We explored the use of continuous waterborne electrical imaging (CWEI), in conjunction with fiber-optic distributed temperature sensor (FO-DTS) monitoring, to improve the conceptual model for uranium transport within the Columbia River corridor at the Hanford 300 Area, Washington. We first inverted resistivity and induced polarization CWEI data sets for distributions of electrical resistivity and polarizability, from which the spatial complexity of the primary hydrogeologic units was reconstructed. Variations in the depth to the interface between the overlying coarse-grained, high-permeability Hanford Formation and the underlying finer-grained, less permeable Ringold Formation, an important contact that limits vertical migration of contaminants, were resolved along ∼3 km of the river corridor centered on the 300 Area. Polarizability images were translated into lithologic images using established relationships between polarizability and surface area normalized to pore volume (Spor). The FO-DTS data recorded along 1.5 km of cable with a 1 m spatial resolution and 5 min sampling interval revealed subreaches showing (1) temperature anomalies (relatively warm in winter and cool in summer) and (2) a strong correlation between temperature and river stage (negative in winter and positive in summer), both indicative of reaches of enhanced surface water–groundwater exchange. The FO-DTS data sets confirm the hydrologic significance of the variability identified in the CWEI and reveal a pattern of highly focused exchange, concentrated at springs where the Hanford Formation is thickest. Our findings illustrate how the combination of CWEI and FO-DTS technologies can characterize surface water–groundwater exchange in a complex, coupled river-aquifer system.
 The objective of this study was to explore the use of high-resolution continuous waterborne electrical imaging (CWEI), in conjunction with fiber-optic distributed temperature sensor (FO-DTS) monitoring, to improve the conceptual model for how surface water–groundwater interaction at the U.S. Department of Energy (DOE) Hanford 300 Area, Washington, regulates uranium (U) transport from the contaminated aquifer into the Columbia River. For over 40 years, starting in 1943, fluids containing radioisotopes and metals, generated during reactor fuel fabrication and chemical separation processes, were discharged to the shallow subsurface of the Hanford 300 Area (Figure 1). The primary chemical inventory included 241 t of copper, 117 t of fluorine, 2060 t of nitrate, and between 33 and 59 t of U, which is the main contaminant of concern. Previous studies have mostly focused on the behavior of U in the terrestrial system and the groundwater. A 1993 modeling study predicted that groundwater U concentrations would decrease to the 30 μg/L drinking water standard in 3 to 10 years [Westinghouse Hanford Company, 1993]. Based on this study, an interim decision for monitored natural attenuation with institutional controls on groundwater was implemented [Environmental Protection Agency, 1996]. A review in 2006, however, showed that, despite source control measures, groundwater U concentrations remained mostly unchanged [Hartman et al., 2006]. A pressing research need at this site is an improved understanding of the potential for long-term discharge of U from this persistent plume into Columbia River surface water. Achieving this understanding requires new studies that capture the spatial distribution of the primary lithologic units along the river corridor as well as spatiotemporal complexity in surface water–groundwater exchange driven by variations in stage levels on the Columbia River.
 Our study considered the hypothesis that lithologic heterogeneity, widely believed to regulate exchange between surface water and groundwater, could be imaged at an unprecedented spatial scale and resolution relative to direct investigation methods performed within the river corridor of this important site. Recent studies have demonstrated how electrical imaging can resolve subsurface heterogeneity impacting surface water–groundwater exchange based on spatial variations in electrical resistivity driven by lithologic variability [Acworth and Dasey, 2003]. FO-DTS is an emerging technology that recently has been employed to determine spatiotemporal dynamics of surface water–groundwater exchange in fluvial, lacustrine, and estuarine environments by exploiting the temperature contrasts between surface water and groundwater [Henderson et al., 2009; Selker et al., 2006a, 2006b; Tyler et al., 2009]. Henderson et al.  recently combined FO-DTS and stationary resistivity imaging to study aquifer-estuary interaction, but no studies have previously integrated CWEI and FO-DTS to fully exploit the complementary information in both data sets regarding spatial and temporal variability in surface water–groundwater exchange.
 Here, we use CWEI to resolve lithologic variability along a major river corridor by determining the spatial variation in the electrochemical polarizability, a property closely related to lithology, in addition to the electrical resistivity as routinely captured in CWEI studies. We adopt empirical petrophysical relations to reconstruct predictions of the variability in specific surface area normalized to pore volume (Spor) of granular material along the river corridor. We also estimate variations in the thickness of the aquifer contributing uranium to the river along the corridor. We use FO-DTS, performed along a line selected based on the images from CWEI, to determine spatiotemporal variability in surface water–groundwater exchange along the river corridor. The FO-DTS confirms the hydrological significance of the spatial variation in lithology imaged with CWEI. The study reveals a spatially and temporally complex pattern of surface water–groundwater exchange, strongly controlled by the thickness of transmissive sediments, with exchange focused at springs that are in some cases associated with paleochannels. Temporal fluctuations in the stage of the Columbia River drive, and may reverse, the flow of water between aquifer and river. Determining this spatial and temporal complexity is a critical requirement for advancing understanding of U transport at the Hanford 300 Area.
1.1. Surface Water–Groundwater Exchange at the Hanford 300 Area
 Hydrogeology at the Hanford site is in large part determined by two formations with distinctly different hydraulic properties (Figure 2). The uppermost unit is the Hanford Formation, containing pebble-to-boulder-size gravels and interbedded sands resulting in high hydraulic conductivity (K) of ∼100 m/d [Williams et al., 2007]. The underlying unit is the Ringold Formation, a highly heterogeneous unit of granule-to-cobble-size gravels interbedded with fine sand and silt resulting in a lower K of ∼0.2 m/d [Williams et al., 2007]. The Hanford-Ringold (H-R) contact is generally considered an important contact that limits vertical migration of contaminants. Identifying the location of the H-R contact is likely critical to determining U distribution along the river corridor as the shape of the confining layer probably regulates contaminant discharge along the shoreline [Fritz et al., 2007]. The topographic surface of the Ringold was modified by high-energy paleofloods and/or Columbia River flows that eroded into the Ringold Formation [Brown, 1960]. These incised channels, with a northwest trending erosional axis, were subsequently filled with younger, more permeable Hanford Formation sediments. Although the location and distribution of paleochannels across the 300 Area is currently poorly defined, paleochannels are expected to be approximately 10–20 m in width, and they likely constrain the U plume by providing hydraulically conductive groundwater flow paths through the region of the plume and into the Columbia River during low river stage conditions. An improved understanding of flow and transport along the river corridor of the Hanford 300 Area critically depends on mapping the locations and lateral extent of these paleochannels.
 The current hydrogeological framework for the river corridor at the Hanford 300 Area is largely based on direct probing techniques aimed at determining the elevation of the H-R contact, and projections of the H-R contact onto the riverbed from identification in boreholes drilled inland [Fritz et al., 2007]. This work supports the presence of a hydraulically resistant layer (the Ringold Formation) underlying the uppermost hydrologic unit (Hanford Formation) through which surface water–groundwater exchange and connectivity between the aquifer and surface water is most focused. Figure 1 includes an estimate of the area of the riverbed (170,000 m2) having the potential to discharge uranium to the Columbia River, outlined in yellow and termed the contributing area [Fritz et al., 2007]. This outlined area represents the area that was determined to be (1) between the average river stage elevation (105 m) and the thalweg (Figure 1), (2) between the 30 μg/L concentration contours for U, and (3) sediments originating from the Hanford Formation based on determination of the top of the Ringold Formation from projections of borehole data and drive point probes and multilevel sampling [Williams et al., 2007].
 Discrepancies between model predictions and field observations of U transport at the site can be attributed partly to the use of oversimplified conceptual models to describe reactive transport in a highly heterogeneous system. These models fail to consider not only subsurface heterogeneity, including the presence of possible preferential flow paths into the Columbia River and heterogeneity in and immediately beneath the streambed but also the effects of short-term and seasonal fluctuations in river stage and the resulting interplay between surface water and groundwater. Hourly and diurnal changes in water levels in the Columbia River of over 1 m are known to give rise to some of the fastest groundwater velocities on the Hanford site and the resulting pressure wave extends as far as 1 km inland [Waichler and Yabusaki, 2005]. These river stage fluctuations appear to correlate with episodic changes in water chemistry, increases in groundwater U concentrations, and elevated U discharges to the river, but cannot be reproduced with the current parameterization of the Subsurface Transport over Multiple Phases (STOMP) [White and Oorstrom, 2006] model used for simulation of transport at the site. Improved modeling of U transport from the aquifer into the river at the Hanford 300 site requires a more detailed characterization of lithologic heterogeneity along the river corridor, coupled with direct knowledge of spatiotemporal variability in surface water–groundwater exchange. Our combined CWEI and FO-DTS study was designed to demonstrate how emerging hydrogeophysical technologies can address such needs at the required measurement scale.
1.2. Survey Layout
 Our study focused on the Columbia River corridor adjacent to the Hanford 300 site (Figure 1), although the CWEI was extended to the south in an effort to place the lithologic variability within a broader context. Figure 1 shows the location of ∼30 km of CWEI survey lines and ∼1.5 km of nearshore FO-DTS cable that is roughly centered on the Hanford 300 Area Integrated Field Research Challenge (IFRC) site. Three additional FO-DTS cables were installed parallel to the nearshore cable shown in Figure 1 (not shown or discussed here). Water depths varied substantially across the surveyed area from a minimum of 2 m close to shore to a maximum of 18 m when crossing the channel thalweg (Figure 3). Divers were used to ensure that the FO-DTS cables were securely positioned on the riverbed, particularly in deeper parts of the channel. Cables were positioned so as to remain below water even at low stage.
1.3. Continuous Waterborne Electrical Imaging
 Electrical methods are based on the injection of current and mapping of the resulting electrical potentials in the earth using combinations of electrodes. CWEI is a recent advancement in electrical methods whereby a string of electrodes is pulled along the water surface while electrical measurements are continuously recorded [Day-Lewis et al., 2006]. This approach results in spatially rich data sets that can be inverted for the subriverbed (or any water body in general) distribution of electrical conductivity (σ = 1/ρ), the reciprocal of electrical resistivity (ρ), and also the polarizability of the subsurface as explored here. In saturated sediments, such as those beneath a riverbed, the electrical conductivity most commonly is modeled using an empirical model with parallel electrolytic (through the interconnected pore space, σel) and surface (along the interconnected pore surface, σsurf) conduction paths, σ = σel + σsurf = σwϕm + σsurf, where σw is the electrical conductivity of the pore-filling fluids, ϕ is the porosity and m is commonly known as the cementation factor and is determined by the connectivity of the pore space [Archie, 1942]. The surface conductivity (σsurf) has been shown to depend linearly on specific surface area normalized to pore volume (Spor) of granular material both in observations [Schon, 1996] and theory [Revil and Glover, 1998]. Here, Spor = S/Vϕ, where S is the sample total surface area, and V is the sample volume. Electrical methods have frequently been used to image lithologic variability in the subsurface, although the interpretation commonly is subject to considerable uncertainty due to the strong dependence on both properties of the fluids (σw) as well as the sediments (ϕ, m, and Spor). The measurements from electrical surveys are commonly reported as a transfer resistance (R), the voltage difference between two electrodes normalized by the current injected into the earth at the current electrodes. The measurements also commonly are presented as an apparent resistivity (ρa), calculated by multiplying R by a geometric factor determined by the relative positions of the four electrodes (units of length). The apparent resistivity is the resistivity of an equivalent homogeneous medium within the support volume of the measurement.
 A valuable extension of the electrical method involves the measurement of induced polarization (IP), which records the transient voltage decay associated with temporary storage of charge in the subsurface. These measurements are often obtained concurrent with resistance measurements. At the low frequencies (<10 Hz) used in electrical surveys, electrochemical polarization is the dominant charge storage mechanism, involving a local redistribution of charge within the electrical double layer at the solid-fluid interface [Slater and Lesmes, 2002a]. The magnitude of this polarization is in large part determined by Spor, an important parameter controlling flow and reactive transport, as it represents the size of the interconnected, polarizable surface [Slater, 2007]. A strong linear relation between imaginary conductivity (σ″), the strength of interfacial polarization recorded with precision frequency domain laboratory instruments, and Spor recently has been shown for 114 unconsolidated soil and sandstone samples derived from seven independent data sets [see Weller et al., 2010, Figure 2]. Field IP instruments, however, usually operate in the time domain and record proxy measures of σ″ such as the normalized chargeability (Mn) parameter [Lesmes and Frye, 2001]. The constant of proportionality between Mn and σ″ can be determined in the laboratory, where measurements on samples are made using a frequency domain dynamic signal analyzer that records σ″ directly, along with measurements of Mn made with the field instrument under identical configuration settings as used during the field survey.
 Unlike electrical conductivity measurements, IP measurements tend to show only a weak dependence on fluid properties (e.g., σw), at least in soils and rocks devoid of metallic minerals [Lesmes and Frye, 2001]. As with electrical conductivity imaging, IP measurements are usually reported as apparent values for a homogenous earth and inverse methods are used to reconstruct the distribution of the subsurface polarizability. Inverse methods now routinely are employed to determine the best estimate of subsurface variability in conductivity that satisfies the measurement and any data (resistance and/or IP) and model constraints [e.g., Binley and Kemna, 2005]. Inversions produce two- or three-dimensional cross sections or volumes of subsurface electrical properties, providing insight into spatially variable lithologic and fluid properties. As stated previously, geophysical parameters commonly are functions of multiple hydrologic properties; moreover, the resolution and quality of inversions is affected by survey geometry, measurement error, and inversion settings [Day-Lewis et al., 2005]. For these reasons, it is critical to interpret geophysical results jointly with other data types, such as direct hydrologic measurements and temperature data.
1.4. Fiber-Optic Distributed Temperature Sensing
 Fiber-optic distributed temperature sensing (FO-DTS) is an emerging technology with applications in fire detection, industrial process monitoring, and petroleum reservoir monitoring, with more recent implementation to obtain spatially rich data sets for monitoring surface water–groundwater exchange in streams, lakes, marshes, and estuaries [e.g., Henderson et al., 2009; Moffett et al., 2008; Selker et al., 2006a, 2006b; Tyler et al., 2009]. The sensor consists of standard telecommunications fiber-optic cable. FO-DTS measurement physics is based on temperature-dependent backscatter mechanisms including Brillouin and Raman backscatter [Selker et al., 2006a]. Most commercially available systems, including the unit used here, are based on analysis of Raman scatter. As laser light is transmitted down the fiber-optic cable, light scatters continuously back toward the instrument from all along the fiber, with some of the scattered light at frequencies above and below the frequency of incident light, i.e., anti-Stokes and Stokes-Raman backscatter, respectively. The ratio of anti-Stokes to Stokes energy provides the basis for FO-DTS measurements. Measurements are localized to a section of cable according to a time-of-flight calculation (i.e., optical time domain reflectometry). Assuming the speed of light within the fiber is constant, scatter collected over a specific time window corresponds to a specific spatial interval of the fiber. Although there are tradeoffs between spatial resolution, thermal precision, and sampling time, in practice it is possible to achieve meter-scale spatial and 0.1°C thermal precision for measurement cycle times on the order of minutes and cables extending several kilometers [Tyler et al., 2009]; thus, thousands of temperature measurements can be made simultaneously along a single cable. This new technology is ideal for generating spatially rich thermal maps that can visualize a large amount of temperature data and quickly identify major trends in surface water–groundwater exchange.
2. Field Data Sets
2.1. Continuous Waterborne Electrical Imaging
 Approximately 30 km of CWEI data were collected on the Columbia River (Figure 1, white lines). Data were collected in July 2008, when relatively high water levels facilitated surveying close to the western bank of the river. All data sets were acquired from a Gregor aluminum hull jet boat, with water depth continuously measured using a 200/50 kHz–10/40 degree depth finder (estimated accuracy of 25 cm based on field tests) and all measurements were georeferenced using a differential GPS (Garmin GPSmap 420s) unit with ±0.5 m horizontal accuracy. CWEI surveys were conducted using a 10-channel time domain resistivity/IP instrument (Syscal Pro, Iris Instruments, France). This time domain instrument records the polarizability as an apparent integral chargeability (Ma) determined from the decay curve after current shutoff. All data were acquired using a 13-electrode cable with graphite electrodes spaced at 5 m intervals. The configuration of the 10 measurement channels was chosen so as to (1) provide a high signal-to-noise ratio for all 10 channels (critical to collecting reliable IP data), (2) result in a desirable sensitivity pattern relative to other common configurations [Mansoor and Slater, 2007], and (3) optimize the 10-channel capabilities of the instrument. Measurements were recorded every 0.5–3.0 m depending on survey speed (in part dictated by strong currents on the Columbia River), resulting in ∼65,000 measurements conducted on ∼30 km of line.
 CWEI surveys (i.e., with the boat and cable in motion) do not permit application of standard methods for assessing measurement error (e.g., repeatability/stacking tests and/or reciprocity). However, surface electrical measurements are inherently low resolution and it is therefore reasonable to expect a smooth distribution of apparent resistivity and apparent chargeability over most natural geologic structures. Furthermore, it is justifiable to remove values inconsistent with the physics of electrical current flow. For example, although it is possible to record negative apparent chargeability in the presence of strong geometrical effects, such effects were not expected here. In the resistivity data set obvious outliers are those measurements with a very small resistivity close to 0 Ohm m, a physically unreasonable result given the soils at this site. These considerations resulted in ∼2% of the data set being removed from the inversion. Confidence in the mapped variability of ρa and Ma along a two-dimensional (2D) line was also established by comparison with ρa and Ma structure observed on adjacent parallel lines (Figure 1). Data quality was very good with only 1.2 percent of the data points removed as likely outliers before further processing. Figure 4 is an example of the distribution of ρa and Ma, with the data showing strong variability both perpendicular (E-W) and parallel (N-S) to the river bank. These plan views of the shore-perpendicular variability in part reflect the substantial increase in water depth toward the channel center (Figure 3). However, it is the shore-parallel variability that is of most interest here as we wish to determine variability of the hydrogeologic framework along the river corridor at the Hanford 300 Area.
 The CWEI data set was inverted for an estimated subsurface distribution of resistivity and chargeability using the RES2DINV package [Loke et al., 2003] as it efficiently handles large data sets and permits a variable thickness water layer of uniform resistivity and zero chargeability (water is nonpolarizable at low frequencies) to be incorporated as an inversion constraint. A uniform resistivity water layer is a valid assumption considering the fact that the survey was run during high stage conditions when we will show that focused groundwater discharge is suppressed such that the physical properties of the water column are likely to be relatively uniform. Measurements made with the shortest electrode separations, almost exclusively sensing the resistivity of the water column, support this assumption. For example, along line 20 m the mean apparent resistivity of the shortest electrode spacing was 99 Ohm m with a standard deviation of 17 Ohm m. Given the significant discrepancy between the sample density shore-parallel (1 measurement every few meters) and shore-perpendicular (1 measurement every ∼20 m), the data set was treated as a series of near-parallel 2D lines. True electrode locations (calculated from the boat location and the azimuth of the electrode cable behind the boat) were first projected onto a best fit 2D line. Each line was then inverted for a 2D electrical structure along the river corridor using the common smooth regularization constraint (whereby model structure is minimized subject to fitting the data to some acceptable tolerance) [de Groot-Hedlin and Constable, 1990]. These 2D inversion results were then interpolated in ArcView to generate a pseudo three-dimensional (3D) data set of ρ and M variation within the surveyed volume of the river corridor.
2.2. Fiber-Optic Distributed Temperature Sensing
 In November 2008, four armored fiber-optic measurement cables were installed semipermanently on the bed of the Columbia River centered on the Hanford 300 Area. Although a zigzag or grid pattern would be an optimal deployment and several bank-perpendicular transects would provide additional information, cables oriented cross flow would more likely snag floating debris and would be subjected to the full force of the current. Given the high currents associated with this river, we opted for a series of four parallel cables at different distances between 2 m and 20 m from the riverbank,. These cables were continuously below the water surface (even at the lowest stage during the data acquisition) throughout the length of the study reach. Here, we focus on analysis of data from a ruggedized, ∼1 cm diameter, 1.6 km long Sensornet EnviroFlex cable that was deployed 2 m from the bank by a team operating on foot in waders, with the cable manually secured to the streambed using stakes, cinder blocks, and native cobbles. GPS data were recorded at discrete locations along the cable (cable location is shown in Figure 1). The DTS unit was programmed to collect data every meter along the cable at a 5 min time interval. A Sensortran eight-channel Gemini control unit was set up and calibrated using ice-and-water baths as well as the stream and air temperatures. The system was remotely accessed with data automatically transferred to a remote server.
2.3. Other Geophysical Surveys
 Waterborne ground penetrating radar (GPR) and seismic methods were also performed along line 20 of the CWEI survey in the hope that they could be used to locate the H-R contact and/or identify paleochannels. Both methods were ineffective, primarily as a result of the excessive signal scattering resulting from the cobble framework making up the armored bed along most of the river. However, GPR surveys along the river corridor during periods of low stage have revealed evidence of paleochannels at some locations, as discussed later.
3. Interpretation and Discussion
3.1. CWEI Imaging of Lithologic Variability
Figure 5 shows the inversions of CWEI data for line 20 (located approximately 20 m from the western shore). The normalized chargeability (Mn = M/ρ), calculated from the inverted ρ and M model space, is plotted in Figure 5a and the ρ model is shown in Figure 5b. The electrical structure shown in Figure 5 is largely reproduced in the other shore-parallel transects (not shown for brevity), although image resolution below the riverbed is reduced due to greater water depths further offshore. We focus on Mn as a lithologic indicator because it is directly proportional to the imaginary conductivity (σ″), an excellent field-scale indicator of lithology [Slater and Lesmes, 2002b]. Furthermore, although lithology exerts a strong control on ρ due to variations in ϕ and clay content, ρ also is dependent on σw. Given that σw varies spatially and temporally in the river corridor as a result of surface water–groundwater exchange, ρ is not a robust indicator of lithologic variability. Figure 5 generally shows a two-layer model composed of a low Mn (high ρ) layer overlying a high Mn (low ρ) layer. The upper layer includes the variable thickness water layer (constrained in the inversion to a layer of constant ρ and zero M). The riverbed is shown as a black line. We attribute the lower, sub–river bottom, part of the upper layer to the Hanford Formation, where the very low Mn is consistent with a coarse-grained aquifer devoid of silt and clay. The lower layer is interpreted as the Ringold Formation, where the much higher Mn is consistent with the substantial silt fraction in this unit.
 The white dashed line in Figure 5a represents our interpretation of the H-R contact based on the two-layer electrical structure of the site. As noted previously, the H-R contact is a critical boundary in any hydrogeologic model of the Hanford 300 Area as it limits vertical movement of contaminants in the overlying permeable Hanford unit. CWEI provides an unprecedented characterization of the hydrogeologic setting along this 2.5 km reach of the river corridor at the 300 site, revealing substantial variability in the depth to the H-R contact. In some places, the Ringold appears to be in contact with the riverbed, e.g., between ∼1300–1550 m (Figure 5a), which is consistent with estimates of the Ringold exposure on the riverbed based on projection of the H-R contact from inland boreholes combined with point probe tests performed on the riverbed (yellow outline in Figure 1). However, CWEI captures many details in the lithologic variability along the river corridor not resolvable from the projections based on boreholes and direct probing techniques. For example, the H-R contact is imaged at 5–10 m below the riverbed in some places, and evidence for paleochannels of coarse sediments incised beneath the H-R contact exists at 100–300 m and 750–900 m along this line. Furthermore, the imaging identifies a second location where the Ringold is in contact with the riverbed between 600 and 700 m that was not resolved from the borehole projections.
Figure 6 shows a plan view spatial distribution of ρ (Figure 6a) and Mn (Figure 6b) at 7 m depth along the river corridor; the estimated contributing area is superimposed on the map (white outline) [Fritz et al., 2007]. Figures 6a and 6b were constructed from interpolating the 2D inversion of the CWEI data sets collected along the multiple shore-parallel lines shown in Figure 1. The inversion slice at 7 m depth was selected as the shallowest (therefore highest resolution) depth that represents the subriverbed sediments across the nearshore before reaching the channel thalweg where depths rapidly drop to as much as 15 m (Figure 3). The plan view images in Figure 6 are cropped east of the edge of the thalweg as they essentially represent only the water layer and contain no useful information. Figure 6 again illustrates the spatial richness of the information that is retrieved from CWEI on the variation in the H-R contact along the river corridor. Figure 6c is a plan view of the variation in Spor at 7 m estimated from the petrophysical relation σ″ = 0.01Spor (σ″ in mS/m, Spor in μm−1) derived by Weller et al. , with Mn converted to σ″ based on laboratory calibration of the linear relations between σ″ recorded with a frequency dynamic signal analyzer and Mn recorded with the instrument used in this study [Slater and Lesmes, 2002b]. Figure 6d shows the variation in Spor estimated at 5 m using the same approach. The region of highest Spor coincides as expected with the location where the borehole projections suggest that the contributing area is narrowest, i.e., little or no exposure of the Hanford Formation, the riverbed instead being composed of Ringold sediments at this location. However, as demonstrated in Figure 5, we capture much spatial variability in the depth to the H-R contact along the river corridor that cannot be defined from existing borehole projections and direct probe sampling. We argue that variation in Spor shown in Figure 6c represents the variation in the distribution of Hanford (low Spor) and Ringold (high Spor) Formations at 7 m depth throughout the contributing area defined by Fritz et al. .
3.2. FO-DTS Investigations of Surface Water–Groundwater Exchange
 The FO-DTS monitoring recorded the spatiotemporal variability in streambed temperature at 5 min intervals from February 2009 to September 2009. Figure 7 shows river temperature, stage, and two subsets of the FO-DTS data collected in March and July–August 2009. Figure 7 shows substantial spatial variability along the cable, with some cable locations exhibiting greater temperature variability than others. Spatial anomalies in Figure 7 appear as horizontal bands, locations where warmer temperatures occur in winter and cooler temperatures occur in summer; these clearly diminish at higher river stages. Winter increases and summer decreases in temperature during low stage are consistent with the effect of groundwater input. During winter months, groundwater temperatures are warmer than surface water temperatures, and groundwater discharges locally increase streambed temperatures. The converse is true during summer months, when groundwater input cools the streambed. We note that Figure 7 contains evidence of a phase lag between stage and temperature response. The hydrological significance of such phase lags between temperature and forcing factors can be investigated using time-frequency analysis [Henderson et al., 2009]. However, this is beyond the scope of our current paper.
 Other factors that could cause temperature anomalies at the streambed include air temperature and solar heating. We dismiss the effects of air temperature (via conduction through the water column), because it would cause temperatures to decrease at low stage in winter and increase at low stage in summer, the opposite of what we observe (Figure 7). Solar heating could cause spatial variability in temperature if substantial variations in river depth occur along the cable. Spot measurements of stream depth, however, do not support a significant contribution from solar heating, and the anomalies observed in Figure 7 persist throughout the night when solar heating is absent. We are therefore confident that anomalies observed in Figure 7 identify locations of enhanced surface water–groundwater exchange adjacent to the 300 Area.
Figure 8 compares the distribution of the average streambed temperature during low-stage conditions on 31 March (representing winter; Figure 8b) and 3 August (representing summer; Figure 8c) with a revised estimate (relative to Figure 1) of the U-contributing area along the river corridor based on the CWEI survey. The correlation coefficients between temperature and stage for data recorded at 5 min intervals over 8 and 12 day periods in winter and summer, respectively, also are shown in Figures 8b and 8c. Whereas the contributing area defined by Fritz et al.  was based on predicted locations where the Hanford Formation is exposed on the riverbed, we utilize the CWEI data to determine variations in the thickness of the Hanford Formation along the river corridor (Figure 8a). This CWEI estimate contains much structure along the corridor that is not captured in the original estimate (Fritz et al.  and Figure 1) based on projections of the H-R contact recorded in boreholes.
 The DTS data set highlights focused areas along the river corridor that are anomalously warm during cold months and anomalously cold during warm months. Furthermore, at these locations the temperature is strongly correlated with river stage, with negative correlation in winter (i.e., higher temperatures as stage falls) and positive correlation (i.e., lower temperatures as stage falls) in summer. In contrast, away from these anomalies the temperatures on the cable are relatively uniform and uncorrelated with river stage. In summer, weak negative correlation (up to about −0.3) between stage and DTS temperature is evident at some locations, which we interpret as the result of cooler deeper water; thus, the effect of water column temperature acts opposite to that of groundwater discharge, supporting our interpretation that strong positive correlation in summer is evidence of surface water–groundwater interaction. The reversal of the anomalies between summer and winter months, along with the strong correlation with river stage at these locations, offers compelling evidence that these are stage-controlled regions of focused surface water–groundwater exchange. Whereas temperature is a function of many factors (air temperature, solar radiation), we are only interested in the temperature variations arising from hydraulic forcing. Our use of the correlation coefficient between stage and temperature offers a simple analysis of the extensive DTS data set bringing us closer to understanding the processes controlling the discharges identified.
 The temperature anomalies appear to be correlated with lithology. Considering the CWEI estimate of thickness of the U contributing area (Figure 8a), we see that the temperature anomalies indicating enhanced groundwater exchange coincide with locations along the river corridor where the Hanford sediments are thickest and the Ringold Formation is at depth. These anomalies are closely associated with five known springs/U seeps [Williams et al., 2007], shown as black crosses in Figure 8. In contrast, temperature anomalies are absent where the Ringold is in contact with or close to the riverbed and no correlation between streambed temperature and stage is observed. The linear correlation coefficient for Hanford thickness as a predictor of temperature-stage correlation is −0.6 and the corresponding Spearman rank correlation coefficient (more appropriate as we have no expectation that the relationship will necessarily be linear) is −0.68. The DTS data sets therefore clearly demonstrate that the lithologic variability imaged with waterborne geophysics is hydrologically important and plays a key role in regulating surface water–groundwater exchange along the river corridor at the Hanford 300 Area.
 The pattern that emerges from the integrated interpretation of data from CWEI and FO-DTS, two emerging technologies in environmental imaging and monitoring, is that of focused surface water–groundwater exchange in locations where the Hanford sediments are thickest and the elevation of the H-R contact is locally depressed by erosion features. As discussed earlier, highly permeable paleochannels are suspected to exist at the contact between the Hanford and Ringold units and likely provide localized preferential flow paths for enhancing surface water–groundwater exchange. Our data lend support to this conceptual model of hydraulic connectivity between the aquifer and river. For example, Figure 8a suggests abrupt depressions in the H-R contact at ∼250 and 800 m along line 20. The cross-sectional expressions of these features are also shown in Figure 5a. We suggest that these are the electrical expressions of coarse paleochannel deposits locally eroded below the H-R contact. The locations of these feature coincides with enhanced surface water–groundwater exchange predicted from the FO-DTS data set (Figure 5).
 Ground penetrating radar (GPR) geophysical surveys were conducted along the westernmost edge of the Columbia River (where field conditions permitted) to look for additional evidence for the presence of paleochannels; anomalies identified in vintage GPR data sets collected to the west of the 300 Area were previously interpreted as paleochannel structures [Kunk and Narbutovskih, 1993]. The GPR survey on the streambed revealed the presence of a coarse sedimentary deposit locally eroded below the H-R contact at the location of the onshore projection of the temperature anomaly at 60 m on line 20 (Figure 8b), with a cross-sectional structure consistent with a paleochannel (Figure 9). The correspondence between the FO-DTS and shoreline GPR data sets provides evidence of paleochannel(s) locally enhancing surface water–groundwater exchange between the U-contaminated Hanford aquifer and the Columbia River at this location. The CWEI and DTS data also support the presence of enhanced exchange focused in depressions of the H-R contact at additional locations along the river corridor at the Hanford 300 Area, most obviously between 800 and 950 m in Figure 5. We note the obvious correspondence with the high contours of U concentration at this location (Figure 8b), lending support to the notion that the focused surface water–groundwater exchange observed with FO-DTS is controlling U transport into the river.
 Our study illustrates the unique spatially rich information on surface water–groundwater exchange at the Hanford 300 Area that comes from integrating CWEI with FO-DTS measurements. These two technologies provide highly complementary information. CWEI images the hydrogeologic framework, primarily the elevation of the H-R contact and the location of possible paleochannel features. In contrast, FO-DTS monitoring identifies areas where surface water–groundwater exchange is focused. The picture that emerges is distinctly different from that obtained from previous studies based on point observations at boreholes drilled at the 300 Area, along with probing and sampling on the riverbed. Fritz et al.  noted that their predicted U-contributing area was likely overestimated. Our study shows that there is much spatial variability in elevation of the H-R contact within this estimated contributing area such that exchange likely is enhanced where the Hanford Formation is thickest, and paleochannels likely are present (Figure 5). Our estimates of regions of focused exchange based on these temperature anomalies coincide with the location of U seeps identified in previous studies. However, the DTS data sets suggest the presence of many additional locations of focused exchange that have not hitherto been recognized from direct sampling.
 We have developed a CWEI- and FO-DTS-based interpretation of hydraulic connectivity between the Hanford aquifer and Columbia River that provides a new framework in which to interpret the complex spatial pattern of U concentration contours developed in past work at the Hanford 300 Area. Our hydrogeologic characterization of this important river corridor provides information that can be used to improve the modeling of the transport of U toward the river. Current flow and transport models for the site assume a limited variation in the elevation of the H-R contact along the river corridor and a spatially invariant solute flux that is clearly not supported by our data set. Furthermore, the identification of zones of focused discharge has implications for the geochemical and microbiological sampling and monitoring efforts directed at understanding the long-term impact of U discharge on water quality in the Columbia River. Ongoing FO-DTS monitoring has potential to be used to identify the timing of groundwater discharge, thus providing critical new information for the design of such sampling campaigns and subsequent estimation of contaminant loading to the Columbia River.
 This research was supported by the Office of Science (BER), U.S. Department of Energy, under the Environmental Remediation Sciences Program grant DE-AI02-08ER64565. Additional funding was provided by the U.S. Geological Survey Toxic Substances Hydrology Program. We thank Eric White (USGS), Chris Curran (USGS), and Jay Nolan (Rutgers-Newark) for valuable assistance in the collection of field data sets. We thank Rory Henderson (USGS) and Brad Fritz (Pacific Northwest National Laboratory) for reviews of our draft manuscript. We are also grateful to Brad Fritz, Bob Peterson, and John Zachara (all at Pacific Northwest National Laboratory) for their insights into the hydrogeologic framework of the Hanford 300 Area. Any use of trade, product, or firm names is for descriptive purposes only and does not imply endorsement by the U.S. government.