Scientific Investigations Report 2007–5216
U.S. GEOLOGICAL SURVEY
Scientific Investigations Report 2007–5216
Temperature is a controlling variable for aquatic life in the water column and in streambed sediments. Exchanges that occur between streams and ground-water systems play a key role in controlling temperatures not only in the stream, but also in the underlying sediments. Heat is transported through stream sediments by advection and conduction where temperature differences exist. Heat as a tracer is a simple yet powerful tool for detecting water movement across the sw–gw interface when this movement is traced by continuous monitoring of temperature patterns in the streambed and subsurface water.
The use of heat as a hydrologic tracer has several distinct advantages over applied chemical tracers. The temperature signal occurs naturally, rather than having to be introduced into the stream setting. The primary measurement is temperature, which is robust and a relatively inexpensive parameter to measure. In contrast to chemical tracers, which often require laboratory analysis before interpretation, temperature data are immediately available for inspection and interpretation (Stonestrom and Constantz, 2003). As a result, analyses of subsurface temperature patterns provide information about sw–gw interactions.
Streams exhibit diurnal temperature fluctuations as a result of solar-driven temperature fluctuations at the land surface, whereas ground water is buffered from these temperature fluctuations below the surface. The difference in temperature between streams and ground water provides a means for tracing exchanges between the two systems. In a gaining reach of a river, water is moving upward into the streambed and carries with it the relatively static temperature signal of the ground water. As a result, the temperatures in and beneath the gaining reach are muted compared with the diurnal fluctuations in the stream. Conversely, if the stream is losing water volume by seeping down through riverbed to ground water, and water is moving downward into the streambed, the diurnal temperature signal of the river is carried by advection and conduction into the surrounding sediments. Subsurface temperature patterns will exhibit the diurnal fluctuations seen in the stream. Figure 18 depicts these concepts.
Heat has been used as a tracer of subsurface water movement for more than 40 years. Analytical solutions to equations that govern the coupled movement of water and heat have been derived and applied to estimate the rate at which water travels from the surface to great depths (Rorabaugh, 1954; Stallman, 1963). Temperature patterns have also been used to study subsurface flow systems ranging from irrigation water in rice paddies to geothermal water beneath volcanoes (Suzuki, 1960; Sorey, 1971). Lapham (1989) used annual temperature records from deep observation wells to identify rates of vertical water flux in several streams in Massachusetts and New Jersey that were based on analytical solutions reported in earlier work (Lapham, 1988). However, these analytical solutions were derived for a few idealized cases and resulted in theoretical, rather than practical, applications because of measurement and computational limitations. Recently, the measurement and modeling of heat and water transport have benefited from significant improvements. Recent innovations in sensor and data-acquisition technology, along with substantial improvements in numerical modeling, present new opportunities for the use of heat as a tracer of stream–ground water exchanges (Stonestrom and Constantz, 2004). Inexpensive and accurate devices are now available for measuring temperature, water level, and water content. These devices, in conjunction with currently available numerical models, provide general solutions of equations for the coupled transport of heat and water.
Sw–gw exchanges can be estimated using heat as a tracer in conjunction with water level measurements (Silliman and Booth, 1993; Silliman and others, 1995; Stonestrom and Constantz, 2003; Anderson, 2005). Temperature has been used as a tracer to identify vertical flux across the sw–gw interface at various locations in the United States. In the Rio Grande near Albuquerque, New Mexico, Bartolino and Niswonger (1999) used the USGS numerical model (VS2DH) to match simulated temperature data to observed temperature, yielding predicted estimates of deep streambed fluxes and spatially averaged hydraulic conductivities. Application of heat as a tracer has been used to examine interactions in alpine streams between stream temperature, streamflow, and ground-water exchanges (Constantz, 1998). The analysis of temperature profiles in ephemeral stream environments has been used to examine percolation characteristics beneath arroyos in the Middle Rio Grande Basin, New Mexico (Constantz and Thomas, 1997), to determine streamflow frequency and duration (Constantz and others, 2001), and to investigate stream losses beneath ephemeral channels (Constantz and others, 2002). Stonestrom and Constantz (2003) provide technical details of the use of heat as an environmental tracer as well as a compilation of seven detailed case studies that use temperature patterns and their interpretation as a hydrologic tool for the assessment of interactions between surface water and ground water in a variety of environmental settings throughout the western United States.
The study site location on the lower Merced River was selected because the reach was within the study area that encompassed the overall USGS’s NAWQA agricultural chemical transport (ACT) study design and objectives. To understand how agricultural chemicals move through and between hydrologic compartments (atmosphere, surface water, the unsaturated zone, ground water, and sw–gw interaction), a gaining reach of river was preferable for the sw–gw component. Reconnaissance of the reach indicated that the reach was gaining and easily accessible for sampling equipment installation and subsequent sampling events. In addition, agricultural chemicals of interest were being applied to crops north and south of the selected reach.
A total of twenty 5-cm (2 in.) polyvinyl chloride (PVC) screened monitoring wells were installed across two transects in the Merced River. Transects were separated by a distance of approximately 100 m. Each transect was equipped with five pairs of monitoring wells: three pairs in the river and a pair on the right and left banks in the riparian zone. Figure 19 depicts a bank and in-stream monitoring well pair. The pairs of monitoring wells at each transect consisted of a shallow and a deep monitoring well screened at approximately 0.5 and 3 m below the streambed for the in-stream monitoring wells, and at approximately 3.5 and 5.0 m below the top of well casing for the riparian zone monitoring wells. The monitoring wells were equipped with temperature loggers and pressure transducers. Figure 20 depicts a cross-sectional view of an equipped transect.
Temperature loggers monitored temperature continuously in both the surface water (above the sediment–water interface) and at three depths within the streambed at both transects. Pressure transducers located in the stream and below the streambed collected water-level data that were used to define boundary conditions. Temperature and pressure-head data were input into the USGS numerical model, Variably Saturated 2-Dimensional Heat (VS2DH; Healy and Ronan, 1996) and its graphical interface VS2DI (Hsieh and others, 2000). This program uses an energy transport approach via the advection–dispersion equation to simulate heat and flow transport. Estimates of streambed hydraulic and thermal conductivity were input into the model until model simulations “fit” observed subsurface streambed temperatures. This inverse modeling method uses a visual best fit and is most sensitive to variations in the input parameter K (hydraulic conductivity).
The monitoring wells were installed by pumping the streambed sediment out while pushing in a 15-cm (6 in.) diameter PVC casing downward to the desired depth. After reaching the desired depth, the smaller 5-cm (2 in.) PVC monitoring well was inserted inside the 15-cm PVC casing, and the 15-cm casing was pulled out of the streambed, allowing the surrounding streambed sediments to collapse around the 5-cm PVC monitoring well. The streambed monitoring wells were installed so that the top of the casing of each of the wells was slightly above the streambed. Once in place, the monitoring wells were sealed with standard pressurized monitoring well caps to prevent stream water from entering the monitoring wells. The same installation procedure was used for the riparian zone monitoring wells; however, a combination of hand augering and pumping was used to install the outer casing to the desired depth. A Monterey sand pack was placed around the screened interval of each of the riparian zone monitoring wells. The wells were then sealed with a bentonite cap and backfilled to land surface. Convection in the riparian zone wells was not of concern because convection does not occur in small-diameter wells under saturated conditions or in the absence of large temperature gradients between temperature probes, as confirmed by reconnaissance and collected data.
A string of three HOBO Water Temp Pro temperature loggers (accuracy: ±0.2°C at 0 to 50°C range) was fastened to a small diameter rope at depths of approximately 0.5, 1.0, and 2.0 m from the monitoring well cap (fig. 21). These depths were chosen because during reconnaissance of the site, static ground-water temperatures were encountered at approximately 2.5 m below the sediment–water interface. The temperature loggers were then weighted with a stainless-steel bolt and placed in each of the deep in-stream monitoring wells. The same procedure was used for the riparian zone monitoring wells; however, the temperature loggers were placed at approximately 3.5, 4.0, and 5.0 m below the top of the monitoring well casing. These depths were chosen because the water table was approximately 3 m below land surface. A temperature logger was also placed in the stream to record stream temperature. A total of 10 pressure transducers were used to record water levels: eight in four pairs of the in-stream monitoring wells, one in the deep riparian monitoring well at the downstream transect, and one in the Merced River.
Temperature and pressure were measured at 15-minute intervals. Manual water level measurements were also taken through the study in the riparian zone monitoring wells and during data downloads for the in-stream monitoring wells. Because the top of the casing for the in-stream monitoring wells was underwater (approximately 5–10 cm above the streambed), a 1.8-m riser was attached to the in-stream monitoring well and allowed to equilibrate prior to water level measurements. Table 8 identifies the wells by name, location, and the type of continuous data (temperature and [or] pressure head) recorded in each of the monitoring wells.
In this study, the application of temperature as a tracer utilized continuous monitoring of water levels, and subsurface temperature, as well as the elevation and temperature of the stream. Data were used to specify the boundary conditions in the numerical model VS2DH, and estimates of hydraulic and thermal conductivity were adjusted in the model until “best fit” simulations matched the observed subsurface stream temperatures at intermediate depth. This method requires high precision in the water-level data and certainty in the elevations (depths) at which the pressure transducers are placed at. However, because of problems associated with the instrumentation used to record continuous water-level elevation, prolonged unexpected high streamflow events, and the resulting scour and burial of the monitoring wells, the results of the recorded total-head distributions and temperature profiles are discussed prior to the modeling results, as these results affect the final model results. Furthermore, results and discussion of total-head distributions and collected temperature data for monitoring wells at the upstream transect are not presented because of vandalism and scour.
Total hydraulic head (meters) is defined as the recorded pressure head plus the known elevation at which each of the pressure transducers was placed. A pressure transducer was placed in the river to record continuous stream elevation data. These recorded stream elevations were used to calculate differences in total head between the ground water and the river, thereby indicating the direction of flow across the sediment–water interface at the instrumented site. Gaining and losing reaches were defined from the perspective of the river, which is standard protocol. Specifically, a positive value indicates a gaining reach and a negative value indicates a losing reach. However, in the present study, the pressure transducer placed in the river did not provide the quality of data that the instrument was intended to record for several reasons. The initial study design included a plan for differences in head between the river and ground water on a very small scale (1–3 cm). Unfortunately, the inherent error associated with the estimated elevation of the location of the pressure transducer placed in the Merced River was greater than the measured head differences between the surface water and ground water. This problem was further complicated with each download of data; when the pressure transducer was removed and downloaded, it was nearly impossible to replace the instrument at an exact height in the well. In addition, the PVC casing that housed the pressure transducer was at approximately a 40º angle from horizontal, and it was difficult to replace the instrument at its intended location because it was not free-hanging under gravity.
Pressure transducers located below the streambed in the paired monitoring wells screened below the streambed did not have these limitations. These pressure transducers were attached to the cap of the monitoring wells and allowed to hang freely within the monitoring well. Manual water-level measurements taken immediately after data downloads agreed well with recorded water levels to within <0.3 cm. As a result, the paired in-stream monitoring wells (MW) (delta H = deep MW – shallow MW) were used to calculate head differences.
Although the pressure head recorded in each of the streambed monitoring wells agreed with manual measure-ments, the minor discrepancy between the recorded water level and measured water level seemed to increase in the latter part of the study, suggesting drift in the pressure transducers. The pressure transducers used in this study were not vented to the atmosphere and were placed in monitoring wells that were sealed at the streambed. As a result, this type of pressure transducer requires an additional pressure transducer (baro-logger) to record and compensate for changes in atmospheric pressure. The data collected from the baro-logger were subtracted from the recorded head data collected in the monitoring wells. The atmospheric pressure and water levels calculated by the baro-logger and the pressure transducers are temperature compensated, though large changes in temperature over a short period can affect how the instrument calculates water pressure (Davies, 2002). ">Although the pressure head recorded in each of the streambed monitoring wells agreed with manual measure-ments, the minor discrepancy between the recorded water level and measured water level seemed to increase in the latter part of the study, suggesting drift in the pressure transducers. The pressure transducers used in this study were not vented to the atmosphere and were placed in monitoring wells that were sealed at the streambed. As a result, this type of pressure transducer requires an additional pressure transducer (baro-logger) to record and compensate for changes in atmospheric pressure. The data collected from the baro-logger were subtracted from the recorded head data collected in the monitoring wells. The atmospheric pressure and water levels calculated by the baro-logger and the pressure transducers are temperature compensated, though large changes in temperature over a short period can affect how the instrument calculates water pressure (Davies, 2002).
Figure 22 depicts discharge and recorded head differences (delta H) between the deep and shallow monitoring wells at the downstream transect. Mean daily discharge data is obtained from the Department of Water Resources stream gaging station, Merced River near Stevinson (site ID 11272500), approximately 16 mi upstream from the study area. An 18-hour travel time was determined and applied using a travel-time equation from Kratzer and Biagtan (1997). The pressure transducer malfunctioned after the September 2004 download in monitoring well pair RW-044 and RW-043 (fig. 22A). Head differences for monitoring well pair RW035 and RW034 (fig. 22B) recorded a generally gaining river throughout the study period with distinct flow reversals (gaining to losing) during high streamflow events. These flow reversals correspond to storm events during winter months, relatively large releases at the New Exchequer Dam to improve downstream salmon habitat from mid-April to mid-May, and smaller releases made in October as part of the Vernalis Adaptive Management Plan (VAMP). The water-level data collected from the monitoring well pair RW044 and RW043 (fig. 22A) did not record a flow reversal across the sediment–water interface during the May 2004 VAMP release.
Agricultural fields upgradient of the transects are irrigated between March and September, which seasonally recharges the surficial aquifer. The rise in water levels of the surrounding aquifer is reflected in the water levels of the monitoring well pairs, indicating a gaining stream during irrigation season 2004. Irrigation season 2005 did not result in gaining conditions because of large streamflow releases from New Exchequer Dam beginning March 2005 and continuing through the end of the study period. These releases were made as a result of spring melt of an exceptionally high snowpack in the upper part of the basin.
During the winter months, the stream temperature becomes much cooler than recorded subsurface temperatures, and conversely during the summer months, the stream temperature becomes much warmer. Figure 23 depicts daily average temperature profiles collected at the downstream transect. The in-stream monitoring wells recorded nearly the same temperature patterns at 0.5, 1.0, and 2.0 m below the streambed. Periods of gaining or losing are recorded, and in general, the subsurface temperatures indicate a slightly gaining-to-neutral reach throughout 2004, with the exception of mid-April to mid-May 2004 during the VAMP flow events when the temperature record indicated a losing scenario.
Temperature records from 2005 coincide with the exceptionally high streamflow. This likely caused scour around the in-stream monitoring wells during the rising limb of the hydrograph and subsequent burial during the falling limb. Field observations support this assumption. As a result, significant uncertainty was created with respect to depth of recorded temperature variations. Despite this concern, the collected temperature data for this time period indicates a losing-stream scenario until June 2005, after which a strongly gaining-stream time period is recorded. Stream elevation approached and overtook bank well BW017 during the spring 2005 dam release and was in direct contact with the river and recorded stream temperature.
The bank wells follow the same temperature patterns as the in-stream wells, but to a much lesser extent. During the summer months, ground-water temperatures at 3.5 m were warmer than those recorded at 5 m, and conversely, during the winter months, recorded temperatures at 5 m were warmer than those at 3.5 m. These changes occurred in June and November, respectively. The bank wells do not record diurnal temperature variations (fig. 24), but do record slight seasonal changes.
One-dimensional modeling of heat and water flow was used to interpret temperature and head observations and to estimate vertical sw–gw fluxes at wells RW044 and RW035 in the downstream transect. Two vertical, one-dimensional models with 2-cm grid-blocks were calibrated for each of the deep monitoring wells at the downstream transect, and each well was modeled separately (fig. 25).
The energy transport and water-flow model, VS2DH, was used to fit simulated temperatures to observed temperatures and heads. At the beginning of the study, the intention was to apply recorded pressure head of the surface water to the top boundary and pressure head of the deep monitoring wells to the bottom boundary of each of the one-dimensional models. However, because of issues with the quality of data of the surface-water pressure transducer as discussed in the total-head distribution section, a new approach to applying pressure-head data to the model was necessary.
The difference in pressure head between the deep and shallow wells was assigned to the bottom boundary, and a pressure head of 0 was assigned at the sediment–water interface (top boundary) for the model. Positive pressure heads indicate discharge from the aquifer to the stream. Figure 25 depicts the model domain at 2 m below the streambed, with a lower model boundary that corresponds to the deepest temperature logger; however, the screened interval of the MW is at 3 m below the streambed. As a result, the head difference between a deep and shallow well is over a vertical distance of 2.5 m and does not match the model domain. It was assumed that the measured head difference over 2.5 m was linear and that K of the materials throughout the model domain was constant as a result of the homogenous streambed material. Therefore, the head difference over 2.5 m was corrected to a head difference over 2 m to match the model domain.
The temperatures applied to the top and bottom boundaries were recorded above the sediment–water interface and 2 m below the streambed, respectively. Estimates of streambed hydraulic conductivity were input into the model until model simulations provided a “best fit” of observed temperatures at depth. The observed temperatures used to match model simulations were measured were 0.5 and 1.0 m below the streambed. Figures 26 and 27 depict the results of the one-dimensional modeling efforts at the downstream transect for in-stream monitoring wells RW-044 and RW-035, respectively.
The streambed of the lower Merced River Basin proved to be a highly dynamic system, with mobile bar forms and substantial bed load transport during periods of low streamflow. High streamflow events also included suspended load. The model results in figures 26 and 27 depict periods where simulated temperatures nearly match observed temperatures. Periods of departure have three explanations: (1) they may be the result of the streambed characteristics changing over time, resulting in varying K values; (2) they may be a result of scour near wells that changes the effective depth of the temperature loggers and alters the model domain; and (3) the hydraulic head gradient that the model calculates is over a 2-m domain; however, changes in effective depth of pressure transducers that are due to scour near wells may not always coincide with the assumed model domain, resulting in calculated head gradients that are not representative of actual gradients. ">The streambed of the lower Merced River Basin proved to be a highly dynamic system, with mobile bar forms and substantial bed load transport during periods of low streamflow. High streamflow events also included suspended load. The model results in figures 26 and 27 depict periods where simulated temperatures nearly match observed temperatures. Periods of departure have three explanations: (1) they may be the result of the streambed characteristics changing over time, resulting in varying K values; (2) they may be a result of scour near wells that changes the effective depth of the temperature loggers and alters the model domain; and (3) the hydraulic head gradient that the model calculates is over a 2-m domain; however, changes in effective depth of pressure transducers that are due to scour near wells may not always coincide with the assumed model domain, resulting in calculated head gradients that are not representative of actual gradients.
Figure 26 depicts the simulated temperatures and modeled flux from March to September 2004 for monitoring well RW-044 and may be an exception to the explanations for departures from simulated temperatures as explained above. The pressure transducer in this monitoring well did not record a flow reversal during the April 2004 VAMP release and showed significant “noise” after the first data download in May 2004. The author of this report believes that this pressure transducer was faulty and failed after the second and last download in September 2004. The noise in the pressure-head data may explain why simulated temperatures were slightly higher than the actual temperatures recorded at approximately 0.5 and 1.0 m below the streambed.
Figure 27 depicts modeling results for monitoring well RW-035. Simulated temperatures match up until the dam release in October 2004. A departure from observed temperatures occurs during and following the dam release. The higher streamflow likely scoured out the fines accumulated over the summer prior to the October release, thereby increasing the streambed K. The departure of simulated temperatures from observed temperatures for this period indicates that for the simulated temperatures to continue to match the observed temperatures, the streambed K (and resultant vertical flux) must be higher than the model input value used prior to this period. Figure 28 depicts the results of a model run in which the streambed K was increased from 1 to 10 m/day for the period of departure (10/18/2004 to 1/25/2005). The resultant simulated temperatures for the departure period provide an improved match-to-observed temperatures and substantiate the interpretation for this departure. The resultant modeled vertical flux increased from 0.4 to 4.6 cm/day.
Although increasing hydraulic conductivity improves the simulated temperatures compared with the observed temperatures, it is difficult to accept that this factor alone explains the departure because it unlikely that vertical flux increased an order of magnitude over such a short period of time. Instead, a combination of factors may better explain the departure. It is likely that changes in effective depth of pressure transducers, attributed to scour near wells, resulted in instrumentation depth not coinciding with the assumed model domain. The result is that the head gradients that the model calculates are not representative of actual gradients. This explanation coupled with increased streambed K resulting from scouring of fines because of higher streamflow may provide a more accurate explanation of the departure.
Because of the data gap that occurs between January and February 2005 for monitoring well RW-035, and departure period following the October 2004 dam release, the model was run as three separate modeling periods. Table 9 lists the well name, modeling period, thermal conductivity, hydraulic conductivity, and average vertical flux input values of hydraulic, thermal conductivity, and average vertical flux.
U.S. Department of the Interior | U.S. Geological Survey
URL: https://pubs.usgs.gov/sir/2007/5216
Page Contact Information: Publications Team
Page Last Modified: Thursday, 01-Dec-2016 20:00:30 EST
