Manfred KOCH, [2] Phatsaratsak ARLAI.

Transcription

1 Deterministic and stochastic modeling of groundwater flow and solute transport in the heavily-stressed Bangkok coastal aquifer, Thailand, and investigation of optimal management strategies for possible aquifer restoration [1] Manfred KOCH, [2] Phatsaratsak ARLAI [1] Department of Geotechnology and Engineering Hydrology, University of Kassel, Germany, [2] Department of Construction Technology, Nakhon Pathom Rajabhat University, Nakhon Pathom, Thailand, Abstract As part of a comprehensive study of the heavily stressed Bangkok coastal multi-aquifer system numerical simulations of the relevant groundwater flow and transport processes under the present-day- and possible future stress conditions have been performed. The major objectives of these investigations, the approaches taken to that regard are as follows: (1) 3D steady-state and transient calibration of the aquifer flow system using MODFLOW, including automatic parameter estimation code UCODE; (2) stochastic MCsimulations to take into account uncertainties of aquifer parameters, observed heads and reported pumping rates and comparison with analytical stochastic theory; (3) MTD3MS solute transport modeling and the determination of the cradles of saline groundwater pollution; (4) analysis of the present-day and future sustainability of the groundwater resources in the aquifer; (5) investigation of feasible aquifer restoration (remediation) schemes through groundwater management strategies and, (7) investigation of density effects of the saline plume concentrations on the results obtained above, using the SEAWAT model. Keywords: Bangkok coastal aquifer, groundwater pollution, stochastic modeling, optimal aquifer restoration 1. Introduction Bangkok and the adjacent provinces have experienced an economic boom over the last decades which have triggered a dramatic increase of population and industrial factories in both Bangkok and its adjacent provinces. As a consequence, groundwater demand has progressively augmented and hence uncountable wells have been drilled arbitrary into the Bangkok multilayered aquifers system - until 1983 often without government permission or control - to supplement the limited amount of surface water available. Eventually this has led to the existing situation where the actual groundwater withdrawal exceeds the natural aquifer yield, so that the piezometric heads in the aquifers system have significantly decreased, especially, in the 2 nd, the 3 rd and the 4 th aquifer layer, and head gradients have built up that are inducing an invasion of saltwater from its sources into the producing aquifers, leading to saltwater contamination there (Gupta, 1986; Gangopadhyay, 1997; Chaowiwat, 1999; Buapeng, 1999). However, notwithstanding several research activities in recent years, the mechanisms of saltwater intrusion in the Bangkok aquifers system are not yet fully clarified, namely, it is still a matter of debate whether the saline pollution encountered in some areas is due to either classical seawater intrusion or to vertical seepage of saline water from the topmost clay layer. The objective of the present paper is to investigate through numerical flow and transport modeling the saline groundwater contamination and, in particular, the cradles of the saline water and then to develop possible effective mitigation plans. Since Thai water authorities are presently considering the implementation of an Aquifer Storage Recovery (ASR) program which relies mainly on artificial recharge, one specific aim of the investigation is to conduct a numerical feasibility study of possible recharge schemes and, if so, to provide some preliminary schemes for their optimal design. The present work complements earlier studies to this regard of Arlai et al., (2006a,b,c,d) such that the best ASR- management scheme found by classical (constant-density) flow and transport modeling will be tested here by including possible density-dependency of the saline contamination plumes into the flow and solute transport process, using the SEAWAT-2000 model (Langevin et al., 2003). As presented by (Arlai and Koch, 2007b), there is some evidence of the importance of such densityeffects for understanding the pollution dynamics in the Bangkok aquifers system. 1

2 2. Hydrogeology and groundwater situation The Bangkok multi-aquifer system is located underneath the lower Chao-Praya basin which is demarcated in the east, north and west by ridges of hills and mountains and, in the south, by the Gulf of Thailand. The multi-aquifer system comprises of a topmost soft/stiff clay layer and eight complex water bearing layers under Bangkok and neighboring provinces (cf. Arlai, 2007; Arlai et al., 2006a) for details) (Fig 1). The prime recharges into the groundwater basin occur at the basin flanks where the 2 nd to 4 th aquifer is outcropping. Groundwater is mainly withdrawn from the Phra Pradaeng (2 nd ), Nakhon Luang (3 rd ) and Nonthaburi (4 th ) aquifers so that water levels in these aquifers have declined from 30 m to 60 m below MSL in some areas in recent decades, which has led to land subsidence of more than 20 mm/year underneath Bangkok metropolis and to saltwater encroachment in some places, either due to horizontal seawater intrusion or to vertical seepage of saline connate water from the marine clay layer, an issue which is the major focus of the present study. Fig. 1. 3D geological map of Bangkok aquifers 3. Model set-up and implementation 3.1 Flow model The numerical flow model used is the 3D finite-difference MODFLOW-96 model (Harbaugh and McDonald, 1996). The model setup has been modified from an earlier MODFLOW model of Kokusai Kogyo (1995) whereby the aquifer is divided in 9 layers, with 55 rows and 52 columns, and grid sizes varying from 2x2 km 2 to 16x16 km 2 (Fig. 2). The top boundary of the model which represents the water table is specified as constant head. The main recharges into the groundwater basin are at the basin flanks where the 2 nd to 5 th modeled layer are outcropping and are simulated also through Dirichlet BC s. Recharges inside the basin are assumed to be zero, as the topmost clay layer (varying from 15 to 30 m thickness) obstructs seepage of surface waters into the deeper aquifer layers. The bottom of layer 9 is specified as a no-flow Neuman BC. Some cells in the 55 th row in the 1 st to 3 rd layer that connect directly to the Gulf of Thailand are fixed as constant head at sea level. Initial transmissivities and vertical leakances are modified from Kokusai Kogyo (1995) and Gangopadhyay (1997), but have been supplemented by new geological profiles. 3.2 Solute transport model The solute transport model used is the MT3DMS model (Zheng and Wang, 1999) which simulates density- independent!! reactive transport, using the groundwater flow field computed by MODFLOW. As for the solute boundary conditions, seawater Dirichlet BC s conditions are set up at cells located in the Gulf of Thailand. Dirichlet BC s of specified (observed) concentrations are also specified for cells in the upper clay layer that acts as a major source of saline pollution (cf. Arlai, 2007). 2

3 Fig. 2. FD grid in modeled layer 5, Nonthaburi aquifer (a); 3D-FD grid of 9-layer aquifers model (b). 4. Steady-state and transient model calibration 4.1 Deterministic steady-state calibration Reasonable calibrated parameters are estimated and evaluated in the MODFLOW flow simulation by using (1) the conventional misfit error (trial-and-error) approach (Anderson and Woessner, 1992), (2) sensitivity analysis by automatic non-linear regression, using UCODE and, (3) pure stochastic analysis with Monte Carlo random simulations. The steady state calibration of the relevant hydraulic parameters has been carried out using the water year 1999 head data from 179 monitoring wells (cf. Arlai, 2007; Arlai et al., 2006a,c). A qualitative evaluation of the calibration success is provided by visual inspection of similar patterns between computed and observed heads in the 3 rd, 4 th and 5 th model layers which are the prime productive groundwater layers and where most of the observation wells are located (Figure 2). A quantitative assessment is carried out by the analysis of the scatter plot of measured against computed heads (Figure 3a) and by a measure of the residual error quantified by (1), the mean error (ME), (2) (a) (b) (c) Fig. 3: Observed versus steady-state computed heads for 1999 in (a) layer 3, (b) layer 4, (c) layer 5. 3

4 Observed Head (m. at MSL) R 2 = Computed Head (m. at MSL) (a) Calibration Measure (m.) Change in Tranmissivity and Vertical Leakance in Percentage ME of T MAE of T RMS of T ME of Vk MAE of Vk RMS of Vk (b) Fig. 4: (a) Observed versus computed heads for 1999, with upper and lower 95% confidence limits; (b) ME, MAE, RMS obtained when transmissivities T and vertical leakances Vk are varied percentally from their optimal calibrated values. the mean absolute error (MAE) and, (3) the root mean squared error (RMS) (Figure 3b). The scatter plot reveals a well-posed calibration, since all points are closed to the diagonal line, with the coefficient of determination R 2 being close to one. Moreover, the diagrams of the sensitivities for ME, MAE and RMS (Figure 3b) disclose that this set of calibrated hydraulic parameters is optimal in the sense that it provides the best result among the manifold of slightly perturbed parameter values. 4.2 Deterministic transient calibration The transient flow and solute transport calibration is based on observations of heads and salinity between 1993 to 1997 and 2000 to From this analysis one gets some important information about the relation among discharge, storage change and recharge with time and, in particular, the equilibrium yield, as illustrated in Fig. 5 and discussed in more detail in Arlai et al. (2006a). Equilibrium System Fig 5. Discharge, storage and recharge changes between 1993 and 2002 and the equilibrium yield 4.3 Stochastic calibration using MC-simulations Although the deterministic modeling approach used previously, namely, the trial-and error has resulted in satisfactory calibrations, the observed objective piezometric head data could not be fitted perfectly, leaving a nonzero residual as quantified by the RMS (see Figure 4b). There are two reasons for this: (1) the objective head and/or the pumping data are not exactly measured and/or (2) the particular, deterministic calibration parameters obtained represent only a local instead of a global minimum of the piezometric response surface. On the other hand, one has to assume that the Bangkok 4

5 aquifer system is more heterogeneous than is pictured by the zonal calibrated transmissivity T and leakance fields Vk obtained so far. Moreover, given that local estimates of T from pump tests and of Vk from geological borehole profiles are available that point to a rather heterogeneous subsurface structure, one would like to condition the model calibration on these subjective model observations. To this avail a stochastic modeling approach has been performed (cf.arlai et al. 2006c, for details). Applying a random field generator, realizations of a logarithmic transmissivity field Y=lnT with 2 various sets of variances σ Y and correlation lengths (λ x, λ y ) for each layer that characterize the possible stochastic range of the lnt field in the multi-aquifer system are simulated. Using these 2 Monte Carlo MODFLOW simulations, we investigated how σ Y contributes to the σ 2 H of the observed head and/or residuals. Stochastic theory, primarily developed by Gelhar and coauthors (cf. Gelhar, ) predicts that σ H and σ Y are related to each other as σ ~ 2 H σ Y *λ 2. The ultimate goal of this analysis has been to understand which factors affect the residual error of the model estimation. Obviously, both transmissivity variations and errors in the head measurements are mostly responsible for a non-zero estimated residual head. Hence, the variances of head that are obtained from stochastically generated transmissivities and the intrinsic errors of the head measurements were determined. The results (cf. Arlai et al., 2006c) show that the stochastically predicted variances of the head are still somewhat lower than the variances of the residual head, indicating additional uncertainties in the fitted model. Indeed, the pumping rates turn out to be very evasive. To investigate the effects of the latter on the residual head variance, Monte Carlo simulations with randomly disturbed pumping rates of varying magnitudes are performed. The results show that pumping plays a smaller but still significant role for the estimation of the residual error, as the residual head variances obtained from stochastic pumping are lower than those of the stochastic transmissivity field. For further details of this stochastic modeling approach we refer the reader to Arlai et al. (2006c). 5. Investigation of the cradles of the saline pollution The steady state water balance of the Bangkok aquifers system for the water year 1999 is shown in Fig. 6 and discloses that 34% of the inflow into the aquifer system originates in the clay layer, while only 6% of the inflow intrudes from the sea, providing evidence that the saline water in the marine clay is indeed the main pollution cradle in the Bangkok aquifers. Additional support of the dominant adverse effect of the vertical saltwater intrusion on the groundwater quality comes from the inspection of observed chloride Cl - concentrations, illustrated in Fig.6. Steady state water budget in

6 (a) (b) Fig 7. Monitoring wells and profiles (a) and sensitive leakage areas using 1990 measured data (b). Figs. 7 and 8. The 2D vertical cross-sections of Cl - (Fig. 8) illustrate firstly that the primary contamination source is the topmost enriched saline clay layer and, secondly, a finger-like structure of the vertical plume movement, most likely due to high-conductivity channels in the heterogeneous aquifer, but also due to local pumping effects in the middle and lower layers of the aquifer. Based on these profiles, there has been delineated an area of sensitive saltwater leakage in the aquifer system (Fig. 7, right panel). Using the flow and transport model-modflow-96&mt3dms, the downward migration of the saltwater emanating from the top clay layer has been simulated, results of which are shown in Fig. 9 and appear to, at least qualitatively, corroborate the observations. These snapshots unveil evidently two different contamination sources, namely, the upper clay layer from which saline plumes are sinking into the aquifer and the seawater in the Gulf of Thailand which intrudes horizontally. Profile 1 Profile 2 Profile 3 Profile 4 Profile 5 Profile 6 Fig. 8. Observed chloride (mg/l) fingerprints for several profiles (see Fig. 7) in

7 Fig. 9. Snapshots of simulated saline intrusion plumes for 1993 (a), 1995 (b) and 1997 (c). The results obtained with this simplified approach, though, must be taken with a grain of salt, as the negative buoyancy effects of the salinity are not taken into account at this stage, with the consequence that the adverse pollution effects might be somewhat underestimated. For more details on the analysis of possible density-effects in the present aquifer system we refer to Section 8 and, in particular, to Arlai and Koch (2007b) 6. Delineation of a contamination risk map Using the results of the saltwater transport modeling of the previous section, we are now able to delineate particular types of contamination zones, with respect to their sources and effects, i.e. to establish a pollution risk map of the Bangkok aquifer system. Such a risk map could be very helpful for a successful future management of the groundwater quality in the area. Four types of contamination zones have been demarcated (Fig.10), namely, (1) a zone of seawater intrusion (c > 4000 mg/l nearby the shoreline), (2) a mixing zone between horizontal seawater and vertical saltwater intrusion, (3) a zone of shallow vertical saltwater intrusion (c > 1000 mg/l in the 3 rd layer) and, (4) a zone of deep saltwater intrusion (c > 1000 mg/l in the aquifer down to the 5 th layer), the latter actually defining a highly sensitive intrusion zone (HSI-Z). This map can obviously clarify the sources and effects of most of the saline contamination in the Bangkok aquifer system. As a matter of fact, it discloses, for example, that seawater intrudes heavily into the coastal aquifer near Samut Sakhon, southwest of Bangkok, but less so along the Samut Prakan shoreline, while vertical saltwater pollution occurs mainly in a band that extends from west of Bangkok in northeastern direction up to Pathum Thani, Lam Luk Ka. Contamination is especially prevalent in the highly sensitive intrusion zone southwest of Pathum Thani. 7

8 Fig. 10. Risk map that delineates four types of contamination zones. Blue color marks the area of seawater intrusion; dark, the green area of mixing between horizontal seawater- and vertical saltwater intrusion; green, the area of vertical, shallow saltwater intrusion; and red, the area of deep saltwater intrusion, i.e. the highly sensitive intrusion 7. Sustainable yield and unmet water demand Arlai et al. (2006a) investigated the possibility of applying the so-called concept of sustainable yield to Bangkok aquifers system. The sustainable yield has been defined there as the maximal groundwater yield that may be withdrawn so that the water levels in the 3 rd, 4 th and 5 th layer do not decrease by more than 25% of their current water levels (Dec, 2002) and/or that the chloride concentrations are < 250 mg/l, i.e. a threshhold value that corresponds to internationally accepted water drinking standards. Fig. 11. Unmet water demand( umd ) in year 2032 in various provinces when projecting the groundwater extraction over the next 30 years. 8

9 Using again the MODFLOW-96/MT3DMS density- independent flow and transport model, future groundwater pumping scenarios are simulated that can fulfill the sustainable conditions above. The results of these simulations indicated (cf. Arlai et al.,2006a) that by diminishing the groundwater withdrawal rate in layers 3, 4 and 5 at the annual rate of 1.2%, 1.2% and 1.9%, respectively, the sustainable yield in 2032 is 4.86 x 10 5 m 3 /d. As this amount of groundwater volume cannot serve the future demand, there will be a difference, the so-called unmet demand (umd) that must be supplemented by surface water. This umd is illustrated in Fig. 11 for each province affected. However, the more detailed numerical analysis (cf. Arlai et al., 2007a;b) shows that, even though the drop of the piezometric heads might be acceptable with regard to the sustainability limit, saline pollution within layers 3, 4 and 5 cannot be significantly alleviated. There are two main reasons for this adverse behavior, namely, (a) saline water leaks downward from the uppermost marine clay layer into the lower ones and (b), because of the principle direction of the groundwater flow being from the model boundaries toward the center of the aquifers, the hydraulic gradients induced by the groundwater pumping cannot push back the saline water to its original source-location 8. Numerical simulation of possible aquifer restoration schemes 8.1 MODFLOW-96&MT3DS constant density flow and transport models Using the (constant-density) flow and transport model-modflow-96&mt3dms, Arlai et al. (2006b, 2007) have investigated 31 possible remedial schemes of possible aquifer restoration that consist mainly of a combination of policies or/and constructive measures. The most salient implications of these schemes on the saline plume distribution in 2032 are illustrated in the left panels of Fig. 12 and can be briefly summarized as follows: (a) The laissez-faire (without scheme, WOS ) scheme leaves the future pump rates in the 3 rd to 5 th aquifer layer as in 2002, but increases future pumping in the 6 th to 9 th layer by the same average growth rate as has been encountered in the last two decades, 1983 to The minimum piezometric heads obtained with WOS in 2032 is -52, -68 and -76 m (MSL) in layers 3, 4 and 5, respectively. Moreover, in some sections of the aquifer the saltwater plume migrates deeply downward, whereas seawater intrudes landward over 12 km in layer 3, clearly limiting the usable aquifer yield. (b) The best policy scheme consists in keeping the present pump rate in all layers until 2011, decrease the pumping thereafter to 60% of today s value in low-sensitive zones, but entirely shut off the pumps in the high-sensitive ones. This scheme has the appeal to give 5 more years for enactment. It turns out that this option is the best among the non-constructive measures, with recovered heads of -29, -37 and -42 m (MSL) in layers 3, 4 and 5, respectively, in Meanwhile the saline water plume area is reduced by 1.6 km 2 and seawater invades 2 km less than with WOS. (c) The best integrating policy- and constructive scheme keeps pump rates at the values of 2002, with the pumps in the deep layer 6 completely shut off in year Recharge wells are set up along the seawater intrusion front, (c > 4000 mg/l), and a clean-up well penetrating to the depth of the saltwater front (c< 1000 mg/l) that extracts saline water which, after treatment, is recharged back. This scheme results in minimum water levels of -49, -65 and -76 m (MSL) in layers 3, 4 and 5, respectively. And the saltwater plume area is decreased by 1.2 km 2 and seawater intrusion is pushed back by 4 km relative to the reference scheme WOS. At this point we would like to mention an extension of the present trial-and-error approach to find an optimal restoration scheme by the authors (Arlai et al., 2007), namely, the use of a constrained optimization groundwater management tool (GWM) to fine-tune the search for optimal rechargedischarge schemes for aquifer restoration. The results obtained with this novel technique (cf. Arlai, 2007, for details) indicate that, for the same hydraulic constraints of aquifer restoration, additional cost-benefits in the implementation can be achieved. 9

10 8.2 SEAWAT variable density models As the constant-density flow and transport model used so far may not be able to accurately represent the real physical mechanism of the transport of dense solutes, such as the vertical saline plume migration and seawater intrusion encountered here, the 3D variable- density groundwater flow and solute transport model SEAWAT-2000 (Langevin et al., 2003) has additionally been used to simulate the same aquifer remediation schemes. The SEAWAT-model results obtained for the remediation schemes discussed are illustrated in the right panels of Fig. 12 and are to be compared directly with the left panels. Most strikingly, one observes from these two figures that the salinity plume migrations for both model-approaches do not show any significant difference, i.e. the inclusion of variable density into the model does not appear to have any noticeable effect, at least, with regard to the future saline plume evolution. However, a more detailed inspection of the piezometric heads obtained for the two model approaches reveals that these are about 5 m lower for the variable- than for the constant-density model in the producing layers indicated in the previous section. For a more detailed discussion of the various aspects of densitydependent flow and transport in the present aquifer system we refer to Arlai and Koch (2007b). Fig 12. Left panels: Snapshots of simulated saline plumes for Dec, 2032, using the constant-density MODFLOW&MT3DMS model and three kinds of remediation schemes: WOS (top), best policy (middle), and best integrating policy- and constructive (bottom). Right panels: Same as left panels, but using the variable-density flow and transport model SEAWAT

11 9. Conclusions As part of a comprehensive study of the Bangkok coastal multi-aquifer system (cf. Arlai, 2007, for details) which, in the wake of a tremendous population- and industrialization increase in recent decades in that major part of Thailand, has come under heavy stress, leading to a decline of both groundwater tables and -quality, numerical simulations of the relevant groundwater flow and transport processes under the present-day- and possible future stress conditions have been performed. The major objectives of these investigations, the approaches taken, and the results obtained to that regard are as follows: (1) 3D steady-state and transient calibration of the multi-layered aquifer flow system using the MODFLOW model, including automatic parameter estimation code UCODE; (2) stochastic MC-simulations to take into account the uncertainties of aquifer parameters, observed heads and reported pumping rates and comparison with results of analytical stochastic theory; (3) MTD3MS solute transport modeling and the determination of the cradles of saline groundwater pollution encountered in most sections of the aquifer system and which are supposed to either the horizontal seawater intrusion from the Gulf of Thailand and/or the widespread upper marine clay layer containing saline formation waters; (4) analysis of the present-day and future sustainability of the groundwater resources in the aquifer, both quantitatively and qualitatively, by calculation of the sustainable yield of the aquifer, and which turns out to be already exceeded nowadays and, much more so, in the future target year 2032, which means that there will be an unmet water demand that is computed for the various provinces; (5) numerical investigation of feasible aquifer restoration (remediation) schemes through groundwater management strategies that include policy - or non-constructive measures, as well as a combination of policy - and constructive (use of recharge- and clean-up wells) measures by trial and error approaches; (6) use of an optimization groundwater management tool (GWM) to fine-tune the search for optimal recharge-discharge schemes for aquifer restoration and which indicates that, for the same hydraulic constraints of aquifer restoration, additional cost-benefits in the implementation can be achieved. (7) investigation of the possible effects of the, hitherto, neglected density-dependency of the flow on the saline plume concentrations on the results obtained above, using the variable-density SEAWAT model and where, surprisingly, the conclusions are that such a, computationally much more burdensome, approach may not be needed in the present groundwater flow and transport application. Putting all things together, the present exhaustive modeling exercise should offer Thai water management authorities some important guidelines for the future management of the Bangkok aquifer, one of the most precious, but also vulnerable groundwater resources of Thailand. References Anderson, M.P. and W.W. Woessner (1992) Applied Groundwater Modeling, Simulation of Flow and Advective Transport, Academic Press, San Diego, CA, 381pp.. Arlai, P. (2007) Numerical Modeling of possible Saltwater Intrusion Mechanisms in the Multiple-Layer Coastal Aquifer System of the Gulf of Thailand, Ph.D. Thesis, University of Kassel, Germany, 148pp. Arlai, P., M. Koch, and S. Kootanakulvong (2006a) Modeling Flow and Transport for Sustainable Yield Estimation of Groundwater Resources in Bangkok Aquifer system, European Geosciences Union General Assembly in the Poster, April 2-4,

12 Arlai, P., M. Koch, S. Koontanakulvong and Weerapol B. (2006b) Numerical Modeling as a Tool toinvestigate the Feasibility of Artificial Recharge to Prevent Possible Saltwater Intrusion into the Bangkok Coastal Aquifers System, In: Proceedings of Groundwater Hydraulics in Complex Environments, Toulouse, France, June 12-14, Arlai, P., M. Koch, and S. Koontanakulvong (2006c.) Statistical and Stochastic Approaches to Assess Reasonable Calibrated Parameters in a Complex Multi-Aquifer System, In: Proceedings of CMWR XVI, Copenhagen, June 19-22, Arlai, P., M. Koch and S. Koontanakulvong (2006d) Numerical Investigation of the Cradle of Saline Contamination and Effective Remediation Schemes for Amending Saline Water Pollution Problem in the Bangkok Coastal Aquifers System, Poster, 3rd APHW Conference, Bangkok, October 16-18, Arlai, P., M. Koch and S. Koontanakulvong (2007a) Embedding an Optimization Module within a 3D Density Dependent Groundwater and Solute Transport Model to determine an effective Groundwater Management Scheme in the Bangkok Aquifers System, In: Proceedings of Asian Simulation and Modelling 2007, Chiang Mai, Thailand, January 9-11, Arlai, P. and M. Koch (2007b) Need for density-dependent flow and transport modeling of horizontal seawater and vertical saltwater intrusion in the Bangkok-multilayered aquifers system?, In: Proceed. 12 th National Civil Engineering Conference; Thailand Advances with Civil Engineering, Pitsanulok, Thailand, May 2-4, Buapeng, S. (1999) Special Lecture on Groundwater Crisis and Land Subsidence in Bangkok Areas and Suburban, at Department of Water Resources Engineering, Chulalongkorn University, Bangkok, Thailand. Chaowiwat, W. (1999) Simulation of Saltwater Intrusion in Nonthaburi Aquifer, M.Eng Thesis, Chulalongkorn University, Bangkok. Gangopadhyay, S., (1997) Deterministic-Stochastic Modelling in a Complex Groundwater System, AIT, D.Eng Dissertation, WM-96-5, Pathum Thani, Thailand. Gelhar, L.W. (1993) Stochastic Subsurface Hydrology, Prentice Hall, Englewood Cliffs, NJ. Gupta, A.D., (1986) Simulated Salt-Water Movement in the Nakhon Luang Aquifer, Bangkok, Thailand, Groundwater, Vol. 23: Harbaugh, A.W. and M.G. McDonald (1996) Programmer's documentation for MODFLOW-96, an update to the U.S. Geological Survey modular finite-difference ground-water flow model: U.S. Geological Survey Open-File Report , 220 pp. Kokusai Kogyo Co.,Ltd. (1995) Study on Management of Groundwater and Land Subsidence in the Bangkok Metropolitan Area and its Vicinity, JICA, Department of Mineral Resources, Bangkok. Langevin, C.D., W.B. Shoemaker, and W. Guo (2003) Documentation of SEAWAT-2000 version with the variable-density flow process (VDF) and the integrated MT3DMS transport process (IMT), USGS Report Sanford, W.E. and S. Buapeng (1996) Assessment of a Groundwater flow model of the Bangkok basin, Thailand, Using Carbon-14-Based Ages and Paleohydrology, J. Hydrology, Vol.4: Zheng, C. and P.P. Wang (1999) MT3DMS: A modular three dimensional multispecies transport model for simulation of advection, dispersion, and chemical reactions of contaminants in groundwater system; Documentation and user s guide. US Army Corp of Engineers, Report SERDP