Water Resources Research

Using radon as an environmental tracer for estimating groundwater flow velocities in single-well tests



[1] Naturally occurring radon-222 was evaluated for its use in estimating groundwater flow velocities using single-well tests. Investigations were carried out for four different well scenarios, which revealed the advantages and limitations of the approach. On one hand, it was shown that radon is useful as an environmental tracer because of (1) the low costs of the method, (2) the avoidance of any artificial tracer injection into the aquifer, (3) the immediate availability of results, and (4) the need for only a single monitoring well. On the other hand, several potential sources of error were identified, including poor sampling, inadequate hydraulic connection of the well because of a clogged screen, and an unsuitable well diameter resulting in excessively long or short well water residence times. The practical approach is supported by in-depth theoretical considerations. General recommendations are presented concerning the use of radon as an environmental tracer for groundwater flow assessment.

1. Introduction

[2] Groundwater flow velocities, i.e., migration rates along the hydraulic gradient from high to low hydraulic head, can generally be obtained by application of Darcy's law. This conventional approach implies that the slope of the piezometric surface between two groundwater monitoring wells can be assumed constant. Hence, assessment of small-scale groundwater migration patterns requires groundwater wells separated by distances comparable to the distance over which significant changes of the hydrological system can be expected. If an aquifer domain is assessed based on widely spaced wells, small-scale spatial variations in the groundwater migration pattern might not be resolved. In such cases, single-well tracer tests can be applied as an alternative to multiwell approaches [Klotz, 1977; Drost and Hoehn, 1987].

[3] Conventional single-well tests for the determination of groundwater flow rates have primarily employed salt tracers, fluorescence dye tracers, or artificial radioactive components [Drost and Hoehn, 1987; Schroth et al., 2001; Azizian et al., 2005; Gouze et al., 2008]. These types of tests rely on the inflow of “fresh” groundwater into a well and the resulting decrease in tracer concentration there. This decrease can be measured as a function of time and used for a quantitative assessment of the magnitude of groundwater flow through the well.

[4] However, two major disadvantages of the conventional single-well approaches can be identified: they require constant water-quality monitoring, i.e., the recording of time series, and (in most cases) time-consuming laboratory analysis. A third major drawback is the necessity of injecting an artificial tracer into the groundwater, which may carry legal/regulatory implications or can change the natural hydrogeochemical conditions. Hence, the use of “environmental tracers,” i.e., of substances that occur naturally in the groundwater, is generally preferable to the use of artificial components.

[5] A prerequisite for using a naturally occurring substance as an environmental tracer in a single-well test is that its concentration in the well must change as a function of the groundwater flow velocity, which determines the residence time of the groundwater (and thus the tracer) in the well. This paper shows that the naturally occurring radionuclide 222Rn, hereafter referred to as “radon,” meets this requirement, thus allowing its use in single-well tests for assessing groundwater migration.

[6] A case study introducing this concept was published by Hamada [2000]; however, the study was restricted to very low groundwater flow velocities of only a few millimeters per day. The monitoring wells had a diameter of 38 mm and were installed in mudstone and its weathered derivatives. As an outcome, the author concluded that the method was applicable for assessing groundwater flow velocities ranging between 1 and 10 mm/d. Another, somewhat simplified, mathematical approach was published by Cook et al. [1999]. The authors assumed a perfectly mixed water column in the screened well section. In contrast to the Hamada approach, which uses infinite elements as described in section 2.2 in more detail, the Cook solution is based on a differential equation assuming that the net concentration inflow equals the net concentration outflow plus decay. The Cook solution shall not be discussed further here; for details, see Cook et al. [1999].

[7] In this study, the approach introduced by Hamada [2000] has been applied to higher velocities, i.e., velocities that can usually be expected in sand and gravel aquifers. To further improve the concept, its theoretical framework was modified in order to account for the convergence of groundwater streamlines toward a well, which is because of the different hydraulic conditions within the well and its immediate surroundings. While Hamada [2000] did not consider convergence, Cook et al. [1999] mentioned it. Cook et al. [1999] semiquantitatively assessed that convergence causes flow velocities through the well that exceed those through the aquifer by factors of between 2 and 4. For an uncased borehole, they quantitatively suggested a factor of 2, an assumption that is in accordance with our theoretically derived result (equation (6), section 2.2).

[8] This paper discusses observed data at two study sites. Results from four wells with significantly different setups are presented to show the suitability of radon as a single-well tracer. In order to reveal not only the general applicability and specific advantages of the method but also its limitations, the wells were chosen in order to represent both well-suited and less well-suited conditions.

2. Theoretical Background

2.1. Natural Occurrence of Radon in Groundwater

[9] A main reason for the suitability of radon as a single-well tracer is its ubiquitous occurrence in groundwater. In every aquifer matrix, radon is constantly produced by the radioactive decay of the ubiquitously occurring radioisotope radium-226. The radium-226 concentration of the mineral matrix, the net radon emanation into the water-filled pore space, the specific porosity of the aquifer matrix, and its bulk dry density determine the natural radon background concentration in the groundwater equation image. Radon equilibrium concentrations in groundwater generally range between about 1 and 50 Bq/l but can be much higher as result of a high radium concentration in the aquifer matrix [Andrews and Wood, 1972; Rama and Moore, 1984; Tanner, 1980]. Its continuous supply by emanation from the aquifer matrix and its constant decay lead to a steady-state equilibrium concentration in all groundwater systems that are not subject to short-term varying interaction with a surface water body such as a lake or river [Schubert et al., 2006a] or the ocean [see, e.g., Santos et al., 2009].

[10] However, as soon as groundwater enters a monitoring well, radon decay is no longer balanced by radon production. This disturbance of the source/decay steady-state equilibrium is because of the loss of immediate contact between groundwater and the solid aquifer matrix. The short radon half-life of 3.8 days results in a fast (and mathematically reproducible) concentration decrease in the well water. The predictable radon deficit can therefore be used as a quantitative measure of its residence time in the well and thus for determining the groundwater flow velocity through it.

[11] The decrease of the initial radon equilibrium concentration in the groundwater (natural radon background equation image) because of the absence of radon production is quantified in equation (1):

equation image

where lambda (equation image) represents the radon decay constant (2.1 × 10−6 s−1) and equation image is the radon concentration in the well water after time t has elapsed. The initial radon concentration in the groundwater and its relatively fast decrease in the well water are the central parameters permitting the use of radon as an environmental tracer for the applications presented in this paper.

[12] In addition to the points discussed above, radon is chemically inert (being a noble gas), which prevents retardation in the aquifer, and is thus beneficial as an aquatic tracer. A further advantage is the relatively simple and low-cost way of detecting and measuring radon in water concentrations. The related practical aspects of sampling and on-site analysis will be discussed later in more detail (see section 3).

2.2. Mathematical Approach

[13] Under steady-state flow conditions (which, as mentioned above, can generally be assumed for natural aquifers), the residence time (and hence the radon deficit) of groundwater in a monitoring well depends on the radius (r) of the well and on the velocity of the groundwater flow (“well water”) passing through it (vww). Since the well radius is known, the radon deficit equation image can be used as a direct quantitative measure for the groundwater flow velocity through the well.

[14] As mentioned above, the first theoretical concepts of this approach were given by Cook et al. [1999] and Hamada [2000]. Cook et al. [1999] used a simplified approach and considered the water volume in the well as perfectly mixed. The approach of Hamada [2000] is more realistic as it accounts for the fact that radon's short half-life and its low-diffusion coefficient in water leads to a heterogeneous distribution of radon in the wellbore. The Hamada [2000] approach, on which our study is based, is illustrated in Figure 1 and is briefly summarized in the following.

Figure 1.

Schematic sketch of a subarea (dA; shaded gray) of the sectional area of the well with length 2x = 2r cos equation image and width y = r cos equation image d equation image (modified after Hamada [2000]).

[15] In the Hamada approach, the cross-sectional area (A) of the well is subdivided into subareas (dA; shaded gray in Figure 1), each with a length 2x and an infinitesimally small width y. From application of elemental trigonometric relations, it follows that each subarea has a length 2x = 2requation image and a width y = dUequation image, with U being the circumference of the well and dU an infinitesimally small segment of it. On the basis of the concept illustrated in Figure 1, the radon inventory corresponding to each individual subarea equation image, which depends on the initial radon concentration of the groundwater entering the well equation image and on the well water flow velocity (vww), can be quantified as

equation image

[16] The water residence time for each subarea can be expressed as t = 2r cos equation image/v (see Figure 1). Accordingly, the integral in equation (2) can be written as equation image

[17] Consequently, by integrating equation (2) for the complete cross-sectional area of the monitoring well equation image, i.e., from equation image to equation image, the radon inventory equation image is given by

equation image

[18] After rearrangement of equation (3), the radon deficit in the well water, being a function of the well radius and the groundwater flow velocity, can therefore be quantified as

equation image

[19] Equation (4) shows that the easily measurable radon deficit equation image detected in the well water (integrated over the whole sectional area equation image) depends quantitatively on the known radius of the well (r), the known decay constant equation image, and the groundwater flow velocity through the well (vww), which is the parameter of interest.

[20] An important aspect, which was not considered by Hamada [2000], is the fact that because of the different hydraulic conditions in the aquifer and well (including filter screen and filter pack), groundwater flow is generally characterized by a convergence of the streamlines toward the well (see Figure 2); [Drost et al., 1968];. Cook et al. [2000] did mention flow convergence around the well, but quantified it only for the case of an uncased borehole. In equation (4), the groundwater flow velocity represents the flow velocity inside the well. Depending on the difference in the hydraulic conditions between the aquifer and well, the velocity of the well water (hereafter referred to as vww) can be substantially higher than the groundwater flow velocity in the host aquifer (hereafter referred to as vgw).

Figure 2.

Schematic horizontal flow pattern (top view) within and around a well (dotted line) (after Englert [2003]).

[21] Several approaches to quantify this effect have been developed. Drost et al. [1968] found an empirical relationship between inflow width and well radius (Figure 2) and introduced a horizontal convergence factor equation image (equation image, equation (5)) that is based on the well geometry and all relevant hydraulic conductivity data (equation (6)). By applying equation image, the groundwater flow velocity in the undisturbed aquifer (vgw) can be quantitatively related to the flow velocity determined inside the well (vww) by using the relationship

equation image


equation image

[22] In equation (6), K1 is the hydraulic conductivity of the filter screen, K2 is the hydraulic conductivity of the filter pack, K3 is the hydraulic conductivity of the aquifer, r1 is the inside radius of the filter screen, r2 is the outside radius of the filter screen, and r3 is the borehole radius, i.e., the outer radius of the filter pack (see Figure 3); [Drost et al., 1968].

Figure 3.

Schematic horizontal cross section of a screened borehole with filter pack (after Drost et al. [1968]).

[23] If the well is constructed without any filter pack, or if the hydraulic conductivity of the filter pack can be assumed equal to the hydraulic conductivity of the aquifer (r3 = r2 and K3 = K2), equation (6) simplifies to equation (7):

equation image

[24] If the well is an open borehole without any filter screen and filter pack (r3 = r2 = r1 and K3 = K2 = K1), equation (6) simplifies to equation image, a relation which was already developed empirically and published by Ogilvi [1958] and applied by Cook et al. [1999].

[25] Another fact that might be of relevance is the occurrence of a vertical flow component because of a vertically heterogeneous distribution of the hydraulic conductivity, leading to a compartmentalization of the well volume. That potential source of error can, if necessary, be at least reduced by using packers that isolate the sections of interest along the well screen. If a packer is used, vertical convergence plays a role in addition to its horizontal component, and the water inside the well flows even faster compared to the groundwater in the undisturbed aquifer. For packer lengths (lP) of up to 0.5 m, Klotz [1977] developed an empirical relationship for a vertical convergence factor equation image, which is given in equation (8) and which has to be applied comparable to equation image (see equation (5)). However, in the study discussed here, no packers were used for water sampling. Thus, the vertical convergence factor does not apply:

equation image

[26] Finally, it shall be mentioned that, even if no compartmentalization is expected, it should be guaranteed that the water samples are taken from a depth that is not influenced by radon degassing into the unsaturated zone of the soil or into the air-filled well tubing. If the water in the well is assumed to flow slowly without any turbulence, radon diffusion is the main driver for degassing. Applying a diffusion coefficient for radon in water of 10−9 m2/s, it can be assumed that diffusion affects only the top 10 cm of the water column. However, the actual sampling procedure brings turbulence into the water, regardless of whether the water is pumped or bailed from the well. On the basis of these considerations, a sampling depth of at least 0.5 m below the water table should be allowed to minimize errors in water-sampling results. The suggested value is not a result of actual depth profile measurements but is rather a professional judgment based on the fact that 0.5 m is almost an order of magnitude greater than the depth potentially affected by diffusion alone.

3. On-Site Activities

3.1. Site Description

[27] In order to evaluate the applicability of the theoretical concept discussed above, two field campaigns were carried out at two different sites. Four groundwater wells were sampled (two at each site). All wells were screened in aquifer sections that had been previously investigated, which allowed comparing results to (yet unpublished) data gained during former hydrological surveys [cf. Dohrmann, 2000]. The wells chosen for the study are relatively new and well maintained, thus, permeability reductions because of clogging of the screened sections could be assumed to be negligible.

[28] At the first site (site A), two monitoring wells were chosen that are installed in two hydraulically dissimilar aquifer sections, an upper one consisting of a homogeneous mixture of gravels and sands, and a lower one consisting of silty fine sands. Well A1 is screened in the upper highly permeable aquifer; well A2 is fed by groundwater discharging from the underlying low-permeability section. The technical parameters of the two wells and the hydraulic conductivities of the two aquifer sections are given in Table 2 (for abbreviations see Table 1).

Table 1. Symbols and Notations Used Throughout the Paper
rwell radius
Awell sectional area
twater residence time in the well
equation imagenatural radon background concentration in the groundwater
equation imageradon concentration in the well water after the time t has elapsed
vwwgroundwater flow velocity through the well
vgwgroundwater flow velocity in the aquifer
equation imagehorizontal convergence factor
equation imagevertical convergence factor
r1, r2, r3inside radius of the filter screen, outside radius of the filter screen and outer radius of the filter pack, respectively
K1, K2, K3hydraulic conductivity of the filter screen, the filter pack, and the aquifer, respectively
Table 2. Well and Aquifer Parameters Applicable for Sites A and B Including Reference Groundwater Flow Velocities Based on Numerical Modeling
 r1 (cm)r2 (cm)r3 (cm)K1 (m/s)K2 (m/s)K3 (m/s)equation imageVgw Reference (cm/day)
Well A112.012.525.010−310−310−32.085–8
Well A212.012.525.010−310−310−63.350.5–2
Well B16.06.2515.010−310−38 × 10−42.2725–50
Well B26.06.2515.010−310−38 × 10−42.2725–50

[29] The second field experiment (site B) was carried out at a site with monitoring wells installed in a highly permeable gravel/sand aquifer with a homogeneous grain-size distribution but containing significant spatial variability in background radium and hence radon concentrations. Again, two wells, similar to each other and about 200 m apart, were chosen for sampling (B1 and B2).

[30] For the construction of all four wells, standard slotted screens with a slot width of 0.5 mm were used. The filter pack consists of fine gravel with a grain size between 2 and 6 mm. All relevant physical and hydraulic data are summarized in Table 2.

[31] All radon-in-water analyses were carried out directly in the field. On-site analysis of either a discrete sample or a running pump stream was accomplished by applying the mobile radon-in-air monitor AlphaGuard (Saphymo, Frankfurt, Germany) as discussed by Schubert et al. [2006b].

3.2. Determination of equation image

[32] The most difficult step of the sampling procedure is taking a sample representative of equation image, i.e., a sample representative of the well water, “uncontaminated” by any “fresh” groundwater drawn from the aquifer. For determination of equation image, one must ensure that (1) the sample is taken from the screened section of the well, (2) the radon concentration in the sampled well section is not influenced by degassing of the water, and (3) no fresh groundwater is sucked in from the aquifer during sampling thus contaminating the equation image sample.

[33] Two methods of accurate sampling for equation image were applied: grab sampling and pumping. In both cases, the sample volume should, on the one hand, be as small as possible in order to achieve a sample representative for equation image (i.e., no contamination with fresh groundwater), but, on the other hand, it should be large enough to guarantee reliable counting statistics even at low concentrations.

[34] If laboratory analysis (Liquid Scintillation Counting) is chosen for radon detection, water samples of about 20 mL are sufficient [e.g., L'Annunziata, 1998]. However, if the well setup allows representative samples of larger volumes, on-site radon detection, as described by Schubert et al. [2006b], can be applied in order to avoid laboratory analysis and additional costs, as well as radon losses because of sample handling.

[35] At site A, samples for determining equation image of 1 L each were taken using a specially designed grab sampler (bailer). With the well dimensions given in Table 2, each meter of filter screen represents about 45 liters of water. Thus, a 1 L water sample taken with a bailer can be considered representative of the stagnant well water. The sampler design allowed sampling from a certain depth in the well (screened section) and sample transfer into the applied radon stripping unit without any air contact, thus avoiding degassing of the sample. In order to obtain reproducible results, two equation image samples from each of the wells A1 and A2 were taken and analyzed.

[36] For determination of equation image at site B, the two wells were sampled by pumping the water (0.25 liters for each sample) from the desired well depth directly into the stripping unit. Avoiding sample contamination with fresh groundwater was accomplished by moving the end of the sampling tube slowly downward in the screened section of the well while applying a very low pump rate of 0.25 l/min. A weight was attached in order to keep the end of the tube (more or less) in the center of the screened well casing. This approach ensured that no substantial turbulent flow was generated in the well. Again, for increasing the reproducibility of the equation image data, two samples were taken from each of the wells B1 and B2.

[37] In each case (sites A and B), the water sample was transferred into a stripping unit for sample analysis, with the stripping units being of appropriate size for the sample volumes chosen. The stripping units were part of the experimental setup allowing determination of the radon concentration directly on site. A defined volume of air was pumped in a closed loop through the sample sitting in the stripping unit and through the radon-in-air monitor attached to it. In the stripping unit, the circulating air is sparged through the water sample. The radon that is dissolved in the water sample partitions into the airstream in accordance with the radon-partitioning coefficient between water and air for the given temperature. The equilibrium radon concentration in the airstream is recorded in the radon-in-air monitor. It can easily be converted into the original radon concentration in the water sample using the radon-partitioning coefficient and the volumes of air and water as described by Schubert et al. [2006b].

3.3. Determination Of equation image

[38] Compared to the difficulties related to representative sampling for equation image, determination of equation image is less error prone. If sampling is carried out not before water has been pumped from the well until all relevant parameters (temperature, redox conditions, pH, electrical conductivity, and radon concentration) are constant in the pumping stream, a 100% fresh groundwater sample can be guaranteed (no “contamination” with stagnant well water). A less complicated, but equally reliable, approach for well purging is replacing the stagnant water with fresh groundwater by pumping three well volumes, assuring complete water replacement. In contrast to the determination of equation image, no discrete sample needs to be taken for the determination of equation image; the running groundwater pumping stream can be analyzed constantly and directly. Continuous fast-response radon measurements were obtained by passing 2.5 l/min water through the commercially available stripping unit RAD-Aqua (Durridge Company, USA). At all four wells, the radon equilibrium concentration in the air loop led through the RAD-Aqua and the radon-in-air monitor was recorded for 45 min, applying a 1 min counting cycle. The resulting 45 concentration values determined at each well allowed a statistically reliable determination of equation image.

4. Results

[39] The radon concentrations determined in the stagnant water equation image of the wells at site A (wells A1 and A2) and site B (wells B1 and B2), the respective radon background concentrations of the groundwater equation image, and the resulting radon deficits averaged for the two samples taken from each of the four wells equation image are summarized in Table 3. The standard deviations of the equation image ratios were calculated according to the law of propagation of uncertainty [ISO/IEC, 2008]. The groundwater velocities in the wells (vww) derived from the respective radon deficits given in equation (4) are also shown in Table 3. The uncertainties of the well water velocities vww can be derived graphically from Figures 4a and 4b. From the values for vww, the groundwater flow velocities (vgw) were derived by applying equations (5) and (6).

Figure 4a.

Radon deficit versus well water flow velocity valid for r1 = 12 cm shown for well A1 (graph a) and well A2 (graph b); uncertainty ranges are shaded gray.

Figure 4b.

Radon deficit versus well water flow velocity valid for r1 = 6 cm shown for B1 (dashed line) and B2 (dotted line); uncertainty range is shaded gray.

Table 3. Radon Concentrations, Radon Deficit, and Flow Velocities Determined at Sites A and B
 222Rn (Bq/l)equation imagevww (cm/d)vgw (cm/d)
Site A
equation image14.00.833 ± 0.0149.64.6
equation image14.40.833 ± 0.0149.64.6
equation image17.10.833 ± 0.0149.64.6
equation image17.00.833 ± 0.0149.64.6
equation image0.30.113 ± 0.016< 0.4< 0.1
equation image0.40.113 ± 0.016< 0.4< 0.1
equation image3.10.113 ± 0.016< 0.4< 0.1
equation image3.10.113 ± 0.016< 0.4< 0.1
Site B
equation image13.60.975 ± 0.00436.316.0
equation image13.80.975 ± 0.00436.316.0
equation image14.00.975 ± 0.00436.316.0
equation image14.10.975 ± 0.00436.316.0
equation image41.90.979 ± 0.00543.519.2
equation image42.10.979 ± 0.00543.519.2
equation image43.00.979 ± 0.00543.519.2
equation image42.80.979 ± 0.00543.519.2

5. Discussion

5.1. Site A

[40] In well A1, a 17% reduction of the radon concentration compared to the concentration background was determined. This indicates an average residence time of the groundwater in the well of about one day (equation (1)). Considering the radius of the screened well section and the calculated radon deficit, a flow velocity of the water in the well of vww = 1.11 × 10−6 m/s (≈ 9.6 cm/d) is obtained (illustrated in Figure 4a, graph b). By taking into account the difference of the hydraulic conditions between the well and aquifer, i.e., by considering the horizontal convergence of the groundwater streamlines toward the well, a groundwater flow velocity of vgw = 5.32 × 10−7 m/s (≈ 4.6 cm/d) is derived by using equations (5) and (6)equation image (see Table 3). This value is in good agreement with data derived from numerical modeling of the sand/gravel aquifer hydraulically connected to well A1. Model parameters such as hydraulic conductivity were obtained by physical sediment analysis during installation of the monitoring wells [Dohrmann, 2000].

[41] The radon deficit determined in well A2 is much more pronounced. Because of the very low permeability of the silty fine sand that is hydraulically connected to A2, the residence time of the water in the well is roughly 2 weeks, causing a radon deficit of 89%. Only 11% of the radon background value was detected in the samples taken from the stagnant well water. With the given well radius, this radon deficit indicates a theoretical flow velocity of the groundwater in the well of about vww = 4.63 × 10−8 m/s (≈0.4 cm/d) (Figure 4a, graph b). However, at such slow groundwater flow velocities, radon diffusion from the aquifer into the well starts to contribute significantly to the radon inventory in the well. This is because of the fact that groundwater migration velocities of less than 10−7 m/s are less effective in moving radon away from its source than diffusion [Tanner, 1980]. The radon inventory in A2 can thus not be attributed to groundwater movement alone. Consequently, sound interpretation of the radon deficit detected in well A2 allows only a semiquantitative estimation of the groundwater flow velocity. By applying a convergence factor of α = 3.35, a value of vgw < 0.1 cm/d can be estimated for well A2 (see Table 2). This example demonstrates a limitation of the discussed method if the groundwater flow velocity is too slow to allow neglecting the influence of radon diffusion.

5.2. Site B

[42] Although both wells are installed in the same aquifer, the background radon concentrations in the groundwater are significantly different, indicating a considerable variance in the radium concentrations within the aquifer (the interstitial distance between B1 and B2 is about 200 m). However, in spite of the difference in equation image, the radon deficits determined in both wells are the same, indicating identical hydrological conditions. The values for equation image determined in B1 and B2 are 97.5% and 97.9% of the background concentrations, respectively. The very small radon deficit of about 2% indicates an average residence time of the groundwater in the wells of only about 3 h.

[43] Considering the well radii and the radon deficits detected in B1 and B2, flow velocities of vww = 36 and 44 cm/d, respectively, are obtained by using equation (4) (Table 3; Figure 4b). The hydraulic conditions that characterize wells and aquifer at site B result in a horizontal convergence factor of α = 2.27, which leads to groundwater flow velocities in B1 and B2 of vgw = 16 and 19 cm/d, respectively.

[44] The values determined at site B are in reasonable agreement with numerical model data for the sampled aquifer [Dohrmann, 2000]. However, they also show a limitation of the method if the residence time of the groundwater in the well is too short, i.e., if the radius of the well is too small with regard to the groundwater flow velocity. In the case of radon deficit values in the well water of less than about 10%, reliable data interpretation becomes difficult even if a very precise on-site analysis of the radon concentration can be guaranteed. This is because at small radon deficit values, small variations in equation image lead to considerably different results for vww and thus vgw (see Figure 4b). Such minor radon deficits result from residence times of the well water of less than about 12 h, i.e., from groundwater entering a well with a relatively small diameter at a relatively high flow velocity. On the other hand, and as discussed for well A2, flow velocities of less than 1 cm/d complicate data evaluation because of the increasing impact of radon diffusion to the radon inventory of the water column in the well.

[45] According to manufacturer's information, the hydraulic conductivity of both the well screen and the filter pack is 10−3 m/s. This information is valid for both sites, A and B (see Table 2). The values are associated with an uncertainty of ±10%, respectively. For the four well setups sampled at the two sites, this uncertainty may impact the horizontal convergence factor (see equation (6)) and thus the estimated flow velocity by at most 4%, i.e., insignificantly.

6. Conclusions

[46] The advantages of using radon as an environmental tracer in single-well tests for estimating groundwater flow velocities are (1) the low costs of the methodical approach, (2) the avoidance of artificial tracer injection into the aquifer, (3) the necessity of only one groundwater monitoring well, and (4) the possibility of immediate availability of results (no time-consuming lab analysis needed).

[47] Potential sources of error or uncertainty include (1) degassing of the water sample during sampling, (2) contamination of the equation image sample with fresh groundwater and of the equation image sample with remaining stagnant well water, and (3) clogging of the screened well section, which can decrease its permeability. The resulting “skin effect” between the aquifer formation and the well can, to a certain degree, prevent flow lines near the border streamline from entering the well. The effect will most likely be found in wells that are not well maintained and have not been purged for long periods of time.

[48] Using radon as a single-well velocity tracer requires a well diameter that allows, at the given groundwater flow velocity, an average residence time of the water in the well of about 0.5 to 2 radon half-lives (i.e., about 2 to 8 days). The four wells discussed in this work were chosen in a way to reflect three different scenarios: (1) conditions well suited for the radon method (well A1), (2) a well diameter less than optimal at the given groundwater flow velocity (wells B1 and B2), and (3) an insufficient groundwater flow velocity for which the impact of radon diffusion becomes non-negligible, allowing only semiquantitative results (well A2).

[49] In addition, it has to be kept in mind that the approach relies, as in many other tracer applications, exclusively on effects within the well, and thus can be affected by this artificial aquifer discontinuity. Future research in the field should therefore focus on the application of tracers that allow assessing the local groundwater flow velocity by using parameters derived independently within the aquifer. A promising step in this direction has been made by Istok et al. [1997, 2002], who applied radon in single-well push-pull experiments.