The width of a mixing zone between freshwater and seawater is important primarily because it directly reflects the extent of mixing and the growth and decay of the mixing zone indicates changes of the flow regime and water exchange between freshwater and coastal seawater. Wide mixing zones have been found in many coastal aquifers all over the world. However, the mechanisms responsible for wide mixing zones are not well understood. This work examines the hypothesis that kinetic mass transfer coupled with transient conditions, which create the movement of the mixing zone, may widen mixing zones in coastal aquifers. The hypothesis is tested by conducting two-dimensional numerical simulations based on a variable-density groundwater model for a scaled-tank model and a field-scale model. Periodic water levels, representing periodic tidal motion and freshwater table fluctuations, are imposed at the seaward and landward boundary, respectively, which cause the movement of the mixing zone. Both the scaled-tank model and the field model show that the combination of the moving mixing zone and kinetic mass transfer may significantly enhance the extent of mixing and create a wider mixing zone than the models without kinetic mass transfer. In addition, sensitivity analyses indicate that a larger capacity ratio (immobile porosity/mobile porosity) of mass transfer leads to a wider mixing zone, and the maximum width of the mixing zone may be reached for a given capacity ratio when the mean retention time scale in the immobile domain (the reciprocal of mass transfer rate) and the period of water level fluctuations are comparable.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 Interaction between groundwater and coastal seawater results in two complementary processes: seawater intrusion and submarine groundwater discharge (SGD). Understanding these processes meets the urgent needs for preserving vital fresh groundwater resources and coastal and offshore environments in highly populated coastal areas worldwide. The mixing zone developed at the freshwater-seawater interface is one of the most important features in complex coastal hydrogeologic systems. As the cumulative effect of many processes and mechanisms, such as periodic tidal activities, seasonal water table change, groundwater withdrawal, transport processes driven by density gradient, diffusion and dispersion, and properties of geological formations, etc., the growth and decay of the mixing zone can (1) directly reflect the extent of mixing in coastal aquifers; and (2) provide extremely useful information to serve as an indicator for and measure of effective management of groundwater resources and sustainable stewardship of coastal and offshore environments. For example, upconing of the mixing zone generally indicates the occurrence of seawater intrusion subject to excessive groundwater withdrawal [Bear, 1972]; and the movement of the mixing zone due to seasonal water table fluctuation is often associated with the seasonal variations of SGD [Michael et al., 2005]. Thus, gaining a better grasp of mixing zone development in coastal aquifers within various hydrogeologic settings is a milestone in our efforts to significantly improve our understandings of flow and transport in complex coastal hydrogeologic systems.
 In general, two types of mathematical models have been used to describe the mixing zone development: sharp-interface approximation and miscible fluid model. In the sharp-interface approximation, it is assumed that there is a stationary and abrupt interface between freshwater and intruding seawater, implying that no mixing takes place between freshwater and seawater. This approach is a major simplification and may allow one to use potential flow theory for describing interface propagation, and provides a useful tool for developing a variety of analytical solutions [e.g., Bear and Dagan, 1964; Strack, 1976; Huppert and Woods, 1995; Naji et al., 1998]. The second approach, based on the density-dependent miscible saltwater-freshwater systems, accounts for the presence of a variable-density mixing zone. The latter model is of particular interest in practical applications where one desires to evaluate salt and other species concentrations in coastal aquifers. In this work, we will focus on this model. Due to its practical significance, several numerical models based on miscible fluid physics have been developed to describe and study the problem of seawater intrusion over the past 20 years [Voss and Souza, 1987; Ataie-Ashtiani et al., 1999; Paniconi et al., 2001; Zhang et al., 2004; Paster et al., 2006; Qahman and Larabi, 2006]. Analytical solutions for seawater intrusion based on miscible fluid systems are only available for steady state, simplified cases [Dentz et al., 2006; Bolster et al., 2007].
 Both narrow and wide mixing zones have been observed in numerical, laboratory, and field studies. With a fine discretization and small dispersion, numerical simulations produced narrow mixing zones [e.g., Benson et al., 1998; Karasaki et al., 2006]. Laboratory experiments also demonstrated narrow mixing zones in homogeneous media [e.g., Zhang et al., 2001; Goswami and Clement, 2007; Abarca and Clement, 2009]. However, many field measurements found wide mixing zones, ranging from hundreds of feet to miles. This finding cannot be simply explained by upscaling small-scale laboratory data. For example, groundwater salinity measurements in Everglades National Park, in Southern Florida, USA, indicated the presence of a wide (6–28 km) seawater mixing zone [Price et al., 2003]; in the Floridian aquifer near downtown Brunswick, GA, USA, the mixing zone of seawater and freshwater has been detected across an area of increasing size [Cherry, 2006]; Xue et al.  reported a wide mixing zone of 1.5–6.0 km in the coastal area of LaiZhou Bay, China, and also found that the increasing extension of the salt water intrusion is a major concern in this area [Wu et al., 1993]; Barlow  summarized groundwater in freshwater-saltwater environments of the Atlantic Coast, in which wide mixing zones were observed in many coastal aquifers, e.g., the Biscayne aquifer near Miami, Florida, the upper Potomac aquifer in Virginia's Inland Wedge, the Floridan aquifer system in South Carolina, Georgia, and Florida, and the lower Tamiami aquifer in southwestern Florida, etc.
 The mechanisms responsible for a wide mixing zone still remain the subject of debate. Local dispersion has been considered as a primary mechanism responsible for the occurrence of the mixing zone. During the movement of the seawater front in either the landward or the seaward direction, elements of each fluid are transferred into the opposite environment by the convection component of dispersion, wherein to a large extent they become inseparably blended with other fluid by mixing and molecular diffusion [Cooper, 1959]. Dagan  pointed out that transverse dispersion is the main mechanism creating mixing in the seawater-freshwater interface, but the presumed small transverse pore-scale dispersion can only create a narrow mixing layer at the interface. The extent of mixing is also influenced by hydrodynamic fluctuations of the groundwater and seawater levels. Volker and Rushton  compared a variety of aquifer parameters and the influence of the flow conditions on the configuration and location of the interface. They concluded that a decrease in the dispersion coefficient leads to the contraction of the dispersion zone for a constant freshwater discharge, while the interface becomes more diffuse as the freshwater discharge decreases provided that the dispersion coefficient keeps invariant. Ataie-Ashtiani et al.  numerically examined the effects of tidal fluctuations on seawater intrusion in an unconfined aquifer, and found that the tidal activity created a thicker interface than would occur without tidal effects. However, Karasaki et al.  failed to reproduce a wide mixing zone by imposing a time-varying sinusoidal boundary condition without using a large dispersion coefficient. Heterogeneity in the hydraulic conductivity of the formation also contributes to the mixing enhancement. Heterogeneous hydraulic conductivities lead to spatially varying specific-discharge fields and thus to nonuniform advection. As a result, the mixing zone becomes increasingly irregular in shape, enhancing mixing caused by diffusion across its surface. However, Abarca et al.  showed that the effects of moderate heterogeneity on increasing the width of mixing zone are small. Thus, the widening of mixing zone width may not simply be attributed to heterogeneity of the formation.
 In the present research, we provide an alternative plausible explanation for wide mixing zones observed in coastal aquifers. The hypothesis is that the movement of the mixing zone combined with kinetic mass transfer effects may significantly widen the mixing zone. This study is motivated by the facts that (1) the mixing zone, in reality, seldom remains stationary, and (2) mass transfer processes, representing mass exchange between relatively mobile phases where advective-dispersive transport occurs and relatively immobile phases including low-permeability zones, stagnation pores, and sorption phases, etc. [Coats and Smith, 1964; van Genuchten and Wierenga, 1976], occur in almost all fractured and porous heterogeneous media over various scales ranging from pore scale to field scale, and significantly enhance solute mixing [Michalak and Kitanidis, 2000]. Previous investigations of the mixing zone width are mostly based on the steady state or tidal conditions. Under these conditions, the mixing zone is nearly stationary [Volker and Rushton, 1982; Ataie-Ashtiani et al., 1999; Robinson et al., 2007a]. In reality, however, large-scale recharge into the aquifer as well as withdrawals from it leads to the movement of mixing zone from one position to another. It is now recognized that seasonal oscillations of inland recharge appear to be widespread, clearly indicating that a seasonal mixing zone movement occurs in coastal aquifers [Michael et al., 2005]. On the other hand, the movement of mixing zone can also be caused by the effects from the seaward boundary. Cartwright and Nielsen , based on field experiments, indicated that the mixing zone movement can be caused by coastal waves. It is worth noting that the movement of mixing zone has been also observed in many other coastal areas [Wu et al., 1993; Cherry, 2006]. To the best of our knowledge, no study focusing on the mixing zone development has considered the combined effect of mass transfer processes and transient conditions. Langevin et al.  conducted a simulation of variable-density flow coupled with dual-domain transport for the Henry problem. Without the consideration of mixing zone movement, they found that the steady state salinity distribution was roughly the same as the salinity distribution for the classical Henry problem.
2. Numerical Model
 The proposed hypothesis will be tested by conducting two-dimensional (vertical cross section) numerical simulations based on the variable-density flow and transport equations for a scaled-tank model and a field-scale model. Transient effects will be introduced by imposing periodic water levels at the seaward and landward boundary. A dual domain transport model with first-order mass transfer will be applied to describe transport processes with kinetic mass transfer between mobile and immobile domains. The numerical model is solved by the density-dependent groundwater flow code SEAWAT-2000 implemented in a graphic user interface software Groundwater Vista 5.20 developed for 3D groundwater flow and transport modeling. SEAWAT-2000 itself was developed by combining MODFLOW and MT3DMS into a single program solving the coupled flow and solute transport equations. MT3DMS is implemented with an optional, dual-domain formulation for modeling mass transport.
2.1. Governing Equations
 The governing equation for saturated variable-density groundwater flow in terms of freshwater head is described by [Langevin and Guo, 2006]
where z [L] is the vertical coordinate directed upward; Kf [L T−1] is the equivalent freshwater hydraulic conductivity; hf [L] is the equivalent freshwater head; ρ [M L−3] is the fluid density; ρf [M L−3] is the freshwater density; Sf [L−1] is the equivalent freshwater storage coefficient; t [T] is the time; θe is the effective porosity; and ρs [M L−3] and qs [T−1] are the density and flow rate per unit volume of aquifer of the source/sink, respectively [Langevin et al., 2003].
 The dual-domain transport model involving advection, molecular diffusion, mechanical dispersion, and first-order mass transfer is described by
where θm is porosity of the mobile domain and is equal to θe; θim is porosity of the immobile domain; Cm [M L−3] is dissolved concentration in the mobile domain; Cim [M L−3] is dissolved concentration in the immobile domain; ξ [T−1] is first-order mass transfer rate between the mobile and immobile domain; D [L2 T−1] is the hydrodynamic dispersion coefficient tensor; and v [L T−1] is the pore water velocity vector.
 The relationship between the fluid density and salt concentration is represented by the linear function of state:
where ɛ is a dimensionless constant with a value of 0.7143 for salt concentrations ranging from zero to 35 kg m−3, a typical concentration value for seawater [Langevin et al., 2003]; and ρ is expressed in kg m−3.
2.2. Scaled Tank Model and Numerical Implementation
 A scaled tank model is designed to simulate the mixing zone development under the considerations of transient conditions and mass transfer effects. Zhang et al.  presented an experimental study of a dense contaminant plume in an idealized coastal aquifer based on a tank model, which was numerically studied by Brovelli et al. . Due to its high computational efficiency, this scaled tank model is used here to investigate mass transfer effects on the development of the mixing zone and sensitivity analysis will also be conducted. A schematic representation of the seawater intrusion problem is shown in Figure 1. The tank is 1.650 m long, 0.6 m high, and 0.1 m wide with a beach slope (vertical/horizontal ratio) 1:6.12. A homogeneous, isotropic hydraulic conductivity of 4 × 10−3 m s−1 is assigned to the domain. The mean local longitudinal and transverse dispersivities are measured to be 6.49 × 10−4 m and 1 × 10−4 m, respectively. The total porosity is 0.37. The mean seawater level and the constant freshwater level are 0.439 m and 0.463 m, respectively. The seawater and freshwater densities are 1025 kg m−3 and 1000 kg m−3, respectively, which represent a salt concentration of 35 kg m−3 for seawater. The parameters for the scaled tank model are summarized in Table 1.
Table 1. Geometry, Hydrogeological, and Transport Parameters Used in the Experimental Study of Zhang et al. 
 Rather than only tidal conditions used by Zhang et al.  and Brovelli et al. , periodic water level fluctuations are imposed at the seaward and landward boundaries, respectively, to create the movement of the mixing zone. First, a triangular, periodic function with a period of 40 min (see Figure 2) is imposed at the seawater boundary to simulate the periodic tidal-like motion, while a constant freshwater level of 0.463 m is defined at the landward boundary. The linear variation of water level can be directly implemented based on the variable head boundary condition in SEAWAT by specifying two values of hydraulic head at the beginning and at the end of the stress period, respectively. The software linearly interpolates between the two values according to defined time step. The use of the triangular function instead of a sinusoid function is to minimize the computational effort because much more pressure periods will be needed to reproduce the sinusoid function. Then, another triangular, periodic function with a period of 80 min and amplitude of 0.04 m is defined at the landward boundary to simulate the water table fluctuations, while the seawater level is kept constant at 0.439 m. In reality, the period of the freshwater table fluctuations may be much greater than that of the tidal motion. Sensitivity analysis will be conducted later to investigate the effects of both periods. A constant salt concentration of 35 kg m−3 is enforced at the seaward boundary.
 The simulation domain is discretized into 9900 cells in order to satisfy the accuracy and convergence requirement for grid spacing in terms of the local Péclet number [Voss and Souza, 1987; Zhang et al., 2001; Volker et al., 2002; Brovelli et al., 2007]. The entire model domain is divided into two zones: a surface water zone and an aquifer zone. To simplify the numerical simulation, a large hydraulic conductivity of 0.4 m s−1, i.e., 100 times of the saturated aquifer hydraulic conductivity, a constant porosity of 1, and a constant saltwater concentration of 35 kg m−3 are assigned to all the cells in free seawater area [Winter, 1976; Anderson et al., 2002; Mao et al., 2006; Brovelli et al., 2007; Robinson et al., 2007a]. In addition, to reproduce the flat surface of the sea, a horizontal strip of cells with a variable-head boundary condition is added onto the seawater surface [Brovelli et al., 2007]. Simulations start from steady state conditions generated by using the mean seawater level and the mean freshwater level. The simulation duration for each case is fifty periods of the corresponding triangular functions, a sufficiently long period for the scaled tank models to reach a dynamic equilibrium state of the concentration distribution, i.e., the tolerance of the maximum concentration variation is satisfied when doubling the computation periods.
2.3. Field-Scale Model and Numerical Implementation
 For the field-scale case, we consider a 2D model domain that is 200 m long and 35 m high with a beach slope 1:10. The aquifer was assumed to be isotropic and homogeneous with Kf = 20 m d−1, ne = 0.4, longitudinal dispersivity αL = 0.5 m and transverse dispersivity αT = 0.05 m. Hydraulic conductivity of 1000 m d−1, ne = 1 and constant salt concentration of 35 kg m−3 are assigned to the cells in free seawater area so that the entire domain can be solved by SEAWAT. The mean seawater level and the mean freshwater level are 28 m and 29 m, respectively. For field-scale applications, transient effects introduced by periodic tidal motion on the movement of the mixing zone may not be as effective as those introduced by the freshwater table fluctuations because (1) tidal motion has a much shorter period than freshwater table fluctuations; (2) the amplitude of freshwater table fluctuations can be much larger than that of tidal motion because of seasonal precipitation and temperature patterns; and (3) the effects of the freshwater table change may be enlarged to 40 times on the freshwater-seawater interface according to the Ghyben-Herzberg law based on potential equilibrium [Bear, 1972]. In the present research, we impose a triangular, periodic head variation with a period of one year and an amplitude of 1 m at the landward boundary, while a constant seawater level of 28 m is specified at the seaward boundary. For the numerical simulation, a mesh resolution of 0.5 m was adopted, yielding 28000 cells. This discretization results in a satisfactory Pe of 1. The dynamic equilibrium state of the concentration distribution is found after 100 periods, i.e., 100 years.
3. Results of the Tank Model
3.1. Steady State Condition
 Steady state cases are first simulated to serve as control cases, which neglect both mass transfer and transient conditions introduced by tidal motion and freshwater table fluctuations. By assuming a constant seawater level of 0.439 m and a constant freshwater level of 0.463 m, a SEAWAT simulation was first run for steady state conditions without considering the mass transfer effect. Figure 3 shows the mixing zone, where the contour lines delineate the normalized concentrations 0.1, 0.5, and 0.9. Rather than a sharp interface, a narrow mixing zone is formed due to density gradient and local dispersion. The salinity distribution simulated in our study matches well experimental [Zhang et al., 2002] and numerical results [Zhang et al., 2001; Brovelli et al., 2007] previously obtained based on the same scaled tank model. We also evaluate the mixing zone by including mass transfer but still neglecting transient effects. Similar to the observation by Langevin et al. , the resulting mixing zone is almost the same as the one neglecting mass transfer. Thus, for steady state analyses, mass transfer does not make significant contributions in altering salinity distributions. In fact, by forcing the transient terms in equation (2) to be zero, the transport model reduces to the case without mass transfer. That is, the steady state salinity distributions will become identical for cases with and without mass transfer, although the time scales to reach the steady state may be different.
3.2. Transient and Mass Transfer Effects
Figure 4 shows the mixing zones with the consideration of seawater level oscillations but neglecting mass transfer. A wider mixing zone, particularly at the toe, is observed compared with the mixing zone shown in Figure 3. Furthermore, due to the seawater level oscillations the interface is pushed seaward. This phenomenon is consistent with the simulation results obtained by Robinson et al. [2007a], who conducted a numerical study on a field-scale domain to investigate the effect of tidal forcing on a subterranean estuary. In addition, their results show that the interface is pushed more seaward with a larger amplitude tide. Seawater level fluctuation forces the seawater back and forth and, thus, the equilibrium state shown in Figure 3 is disturbed, yielding a transient velocity field and a fluctuated concentration distribution, which result in enhanced mixing and a slightly wider mixing zone due to hydrodynamic dispersion. This phenomenon has been demonstrated by the laboratory experiment of Zhang et al. . Ataie-Ashtiani et al.  also showed wider mixing zones caused by tidal motion, but observed that a larger tidal amplitude may force the seawater to intrude further inland. The difference may result from a different domain setting, in which an additional unsaturated zone was assumed above the groundwater table.
 However, previous studies including numerical work conducted by Robinson et al. [2007a, 2007b] and Ataie-Ashtiani et al.  and field experiments by Cartwright et al.  indicate that the mixing zone of the saltwater wedge does not fluctuate over the course of a tidal cycle because the forcing time scale is too short. Hence, the movement of the mixing zone in our study can be attributed to the small tank scale and relatively large time scale of seawater level fluctuation. In reality, however, the mixing zone may be forced landward by a combination of increasing tidal range, wave height and infiltration of wave runup [Cartwright and Nielsen, 2001a, 2001b]. Once the wave forcing decreased the contour gradually moved seaward [Cartwright and Nielsen, 2003]. Therefore, the movement of the mixing zone of our scaled tank model may be regarded as a result of complex effects from the seaward boundary.
Figure 4 also shows that the position of the mixing zone varies at different seawater level stages, i.e., the hydraulic gradient determines the position of the mixing zone. Thus, we can only define a dynamic equilibrium state instead of a steady state for the transient case. As mentioned before, dynamic equilibrium is defined as the state where the mixing zone position has no significant variations by doubling the simulation duration. In addition, although the position of the mixing zone varies, the width does not change noticeably over the course of one periodic cycle.
Figure 5 shows the mixing zones with the consideration of both mass transfer effects and seawater level oscillations, where both mobile porosity and immobile porosity are set to be 0.185 and the first-order mass transfer rate coefficient is 0.025 min−1. Figure 5 clearly shows that the mass transfer effect leads to significantly wider mixing zones at all stages of the seawater level compared with those shown in Figure 4. In particular, it is more pronounced at the low and falling water level stages. As already mentioned, in the absence of seawater level oscillations, i.e., the mixing zone is stationary, mass transfer has no effect on the steady state salinity distribution because there is no concentration gradient between the mobile and immobile domains and equation (2) can be simplified to the classical advection-dispersion equation, although the time scale to reach the steady state may be changed. However, in transient cases, the mixing zone is pushed back and forth by complex effects from the seaward boundary, resulting in nonequilibrium in the salt concentrations in the mobile and immobile domains and an enhanced mass exchange between them. The immobile domain here essentially acts as a sink or source for solutes in the mobile zone, determined by the direction of concentration gradient between two domains. Specifically, salts in the mobile domain diffuse into the immobile domain as the mixing zone is dragged inland, while salts are released from the immobile domain to the mobile domain driven by reversed concentration gradients when the mixing zone is towed toward the sea. The disturbed concentration and density gradient field leads to enhanced mixing and a wider mixing zone than would occur in the absence of mass transfer. Moreover, Figure 5 shows that the combination of mass transfer and seawater level fluctuations has greater influences on the high concentration contour lines (see the contour lines of normalized concentration 0.9), which become closer to the seaward boundary.
 Freshwater level fluctuation is observed in many areas, which has been found as a main reason leading to the movement of the mixing zone [Michael et al., 2005]. Figure 6 shows the mixing zones under periodic oscillations of the freshwater level without mass transfer effects. Like the effect from seawater level fluctuations shown above, the mixing zone is pushed seaward, and wider mixing zones are formed. However, the mixing zone moves within a broader range due to a larger period of the freshwater level fluctuation. The width of the mixing zone caused by freshwater variation is expected to increase when the mass transfer effect is taken into account. Figure 7 exhibits the mixing zones with the consideration of both mass transfer and freshwater level oscillations, where both mobile and immobile porosities are set to be 0.185 and the first-order mass transfer rate coefficient is 0.0125 min−1. Likewise, the introduced mass transfer effect significantly increases the width of the mixing zone, especially at the rising level stages. Similarly, one may expect that a larger fluctuation amplitude will lead to a wider mixing zone.
3.3. Sensitivity Analysis
 Parameters of kinetic mass transfer, including mobile and immobile porosity and the first-order rate constant, will be varied in order to investigate the effects of kinetic mass transfer. In order to interpret the results concisely and produce a meaningful generalization, the following dimensionless variables are defined:
where Tf is the water level fluctuation period; ξ−1 represents a characteristic mass transfer time in the immobile domain; β is known as the capacity ratio; Wss is the mixing zone width under steady state condition; and W is the width of the mixing zone under the coupled effect of mass transfer and water level fluctuation. For simplicity, W is represented by the horizontal distance between concentration contour lines of 0.1 and 0.9. Here, we choose the width of the mixing zone at the height of 200 mm at the rising tidal moment to calculate W. Similar results will be obtained for the width of the mixing zone at other heights and tidal moments. By assuming a constant total porosity for the mobile and immobile domain, the effects of mass transfer parameters and water level fluctuations on the dimensionless width of the mixing zone, w, can be investigated by varying the dimensionless variables, τ and β.
Figure 8 shows the simulated results for the sensitivity analysis for the tank model. The width of the mixing zones formed by varying the freshwater level is somewhat wider than that by seawater level fluctuation since the period of the former is assumed larger than the latter. For both cases, with a given mean retention time, i.e., a constant first-order mass transfer rate coefficient, the width of the mixing zone increases with the capacity ratio, indicating that a larger immobile domain may cause a wider mixing zone. With a given capacity ratio, i.e., a constant porosity of the immobile domain, the width of the mixing zone is maximized when the retention time scale of the mass transfer and the period of the water level fluctuation become comparable, i.e., at the same order of magnitude. In such cases, the effects of the capacity ratio will also be maximized. In addition, the left and right tails of the curves shown in Figures 8a and 8b indicate that the mass transfer may not have significant impacts on widening the mixing zone when there is a several orders of magnitude difference between the retention time scale and the water level fluctuation period. In fact, both the limiting cases of very small and large mass transfer rate coefficients can be simplified to a classical advective-dispersive transport problem. For a small mass transfer rate coefficient, it is equivalent to the transport problem in a medium with a smaller total porosity, i.e., practically no mass transfer occurs within a period. For a large mass transfer rate coefficient, the kinetic mass transfer may be considered as an instantaneous process, which simplifies the two-domain model into a one-domain model with a retardation factor, 1 + β. Thus, for both limiting cases, the width of the mixing zone will approach the dynamic equilibrium state in the absence of mass transfer. In our tank model, the mixing zone is significantly widened for τ between 0.1 and 100, and the width reaches maximum for τ to be about 1, i.e., ξ−1 is equal to Tf. For example, the width of the mixing zone is approximately 3.7 times as wide as that under steady state condition for the freshwater level fluctuation case with β = 1 and τ = 1 (see Figure 8b). If other parameters are kept constant, we may expect that the width of the mixing zone will become much larger for a higher β.
4. Field-Scale Modeling Results
 The field-scale model described in section 2 corroborates the results obtained based on the tank model. Figure 9 shows the mixing zone under steady state condition, where a narrow mixing zone is generated. With the introduction of freshwater level fluctuations, the steady state system is disturbed and the mixing zone is slightly widen (see Figure 10). Mass transfer effect is then introduced, where mobile and immobile porosities both are set to be 0.2, namely, β = 1. Three mass transfer coefficients including 0.027 d−1, 0.0027 d−1 and 0.00027 d−1 are employed, which correspond to dimensionless variable τ as 0.1, 1, and 10, respectively. The corresponding mixing zones at the rising water level stage for these three cases are exhibited in Figure 11, which clearly shows wider mixing zones than those shown in Figure 10. In particular, the mixing zone width in the case with τ = 1 is maximal, consistent to the result found in the tank model.
 In the absence of mass transfer, dispersivities, particularly transverse dispersivity, is considered to be the primary factor affecting the width of the mixing zone [Ataie-Ashtiani et al., 1999; Dagan, 2006]. In order to reproduce a wide mixing zone in a real case, the common method is to assume a large, perhaps unwarranted, value of dispersivities [Dagan, 2006]. In this section, we briefly compare the effects of dispersivities and mass transfer on the mixing zone width. In addition to the dispersivities assumed in the cases discussed above, two more groups of longitudinal and transverse dispersitivies are adopted in the field-scale model: αL = 2.5 m and αT = 0.25 m, and αL = 0.1 m and αT = 0.01 m.
Figure 12 shows the mixing zones at the rising freshwater level stage with the consideration of the freshwater level fluctuation alone. It is obvious that larger dispersivities yield a wider mixing zone. However, the maximum mixing zone shown in Figure 12c is still not as wide as that in Figure 11b, although both longitudinal and transverse dispersivities are 25 times of those in the previous case. Thus, in order to generate a wider mixing zone, larger dispersivities must be accepted.
Figure 13 shows the results by further introducing mass transfer with β = 1 and τ = 1 into the three cases with different dispersivities. It is shown that all mixing zones are significantly widened compared with those shown in Figure 12. Furthermore, with the mass transfer effect, all the mixing zone widths become similar, although different dispersivities are used. The mass transfer effect dominates the mixing zone width change. Therefore, in our cases, the effect of kinetic mass transfer is more pronounced than the dispersivities on widening the mixing zone.
 Wide mixing zones have been observed in many aquifers all over the world. However, no agreement has been reached in terms of the responsible mechanisms. In the present work, we propose the hypothesis that kinetic mass transfer combined with movement of mixing zones may significantly widen mixing zones in coastal aquifers. The hypothesis is tested by conducting numerical simulations based on the variable-density groundwater model for both a scaled-tank model and a field-scale model. The movement of the mixing zone may be caused by complex effects from both the seaward boundary (e.g., wave runup) and the landward boundary (e.g., seasonal fluctuation of fresh groundwater head). In our simulations, the movement of the mixing zone is created by assuming triangular, periodic functions for water level oscillations at the seawater and landward boundaries, respectively. In the absence of kinetic mass transfer, the created transient effects slightly widen the mixing zone compared with that in steady state. With the introduction of kinetic mass transfer, mixing zones are significantly widened at all stages within the period.
 Furthermore, sensitivity analyses of dimensionless variables based on the tank model yield the following observations: (1) the mixing zone may be significantly widened by the mass transfer effect regardless of which boundary causes the movement of the mixing zone; and (2) a larger capacity ratio of mass transfer leads to a wider mixing zone, and the maximum width may be reached when the mean retention time scale in the immobile domain and the water level fluctuation period become comparable. Our simulations also investigate the effects of dispersivities on the mixing zone development. Larger dispersivities always yield wider mixing zones. However, dispersivities may not be as effective as kinetic mass transfer on widening the mixing zone in our cases. More importantly, larger dispersitivies and mass transfer are based on different physical interpretations of the transport processes and formation properties. Large, perhaps unwarranted dispersitivities are often considered as the misrepresentation of aquifer heterogeneities [Dagan, 2006]. On the other hand, as is well known, no natural geological media are truly homogenous, and mass transfer occurs in almost all fractured and porous heterogeneous media over various scales ranging from pore scale to field scale. Our findings provide a plausible explanation for wide mixing zones in coastal aquifers which may consist of low-permeability zones, dead-end pores, porous particles, aggregates, and rock matrix between fractures. In such aquifers, the effects of kinetic mass transfer and the movement of mixing zone caused by tidal motion, freshwater table fluctuations, groundwater withdrawal, etc., must be considered to evaluate the growth and decay of the variable-density mixing zone. Certainly, other parameters, such as the amplitude of the periodic stimulation, the hydraulic conductivity, the rate of freshwater flow, the heterogeneity of the geological formations, may influence the growth and decay of the mixing zone. The research of effects of these mechanisms on mixing zone development is continuing.
 This research was sponsored by the National Institutes of Water Resources (NIWR) and U.S. Geological Survey (USGS) under project ID 2007GA165G. This article has not been reviewed by the agency, and no official endorsement should be inferred. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of NIWR and USGS. We thank David Hochstetler for his comments on a preliminary version of the manuscript. We thank three anonymous reviewers for their constructive comments which significantly improved the manuscript quality.