimates of eddy size approximated tracer-derived storage zone areas, and low streambed permeability also limited exchange with subsurface storage zones. In a comparative study of four reaches, Jin et al. [2009] explained variations in transient storage metrics with correlated pool volumes (surface storage) behind debris dams. Stofleth et al. [2008] attributed an increase in solute retention after beaver dam construction to surface storage because the measured hyporheic exchange rate was a small component of the tracer-derived storage zone exchange rate. In contrast, Lautz et al. [2006] showed that debris dams and meanders both increase transient storage and attributed hyporheic exchange. Though large wood clearly influences transport of solutes and heat in streambeds [Lautz and Fanelli, 2008; Mutz and Rohde, 2003; Sawyer et al., 2012], hyporheic exchange rates near large wood are often too slow to influence reach-scale surface water chemistry and temperature. More generally for streams with or without large wood, Wondzell [2011] similarly argued that hyporheic exchange rates are often a small proportion of channel discharge, and the most important ecological benefit of hyporheic exchange may be the habitat complexity it promotes across the streambed, rather than its often minor impacts on surface water quality.

5.3. Restoration Considerations

[36] Large wood is a common element in restoration structures, including constructed steps and J-Hook Vanes. The goal of these structures is to enhance natural features such as pools and riffles, increase biophysical heterogeneity, reduce bank erosion, and provide fish habitat [Rosgen, 1997]. Though the channel-spanning logs in this study are simplified versions of larger restoration structures, they induce similar head drops that drive hyporheic exchange. Below, we address implications of our manipulation study for stream restoration, acknowledging that this study only addresses simple in-stream structures and short-term hyporheic responses.

[37] Large wood additions may improve streambed habitat quality by enhancing hyporheic exchange but may rarely improve surface water quality. Restoration outcomes depend strongly on ambient groundwater discharge or recharge rates, streambed permeability, the blockage ratio of large wood, and spacing between structures. As ambient groundwater recharge or discharge increases, hyporheic exchange beneath large wood decreases, like exchange through bedforms and meanders [Boano et al., 2008; Cardenas, 2009; Cardenas and Wilson, 2006]. Similarly, Lautz et al. [2006] showed that debris dams in losing reaches divert water to long subsurface flow paths that would otherwise return to the channel immediately downstream. Large wood additions to losing or gaining reaches will not enhance exchange unless blockage ratio, channel Froude number, and permeability are sufficiently large. For a given reach, the blockage ratio and spacing of large wood structures can be optimized to maximize hyporheic exchange volumes and rates [Sawyer et al., 2011] for an acceptable flow conveyance [Shields and Gippel, 1995]. Exchange rates sufficient to influence surface water chemistry and temperature may not be attainable in many reaches.

[38] Though hyporheic responses to large wood addition are greater in neutral reaches, streambed habitat responses may be greater in weakly gaining or losing streams. Baxter and Hauer [2000] showed that small downwelling zones in gaining reaches are favorable spawning sites for bull trout. Downwelling water delivers oxygen to embryos while large-scale groundwater discharge buffers surface water temperatures. Large wood additions to gaining streams may create downwelling zones ideal for spawning. Inversely, large wood additions to losing streams may create small upwelling zones that provide thermal refuge for benthic and hyporheic fauna. In this study, large wood additions created small thermally buffered zones across <8% of the streambed. Ambient groundwater recharge reduced the size of upwelling zones near logs but did not significantly reduce thermal buffering in the sediment—simulated temperatures were almost identical for an ambient groundwater recharge rate of 3 cm d−1 and neutral conditions (Figure 8). Modest streambed permeability limited the impact of ambient losing conditions on thermal dynamics [Cardenas and Wilson, 2007; Sawyer et al., 2012]. Where streambed permeability is low, thermal conduction dominates advection throughout much of the hyporheic zone, and thermal dynamics are therefore insensitive to slower and longer hyporheic exchange patterns. As a result, regions of the streambed that would host long, slow exchange paths under neutral conditions have a similar diel temperature pattern under weak downwelling conditions when exchange is absent.

[39] Over long timescales, hyporheic responses to large wood addition will adjust with channel morphology [Wondzell et al., 2009]. Potential morphological responses to large wood include increased sediment storage, gravel bar construction, bank erosion, channel avulsion, and braiding [Montgomery et al., 1996; Wohl, 2011]. These adjustments depend on diverse and site-specific factors such as bank cohesion, substrate and grain size, channel slope, and stream power. In a meadow stream like San Antonio Creek, vegetated cohesive banks would inhibit avulsion and braiding, but some channel widening could occur, and sediment storage would likely increase. Both of these effects would influence hyporheic exchange over decadal timescales through changes in streambed area and permeability.

6. Conclusions

[40] The addition of channel-spanning logs to a losing stream created hyporheic exchange cells at the meter-scale where hyporheic exchange was previously minimal or absent. The added exchange drove heterogeneity in diel thermal dynamics within the streambed. In particular, diel temperature ranges were reduced in upwelling zones downstream of logs. Though limited in size (<8% of total streambed area), these upwelling zones could provide refuge for invertebrates during thermal extremes. The creation of upwelling zones did not measurably impact the diel temperature range of surface water exiting the 175-m reach because hyporheic exchange rates were small relative to stream discharge (∼0.1% or less). However, eddies near large wood may have retarded downstream heat transport in the channel.

[41] Large wood reintroduction is a promising tool for increasing patchiness in biophysical parameters such as temperature within hyporheic habitats, but the impact of hyporheic enhancement on surface water quality may be negligible in many cases. Areas and rates of hyporheic exchange near large wood depend on ambient groundwater flow patterns, in addition to the blockage ratio, channel Froude number, and sediment permeability. Large wood reintroduction maximizes the spatial extent and rates of hyporheic exchange in neutral reaches. Large wood reintroduction drives a smaller hyporheic response in gaining or losing reaches, but the additional exchange may have important ecological consequences. In losing reaches, for example, the additional exchange creates thermally and chemically distinct upwelling zones that may not have existed otherwise due to prevailing downward flow across the bed. More long-term studies are needed to address how historic removal of large wood and current restoration practices may impact hyporheic zone dynamics and habitat.

Acknowledgments

[42] This research was supported by the National Science Foundation (EAR-0836540), an AGU Horton Research grant to AHS, and the Geology Foundation at the University of Texas at Austin. Wade Chaney, Katy Gerecht, Jesus Gomez, Frank Hucks, Melanie Laughlin, Lani Rosenthal, and Derek Sawyer provided assistance in field experiments. We thank Bob Parmenter for providing logistical support for research in Valles Caldera National Preserve. We thank Adam Ward, two anonymous reviewers, and the Associate Editor for their constructive comments.