Assessing the Effects of Fluctuating River Stage on an MTBE Containment Plume Discharging to Surface Water, Pascoag, RI

Ground-water/surface-water interactions and MTBE contaminant plume discharge were investigated in a low-order river that experiences episodic river stage fluctuations. Results show that the hydraulic gradient fluctuates hourly to monthly due to river stage changes, water table recharge events, and reservoir gate adjustments. Hyporheic exchange driven by channel morphology creates small-scale gaining, losing, and parallel flow systems along the mostly gaining reach. During precipitation events, infiltrating rainfall rapidly saturates the extended capillary fringe and the shallow floodplain water table rises forming a ground-water ridge or mound and causing a steepened hydraulic gradient towards the river. The system response is magnified by watershed characteristics which control the river stage hydrograph including stormflow lag and flashiness. Results of this study suggest that a ground-water plume discharging to surface water may have several discharge locations related to transient water-table configurations. Under conditions of a low hydraulic gradient, the MTBE plume is deflected away from the river by hyporheic flow toward a downstream discharge location. When the gradient toward the river steepens in response to precipitation and gate closure, the small-scale hyporheic exchange systems are overcome and the plume discharges along the entire reach. Under these conditions, a high influx of contaminated ground water is discharged from the floodplain to the river, temporarily elevating river contaminant concentrations. During site investigation and monitoring, these transient spatial and temporal relationships could easily be missed by traditional site monitoring strategies.


Background
Rising population and urban development have placed increasing pressure on the availability and quality of ground-water and surface-water resources (Einarson and Mackay 2001;Winter et al. 1998). With escalating concern and knowledge that ground-water and surface-water systems are intrinsically linked, current research is focusing on improving the conceptual model of ground-water/surface-water (GW/SW) interactions. Model accuracy is of even greater importance where contaminated ground water discharges to surface water. The EPA estimates that 75% of Superfund and RCRA sites are located within a half mile of a surface-water body and nearly half have impacted surface-water quality . How these ground-water plumes interact with surface water, both spatially and temporally, is therefore of paramount interest.
Study of GW/SW interaction crosses the boundaries of hydrology, biology, geomorphology, and aquatic chemistry. Research has sought to increase understanding in the contexts of water resources management (riverbank filtration and stream depletion), stream and riparian health (nutrient cycling and benthic biota), flood modeling (bank storage and floodplain hydrology), and ground-water contamination (point and non-point source). Research on these topics has focused primarily on the hydraulic relationships between the coupled systems. For example, early researchers developed the conceptual framework for the hydraulics of flow, transport, and exchange across the GW/SW interface. Subsequent researchers have modified and expanded these concepts with studies that have helped explain the controls on the hydraulic gradient. Current understanding states that GW/SW interaction is controlled by the distribution and magnitude of hydraulic conductivity, the relation of stream stage to the adjacent ground-water gradient, and the geometry and position of the channel within the floodplain (Woessner 2000). In GW/SW systems (i.e. , rivers and adjacent aquifers) the following hydraulic relationships are possible: 1) ground water can enter the channel (gaining), 2) ground water can exit the channel (losing), 3) ground water can travel parallel to the channel (parallel-flow), or 4) ground water can simultaneously enter and exit the channel (flow-through).
Early work through the 1980s on GW/SW interaction was conducted broadly on the topic, including quantifying stream depletion due to irrigation well pumping (Sophocleus et al. 1988) and early modeling of flood-induced bank storage (Gill 1985). As understanding increased and the practical importance of this coupled system was realized, studies on floodplain hydraulics grew more specialized and began to include stormflow and riparian processes.
Most conceptual models of stormflow events for gaining rivers show that as river stage rises, the hydraulic gradient adjacent to the channel reverses from gaining to losing and river water is driven into the aquifer. This process, termed bank storage, has been shown to help to attenuate stormflow ).
Following the passage of peak flow, the gradient reverses back to gaining and the stored water gradually returns to the river channel as return-flow. Others have shown that the shape of the flood hydro graph, geometry of the channel and floodplain, and aquifer properties can affect bank storage and return flow (Chen and Chen 2003 ;Girard et al. 2003 ;Hantush 2005;. A suite of other factors have also been identified that affect the degree and timing of the hydrologic response in the floodplain. For example, floodplain recharge by precipitation and surface runoff was studied by  and shown to increase return flow discharge by steepening the hydraulic gradient towards the river. Zhang and Schilling (2006) showed that vegetation can affect shallow floodplain water tables and stream flow by controlling soil moisture. The degree of forestation has also been shown to be a significant factor controlling large-scale watershed recharge and stream flow . High floodplain water tables can also affect the response to recharge and overbank inundation by increasing antecedent soil moisture Girard et al. 2003). In addition, stormflow and bank storage is also a concern at riverbank filtration sites due to the threat of surface water contamination and gradient alteration .
Besides the overall GW /SW hydraulic relationship within a river reach, smallscale variations that govern the hydraulics beneath river channels are driven by hyporheic exchange processes.. The hyporheic zone is broadly defined as the area beneath and adjacent to a river channel that contains some proportion of surface water and ground water Woessner 2000). Research has shown that the hyporheic zone plays a key role in ecosystems due to the many physical, geochemical, and biological processes occurring therein which often control stream and riparian health (Castro and Hornberger 1991;Hayashi and Rosenberry 2002;Hunt et al. 2006; Wroblicky et al.1998). These processes are sensitive to hyporheic exchange variability which can create spatially limited gaining, losing, and parallel-flow river conditions within a single reach ( Figure   1) Woessner 2000).
The influence of channel morphological features on the development, location, and characteristics of the hyporheic zone has received considerable attention. Studies have demonstrated that channel bedforms, pool-and-riffle sequences, meander bends, changes in slope, and debris dams play a considerable role in determining the presence and type ofhyporheic exchange (Boano et al. 2006;Cardenas et al. 2004;Harvey and Bencala 1993;. Conant Jr. (2004) showed that streambed heterogeneity can also result in spatially-limited discharge zones responsible for the majority of reach ground-water discharge along a reach. The dominant hyporheic control acting on a stream is related to stream gradient, sinuosity, and other geomorphic factors, which also control the path lengths and residence time of hyporheic exchange.  showed that compared to fifthorder rivers, exchange in second-order rivers tends to have shorter residence time due to the abundance of pool-and-riffle sequences and debris dams. Others have shown that hyporheic exchange in low-order rivers is highly transient and can either appear, contract, expand, or relocate with small modifications to the dynamic system (Hunt et al. 2006;Winter et al. 1998;Wroblicky et al. 1998). While higher order rivers do exhibit hyporheic variability, flow path lengths and residence times are considerably longer .
Surface-water discharge is commonly the ultimate fate of ground water contaminated by point or non-point sources. The transport and behavior of these contaminants across the GW/SW interface includes multiple dynamic processes that control the concentration, distribution, and location of discharge. Research has shown that riparian and hyporheic zone processes play an important role in controlling the quality of this water prior to discharge. Non-point source contamination, such as nitrate introduced through regional fertilizer usage, can be largely removed from ground water with a healthy riparian buffer zone (Hayashi and Rosenberry 2002).
However, alteration of the vegetation can impact the water table and can be detrimental to nutrient uptake and cycling .
A pan-European study by BW1 et al. (2002) identified geomorphic and climate factors as important processes in controlling riparian zone water tables, hydraulic gradients, and the position of the GW/SW interface.
Typical point-sources of contamination include landfills, leaking underground storage tanks (LUST), military bases, and industrial facilities with contaminants varying from metals to volatile and semi-volatile organic contaminants (VOCs and SVOCs). Major point sources of ground-water contamination are commonly located adjacent to surface water, presumably due to current or former usage for transportation and power . It is therefore surprising that there has been relatively little research investigating the behavior of ground-water contaminant plumes discharging to surface water.
In a GW/SW exchange system, the hydraulic gradient has the strongest influence on the ultimate discharge location relative to the source zone. For example , Hinzman et al. (2000) showed that increased river stage and bank storage can affect plume discharge and also natural attenuation due to dispersion, dilution, and smearing. They also suggest natural discharge as a remediation alternative due to potentially high contaminant mass discharged from the ground-water system to the surface-water system. Others have shown that fluctuating river stage has the potential to affect contaminant discharge rates and the position of the GW/SW interface .
In what the authors described as the first assessment of a perchloroethylene (PCE) plume discharging to surface water, Conant Jr. et al. (2004) explained observations by pointing to heterogeneous hydraulic conductivity, organic carbon sorption, biodegradation, and variations.in source zone contributions. Their study provided a snapshot of contaminant distributions and hydraulic relationships rather than a transient analysis. Fryar et al. (2000) included transient discharge in a study of a VOC contaminant plume entering a river. They identified temporary reversals in hydraulic gradient and riverbed discharge associated with local storms, flooding, and dry periods that altered contaminant discharge and shifted the discharge location of the plume.
They also identified increased return-flow ground-water discharge following flooding, possibly associated with increasing voe concentrations in the river.
These studies demonstrate the complex transient relationships affecting discharge of contaminated ground water to surface water. Currently, no clear conceptual model exists for describing mechanisms and spatial and temporal variability of ground-water plumes discharging to rivers. This study will therefore address this need by examining the conceptual model of transient GW/SW hydraulic relationships and their resulting affects on discharging plumes.

Objective
This thesis examines the interaction of a gasoline contaminated ground-water plume with the Pascoag River, a low-order river that experiences episodic river stage fluctuations due to controlled reservoir discharge and stormflow. My hypothesis was that ground-water and surface-water hydraulic relationships change in response to river stage fluctuations resulting in spatial and temporal alterations to the fate and transport of the discharging contaminant plume. Specifically, spatial and temporal changes in the hydraulic gradient and plume discharge in relation to river stage were investigated. These changes alter the dynamic equilibrium that occurs between the ground-water and surface-water systems and the established location and groundwater flux into the river channel. Changes in degradation rates, sorption, and the solubility of gasoline compounds were not specifically addressed in this study. In an effort to capture plume transience, data collection incorporated high resolution water  Figure 3. Conceptual cross-section of site looking north-northeast. Shows the general surface and bedrock topography based on refusal depths and geophysics. Contaminants flow through bedrock fractures toward the floodplain and are believed to discharge at the base of the stratified sand, silt, and gravel aquifer. The water table in the floodplain is shallow and frequent ponding occurs on the surface. elevation ranges from 111 m above sea level in the study area to 122 m at the source zone ( Figure 3). The topography of the bedrock surface varies significantly across the aquifer and controls local surface topography. Bedrock depth ranges from less than 3 mat the source zone to 8 m below ground surface at the study site, deepening toward the river.
The bedrock is augen granite gneiss with predominant fractures striking nearly northsouth with an average dip of 65° E and a second set striking N 75°W with a dip of 75° s . A complete description of bedrock lithology and fractures is provided in Appendix I.

Surficial Materials
The thickness of the stratified sand and gravel aquifer at the site varies from 3 mat the source zone to 8 min the study area, increasing towards the river (Figure 3). The aquifer is believed to be glacial deltaic in origin and contains heterogeneous sediments ranging from silt to sand and gravel (Allen unpublished report 2005). Based on sediment borings from adjacent to the study area, the base of the 8 m unconsolidated aquifer is dense till overlain by silt and well-sorted very fine sand ( Figure 3). This is followed by poorly sorted sand and capped by sand and gravel. The soil is mapped as Canton and Charlton extremely stony fine sandy loam and has moderately rapid permeability and moderate water capacity . The soil horizons may be significantly disturbed and some areas may be covered with fill material related to a former textile factory. H. istorical photographs (Appendix I) show the factory and suggest that building foundation debris and a buried channel remain in the study area.
Due to the heterogeneous nature of the aquifer, there is large range in hydraulic conductivity (K) of the aquifer material. Bouwer-Rice falling head slug tests for the aquifer indicate a K range of 10-2 to 10-5 cm/sec (RIDEM unpublished data) with a somewhat narrower range of 10-3 to 10 4 cm/sec on the study site along the river as determined in this thesis (Appendix III). This range is consistent with fine to coarse sands, with the degree of sorting having a significant impact on the K value.

Surface-Water Hydrology
The Pascoag River is a low-order river in the Clear River Subbasin of the Blackstone River Basin . The river begins at the Pascoag Reservoir and flows 1.37 km through Pascoag center prior to the confluence with the Clear River (Table 1 ). The watershed above the study site is 22.6 km 2 and is primarily forested ( Figure 4). A water-powered factory once existed at the study site but was destroyed in the 1960s. In order to maximize river power, the river channel was straightened, deepened, and lined with stone to create a race. A small retention pond was located behind a small earthen dam with a spillway that discharged water through the modified channel (Figure 2). At the study site, the channel is 3 to 4 m wide, between 2 and 3 m deep, and has a gentle gradient (0.002 m/m) below the earthen dam. The riverbed contains large amounts of organic and manmade debris, including piles of bricks and portions of collapsed channel. Fine-grained sediment and larger rounded clasts are as not common in this channel as in typical river channels.
Pascoag Reservoir (also known as Echo Lake) has a total area of 1.41 km 2 , and average and maximum depths of 3.2 and 5.8 m, respectively ( Figure 4) (Plouffe pers. comm. 2006). The gate is usually opened in October to drain the reservoir and closed in March to raise it approximately 1.5 m to its summer level.
Additional gate adjustments are made in response to individual storm events.
Discharge from upper Pascoag Reservoir also occurs over an overflow spillway that becomes active when the reservoir level reaches 2.8 m (above an arbitrary datum).
When the gate is closed, the watershed area of the Pascoag River above the study site is reduced from 22.6 km 2 to 0.56 km 2 with high residential landuse and a considerable amount of impervious surfaces ( Figure 4).
Discharge in the Pascoag River is therefore highly variable due to episodic gate adjustments and rapid response to stormflow. Sustained discharge in the river ranges from near zero in the summer to over 1.5 m 3 /sec during the winter, with stormflow discharges of up to 7.0 m 3 /sec. River stage rise and stormflow is typically confined to the steeply walled channel; however, overbank flooding can occur. Stormflow discharge is flashy with peak Md recession occurring rapidly. Changes in discharge related to gate adjustments at Pascoag Reservoir also occur rapidly and can increase or decrease river stage by more than 0.5 m in minutes. Prior work has also shown that the river channel is hydraulically well-connected to the aquifer with river stage driven water table fluctuations propagating through the aquifer within minutes to a distance of at least 30 m from the channel (Allen and Boving 2006).

Hydrogeology
The hydraulic gradient is relatively steep between the source zone and the floodplain due to steeper surface and bedrock topography, with gradients of approximately 0.06 m/m from the source zone to the floodplain and 0.01 m/m or less within the floodplain. There are no nested multi-level wells in the source zone to identify vertical gradients; however, a downward gradient is assumed. Ground-water flow to the floodplain occurs through bedrock fractures and through the sand and gravel aquifer. Limited historical data from nested multi-level wells in the floodplain indicate an upward vertical hydraulic gradient steepening closer to the river. The water table depth is generally between 1 and 2 m below ground surface throughout the floodplain; however, during some periods it may be less than 1 m below the surface ( Figure 3).

Ground-water/Surface-Water Interaction
No prior studies have been conducted at the site to determine if the river reach is gaining or losing. Attempts to determine verticar gradients using riverbed piezometers were unsuccessful due to loss during high discharge in the winter or by tampering.
Given the presence of contaminant discharge in sections of the river reach and a general model of gaining rivers in the northeast, it is believed that this river reach is predominantly gaining RIDEM unpublished data;Winter et al. 1998).

Contaminant Fate and Transport
The current distribution of contaminants outside of the source zone has been significantly affected by induced ground-water flow from pumping of the former supply well ( Figure 5). It is believed that this has created a secondary contaminant source zone within bedrock fractures and is providing a significant proportion of current contamination . As a result, contaminated ground water may flow between the source zone and the floodplain by traveling northward through bedrock fractures until discharging into the adjacent sand and gravel aquifer at depth. A second possible flowpath suggests that the contaminant plume discharges from the bedrock and migrates towards the river closer to the source zone and then follows the buried channel introduced above. Regardless of the exact ground-water flowpath, bedrock fractures, former channels, topography, and the river combine to control ground-water flow and contaminant transport to the river.
Initially, the contaminant plume, consisting ofMTBE, benzene, toluene, ethyl benzene, xylene (collectively termed BTEX) and other gasoline compounds, extended north-northeast from the source zone in both the bedrock and the overlying unconsolidated sand and gravel aquifer. Maximum dissolved phase MTBE concentrations were over 1,000 mg/Land low-level contamination of 0.04 mg/L extended over an area of 80,000 m 2 (RIDEM unpublished data). After the supply well was shut down, the water table returned to a natural gradient consistent with local topography and surface-water hydrology, shifting the contaminant plume orientation to a north-northwest flowpath. As a consequence, the plume began to discharge to the Pascoag River which flows north along the western extent of the site. Onsite remediation has significantly reduced contaminant concentrations in the source zone; however, the plume continues to discharge to the river and impact surface-water quality along this reach. The early 2006 distribution of MTBE in the floodplain from ground-water monitoring is shown in Figure 5. Source zone MTBE concentrations have been reduced to less than 100 µg/L, while concentrations near the river remain close to 10,000 µg/L (RID EM unpublished data). BTEX concentrations still exceed 2,000 µg/L for individual compounds in the source zone and along the river (RIDEM unpublished data). As described by , multilevel wells and contaminant distributions in the study area indicate that contaminants are discharged from the sloping bedrock into the adjacent sand and gravel aquifer ( Figure 3).
Significant discharge of contaminated ground water to the river occurs downstream of the study site near MW 48 in a short stretch of channel that has been partially filled in and is not part of the active river channel ( Figure 5). Discharge of contaminated ground water is evident by the presence of gasoline odor, orange biofilm, gasoline sheen on the surface water, and MTBE surface-water concentrations up to 2,000 µg/L (RID EM unpublished data). A second area of discharge is located in the main river channel near 3D upstream from the study site closer to the source zone ( Figure 5). Discharge here was identified by an MTBE concentration in the riverbed of up to 2,000 µg/L which is diluted rapidly in the river channel resulting in a low river concentration (RIDEM unpublished data).

METHODS
Research methods and wellfield design were chosen in order to fulfill several obj ectives including, I) to identify spatial and temporal changes in the vertical and horizontal gradients between individual wells and between the aquifer and the river, 2) to identify spatial and temporal changes in MTBE and BTEX concentrations, and 3) to identify and correlate the above observations with river stage fluctuations. In order to accomplish the goals presented above, a well transect was installed perpendicular to the river approximately 550 m downgradient from the source zone.
The wells monitored included five preexisting monitoring wells and I2 wells installed specifically for this project (Table 2 and Figure 6). Three surface water locations were also monitored to determine inflowing river chemistry upstream of 2D, downstream at MW 49, and at JA IR. Except for JA IRB, JA IR, and the riverbed wells which were installed by the hand, all wells were installed by RIDEM using a Geoprobe®. Well Wells were sampled approximately every two weeks unless specific events such as precipitation or gate events warranted more frequent sampling. Periods of less frequent sampling also occurred due to relatively stable conditions (i.e. dry periods).
Continuous measurements of hydraulic head and temperature were collected in several wells for the entire or part of study period. The following is a summary of the field data acquisition methods. Additional information regarding sampling, field instruments, slug tests, and river discharge is available in Appendix II.

Sampling and Field Parameters
Well purging and sampling was performed using a peristaltic pump utilizing EPA low-flow ground-water sampling principles when possible .
In order to maintain pump circulation, the lowest pump rate used was 150 to 250 mL/min, depending on the well depth. Initially wells were purged until the field parameters, dissolved oxygen (DO), pH, electrical conductivity (EC), and temperature stabilized. After several months, the time required for parameter stabilization grew longer probably due to aquifer stratification associated with precipitation, or possibly downward flow along the casing. The method, therefore, was modified to include the monitoring of pumping time to establish a consistent ground-water contribution zone for each sampling events.
Field measurements of DO, pH, EC and temperature were recorded after well purging and satisfactory stabilization. Field "instruments were calibrated and operated according to manufacturers specifications. voe samples were taken by slowly filling duplicate 40 mL VOA vials. The vials were preserved with four drops of 6N hydrochloric acid with zero headspace and stored at approximately 4°C until analysis.
Ground-water and surface-water samples for ion analysis were collected in 125 mL HDPE bottles. The unpreserved samples were stored at 4°C prior to filtering and analysis.

Water Table Measurements
Continuous measurements were collected with In-situ® dataloggers installed in five wells. For quality control purposes and in order to convert relative elevation changes to water table elevations, manual measurements were collected when dataloggers were installed in wells and when they were removed.

Temperature Measurements
Both manual and continuous measurements were also collected to monitor temperature variation in each well. Manual measurements were collected during each sampling event while continuous measurements were collected by datalogger temperature sensors and by Thermochron iButtons® (Dallas Semiconductor). The iButtons were placed in small plastic bags, attached to pump tubing or cable, and placed along the screened interval.

Volatile Organics Analysis
Volatile organic analysis (VOA) samples were stored at 4°C and allowed to equilibrate to room temperature prior to preparation and analysis by a purgeable volatiles method similar to EPA Method 624. Most samples were analyzed within the EPA specified holding time of 14 days; however, several samples were analyzed outside of this holding time but within 18 days. It is not believed that this significantly affected the results.
VOA sample introduction was performed with a 01Analytical4660 purge and trap and analysis completed using a Shimadzu GC-17 A gas chromatograph equipped with QP5000 mass spectrometer (GC/MS). A 6-point external calibration of target compounds (Table 3) was performed from 2 to 160 µg/L and samples were spiked

Inorganic Analysis
Samples for anion and cations were stored at 4°C and filtered with a 0.45 µm syringe filter prior to analysis. Anion samples were not analyzed within the Method 300.0 recommended holding time of 28 days or 48 hours (for nitrate and phosphate). The holding time for cations is 6 months and was fulfilled by a limited number of analyses.
Both anions and cations were analyzed on a Dionex DX-120 ion chromatograph.

Precipitation and Temperature
Between January 21 and August 9, 2006 a total of 81.3 cm of precipitation was measured over several extended periods ( Figure 7). Refer to Appendix I for local historical weather and for measurement station locations. The average monthly temperature was similar to historical averages with a low in February (-2°C) and the high in July (23°C). Daily temperature ranges were dependent upon precipitation, with low diurnal variation occurring during storms.

Reservoir Dischar ge and River Stage
Pascoag Reservoir water level and reservoir discharge to Pascoag River varied significantly related to gate adjustments driven by precipitation and reservoir levels ( Figure 8). In response to these adjustments, river stage varied by approximately 1.0 m for the Pascoag River at the study site ( Figure 9). Gate adjustments at Pascoag ...,. ...,.
;:::: Figure 9. River stage hydrograph and controlling factors, precipitation a nd gate opening. Hydrogra ph correlation with reservoir gate opening confirms that t his is the primary factor controlling r iver stage (A & F). During per iods where the gate is closed and river stage is low (B), fluctuations, alt hough infrequent a re only caused by individual precipitation events and river stage effects are short-term (C). The influence of the reservoir spillway (D) results in a slow increase or decrease in river stage. When the river stage and precipitation are high, bankfull a nd flooding a re common (E).
Reservoir immediately impacted river discharge and river stage, indicated by A and F on the figure. When river stage is low and the weather is dry, variability is also low (B). During these periods, individual precipitation events do not have a significant impact on river stage, but rather their effect is short-term (C). Fluctuating spillway discharge creates gradual increases and decreases in river stage (D) as the reservoir level rises and falls in response to precipitation. When the reservoir gate is open, river stage response to precipitation events is overwhelmed by the discharge from the reservoir (E). Two large precipitation events in June resulted in high reservoir discharge to Pascoag River and river stage increased rapidly to flood stage. During this event, overbank flooding occurred at the site and submerged the monitoring equipment forcing river stage estimation from the top of the well casing.

Water-Table Elevations
Water table elevations in the study area range from approximately 110.9 m to 11 2.6 m with higher minimum and maximum heads at the upgradient wells closer to the source zone (Table 4 and Figure 10). The water table quickly responded to ;::::  Table 2.
fluctuations in river stage. This indicates that the river is hydraulically well connected to the aquifer and the dominant process controlling the position of the floodplain water table is river stage. During precipitation events and following gate closures, river stage elevation was significantly lower than most water table elevations.

Hydraulic Gradient
Overall ground-water flow and hydraulic gradient in the floodplain are directed toward the river both horizontally and vertically but vary considerably spatially and temporally (Table 5). Results indicate that immediately adjacent to the river channel between JA 2S and the river, flow direction is much more variable and reverses   Concentrations in JA 2D and JA 3D behaved similarly, with significant MTBE fluctuations of 2300 µg/L and 4 700 µg/L, respectively ( Figure 15). BTEX concentrations also fluctuated but were considerably lower in magnitude. In    during late May and June; however, the average and maximum concentration was lower than JA 1 RB throughout monitoring.

Field Parameters and Ions
In addition to the following brief summary, a complete list of results and discussion of field parameters, ions, and water temperatures is provided in Appendix III. Changes in EC and DO appear to be influenced by precipitation which dilutes shallow ground water and provides an oxygen source to the relatively oxygen depleted ground-water plume. In the deeper wells, this trend is replaced by increasing EC concurrent with the steepening hydraulic gradient and increasing MTBE concentrations. No trends in pH were identified throughout the monitoring period.
EC correlated well to measured anion and cation concentrations, except where DO influenced the oxidation state of sulfate and nitrate. Table 7 summarizes the average ion concentration for several well groupings and shows some overall trends. In general, floodplain wells did not show any significant trends in water temperature other than deeper wells having lower temperature ranges. General temperature statistics are presented in Table 8 and graphs of well temperatures are located in Appendix III. Riverbed water temperature was useful in showing slight shifts in ground-water discharge and changes in hyporheic exchange. For example, the gate closure in July was accompanied by a decrease in riverbed water temperature in all wells. This indicates a cooler ground-water rather than surface-water source and suggests that more significant ground-water discharge is occurring than at other times.

DISCUSSION
The objective of this study was to assess the effect of fluctuating river stage on an MTBE plume discharging to Pascoag River by investigating variability in the hydraulic gradient. Results of this study indicate that the hydraulic gradient in the floodplain varied hourly to monthly creating nearly continuous fluctuation of the river between gaining, losing, and parallel-flow. These fluctuations also appear to affect the position of the MTBE plume within the floodplain and may impact contaminant discharge to the river.
The observed hydraulic behavior of the river and the hydraulic response of the floodplain to river stage and precipitation contradict the original conceptual model.
First, short subsections of the river reach appeared to be losing while others have significant discharge of contaminated ground water and are gaining. And second, river stage increases and precipitation promoted a steepening of the hydraulic gradient towards the river. The factors responsible for these responses will be discussed as well as the implications to the contaminant plume and to the understanding of GW/SW interactions in general.

Hyporheic Exchange
Variations in hyporheic exchange created small-scale transient gaining, losing, and parallel flow systems along the reach by altering the hydraulic gradient immediately adjacent to the channel. Although a detailed analysis of the riverbed and riverbed exchange was not included with this study, limited hydraulic head, chemistry, and temperature data from the riverbed wells can be used to make inferences about small-scale exchange systems. The predominance of pool and riffle sequences, debris dams, and breaks in slope suggests short hyporheic flow paths of several meters ( Figure 18).
Upstream of 2D and 3D the river is shallow ( <0.5 m) and gaining conditions exist as shown by low riverbed MTBE concentrations and seeps in the riverbed when river stage is low. A break in slope and a debris riffle at 3D result in downward flow at the head of the riffle and upward flow at the toe (Conant Jr. 2000;Conant Jr. 2004;Harvey and Bencala 1993;Woessner 2000) which also marks the head of a deep pool (> 1.0 m). Upward flow at 3D results in significant discharge of contaminated ground water as indicated by consistently high MTBE concentrations. Downstream at JA lRB and 5D, the channel shallows and a debris dam create losing and parallel-flow conditions as river water enters into the hyporheic zone (Conant Jr. 2000;. This is enhanced by the channel bend at 4D and river meandering downstream of 5D which causes water to enter the riverbank and flow parallel to the channel (Cardenas et al. 2004;Woessner 2000) or pass through a lowland riparian area adjacent to the ground-water discharge lo~ation at MW 48. The structure and characteristics of the stone lining in the channel may also affect hyporheic exchange but the effects, if any, are unknown.
Areas of ground-water discharge and recharge and areas of no exchange to the river channel are controlled by hyporheic exchange can also be inferred from riverbed well temperature and chemistry data (Appendix III). Most riverbed temperatures are similar to surface-water temperatures making interpretation of water source difficult.
Nevertheless, in all riverbed wells gradient steepening in the floodplain appears to  Figure 18. Channel morphology along river reach. Cross section from A and A' as shown in Figure 6. The river is gaining upstream of 2D and 3D before a debris riffle creates losing conditions. Gaining conditions coincided with high MTBE concentrations in 3D. Parallel-flow and losing conditions persisted through the deep pool by 40 prior to channel shallowing and a debris dam creating losing conditions at JA IRB and SD.
correspond to decreases in riverbed and river water temperature, regardless of air temperature. This indicates ground-water flow toward the river along the entire reach and an influx of cooler ground water through the riverbed. Ground-water discharge through the riverbed is also generally corroborated by chemistry data including and increase in EC and MTBE, and a decrease in DO. For example, gaining conditions at JA lRB in May coincided with higher MTBE concentrations and EC. In riverbed well 3D, high MTBE concentrations, water chemistry in the riverbed, and channel morphology support the inference that under stable conditions (i.e. , no river stage fluctuations) significant contaminated ground-water discharge occurs at only this location along the reach. These observations indicate that along the entire reach, contaminated ground water is proximal to the channel but is able to discharge only when the gradient is sufficient to overcome small-scale hyporheic exchange.

Floodplain Processes
Variability of the floodplain hydraulic gradient is directly related to a dynamic interaction of several processes affecting the shallow water table: 1) water table recharge, 2) evapotranspiration, 3) ground-water flow, and 4) river stage fluctuations driving oscillation of the water table. The magnitude and impact of each of these processes is dependent on the reservoir gate which also controls the characteristics of the river stage hydro graph and the hydraulic response of the aquifer (Table 9).
Additionally, successive hydraulic gradient steepening periods have a compounding effect that ultimately have a more significant impact on water quality than individual events.  , and proximity to surface-water runoff.
The following is a discussion analyzing floodplain processes during several periods of this study referring back to the hydraulic gradient variability between JA 2S and the River (Figure 11) which had the largest hydraulic gradient variability and steepening in the floodplain. The hydraulic gradient curve can be separated into six periods (A though F) based on river stage elevation. This includes the low-stage periods in early spring (A & B) and late summer (F), the high-stage period in late spring to early summer (C & D), and the transitional periods in mid-February, mid-May, and early July corresponding to sudden river stage changes due to gate adjustments (E).

Spring Dry Period (A)
Precipitation in this period was relatively infrequent, river stage was low, and there was a low hydraulic gradient away from the river channel at JA 2S. Episodic shortduration precipitation events created a significant transient effect on the otherwise stable hydraulic gradient. During a two-part precipitation event on April 23 and 24, 2006, river stage and the floodplain water table rose nearly concurrent with precipitation ( Figure 19). In the two hours between individual precipitation events, river stage receded and then increased with the subsequent rainfall. Flashiness was magnified with the gate closed, preventing sustained discharge and decreasing stormflow lag time. All wells in the study area responded similarly to river stage during the first event, but the water table response to the second event was exacerbated by saturated soil conditions and the water table rose more than river stage. As a result, the river which was initially slightly losing, reversed and became strongly gaining as the head in JA 2S was well above river stage. The greatest floodplain well response came from the wells closest to the river and diminished with increasing distance from ,...... .....  process where rapid water table recharge forms a ridge or mound adjacent to the river channel. The exact shape of the water table irregularity is related to site specific properties of the landscape and the aquifer material. According to this theory, the ridge forms due to almost instantaneous saturation of the capillary fringe and rapid rise in the water table. In Pascoag, the capillary fringe may extend to between 50 and 100 cm above the water table, depending on specific soil properties (Fetter 1994). Ridge formation temporarily steepens the hydraulic gradient and results in increased discharge and contribution to stormflow. According to Figure 19, gradient steepening both towards the river and the topographical high near the source zone may occur; however, a flat hydraulic gradient between JA 3D and JA 2D does not indicate occurrence of flow towards the slope. Following precipitation events, the groundwater ridge dissipates due to lateral drainage and the slightly losing gradient is reestablished until the next precipitation event.
This hydraulic gradient response is also related to antecedent soil moisture conditions (field capacity), which is similar to but independent of capillary fringe processes. The field capacity of a fine sandy loam is approximately 0.25 and decreases during dry periods (Fetter 2001). During large or successive precipitation events, infiltration elevates soil moisture and soil may reach saturation resulting in a rapid rise of the water table.

Spring Wet Period (B)
During the middle of May there were several days of sustained precipitation that kept soil moisture high, the water table high, and ET low. The gate was still closed resulting in low river stage extending through these precipitation events. Direct water-

Effect of Gate (E)
The transition between low river stage and sustained high river stage occurred rapidly and was generally only related to opening or closing of the reservoir gate. In this system, natural seasonal variability in stream flow is muted by the dominance of the reservoir effect. The transitional period immediately after a gate adjustment is perhaps the most dynamic period as the floodplain aquifer and water table must equilibrate to a new hydraulic relationship with the river. This is best exemplified after the gate closure in July, where immediately following gate closure a relatively steep gradient (0.03 m/m) was established between the aquifer and the river and slowly diminished as the water table declined. If not for the precipitation events, the system presumably would have stabilized and the river would have reverted back to a slightly losing gradient. This process results in a flushing effect in the aquifer as it releases a large volume of ground water from storage.
Gate opening and rapid river stage increase in the absence of precipitation on May 18 and June 1, 2006, may result in limited traditional bank storage. In this case, because river stage rise is not accompanied by precipitation, ground-water ridging does not occur and the hydraulic gradient steepens away from the channel, enhancing losing river conditions.

Comparison to other GW/SW interaction studies
Literature review indicates the results of this study are similar to some but generally different from others. Although the process has been described in textbooks , few studies identified a similar process of ground-water recharge leading to strengthening of gaining conditions (Wroblicky et al. 1998  position. Taking a closer look reveals that high-order river systems typically respond with bank storage while low-order river systems typically respond with alteration of small-scale transient hyporheic processes (Table 10). This is similar to the process described by  regarding the importance of channel morphology controls on second-versus fifth-order rivers. This response may be related to stormflow lag time which is dependent on several watershed factors, including basin size, discharge per unit area, and other basin characteristics . Large river systems tend to have longer lag times with disconnected precipitation and stormflow. Pascoag River has only a slight lag between precipitation and stormflow which is consistent with a low-order river with a small watershed that responds quickly to recharge. River stage is also sensitive to changes in predominate landuse, with vegetated uplands providing sustained discharge to the river with the gate open and increased proportion of impervious surfaces generating flashy storm runoff with the gate closed.
Basin lag is important to consider in ground water/surface water interaction studies due to bank storage and gradient reversal processes. Typically in these studies, river stage increases leading to a temporary gradient reversal and bank storage are expected and even assumed. This model seems valid in high order systems but in low order systems, such as Pascoag River, the timing of the river stage increase relative to watertable response precludes the reversal. Instead, precipitation quickly recharges the shallow floodplain water table creating a ground-water ridge or mound. River stage also increases, but due to the over-response from the water table, bank storage can not occur and instead steepening of the hydraulic toward the river is promoted. Thus, river order appears more important than typically realized when considering the expected response of a coupled ground-water/surface-water system to recharge and to river stage and water table fluctuations.

Contaminant Fate and Transport
This study marked the first extensive ground water/surface water interaction investigation at the Pascoag site. The data suggests that steepening of the hydraulic gradient shifts a portion of contaminated ground-water discharge to an upgradient location. Prolonged and successive gradient steepening towards the river likely affects the plume discharge more significantly than individual or short-term steepening. The results indicate that a more complex and transient-ground-water interaction with surface water exists than originally perceived. This altered conceptual model of the plume behavior has important implications for this site and other sites where similar complex interactions may occur.

Plume Dynamics
Combining the floodplain hydraulic gradient with contaminant distributions allows for development of a conceptual model of transient contaminant fate and transport.
Steepening of the gradient between MW 56 and MW 18 results in increased ground-water discharge in the floodplain but the overall flow regime does not change significantly. There appears to be no significant impact on MTBE concentrations due to the fluctuating gradient at MW 18 and decreasing concentrations may indicate a diminishing source zone. Relatively high concentrations at MW 44 indicate that the plume discharges to the floodplain at this location but is only consistently able to discharge to the river near 3D. When the gradient is low, the center of the plume is  (right). When the gradient is low, upstream discharge occurs only near 30 due to plume deflection toward MW 48 by small-scale hyporheic exchange. Gradient steepening allows ground-water to discharge along the reach as the hyporheic zone shrinks or disappears. deflected by hyporheic processes parallel to or away from the river channel towards the downgradient focused ground-water discharge location near MW 48 ( Figure 20).
When the gradient steepens between JA 2S and the river, the hyporheic processes are overcome and significant discharge occurs along the entire reach ( Figure 20). Further ground-water and surface-water flow alteration occurs due to a channel cutoff that develops over the lowland riparian area and into the discharge are near MW 48. The increase in MTBE concentration in JA 2D without a steepening gradient between JA 3S and JA 2S indicates that the plume must migrate from upgradient near MW 44.
Ground water transport along this flow path may be enhanced by a former channel that has subsequently been filled. Historical photographs show this channel was 4 to 5 m wide and was probably 1 to 2 m deep (Appendix I). This buried channel coupled withother buried structures has created complicated subsurface geology and has precluded a detailed investigation by RIDEM in the area around MW 18, 44, and 56.

Ground-water Monitoring and Natural Attenuation
Ground water at the Pascoag site and most other sites is monitored quarterly to semi-annually as specified by individual site operation and management plans (Hazardous Waste Clean-Up Information 2007). This plan is usually adequate for sites not in proximity to surface water. The results of this study indicate that a GW/SW interaction survey with high resolution spatial and temporal water-quality sampling and hydraulic gradient data may be required to accurately delineate and monitor a plume discharging to surface water. While this may be a costly upfront proposition, it may help avoid higher long-term operation and management costs.
Although Pascoag has undergone extensive pump-and-treat remediation, biological degradation and discharge to the river has aided in natural attenuation of the plume. Hinzman et al. (2000) showed that compared to active remediation, natural discharge of contaminants to surface water can remove an equal mass of contaminants from the system. In Pascoag, the contaminant mass discharged to the river varied considerably and was not solely related to dilution but probably a combination of dilution and variable ground-water discharge. Downstream from the study area, MTBE mass discharged by the river varied between 3 g/day and 1100 g/day and BTEX 0 g/day and 220 g/day. Periods of high ground water and contaminant mass discharge coincided with aquifer flushing during the steep gradient periods induced by precipitation and gate closure. This can potentially affect stream health and benthic habitats with pulses of high-level contaminants  rather than low-level continuous exposure and could easily be missed by quarterly monitoring. It may also complicate estimation of long-term contaminant discharge and total maximum daily loads.
Water table fluctuations an~ the continuous shifting of the ground-water plume may affect biodegradation and smear contaminants above the water table and into different areas of the floodplain. Degradation may also be influenced by the alternating of gaining and losing river conditions and the resulting effect on dissolved oxygen delivery to the riverbed and riparian zone. This impact was not assessed in this study, nor was a more detailed analysis of contaminant fate.

Limitations and Sources of Error
There are several limitations ofthis study and to the associated data. For instance, monitoring of the water levels and subsequent water- This study is also limited by the scale of this thesis, to the original conceptual model, and to the thesis question at hand. Because of this, there was certain field data not collected or included in this work. For example, this thesis is not able to identify the exact location of discharge when the plume discharge location apparently shifts upstream. It should also be noted that ground-water and contaminants may still be

SUMMARY AND CONCLUSIONS
This study demonstrates the complex and transient relationships that occur where ground water and surface water interact. The research objective of this study was to assess the effects of fluctuating river stage on an MTBE contaminant plume discharging to the Pascoag River. Results show that the hydraulic gradient fluctuates hourly to monthly, due to river stage and floodplain water table response to precipitation and reservoir gate adjustments. The river along this reach is predominantly gaining; however, hyporheic exchange processes vary spatially and temporally, forming small-scale gaining, losing, and parallel flow systems.
Precipitation events rapidly recharge the shallow floodplain water table, causing ground-water ridging and steepening of the hydraulic gradient towards the river. Even though this response also occurs when river stage is high, ultimately, river stage fluctuations have the largest influence controlling the response of the aquifer. The magnitude of these fluctuations is directly related to the size of the watershed which drives river flashiness, prevents sustained stormflow, and enhances the gradient toward the river.
The gasoline contaminant plume is significantly affected by shifting and steepening of the hydraulic gradient. Under stable hydraulic conditions, the plume enters the floodplain adjacent to MW 44, but only a small portion is able to discharge to the river. Instead, hyporheic flow deflects the plume downgradient parallel to the river where it can discharge. When the hydraulic gradient steepens towards the river following a precipitation event, the hyporheic exchange systems shrink or disappear and the plume enters the river channel farther upgradient. Gradient steepening following precipitation and gate closures results in a flushing effect of the floodplain generating a high influx of contaminated ground water to the river. These episodic releases may impact the stream and benthic habitats more dramatically than long-term low exposure.
Interpretation of these results indicates that the current understanding of G W /SW interactions is able to explain many of the significant processes occurring in floodplains but work still remains. Several questions, comments, and conclusions arise from this work: • Floodplain processes, including water-

Bedrock
The bedrock augen granite g~eiss is typically medium-to-coarse-grained with large feldspar porphyroclasts, and is generally variable in compositions (mostly quartz, feldspar, biotite, hornblende, and other accessory minerals). The second, less common rock unit is generally fine-to-medium-grained granite gneiss that lacks porphyroclasts, and is more quartz-rich than the augen granite gneiss (Hermes et al. 1995;. This unite forms a narrow gradational lens that extends from the source zone into the middle of the site. Both rock units are typically massive but display lineation and foliation that is locally very strong. Both units are highly fractured with fracture locations dependent on the zones of lineation and foliation. A fracture study by  found the dominant trend of mineral lineation is approximately N 2° E and plunges 10° north. Dominant fractures strike nearly north-south with an average dip of 65° E (Figure 21 ). Another less dominant fracture orientation trends N 75°W and dips 75° south. The frequency of these fractures may be localized into fracture zones, with the rock units being more massive between these zones. Other fracture orientations occur, but their frequency and importance appear to be less significant. Also, orthogonal fractures that trend along the same dominant strike direction but dip much more shallowly were detected. c c llO Figure 21. Rose diagram of bedrock fracture trends collected from outcrops in and around Pascoag (Allen and Boving 2006).

Historical Photographs
Photographs of the textile factory that was located at the study site from the 1800s through around 1960 (Figure 22 and Figure 23).

Well Installation and Surveying
The four wells already present in this area prior to this study, MW 18, MW l 8D, The wells in the river, JA IR and JA lRB, had to be installed securely to prevent destruction during high river stage. The well consisted of a 1.5 m long by 10 cm steel casing fitted with two interior wells and was sealed at the bottom. The riverbed well was inserted through a sealed fitting and a 0.3 m screen extended into the riverbed.
The purpose of the river well was to house the pressure transducer monitoring river stage and was only screened through the river. The steel casing was perforated with approximately fifty 1 cm holes to allow interaction with the river. The well unit was installed in the river by digging a hole in the riverbed and was attached to the river channel wall with steel rods. Due to the extremely high river stage, well extensions were installed to bring the height of the wells above the river surface. Other riverbed wells were installed by pushing the PVC well into the riverbed and hitting with a rubber mallet until refusal.
The four wells already present in the study area were previously surveyed for top of casing elevation by RID EM. The additional wells installed for this study were surveyed using the known elevation from MW 44. All wells, including the already existing wells, were surveyed with a Topcon AT-G6 Auto Level using a leveling rod.
Instrument error is ± 2 mm at 1 km, and total error is probably less than 1 cm (Laserbeams.com 2006).

Slug Testing
Bouwer-Rice falling head slug tests were performed on all wells to determine hydraulic conductivity. The test was performed by placing a pressure transducer at the bottom of the well and then inserting a solid slug. Manual depth to water measurements were also taken before and after the test. The test measured the response of the slug insertion and the slug removal after waiting for the water table to respond. Generally, the slug removal data provided a better response curve. Data was transferred to AQTESOL V software to calculate hydraulic conductivity using the slug testing wizard and the appropriate input parameters. Table 12 lists the equipment brand and specifications for field instruments and data collections devices.

River Discharge
River discharge was measured during the study by the float method due to the simple dimensions of the river channel and regular flow (Hudson 1993). Also, only an approximate measurement of discharge was required. The method consisted of first measuring a straight 10 m section of the river at the study site. Average channel width and depth were measured at 0 m, 5 m, and 10 m by taking six depth measurements across a river channel transect. Average velocity was determined by measuring ten This procedure was repeated at various river stages in order to construct a river discharge ratings curve. All GC/MS and purge and trap operating parameters are listed in Table 13. The overall method was similar to EPA Method 624 which is a purgeable volatiles method capable of low µg/L MTBE detection. The GC/MS was internally tuned daily with PFTBA tuning standard prior to running samples to verify mass resolution. The tune was then checked by injecting a BFB external tuning standard to check relative peak intensities. Instrument calibration was performed with six external standards at the following concentrations 2, 10, 20, 40, 80, and 160 µg/L. Each standard and all samples were prepared and introduced in the same manner through the purge and trap device. Deionized water or the sample was carefully transferred from the VOA vial into a clean 5 mL purge and trap luer-lock syringe and adjusted to 5 mL. For each standard, an appropriate amount of 10 mg/L volatiles working standard and BFB surrogate standard (for samples only surrogate was added) was added to the syringe through the luer-lock opening (Table 14). The standard or sample was then transferred though the injection port on the purge and trap into the 5 mL sparge vessel.

Volatile Organics Analysis
After helium purging and purge and trap method completion, the extracted volatiles were transferred from the trap to the GC column for the start of the analytical method.
The sparge vessel and the purge and trap syringe were rinsed thoroughly between samples. Standards were prepared from purchased certified standards in Purge and Trap Grade Methanol. Working standards were stored in the freezer and replaced when quality control indicated standard degradation. Calibration was performed by linear regression and was accepted if R 2 was >0.995. Calibration verifications were analyzed after calibration, after every ten samples, and at the end of a run. Acceptance criteria range was ± 15% in general, but a daily precision of± 5% was preferred in order to minimize relative error associated with response variability. Method blanks spiked with surrogate standard were analyzed at the start of each run. Many dilutions were required and were prepared in 10 mL volumetric flasks.
A minimum detection limit (MDL) study was completed to determine the analytical capabilities of the instrument. The study was completed by analyzing 7 replicates of deionzed water spiked with 5 ug/L of the volatiles standard. The MDL was calculated by multiplying the standard deviation of the replicates by the corresponding student's t-statistic. The MDL is the concentration at which there is 95% confidence that the concentration is not zero. If quantified, values below this must be reported as "not detected". Values between the MDL and the lowest standard (the reporting limit) must be reported as tentative values but can be considered reliable if a blank passes.

Ion Analysis
Cations and anions were analyzed by ion chromatography on a DX 120 (Dionex Corporation). All calibration standards were prepared from 250 mL intermediate standards which were prepared from 1000 mg/L stock standards. Stock standards were made from certified granular salts containing the analyte of interest. All samples were filtered with a 0.45 µm filter prior to analysis. Anion eluent was made by dissolving 0.19 g of sodium carbonate and 0.142 g of sodium bicarbonate to 1 L.
Cation eluent was made by diluting 28 mL of 1 N sulfuric acid to 1 L.
Calibration standard levels for anions and cations are listed in Table 15. Anion calibration was four-point while cation calibration was three-point. Both calibrations were accepted if the R 2 was >0.992 or better and an initial calibration control standard passed (±15%). Addition calibration verification standards were analyzed after ten samples and as the last sample of every run. Periodic method blanks and duplicates were analyzed to determine reproducibility and system performance. Blanks determined any background system contamination and the method reporting limit. An MDL study was also completed for both anions and cations (Table 16).
The calculated MDLs for chloride, nitrate, and magnesium may be lower than instrument analytical capabilities and can only be used as rough guidelines.

Precipitation and Temperature
A combination of weather stations was required to accurately determine the average daily precipitation and temperature. This was because of inconsistencies in individual datasets and because the Pascoag Weather Underground station (0.5 km from study site) came online April 5, 2006(Weather Underground 2006. Once online, there were times when the station was offline for short periods and no data was collected. When service returned, any precipitation that had accumulated was registered as a single event. This also skewed the average daily temperature. Accompanying datasets were obtained from the National Oceanic & Atmospheric Administration (NOAA) website (NCDC 2006) for Woonsocket, RI, North Foster, RI, and Worcester Regional Airport, MA, at respective distances of 17 km, 13 km, and 42 km from the central study site in Pascoag. Values obtained from these stations were averaged to obtain an approximate value for Pascoag. River hydrographs were checked for single events and where inconsistencies occurred, the Pascoag Weather Underground data was typically used.

24.
Time (sec)      DO levels in the riverbed w~lls provides insight into well depth relative to the hyporheic zone and indicates the depth of GW/SW exchange. Higher DO levels suggest that river water is present beneath the channel at the well screen, whereas a lower DO level indicates that ground water is discharging. In the river, lower DO levels show that ground water contributions have increased relative to surface water.

Hydraulic Gradient
The pH results were limited in their use as an indicator of GW/SW interaction.
Results from June 13 may be too low due to a loose wire on the pH meter. Overall, pH was relatively consistent at most locations and may have been most sensitive to rainfall resulting in a decrease in pH.

Ions
Ion concentrations were related to dilution, concentration, DO abundance, and ground-water source (Figure 28) ... The.conservative ions (sodium, potassium, magnesium, calcium, and chloride) behaved similarly at all wells except for slight differences between shallow and deep wells. In the shallow wells ( Nitrate and sulfate concentrations were mostly dependent on DO which controlled oxidation-reduction state (Hem 1985). Because the contaminant plume is typically anoxic, conditions are suitable for nitrate and sulfate reduction. This resulted in low concentrations during anoxic conditions in March and April and higher concentrations when dissolved oxygen was plentiful and reduction was prevented. This is especially important for sulfate which is present at relatively high concentrations upgradient near MW 56 but is reduced rapidly in the floodplain before reaching the river. Phosphate levels were generally low throughout the study due to removal through plant uptake (Hem 1985).
Riverbed and the river water samples also show trends in ion concentrations correlating well with hydraulic gradient ' steepening and increases in MTBE. In the river upstream of the study area, all ions decreased throughout the study period perhaps in response to dilution or decreasing contribution following winter road salt application. Regardless, the ion concentrations in the river at JA 1 R responded independently of the upstream location. The river at the site responded with higher ion concentrations immediately following the gate closure in February, probably due to an increased proportion of ground-water from return-flow. During all other periods, no significant trends were observed.

Water Temperature
Water temperature data can be separated into several groups based upon general trends and relationships. Because of instrument difficulty, accidental removal, and a late monitoring start date in some wells there was some variability in the period of data collection (Table 8). The shallow wells JA 2S and JA 3S had larger temperature ranges, more extreme minimum and maximum temperatures, and a faster response than their corresponding deep wells ( Figure 29). JA 2D and JA 3D both experienced their respective seasonal low temperatures in May and were otherwise smooth temperature curves. The iButton in JA 2D failed in June and was replaced with a less precise substitute resulting in less resolution throughout July.
MW 18 and MW 18D temperature curves ( Figure 29) were punctuated by many increases and decreases in temperature that appear to correspond to precipitation events or changes in river stage related to both precipitation events and reservoir gate adjustments. Like the shallow wells above, MW 18 had a steeper slope, and a larger temperature range, while MW 18D also had its seasonal low in May.
The four riverbed wells were monitored for various time periods but all show the same short-term and long-term trends ( Figure 30). Specifically, temperatures remained low until the middle of May and then increased until stabilization June 1 through mid-June. Temperatures increased again and then rapidly decreased in early July when river stage declined before gradually increasing again in August. River temperature covaried with air temperature and appears to experience smaller diurnal variability when flow is greater, presumably in response to greater heat capacity (Constantz 1994). There were several periods when river temperature and air temperature trends diverged and may indicate a different water source and GW/SW exchange.