Simulations to Verify Horizontal Flow Measurements from a Borehole Flowmeter

Transcription

1 Simulations to Verify Horizontal Flow Measurements from a Borehole Flowmeter by Scott C. James 1,RichardA.Jepsen 2,RichardL.Beauheim 3, William H. Pedler 4, and Wayne A. Mandell 5 Abstract This paper reports on experiments and simulations of subsurface flow from a slotted acrylic tube deployed in a sand-tank flow chamber for two different purposes. In the first instance, the slotted tube is used to represent a single fracture intersected by an uncased well. In the second instance, the slotted tube is used to represent a multislot well screen within a porous medium. In both cases, the scanning colloidal borescope flowmeter (SCBFM) measures ground water velocity within the well by imaging colloids traveling through a well to measure their speed and direction. Measurements are compared against model simulations. For the case of a slotted tube representing a single fracture, SCBFM and model results agree with respect to the flow direction and to within a factor of 1.5 for the speed near the well s center. Model and experimental agreement lend confidence that for an uncased well drilled in a fractured-rock medium, a calibrated SCBFM could be used to identify and quantify flowing features. Next, the SCBFM was deployed in a four-column multislotted casing with slots aligned with the flow direction. Another numerical model was developed to estimate the flow field within this well screen to evaluate the potential usefulness of employing the SCBFM in a screened well to estimate flow speed and direction in the surrounding porous medium. Results indicate that if the slots are not aligned with the flow, the SCBFM may only provide order-of-magnitude speed measurements and direction measurements with an uncertainty of approximately ±25. Introduction Hydraulic properties of aquifers traditionally have been estimated from laboratory measurements of core samples or from field estimates such as slug tests or pumping tests for composite sections of the aquifer. Although these approaches provide data that are valuable 1 Corresponding author: Sandia National Laboratories, Geohydrology Department, P.O. Box 5800, Albuquerque, NM ; scjames@sandia.gov 2 Sandia National Laboratories, Mechanical Environments, P.O. Box 5800, Albuquerque, NM ; rajepse@sandia.gov 3 Sandia National Laboratories, Repository Performance Department, 4100 National Parks Highway, Carlsbad, NM 88220; rlbeauh@sandia.gov 4 RAS Inc., 311 Rock Avenue, Golden, CO 80401; bpedler@ rasinc.org 5 U.S. Army Environment Center, SFIM-AEC-ERA, 5179 Hoadley Road, APG-EA, Aberdeen, MD 21010; wayne.mandell@ eac.apgea.army.mil Received March 2005, accepted July Journal compilation ª 2006 National Ground Water Association. No claim to original US government works. doi: /j x 394 Vol. 44, No. 3 GROUND WATER May June 2006 (pages ) for site assessment, they may undesirably average zones of preferential flow that are chief conduits of ground water contaminants. Similarly, horizontal velocity has traditionally been estimated on the basis of Darcy s equation (using measured gradients and hydraulic conductivity) or from transport of tracers under natural or forced-gradient conditions. Borehole tools capable of directly measuring horizontal ground water flow and direction in narrow intervals of the aquifer may provide vital information to hydrogeologists. That is, once calibrated in the laboratory, they can be effectively deployed in the field. Borehole flowmeter data indicate where ground water is entering and exiting the borehole and can assist in estimation of subsurface transport. Cross-borehole flow-logging tests can indicate the degree of connectivity of fractures beyond the wellbore, and transient tests can be used to estimate hydraulic properties (for example, transmissivity and storativity) of fractured media, but these data are spatially averaged and not locally detailed. Drost et al. (1968) developed and tested a scintillationcounter probe able to fit down a borehole that determined flow direction by tracing radioisotopic elements injected

2 into the borehole. More recently, interest has focused on directly measuring ground water speed and direction using downhole instrumentation that requires no tracer (e.g., Cronk and Kearl 1990; Kerfoot et al. 1991; Kearl et al. 1992, 1999; Kearl 1997; Ferriz and Pedler 1999; Kearl and Roemer 1998; Wilson et al. 2001; Beauheim 2000; Korte et al. 2000; Pedler and Jepsen 2003). Examples of these types of instruments include the horizontal heat-pulse flowmeter (KVA flowmeter) (Kerfoot 1982; Kearl et al. 1999; Melville et al. 1985), the acoustic Doppler velocimeter (Kraus et al. 1994), the laser Doppler velocimeter (Momii et al. 1993), the fixed-point colloidal borescope (Kearl and Case 1992), and hydrophysical logging (Anderson et al. 1993), which directly measure the horizontal component of velocity inside a wellbore. Applications of flowmeters to field assessments and controlled laboratory evaluations have led to a better understanding of the distribution of ground water flow in a borehole and the operational mechanics of various flowmeters. Nevertheless, the applicability of these instruments is uncertain when they are used in existing wells with standard slotted well casings. Borehole and well casing effects can negatively impact the measured flow speed and direction (Kerfoot 1988). For example, Dinwiddie et al. (1999) and Halford (2000) show that the presence of the flowmeter itself can significantly influence calculations of aquifer characteristics because of flow redistribution, especially in gravel-packed wells. Boman et al. (1997) provide an excellent overview of borehole flowmeter applications for aquifer characterization. Sandia National Laboratories, RAS Inc., and the U.S. Army Environment Center developed a sand-tank laminar flow test chamber to facilitate testing and calibration of the scanning colloidal borescope. In this work, a computational fluid dynamics code is used to qualitatively assess the likelihood that correct estimates of flow speed and direction in the surrounding medium can be inferred from speed and direction measurements made within a well screen using a colloidal borescope. First, the results from a numerical and experimental investigation for flow through a single slot in a well (approximating a fracture) are presented. Then, the borescope is deployed in a standard, slotted, fiberglass well screen and data are compared to a second model developed for this application. Further, the model is extended to examine different orientations of the well-screen slots with respect to the flow direction. Wood et al. 1997). The LLNL scanning colloidal borescope flowmeter (SCBFM) has a variable focal length with a 0.5-m range. The SCBFM is used to evaluate horizontal ground water speed and direction in a well. It comprises a charged-couple device (CCD) video camera, magnetometer, light source, and a remotely controlled, variable focal point lens mechanism to track colloidal-sized particles (1 to 5 lm) passing through a mm 2 field of view. Figure 1 is a schematic of the SCBFM used in this study. By recording the output of the CCD video camera and using advanced particle-tracking software, the speed and compass direction of colloidal particles advected by horizontal ground water flow within a well can be measured. The scanning feature allows a 0.5-m interval to be investigated without relocating the instrument, while facilitating a three-dimensional evaluation of the flow field within the well that can detect and characterize fast pathways, swirling, and secondary flow cells. The upper limit on speeds measurable by the SCBFM is defined by the width of the field of view ( mm 2 ) and the time interval between consecutive video frames (1/30 s). If a particle is traveling so fast that it appears in only a single video frame, its speed cannot be measured (the The Colloidal Borescope The first colloidal borescope was developed at Oak Ridge National Laboratories (ORNL). Kearl and Case (1992), Kearl (1997), and Kearl and Roemer (1998) have described the tool, its applications, and associated data analysis. The ORNL borescope had a fixed focal plane, requiring that the tool be repositioned whenever observations at different depths were desired. An improved colloidal borescope was developed at Lawrence Livermore National Laboratory (LLNL) (Ferry et al. 1995; Figure 1. Schematic of the SCBFM. S.C. James et al. GROUND WATER 44, no. 3:

3 theoretical upper limit is ~48 mm/s). Typically, a particle moves less than half the distance of the field between frame grabs, and the best results are for particles that move less than a quarter of the field (or 0.4 mm). It is not uncommon for a particle to disappear before reaching the edge of the field, indicating that it has some vertical speed. Thus, if a particle appears in only two grabbed frames and only moves 0.4 mm horizontally, then its vertical speed must have been no greater than ~25% of its horizontal speed, and no less than 12.5%. The absolute limit on vertical speed (assuming that a particle is counted on successive frames) is 0.1 mm/(1/30) s, or 3 mm/s. In practice, the vertical speed should be less than half the horizontal speed to allow a particle to cross an eighth of the horizontal field while passing vertically through the focal plane. Experimental Set-Up The Test Chamber Based on the work of Drost et al. (1968), a laboratoryscale horizontal flow test chamber was deployed at the Sandia Soil and Sediment Transport Laboratory in Carlsbad, New Mexico. This test chamber establishes a stable, horizontal flow in a porous medium with an observable gradient and pore velocities ranging from to m/s. The dimensions of the test chamber are 0.9 m (3.0 feet) wide, 1.2 m (4.0 feet) deep, and 2.1 m (7.0 feet) long. Two 15-cm-long (6 inch long) reservoirs are located at each end, yielding an overall length of 2.4 m (8.0 feet). A pump connecting the two reservoirs recirculates water through the test chamber. Once the test chamber was filled with sand to a depth of 1.2 m, the hydraulic conductivity of the porous medium was measured. The well-rounded, well-sorted, coarse sand (8/16) has a measured bulk porosity of Hydraulic conductivity was calculated by pumping m 3 /s (9.0 gal/min) through the tank, yielding a m head difference between the reservoirs. This is still within the laminar flow regime. Darcy s law states that 1.5 mm wide with 90 cutouts. The openings were oriented perpendicularly to the long axis (flowing direction) of the chamber. This was designed to simulate a simple horizontal fracture with a partial (50%) opening as shown in Figure 2. Three recirculation rates were employed during horizontal flow simulations, and the corresponding pore velocities were used in the numerical modeling. Single-Slot Well Casing Numerical Model The numerical models presented here were developed with CFD2000 (Adaptive Research 2002a). This general-purpose computational fluid dynamics software offers an integrated environment to build the model geometry, generate the computational mesh, specify boundary conditions and fluid properties, stipulate the solution method, and visualize the solution. CFD2000 solves the three-dimensional Reynolds-averaged Navier- Stokes equations, which provide for conservation of mass, momentum, energy, reacting species, and turbulence if necessary (Adaptive Research 2002b). The model provides procedures to optimize mesh spacing and orthogonality. The first model is of the single-slot, 10.2-cm-diameter well in porous media that is intended to simulate flow in an unscreened well through a single, water-conducting K ¼ QL yzh ð1þ where K is the hydraulic conductivity (m/s), Q is the discharge (m 3 /s), y is the width of the flow simulator (m), z is the saturated depth (m), and h is the head difference (m). L is the length of the test chamber (cm). The test chamber is 0.9 m wide with an average saturated depth of 0.9 m. Therefore, for L ¼ 2.1 m and h ¼ m, the hydraulic conductivity, K, is m/s. Hydraulic conductivity was calculated for a second pumping rate of m 3 /s (4.0 gal/min), also resulting in K ¼ m/s. The corresponding permeability is m 2. To simulate a simple horizontal fracture, two slots were cut in a 10.2-cm (4 inch)-o.d. acrylic tube, which was placed in the center of the test chamber as it was filled with sand. These diametrically opposed slots were Figure 2. Schematic of the simulated fracture. 396 S.C. James et al. GROUND WATER 44, no. 3:

4 fracture. This model is not perfectly representative of flow through a single fracture because flow through the porous medium will be focused on the single slot from all three dimensions, rather than just within the plane of the fracture. The model does, however, accurately simulate flow conditions in the laboratory test chamber, thereby facilitating calibration of the SCBFM. Fracture inlet velocities were estimated with a preliminary, coarsely meshed model using Cartesian coordinates with a total of 24, 19, and 18 cells in the x-, y-, and z-directions, respectively, yielding 8208 cells that model a m 3 domain within the test chamber. Boundary conditions include frictionless walls (that effectively model the much larger test chamber by removing model-domain no-slip boundary layer effects) on all sides except for the inlet, which has specified pore velocity, and outlet, which allows free flow dictated by conservation of mass. Outside the well screen is a porous medium with a specified permeability of m 2. The well casing is defined by (no-slip) walls everywhere except at the single slot. Slot inlet velocities are recorded for each modeled pore velocity (velocity specified at the inlet slots shown in Figure 2) and used in the more refined model of the single-slot well containing the borescope. A more detailed model of the interior of the well screen using a cylindrical coordinate system with a finely gridded mesh is shown in Figure 3. Boundary conditions for the cylindrical model include no-slip walls at the well casing, constant inlet velocity at the slot as specified by output from the preliminary model, and a free-flow, mass-conserving outlet slot. This model within a model decreases computational time because the refined grid necessary to model the borescope geometry accurately need not be carried throughout the entire flow simulator domain. The well slot (fracture) was centered in the test chamber, with the focal point of the borescope set at the middle of its 0.5-m range as shown in Figure 3. A total of 89, 14, and 36 cells were used in the z-, r-, and h-directions, respectively, yielding a model with 44,856 cells. The grid was refined in the vicinity of the fracture to increase model precision. Although the model was developed in cylindrical coordinates, the postprocessing tools for CFD2000 provide results in a Cartesian coordinate system. Nevertheless, because of the geometry of the test chamber, results presented in the Cartesian coordinate system make intuitive sense and can easily be compared to experimental data. Single-Slot Experimental and Model Results Several simulations were run at various inlet velocities with and without the borescope inside the single-slot well casing. Figure 4 is an example of the model results using the single-slot acrylic tubing. It shows a contour plot of the x-direction velocities (u velocities) along a horizontal plane, coplanar with the center of the slot (fracture). The fracture inlet is indicated in blue and the outlet in red. The inlet and outlet slots were simulated with a single cell in the z-direction, and those cells have both a flow condition and a wall condition associated with them. The postprocessing software gives precedence to the wall condition, and the red rim inside the well Figure 3. An isometric view of the model geometry and grid cells for flow in the well interior. The SCBFM is included in the model by inactivating (blocking) the model cells that it physically occupies. The plan view is shown in the lower right corner. S.C. James et al. GROUND WATER 44, no. 3:

5 Figure 4. The color scale is linear for velocity in the x-direction in m/s. The fracture inlet is on the left (indicated in blue) and flow is from left to right (negative x-direction) with the outlet indicated in red. The three support rods are shown. The inset in the lower left corner is a vertical cross section of the velocity field and that in the lower right corner is the horizontal cross section. Here, the slot inlet velocity is m/s. that crosses the inlet and outlet boundaries is a postprocessing artifact and thus displays a zero velocity for cells adjacent to the wall. Although significant dissipation of the u velocity is seen as the flow proceeds toward the center of the well, this does not imply a violation of conservation of mass. The lower left inset in Figure 4 is a vertical cross section of the velocity (vector) field that reveals circulation cells that appear both above and below the fracture horizon and slightly enhance the velocity in the fracture plane because both cells converge in that region. Toward the center of the well, flow is principally a narrow jet where the u velocity is progressively reduced in this jet because the flow begins to turn toward the vertical, thereby dissipating momentum in the horizontal direction. Furthermore, comparatively weak circulation cells are 398 S.C. James et al. GROUND WATER 44, no. 3: observed in the horizontal plane of the inlet on either side of the jet as shown in the lower right inset of Figure 4. Flow is focused (relative to the inlet velocity) just inside the central portion of the inlet by these circulation cells converging on this region. As shown in Figure 5, in the vertical direction, circulation cells are observed extending up to 15 cm above and below the plane of the slot. Model results indicate that the three support rods that connect the light source to the body of the borescope do not significantly interfere with the flow field near the center of the casing and, therefore, do not impact borescope measurements. Model runs were continued until steady state was achieved, typically at ~20 s of simulated time (~4 CPU hours on a Pentium GHz PC).

6 Figure 5. The x-z planar view is tilted 15 in the y-direction with m/s slot inlet velocity. Color scale is for velocity in the x-direction and flow is from right to left (negative x-direction in this figure). In addition to measuring flow at the plane of the fracture, the scanning capability (i.e., focusing at different depths in the borehole) of the SCBFM was used to investigate the flow field above and below the fracture at a flow rate yielding a pore velocity of m/s. Virtually no horizontal flow was measured by the SCBFM ~1 cm above and below the fracture horizon near the center of the well. Although there may have been a horizontal component to the flow in this region as shown by the model results, the vertical component was large enough to force colloids through the vertical viewing range of the SCBFM so quickly that a horizontal velocity was not detected. However, as shown in Figure 5, at distances between 10 and 15 cm above and below the fracture horizon, both the SCBFM and the model observed low-speed flow of a lesser magnitude in a direction opposite to that at the fracture. This delineates the outer limit of a circulation cell within the well caused by the relatively high flow velocity entering the well at the fracture inlet. Note that the top and bottom of the figure do not correspond to model-domain boundaries and that flow circulates beyond the range shown in Figure 5. Also, while model simulations ultimately yield a steady flow field, both laboratory and field measurement always S.C. James et al. GROUND WATER 44, no. 3:

7 12.58 cm (4.954") 10 Slots: Aperture = cm (0.002") Width = 6.35 cm (2.5") Vertical spacing = 0.8 cm (0.3125") Rib width = 3.5 cm (1.378") Solid pipe to bottom 10 cm (3.9") Figure 6. Schematic of the multislot well casing used in this work. show some level of transients (chaotic flows). These characteristics are inherent to the systems under study, and additional investigation is beyond the scope of this work. As expected, the SCBFM showed that the flow direction through the well was consistent with the flow direction through the test chamber as a whole. The velocities measured by the SCBFM near the center of the well and the modeled velocities near the center of the well are compared in Table 1. The model generally predicted velocities 67% to 82% of those measured by the SCBFM. This suggests that the SCBFM provides a reasonably accurate measurement of the flow in a well near an intersection with a single, water-bearing fracture. Table 1 also shows that flow is focused through the fracture inlet because the inlet velocity calculated by the model is over an order of magnitude greater than the pore velocity in the test chamber. (Note that because of the three-dimensional flow focusing discussed previously, comparing velocities in the well to pore velocities in the test chamber may not provide information relevant to flow in a single fracture.) By the time the flow reaches the center of the well, however, it has slowed significantly from the inlet velocity. Unfortunately, neither the velocity measured by the SCBFM nor the velocity predicted by the model near the center of the well appears to vary linearly from the inlet velocity (or the pore velocity). Consequently, fluid velocities in the fracture some distance from the well may prove difficult to quantify based only on SCBFM measurements in the well. Overall, for a single-slot acrylic tube aligned with the flow direction, the SCBFM provides accurate measurements of flow speed and direction in the well. Knowing how that flow velocity relates to the speed in a fracture, however, is problematic. Additional calibration and modeling may yield estimates of fracture flow rates through inverse modeling. The SCBFM may be better suited to indicate differences in flow velocities in different fractures than to determine absolute velocities. Table 1 Comparison of Model and Experimental Results for Inlet Flow Speeds and Flow Speeds Near the Center of the Well for the Single-Slot Case Pore Velocity in Test Chamber (m/s) Modeled Fracture Inlet Speed (m/s) Modeled Speed at Center of Well (m/s) Measured Speed at Center of Well (m/s) Not measured 1 1 Rate exceeded the capability of SCBFM software. 400 S.C. James et al. GROUND WATER 44, no. 3:

8 Multislot Well Casing Numerical Model Unlike the single-slot modeling that used the preliminary model shown in Figure 3 to define the inlet velocities for a second, more detailed model of the well, multislot modeling was performed using a refined version of the preliminary grid. The multislot well screen model geometry is of a cm-O.D. fiberglass tube with four columns of 58 slots as shown in Figure 6. Although a well screen can have several hundred slots along its length, computational limitations allowed only 10 rows of slots to be modeled. The overall model geometry is similar to that shown in Figure 3, with the well casing placed within a m 3 porous medium with a permeability of m 2. Frictionless walls are used on all sides of the model except at the specified velocity inlet and free-flow outlet to approximate an infinite medium surrounding the casing. Moreover, the frictionless upper boundary condition at z ¼ m effectively creates a plane of symmetry about the 10 rows of slots (only 0.4 cm of blank casing were modeled between the top slot and the boundary as opposed to 0.8 cm between consecutive slots), yielding a model equivalent to one that estimates flow inside a slotted well screen with 20 rows of slots. Because large aspect ratio cells are numerically unstable, 24, 24, and 104 cells were required in the x-, y-, and z-directions, respectively, yielding 59,904 cells. Grid refinement is included inside the well screen to improve model accuracy. Uniform pore velocity boundary conditions of and m/s were applied in the x-direction at the inlet (refer to Figure 3), corresponding to and m 3 /s pumping rates in the flow chamber, respectively. The orientation of the casing is described more usefully by referring to the directions as east, north, and high instead of x, y, and z, respectively. The model was run with: (1) the slots oriented with the flow direction (east); (2) the ribs oriented with the flow direction; and (3) an intermediate case. Model simulation was continued until steady state was achieved, typically at ~20 s of simulated time (~6 CPU hours on a Pentium GHz PC). The modeled flow fields inside the well screen are described subsequently. Multislot Experimental and Model Results Flow Field When Slots Align with the Flow Direction Figure 7 shows model results in a centered vertical cross section parallel to the imposed flow direction in the multislot well. The u velocity contours reveal hot spots at the lowest two slots. These hot spots represent end effects due to flow focusing where fluid in the porous medium below the slots has an upward velocity component as it approaches the well casing (path of least resistance), thereby increasing the flow into these slots and directing it slightly upward. This flow focusing is consistent with what was observed in the single-slot model as well. In addition, the superimposed velocity field shows a small recirculation eddy below the lowest slot that serves to enhance the u velocity near the bottom slot and also direct it slightly upward. These results suggest that the SCBFM should focus on points in the well casing at least eight slots from the ends of the screened interval where the magnitude of the z-component of the velocity is <10% of the x-component. A vector plot of the flow field inside the well screen at a horizontal cross section corresponding with the eighth slot (z ¼ m) when the slots are oriented with the flow is shown in Figure 8 for a pore velocity of m/s. The color of the arrows represents the Figure 7. Contour plot of the u velocity (x-direction) along a vertical cross section at y = 0. Velocity vectors are overlain and indicate that by the eighth slot from the bottom, nearly uniform flow in the x-direction is established within the casing. S.C. James et al. GROUND WATER 44, no. 3:

9 Figure 8. Vector plot of the flow field in a horizontal z cross section at the eighth slot from the bottom for a pore velocity of m/s when the slots are aligned with the flow direction. The arrows indicate flow direction and the color indicates the magnitude of the x-component of the velocity (u velocity). Note that near the center of the well and away from end effects, the u velocities are fairly uniform at ~ m/s. magnitude of the velocity in the x-direction (u velocity). For this orientation, the model results indicate maximum centerline velocities on the order of the pore velocity in the surrounding medium. Note the circulation cells at the north and south slots where flow both enters and exits the casing. A limited number of experiments was performed with the SCBFM in a multislot well screen with the slots aligned with the direction of flow. One measurement was made with a flow rate of m 3 /s (pore velocity of m/s), and six measurements (at different depths) were made with a m 3 /s flow rate (pore velocity of m/s). These results are compared to model results for the same pore velocities in Table 2. These results indicate that an SCBFM can be used to measure the speed and direction of ground water flow through a screened well if it is positioned a sufficient distance from well screen end effects and if the slots are Table 2 Comparison of Model and Experimental Results for Flow Speed Near the Center of the Well for the Multislot Well Screen Pore Velocity in Test Chamber (m/s) Modeled Speed at Center of Well (m/s) Measured Speed at Center of Well (m/s) to alignednormaltotheprincipal flow direction. Although the results in Table 2 show good agreement between the pore velocity in the test chamber and the velocities modeled and measured in the well, quantifying the relationship between the velocity observed by an SCBFM and the pore velocity in the surrounding formation may be problematic. The amount of flow focusing through individual slots would be a function of borehole size, ratio of formation permeability to sand/gravel pack permeability, and the particular screen geometry (slot width and aperture). Careful calibration using a laboratory flow chamber designed as closely to the field configuration as possible would be necessary to build confidence that SCBFM results can be extrapolated to actual pore velocities. Flow Field When Slots Do Not Align with the Flow Direction No experimental results were obtained with the SCBFM when the well screen ribs were aligned with the flow direction because we were unable to maintain the flow of colloids through the test chamber. The number of visible colloids decreased significantly during the experiments with the slots aligned with flow, and no colloids were visible by the time the experiments with the ribs aligned with flow were attempted. We believe that the sand in the test chamber progressively filtered colloids out of the water. When the ribs are aligned with the flow direction, significantly different modeled flow conditions are established inside the well casing. Modeled flow speeds near the center of the well are about one-third of those found in the pores of the surrounding medium with u velocities 402 S.C. James et al. GROUND WATER 44, no. 3:

10 of ~ m/s. Recirculation eddies are established in the northern and southern quadrants of the well. In addition, the velocity estimated near the center of the well casing, while uniformly flowing in the positive x-direction, is approximately half the magnitude when the slots are aligned with the flow. Only if it is ascertained a priori that the ribs are aligned with the flow could the SCBFM be properly calibrated to measure both flow speed and direction. Again, correlations between the pore velocity and measured and modeled well velocities could be inferred given sufficient calibration and known slot orientation. One additional configuration with the slot/rib interface aligned with the principal flow direction is investigated (the well screen is rotated 16 counterclockwise from the case when the ribs are aligned with the flow). In this case, flow near the center of the well screen at the eighth slot is directed 23 south of east, which is not surprising considering the configuration. Modeled flow speeds at the center of the well are about two-thirds as fast as the pore velocity in the surrounding medium, with u velocities near m/s and v velocities about m/s. Flow enters primarily through the northwest slots and exits through the southeast slots. Several small horizontal recirculation cells are present. What is most notable is that the flow direction near the center of the cell is not aligned with the principal flow direction but is shifted toward the axis of the slots. While the u velocity is between that seen in the previous cases, there is an additional v velocity in the negative y-direction (south) approximately half the magnitude of that in the x-direction (east). It is evident that the orientation of the slots with respect to the flow can significantly alter the modeled (and likely the measured) flow direction inside the well casing. Thus, not knowing the orientation of the well screen and the principal flow direction could lead to errors in the estimated flow direction of ±23 for this orientation of the slots. As was the case with the ribs aligned with the flow direction, no SCBFM results could be obtained when the slots and ribs were misaligned with the flow direction because of the apparent absence of colloids in the flow chamber. Results and Conclusions Numerical modeling demonstrates that the flow fields established within either single-slot (fracture) or multislot well screens is complex. Recirculation eddies are observed in different locations depending upon the orientation of the well screen with the principal flow direction. Also, within the well screen, flow may decrease by up to an order of magnitude from just inside the slots to the well center. Hence, flow in a screened well is focused both by the dimensions and properties of the sand/gravel pack and by the specific configuration of the well screen. Therefore, to relate flow velocity measurements made within a screened well to the pore velocity in the surrounding formation, calibration under laboratory conditions that are as close to the field conditions as possible is required. Fortunately, in a well calibrated system, numerical modeling may help to estimate both the slot inlet velocity as well as velocity in the surrounding medium. For the case of a well with a simulated single fracture in the test chamber, the SCBFM measures the correct flow direction. The speeds measured in the center of the well are as much as 33% higher than those modeled. This difference may be attributed to imprecise centering of the SCBFM in the borehole and/or differences between the modeled inlet speed and the actual speed in the test chamber, which are likely somewhat variable. Both the SCBFM and the model indicate that the flow speed measured at the center of the well is less than the inlet velocity, and this measured speed is sensitive to the centering of the SCBFM in the well. Care must be taken that the SCBFM is properly placed adjacent to a single fracture because misplacement could result in measurement of a vertical recirculation eddy that would yield lower speeds directed opposite to the local fracture flow. Future development could focus on improving and subsequently calibrating the SCBFM by incorporating scanning in the horizontal plane in addition to the vertical. This would be particularly important for single-slot or fracture cases when flow through the center of the well is concentrated in a narrow jet (Figure 4). Although a variety of screens is commercially available, only the five-slot type was examined. In practice, wirewrap-type screens would be the best to use in conjunction with an SCBFM. All field investigations for horizontal flow characterization should include a video survey to evaluate the type and condition of the screen. For a multislot well screen, when the slots are aligned with the flow direction and assuming sufficient calibration, SCBFM measurements near the center of the well can be used to estimate both flow speed and direction in the surrounding medium. However, based on model results only, not knowing the rib orientation with respect to the principal flow direction could lead to errors in the speed estimate of more than a factor of 2. Local flow direction could be estimated to within ±23 for the particular rib-slot configuration studied here. A multislot well casing may have small eddies above the uppermost and below the lowest slots, suggesting that the SCBFM should be deployed at least eight slots from the top or bottom of the screen. It should also be noted that recirculation eddies form in the horizontal plane and the SCBFM should be located (centered) to avoid them. Unfortunately, horizontal eddies develop in different magnitudes and locations depending upon the orientation of the ribs with respect to the flow direction, which could yield erroneous SCBFM approximations of the local flow velocity. Again, horizontal scanning might improve calibration. The SCBFM is a useful tool for evaluating flow speed and direction from individual fractures intersecting an unscreened well. One should keep in mind, however, that the speed and direction of flow within a single fracture at a single location is not likely to be representative of the average overall flow in the surrounding medium. In addition, the SCBFM can also be a useful velocimeter for S.C. James et al. GROUND WATER 44, no. 3:

11 multislot well screens. In this case, knowledge of the type of screen construction and how the slots align with the principal flow direction is important. Without this knowledge, which is difficult to obtain, the SCBFM can be relied on for order-of-magnitude speed measurements and direction measurements with an uncertainty of approximately ±25. Additional modeling and experimentation would be needed to assess the utility of the SCBFM in well screens with more columns of slots or wider slots than were used for the experiments described herein. The results presented here should underscore the necessity of performing computational fluid dynamics modeling for any device intended to measure flow inside a wellbore. It is imperative that this type of modeling be conducted for screened wells because the screen may significantly impact the internal flow field. Without understanding how the well screen, and potentially the measurement device, is affecting the flow in the well, measurements of flow speed and direction cannot be related to flow in the surrounding formation. Controlled laboratory experiments and detailed modeling should be used to calibrate any device intended to measure flow within a well. Acknowledgments Sandia is a multiprogram laboratory operated by Sandia Corporation, a Lockheed Martin Company, for the United States Department of Energy s National Nuclear Security Administration under contract DE-AC04-94AL Funding for this project was provided by the U.S. Army Environmental Center, DOE and RAS Inc. The authors acknowledge LLNL for contributing the SCBFM for this and other horizontal flow studies. The manuscript was improved by the helpful comments of Keith Halford and two anonymous reviewers. References Adaptive Research. 2002a. STORM/CFD2000 User Guide. Alhambra, California: Simunet Corporation. Adaptive Research. 2002b. STORM/CFD2000 Theoretical Background. Alhambra, California: Simunet Corporation. Anderson, W.P., D.G. Evans, and W.H. Pedler Inferring horizontal flow in fractures using borehole fluid electrical conductivity logs. EOS, Transactions of the American Geophysical Union Fall Meeting 74, no. 43: 305. Beauheim, R.L Evaluation of the colloidal borescope as a monitoring tool at the waste isolation pilot plant. Sandia Report SAND Albuquerque, New Mexico: Sandia National Laboratories. Boman, G.K., F.J. Molz, and K.D. Boone Borehole flowmeter application in fluvial sediments: Methodology, results, and assessment. Ground Water 35, no. 3: Cronk, T.A., and P.M. Kearl Colloidal borescope A means of assessing local colloidal flux and groundwater velocity in porous-media. In Manipulation of Groundwater Colloids for Environmental Restoration, ed. J.F. McCarthy and F.J. Webber, Boca Raton, Florida: Lewis Publishers. Dinwiddie, C.L., N.A. Foley, and F.J. Molz In-well hydraulics of the electromagnetic borehole flowmeter. Ground Water 37, no. 2: Drost, W., D. Klotz, A. Koch, H. Moser, F. Neumaier, and W. Rauert Point dilution methods of investigating ground water flow by means of radioisotopes. Water Resources Research 4, no. 1: Ferriz, H., and W.H. Pedler Borehole geophysics applied to the study of landfill sites in fractured bedrock terrains. In Proceedings of the Symposium on the Application of Geophysics to Engineering and Environmental Problems, ed. M.H. Powers, L. Cramer, and R.S. Bell, Wheat Ridge, Colorado: Environmental and Engineering Geophysical Society. Ferry, R.A., L.R. Rueth, R.K. Landgraf, B.J. Qualheim, and P.M. Kearl Direct ground water flow direction and velocity measurements using the colloidal borescope. UCRL-JC Livermore, California: Lawrence Livermore National Laboratory. Halford, K.J Simulation and interpretation of borehole flowmeter results under laminar and turbulent flow conditions. In Proceedings of the 7th International Symposium on Logging for Minerals and Geotechnical Applications. Kearl, P.M Observations of particle movement in a monitoring well using the colloidal borescope. Journal of Hydrology 200, no. 1 4: Kearl, P.M., and C.M. Case Direct field measurement of ground water velocities. In Interdisciplinary Approaches in Hydrology and Hydrogeology, ed. M.E. Jones and A. Laenen, Minneapolis, Minnesota: American Institute of Hydrology. Kearl, P.M., N.E. Korte, and T.A. Cronk Suggested modifications to ground water sampling procedures based on observations from the colloidal borescope. Ground Water Monitoring and Remediation 12, no. 2: Kearl, P.M., and E.K. Roemer Evaluation of groundwater flow directions in a heterogeneous aquifer using the colloidal borescope. Advances in Environmental Research 2, no. 1: Kearl, P.M., E.K. Roemer, E.B. Rogoff, and R.M. Renn Characterization of a fractured aquifer using the colloidal borescope. Advances in Environmental Research 3, no. 1: Kerfoot, W.B Monitoring well construction and recommended procedures for direct ground water flow measurements using a heat-pulsing flowmeter. In Ground Water Contamination: Field Methods, vol. 963, ed. A.G. Collins and A.I. Johnson, Philadelphia, Pennsylvania: American Society for Testing and Materials. Kerfoot, W.B Comparison of 2-D and 3-D ground water flowmeter probes in fully-penetrating monitoring wells. In Proceedings of the Second National Symposium on Aquifer Restoration and Ground Water Monitoring, ed. E.D. Nielson, Worthington, Ohio: National Water Well Association. Kerfoot, W.B., G. Beaulieu, and L. Kiely Direct-reading borehole flowmeter results in field applications. In Proceedings of the Fifth National Outdoor Action Conference on Aquifer Restoration, Ground Water Monitoring and Geophysical Methods, Dublin, Ohio: National Water Well Association. Korte, N., P.M. Kearl, R.L. Siegrist, M.T. Muck, and R.M. Schlosser An evaluation of horizontal recirculation using single-well tests, pumping tests, tracer tests, and the colloidal borescope. Ground Water Monitoring and Remediation 20, no. 1: Kraus, N.C., A. Lohrmann, and R. Cabrera New acoustic meter for measuring 3D laboratory flows. Journal of Hydraulic Engineering 120, no. 3: Melville, J.G., F.J. Molz, and O. Güven Laboratory investigation and analysis of a ground water flowmeter. Ground Water 23, no. 4: Momii, K., K. Jinno, and F. Hirano Laboratory studies on a new laser Doppler velocimeter system for horizontal groundwater velocity measurements in a borehole. Water Resources Research 29, no. 2: S.C. James et al. GROUND WATER 44, no. 3:

12 Pedler, W.H., and R.A. Jepsen Laboratory and numerical evaluation of borehole methods for subsurface horizontal flow characterization. Sandia Report SAND Albuquerque, New Mexico: Sandia National Laboratories. Wilson, J.T., W.A. Mandell, F.L. Paillet, E.R. Bayless, R.T. Hanson, P.M. Kearl, W.B. Kerfoot, M.W. Newhouse, and W.H. Pedler An evaluation of borehole flowmeters used to measure horizontal ground-water flow in limestones of Indiana, Kentucky, and Tennessee, Water- Resources Investigations Report Indianapolis, Indiana: USGS. Wood, D.K., R.A. Ferry, and R.K. Landgraf Direct ground water flow direction and velocity measurements using the variable-focus colloidal borescope at Sandia National Laboratories, Albuquerque, New Mexico. UCRL- AR Livermore, California: Lawrence Livermore National Laboratories. S.C. James et al. GROUND WATER 44, no. 3: