Variations in hydraulic conductivity with scale of measurement during aquifer tests in heterogeneous, porous carbonate rocks

 Previous studies have shown that hydraulic conductivity of an aquifer seems to increase as the portion of the aquifer tested increases. To date, such studies have all relied on different methods to determine hydraulic conductivity at each scale of interest, which raises the possibility that the observed increase in hydraulic conductivity is due to the measurement method, not to the scale. This study analyzes hydraulic conductivity with respect to scale during individual aquifer tests in porous, heterogeneous carbonate rocks in southeastern Wisconsin, USA. Results from this study indicate that hydraulic conductivity generally increases during an individual test as the volume of aquifer impacted increases, and the rate of this increase is the same as the rate of increase determined by using different measurement methods. Thus, scale dependence of hydraulic conductivity during single tests does not depend on the method of measurement. This conclusion is supported by 22 of 26 aquifer tests conducted in porous-flow-dominated carbonate units within the aquifer. Instead, scale dependency is probably caused by heterogeneities within the aquifer, a conclusion supported by digital simulation. All of the observed types of hydraulic-conductivity variations with scale during individual aquifer tests can be explained by a conceptual model of a simple heterogeneous aquifer composed of high-conductivity zones within a low-conductivity matrix. Received, January 1997 Revised, August 1997, November 1997 Accepted, November 1997     

Interpretation of pumping test data from large diameter wells on an oceanic island

 Estimation of aquifer parameters is vital for the assessment of groundwater potential and groundwater flow regime. On an oceanic island where fresh water lens is fragile and sensitive to various stresses, it is even more essential that in order to assess the potential of fresh water lens and the effect of various stresses on the fresh groundwater regime the parameters should be representative to the field hydrogeological set up. Pumping tests conducted on existing large diameter wells on an oceanic island have been analyzed. A finite difference method has been used to take into account the well storage, partial penetration and upconing effect into the aquifer. Forward modeling has been carried out to estimate aquifer parameters from the pumping test data. Field examples are described. Received: 15 June 1998 · Accepted: 25 August 1998     

Reactive transport impacts on recovered freshwater quality during multiple partially penetrating wells (MPPW-)ASR in a brackish heterogeneous aquifer

The use of multiple partially penetrating wells (MPPW) during aquifer storage and recovery (ASR) in brackish aquifers can significantly improve the recovery efficiency (RE) of unmixed injected water. The water quality changes by reactive transport processes in a field MPPW-ASR system and their impact on RE were analyzed. The oxic freshwater injected in the deepest of four wells was continuously enriched with sodium (Na⁺) and other dominant cations from the brackish groundwater due to cation exchange by repeating cycles of ‘freshening’. During recovery periods, the breakthrough of Na⁺ was retarded in the deeper and central parts of the aquifer by ‘salinization’. Cation exchange can therefore either increase or decrease the RE of MPPW-ASR compared to the RE based on conservative Cl⁻, depending on the maximum limits set for Na⁺, the aquifer’s cation exchange capacity, and the native groundwater and injected water composition. Dissolution of Fe and Mn-containing carbonates was stimulated by acidifying oxidation reactions, involving adsorbed Fe²⁺ and Mn²⁺ and pyrite in the pyrite-rich deeper aquifer sections. Fe²⁺ and Mn²⁺ remained mobile in anoxic water upon approaching the recovery proximal zone, where Fe²⁺ precipitated via MnO₂ reduction, resulting in a dominating Mn²⁺ contamination. Recovery of Mn²⁺ and Fe²⁺ was counteracted by frequent injections of oxygen-rich water via the recovering well to form Fe and Mn-precipitates and increase sorption. The MPPW-ASR strategy exposes a much larger part of the injected water to the deeper geochemical units first, which may therefore control the mobilization of undesired elements during MPPW-ASR, rather than the average geochemical composition of the target aquifer.     

Detecting a salinity plume in an unconfined sandy aquifer and assessing secondary soil salinization using electromagnetic induction techniques, North Dakota, USA

 Land-use changes on the Sheyenne Delta in southeastern North Dakota, USA, have prompted research on impacts to the unconfined Sheyenne Delta aquifer (SDA). This study examines effects of the saline discharge of a flowing artesian well that taps the Dakota aquifer (DAK) on SDA groundwater chemistry and soil salinity. Objectives were to map the saline plume in the SDA using induction techniques, to assess chloride migration in the SDA, and to evaluate induction sensitivity to moderately saline sands. Induction data, collected in a 2.9-ha grid, were compared to 31 soil profiles analyzed for gravimetric moisture, electrical conductivity, and chloride. Soil salinization is widespread, but only 7% of the area meets the 4-dS/m threshold for saline soils. SDA chloride distribution was determined on transects oriented with and perpendicular to the flow path determined from induction readings. Chloride was detected in the aquifer 550 m from the source, indicating a transport rate of 21 m/yr. Complex recharge and discharge patterns and hummocky relief contribute to a wide chloride plume at this site. A mass balance based on soil-water content and chloride concentration shows that only 4% of the chloride from the DAK well remains in the grid volume. Received, January 1998 · Revised, January 1999 · Accepted, March 1999     

A method for designing configurations of nested monitoring wells near landfills

 A method was devised for designing configurations of monitoring wells, consisting of vertically nested intakes in boreholes. The network-design method involves analyzing a subset of potential contaminant plumes emerging from the downgradient margin of a landfill. Plume widths are evaluated along selected equipotential lines and compared to the lengths of those lines. The method was applied to a 32-ha solid-waste landfill in Tarrant County, Texas, USA. Sixtynine potential source nodes were considered. A 15-borehole network devised by the method registered 93 detections in total, detecting all 69 model-generated plumes by at least one borehole. Based on an enumeration procedure, a minimum of 10 boreholes was needed to detect all of the model-generated plumes. However, the less conservative 10-borehole network had little capability for backup detection. An existing monitoring network of seven downgradient wells detected only 38 model-generated plumes. Results of this study illustrate a practical need for structured approaches to designing detection-based groundwater-monitoring configurations. Received, February 1997 · Revised, July 1997, December 1997 · Accepted, November 1997     

Digital simulation of the scale effect in hydraulic conductivity

 Measured hydraulic conductivity increases with the scale of testing, but the reason for this increase is not clear. Possibly, high-conductivity heterogeneities are more effective in raising hydraulic conductivity over the regional scale than at the local scale. Alternatively, borehole skin and storage effects, among others, can systematically bias the results of small-scale tests; thus, the increase may simply be an artifact of the test method. Radial-flow tests were simulated at various scales in digital models of fractured double-porosity media. The mean hydraulic conductivity increases until the radius of influence in the test exceeds the fracture spacing. Therefore, under radial flow, hydraulic conductivity is indeed dependent on measurement scale. The increase in hydraulic conductivity with scale is a natural consequence of heterogeneity. Over short distances, water converging toward a borehole must generally flow across heterogeneities. Therefore, small-scale tests tend to measure a weighted harmonic mean of the hydraulic-conductivity field. Over a large area, however, flow is primarily along high-conductivity heterogeneities. Therefore, large-scale tests approach a weighted arithmetic mean where high-conductivity heterogeneities have a greater influence. Received, April 1997 Revised, January 1998 Accepted, December 1997     

Regional characterisation of hydraulic properties of rock using air-lift data

 Hydrogeologic studies are commonly data-intense. In particular, estimations of hydraulic properties of hard rock often require large amounts of data. In many countries, large quantities of hydrogeologic data have been collected and archived over the years. Therefore, the use of existing data may provide a cost-efficient alternative to collecting new data in early stages of hydrogeologic studies, although the available data may be considered imprecise. Initially, however, the potential usefulness, i.e., the expected accuracy, of the available data in each specific case must be carefully examined. This study investigates the possibilities of obtaining estimates of transmissivity from hard-rock air-lift data in Sweden within an order of magnitude of results obtained from high-quality injection-test data. The expected accuracy of the results was examined analytically and by means of statistical methods. The results were also evaluated by comparison with injection-test data. The results indicate that air-lift data produce estimates of transmissivity within an order of magnitude compared to injection-test data in the studied examples. The study also shows that partial penetration and hydrofracturing may only affect the estimations approximately half an order of magnitude. Thus, existing data may provide a cost-efficient alternative to collection of new data in early stages of hydrogeologic studies. Received, January 1998 Revised, November 1998, December 1998 Accepted, December 1998     

Competition between topography- and compaction-driven flow in a confined aquifer: Some analytical results

 An analysis is presented of the competition between topography- and compaction-driven flow in a confined aquifer. The aquifer crops out and is unconfined in an area with an elevated water table. In the absence of compaction, the elevated water table drives the flow toward areas with a lower water table, and leakage occurs from the aquifer through an overlying aquitard. Under these conditions, recharge occurs in the outcrop area. When sedimentation occurs above the aquifer away from the outcrop area, compaction tends to reverse the flow direction in the aquifer, causing expulsion of groundwater in the outcrop area. The actual flow field arises from the competition between the two driving mechanisms. An analytical expression for the combined flow field is derived in which the contribution from both mechanisms is apparent. The expression is useful as a practical guideline for estimating whether or not compaction can be safely ignored in regional groundwater-flow studies of confined aquifers. Received, July 1998 / Revised, February 1999 / Acccepted, February 1999     

Numerical analysis of the hydrogeologic controls in a layered coastal aquifer system, Oahu, Hawaii, USA

 The coastal aquifer system of southern Oahu, Hawaii, USA, consists of highly permeable volcanic aquifers overlain by weathered volcanic rocks and interbedded marine and terrestrial sediments of both high and low permeability. The weathered volcanic rocks and sediments are collectively known as caprock, because they impede the free discharge of groundwater from the underlying volcanic aquifers. A cross-sectional groundwater flow and transport model was used to evaluate the hydrogeologic controls on the regional flow system in southwestern Oahu. Controls considered were: (a) overall caprock hydraulic conductivity; and (b) stratigraphic variations of hydraulic conductivity in the caprock. Within the caprock, variations in hydraulic conductivity, caused by stratigraphy or discontinuities of the stratigraphic units, are a major control on the direction of groundwater flow and the distribution of water levels and salinity. Results of cross-sectional modeling confirm the general groundwater flow pattern that would be expected in a layered coastal system. Groundwater flow is: (a) predominantly upward in the low-permeability sedimentary units; and (b) predominantly horizontal in the high-permeability sedimentary units. Received, October 1996 Revised, August 1997 Accepted, September 1997     

Simulation of seawater intrusion into the Khan Yunis area of the Gaza Strip coastal aquifer

 The Gaza Strip coastal aquifer is under severe hydrological stress due to over-exploitation. Excessive pumping during the past decades in the Gaza region has caused a significant lowering of groundwater levels, altering in some regions the normal transport of salts into the sea and reversing the gradient of groundwater flow. The sharp increase in chloride concentrations in groundwater indicates intrusion of seawater and/or brines from the western part of the aquifer near the sea. Simulations of salt-water intrusion were carried out using a two-dimensional density-dependent flow and transport model SUTRA (Voss 1984). This model was applied to the Khan Yunis section of the Gaza Strip aquifer. Simulations were done under an assumption that pumping rates increase according to the rate of population growth, or about 3.8% a year. Model parameters were estimated using available field observations. Numerical simulations show that the rate of seawater intrusion during 1997–2006 is expected to be 20–45 m/yr. The results lead to a better understanding of aquifer salinization due to seawater intrusion and give some estimate of the rate of deterioration of groundwater. Received, September 1997 Revised, January 1998, July 1998 Accepted, August 1998     

侧向有限越流承压含水层中非完整井非稳定流模型及解析解

工程建设中当距离抽水井r=rb处水位基本没有变化或不受抽水影响时,或当此处存在止水帷幕时,含水层系统视为侧向有限延伸,rb为有限半径。为此,构建更加符合工程实际的侧向有限延伸的典型弱透水层-承压水层系统中非完整井非稳定流计算模型,同时考虑井径和井储效应的影响,应用Laplace变换和分离变量法得到了水位降深在拉氏空间下的解析解,并应用拉氏数值逆变换Stehfest法得到真实空间下的水位降深。新建立的解析解可以进一步退化为诸多已有解,并进一步将其与已知解和有限元数值解进行对比,验证了所得解的正确性和可靠性。基于新建解重点分析了侧向边界和井的完整性对承压水层水位降深的影响。结果表明:含水层系统的侧向有限边界仅对抽水后期的水位降深影响明显,含水层系统侧向无限延伸情况下的水位降深要大于情形1(在r=rb处为定水头边界)且明显小于情形2(在r=rb处为不透水边界)下的水位降深,rb越小,两者之间的误差越大;抽水井的完整性对整个抽水期间不同情形下的水位降深均有明显的影响,承压含水层顶板处的水位降深随着抽水井滤管的长度和埋深的增加而减小。     

Assessing aquifer susceptibility to and severity of atrazine contamination at a field site in south-central Wisconsin, USA

 A field study from October 1989 through July 1992, conducted on a 4.1-km² area in south-central Wisconsin, USA, examined the distributions of atrazine and its chlorinated metabolites in groundwater and related those distributions to the groundwater flow system. MODFLOW and PATH3D were used to assess bedrock-aquifer susceptibility to contamination. Estimated travel time from water table to bedrock surface ranges from <0.25 to >512 yr. Spatial distribution of the estimates demonstrates that increased travel time to bedrock can result from the presence of shallow surface-water bodies, greater depths to bedrock, and smaller hydraulic conductivities. Estimated travel times to local domestic wells are inversely related to atrazine and desethylated atrazine concentrations observed in water from those wells. The potential impact of long-term atrazine use on aquifer water quality was investigated using MT3D in two best-case scenarios. Uncertainties associated with predicted atrazine concentrations at various depths and times were estimated. For shallow groundwater, widespread violations of Wisconsin's current preventive action limit were predicted, but with large uncertainty stemming from uncertain estimates of input parameter values. The simulations indicate, however, that moderate inputs at the water table are very unlikely to produce violations of Wisconsin's standards deeper in the aquifer. Received, October 1997 Revised, July 1998 Accepted, July 1998     

Effects of horizontal strain in estimating specific storage and compaction in confined and leaky aquifer systems

 Mathematical and numerical distinctions are made between developments of governing equations involving groundwater flow and granular displacement where (1) only vertical strain and no change in total normal load are assumed, and (2) such limitations are eliminated. The former is referred to here as the Jacob-Terzaghi method; the latter is referred to as the Biot method. Numerical comparisons of drawdown, compaction, and strain are made for hypothetical confined and leaky aquifer systems for one- and three-dimensional strain problems. Simulation results indicate that incorporation of horizontal strain results in reduced drawdowns. Vertical compaction for a specified stress distribution is greatly reduced when horizontal strain is invoked. The distribution of strain components indicates that approximately two-thirds of the total volume strain originates from horizontal compaction in pumped aquifers. Incorporation of horizontal strain also affects the distribution of flux through the overlying confining layer as a function of distance from the pumping well and may ultimately affect the interpretation of aquifer-test data. Received, January 1999 / Revised, May 1999, August 1999 / Accepted, August 1999     

Groundwater capture processes under a seasonal variation in natural recharge and discharge

 "Capture" is the increase in recharge and the decrease in discharge that occurs when pumping is imposed on an aquifer system that was in a previous state of approximate dynamic equilibrium. Regional groundwater models are usually used to calculate capture in a two-step procedure. A steady-state solution provides an initial-head configuration, a set of flows through the boundaries for the modeled region, and the initial basis for the capture calculation. The transient solutions provide the total change in flows through the boundaries. A difference between the transient and steady-state solutions renders the capture calculation. When seasonality is a modeling issue, the use of a single initial hydraulic head and a single set of boundary flows leads to miscalculations of capture. Instead, an initial condition for each season should be used. This approach may be accomplished by determining steady oscillatory solutions, which vary through the seasons but repeat from year to year. A regional groundwater model previously developed for a portion of the San Pedro River basin, Arizona, USA, is modified to illustrate the effect that different initial conditions have on transient solutions and on capture calculations. Received, September 1996 · Revised, October 1997 · Accepted, October 1997     

Validity of a sharp-interface model in a confined coastal aquifer

 The problem of seawater intrusion is considered for the case of a confined coastal aquifer in which there is steady seaward flow of fresh water. Using the GWCH2O model, the problem is solved first for the case of no dispersion where a distinct interface exists separating the fresh water from the salt water. The problem is solved next by taking into account dispersion and diffusion of the salt-water component, along with the density effect. In this respect, a two-dimensional finite-element model, 2D-VDTRAN, is developed to simulate density-dependent solute transport. To investigate the limitation of the sharp-interface approach in coastal aquifers for conditions of both steady state and unsteady state, the problem is solved twice using the two models with different parameter values. These parameters are combined in dimensionless form, resulting in four named parameters: seepage factor (A); dispersion-to-advection ratio (B); geometry ratio (C); and time-scale factor (T). Using the density-dependent model, the dimensionless width of the transition zone (W/L) is determined for different values of A, B, C, and T. Steady-state simulations show that the sharp-interface approach is valid only when the system is dominated by advection, i.e., when 0<B≤5% for all values of A and C, or when A≥65% for all values of B and C. However, the unsteady-state analysis shows that the applicability of the sharp-interface approach is sufficiently accurate at early times. Received, October 1997 Revised, June 1998, October 1998 Accepted, November 1998     

Evaluation of the streambed leakage concept in analytical models using data from three pumping tests

Two-dimensional analytical models of pumping induced drawdown and stream depletion account for the streambed properties and stream geometry. Using data from three pumping tests performed under various hydrologic conditions, the following hypotheses were tested: (1) a partially penetrating stream at a certain distance from the pumping well can be represented by the streambed leakage term in the analytical models; (2) in the streambed leakage term, the streambed conductance coefficient λ=(WK′)/m′ accounts for the average stream width W, the streambed thickness m′, and hydraulic conductivity K′. The first hypothesis was tested by comparing results from tests under flow and no-flow stream conditions. The second hypothesis was tested by comparing results from two tests under low and high streamflow conditions. Similar estimates of the hydraulic conductivity and storativity for all tests indicate the validity of the streambed leakage term. The drawdown data of the test under low- and high-flow conditions (varying W) do not follow the predictions of the analytical models, which results in inconsistent λ estimates. Thus, for different hydrologic conditions, λ cannot be scaled solely by the stream width W. One possible explanation for this result is streambed dynamics caused by the changes in the stream stage.

---

Résumé Les rabattements dans les nappes et cours d’eau générés par des modèles analytiques de pompages en deux dimensions dépendent des propriétés du lit du cours d’eau considéré et de sa géométrie. En utilisant les données de trois essais de pompages effectués en conditions hydrologiques contrastées, les hypothèses suivantes ont été testées: (1) un ruisseau à pénétration partielle situé à une certaine distance du puits de pompage peut être exprimé par le terme de drainance dans les modèles analytiques; (2) dans le terme de drainance du lit du cours d’eau, le coefficient de conductivité du lit λ=(WK′)/m′ représente la largeur moyenne du cours d’eau W, l’épaisseur du lit m′, et la perméabilité K′. La première hypothèse a été testée en comparant les résultats des tests en conditions d’assèchement et d’écoulement. Pour éprouver la seconde hypothèse, les résultats de deux tests en conditions d’étiage et de hautes eaux ont été comparés. Sur l’ensemble des tests, la concordance entre les perméabilités calculées a permis de valider le terme de drainance du cours d’eau. Les rabattements observés en conditions d’étiage et de hautes eaux (W variable) ne suivent pas les prédictions des modèles analytiques, fournissant au final des estimations discordantes de λ. Aussi, dans le cas de conditions hydrologiques différentes, la seule largeur du cours d’eau ne suffit pas à estimer λ. Une explication possible de ce résultat réside dans la dynamique du lit causée par les variations de niveau du cours d’eau.

---

Resumen Modelos analíticos en dos dimensiones de descenso inducido por bombeo y de agotamiento de la corriente del río representan las propiedades del lecho del río y la geometría del mismo. Usando datos de tres ensayos de bombeo llevados a cabo bajo condiciones hidrológicas variadas, se probaron las siguientes hipótesis: (1) puede representarse un río parcialmente penetrante a una cierta distancia del pozo de bombeo por el término de goteo a través del lecho del río en los modelos analíticos; (2) en el término de goteo a través del lecho del río, el coeficiente de conductancia del lecho del río λ=(WK′)/m′ representa el ancho promedio del río W, la potencia del lecho del río m′, y la conductividad hidráulica K′. La primera hipótesis fue comprobada comparando los resultados de pruebas bajo condiciones de flujo y no flujo de corriente. La segunda hipótesis fue probada comparando los resultados de dos pruebas bajo condiciones de corriente altas y bajas. Estimaciones similares de la conductividad hidráulica y el almacenamiento para todas las pruebas apuntan hacia la validez del término de goteo a través del lecho del río. Los datos de descensos de la prueba bajo condiciones de flujo altos y bajos (variando W) no siguen las predicciones de los modelos analíticos, lo que resulta en estimaciones inconsistentes de λ. Así, para condiciones hidrológicas diferentes, λ no puede ser escalada solamente con el ancho de la corriente W. Una posible explicación para este resultado es la dinámica del lecho del río causada por los cambios en el nivel del río.

     

Influence of groundwater on the degradation of irrigated soils in a semi-arid region, the inner delta of the Niger River, Mali

The problem of soil degradation through alkalinization/salinization in an irrigated area with a semi-arid climate was examined in the inner delta of the Niger River, Mali, by the study of groundwater hydraulics and hydrochemistry in an area recharged by irrigation water. On the basis of data analysis on various scales, it is concluded that the current extent of the surface saline soils is due to a combination of three factors: (1) the existence of ancient saline soils (solonchaks) resulting from the creation of a broad sabkha west of the former course of the Niger River, now called the Fala of Molodo. These saline crusts were gradually deposited during the eastward tilting of the tectonic block that supports the Niger River; (2) the irrigation processes during the recent reflooding of the Fala of Molodo (river diversion in 1950). These used very poorly mineralized surface water but reintroduced into the alluvial groundwater system – generally of a low permeability (K=10–6?m?s–1) – salts derived from the ancient solonchaks; and (3) the redeposition of the dissolved salts on the surface due to the intense evapotranspiration linked to the present Sahelian climate. In this context, only efficient artificial draining of subsurface alluvial groundwater can eliminate most of the highly mineralized flow and thus reduce the current saline deposits.     

Inverse modeling of groundwater flow in the semiarid evaporitic closed basin of Los Monegros, Spain

 Only minor attention has been given in the past to the study of closed-basin hydrogeology in evaporitic environments, because these basins usually contain poor-quality groundwater. The motivation for hydrogeological research in the Los Monegros area in northeastern Spain was the approval in 1986 of a large irrigation project in the Ebre River basin. The irrigation of 60,000 ha is planned, partly in an evaporitic closed basin containing playa lakes. The project has given rise to environmental concerns. The evaluation of the hydrologic impacts of irrigation requires quantifying properly the hydrogeology of the area. With the available information, a conceptual hydrogeological model was formulated that identifies two main aquifers connected through a leaky aquitard. On the basis of the conceptual model, a numerical model was calibrated under steady-state conditions using the method of maximum-likelihood automatic parameter estimation (Carrera and Neuman, 1986a). The calibrated model reproduces the measured hydraulic heads fairly well and is consistent with independent information on groundwater discharge. By the solution of the inverse problem, reliable parameter estimates were obtained. It is concluded that anisotropy plays a major role in some parts of the lower aquifer. The geometric average of model conductivity is almost two orders of magnitude larger than the average conductivity derived from small-scale field tests. This scale effect in hydraulic conductivity is consistent with the findings of Neuman (1994) and Sánchez-Vila et al. (1996). Received, December 1997 · Revised, December 1997 · Accepted, January 1998     

Flow of groundwater with variable density and viscosity, Atikokan Research Area, Canada

 Flow of groundwater with variable density and viscosity was simulated at the Atikokan Research Area (ARA) in northwestern Ontario, Canada. An empirical viscosity–concentration equation was modified to include total-dissolved-solids (TDS) data from the ARA. The resulting equation was used successfully to estimate reasonably accurate viscosity values over the expected range of temperature and concentration, in comparison with experimental values derived for sodium chloride solutions. A three-dimensional finite-element code, MOTIF, developed by Atomic Energy of Canada Limited, was used in the simulations. The inclusion of the effects of depth-increasing temperature and TDS-dependent fluid-density distribution, while maintaining only a temperature-dependent viscosity relationship in a simulation, resulted in a more penetrative flow against expected buoyancy effects (i.e., the physics of the system was not honored). Accounting for concentration in the viscosity equation caused water to be less penetrative and more in accordance with the expected physics of the system. A conclusion is that fluid concentration should be considered simultaneously in calculating the density and viscosity of a fluid during modeling of variable-density flow in areas underlain by fluids with high TDS. Results of simulations suggest that both flow directions and magnitudes should be employed simultaneously during the calibration of a model. Large-scale groundwater movement in the ARA may be analyzed with carefully selected vertical no-flow boundaries. By incorporating the geothermal temperature gradient, groundwater recharge increases by 12%; thus, this gradient plays a significant role in groundwater flow at the ARA. Variability in the fluid concentration at the ARA neither decreases nor increases recharge into the groundwater system. The hypothesis that an isolated continuous regional flow system may exist at depth in the ARA is not supported by these simulations. Received, September 1996 Revised, September 1997, February 1998 Accepted, February 1998