Sensitivity analysis of Seawater intrusion model in Tra Vinh province, Mekong Delta

Seawater intrusion into coastal aquifers is a serious problem, leading to shortage of water supply in Tra Vinh province which is a typical coastal province in Mekong Delta Vietnam. One of the main reason for salinity intrusion is excessive abstraction from groundwater. In addition, climate change is an uncertain challenge which may affect the situation in the future. GMS groundwater modelling is an effective tool to simulate groundwater �ow and salinity transport processes in this research. A conceptual and numerical model were built and calibrated to simulate groundwater �ow in the case study area located in Tra Vinh with the use of hydraulic head data accumulated in the period 2006 to 2017, a sensitive analysis has been conducted on hydraulic conductivities, speci�c storage and conductance of boundary condition in order to understand model behaviour. Model results show that groundwater is overexploited in Tra Vinh province. Groundwater levels were continuously decreased in the whole area and in all aquifers. Groundwater recharge is limited due to impermeable layers on surface and is recharge is only possible in sand dune areas. Groundwater abstraction supplies domestic water demand which is much more than recharge. The sensitivity analysis shows that hydraulic conductivity is the most sensitive factor impacting the vales of the groundwater heads. Hydraulic conductance of the boundary is the most sensitive parameter under the head-dependent �ow boundary, and hydraulic conductivity is the least sensitive parameter. However, speci�c storage is more sensitive to impact the values of change of storage. Thus, Boundary conductance, hydraulic conductivity and speci�c storage coe�cient are important parameters for the accuracy of the model results. Saline groundwater occurs in the coastal zone and in the north area. Saltwater intrusion occurs in these area. The salinity data availability is too limited to provide a clear description of salinity distribution in the area.


Introduction
Groundwater resources are important freshwater resources which are often used as drinking water, agricultural water and industrial water in most of regions (Vengadesan and Lakshmanan 2019).Coastal areas belong to one of the richest environment of agriculture, industry and shery, more than half of the word's mega-cities lie within 50 km of the coast, and population densities here are 2.6 times greater than island (Montagna 2013), so population is often dense in coastal areas (McLusky and Elliott 2004, Wolanski and Elliott 2015, Dang 2019).Dense population, agricultural and industrial activities demand a lot of freshwater in costal zones, while fresh surface water was polluted, effected by climate change and overexploitation (Post 2005).Groundwater resources often are the main freshwater resources in most of costal resigns, but groundwater resources are vulnerable to human activities and climate change especially in coastal area (Unsal 2014, Michael 2017).
Demands of freshwater is increasing because of population growth and economic development.Fresh groundwater is excessively abstracted in coastal regions due to a growth in water demanding activities.
Aquifers of coastal regions connect fresh groundwater in elds to sea water (Janardhanan 2012).
Excessive exploitation of groundwater results in some negative consequences such as seawater migrating into aquifer systems which is called as saltwater intrusion, increased freshwater pollution and in uences for local ecosystem (Post and Werner 2017).
Saltwater intrusion obviously affects the quality of the groundwater, over 1% of seawater (250 mg/l chloride) turns fresh water in unsuitable drinking purpose (Werner 2013).Not only overexploitation but also global warming which cause sea level raise and other climate change impacts affects the groundwater quality (Janardhanan 2012).
Simulating saltwater intrusion with the use of groundwater ow and salinity transport model is challenging, so sensitive analysis have important signi cant to model calibration and provide recommendation for data collection (Razavi 2015).Sensitivity analysis can help to analysis the robustness of the model and simplify model (Pannell 1997, Saltelli, Ratto et al. 2008).few sensitive analysis studies addressed parameter sensitives in saltwater intrusion modelling process.Shoemaker (2004)reported that dispersivity is important parameter to simulating and observing hydraulic head and salinity through parameter sensitives analysis for SEAWAT model (Shoemaker, W. B 2004).
In this study, the major challenge is simulating groundwater ow and salinity transport process.Output of the GMS model is variable because of the varying input factors during the simulation process.But the model which we need is an accurate simulation between model's output and real-word phenomena.
Parameter sensitives are applied to calibrating and assess the model and improve models input value to obtain more accurate model.

Material And Methods
GMS is an integral groundwater modelling environment, it is often used to simulate and construct groundwater models, for instance groundwater ow, groundwater transport and salinity intrusion models.
GMS provides pre-and -post processors for MODFLOW, SEAWAT and MT3DMS which are needed to be used in the research.
The GMS model has been built to evaluate the condition of groundwater abstraction and salinity intrusion in Tra Vinh province.Salinity intrusion as the main problem limiting groundwater development.
Groundwater ow and saltwater transport models were chosen to assess groundwater conditions.In order to simulate groundwater system with the use of GMS tools, some basic steps are necessary in GMS modelling process.

Research site
The Mekong river ows into the sea through a number of branches (Hoc 2005).Mekong Delta is located in the southwestern of Vietnam the downstream of the Mekong River.It is a very at area.Tra Vinh province is located in Mekong delta costal resign, which is separated with other province by Co Chine River (boundary between Ben Tre, Vinh Long and Tra Vinh province on the north) and Hau River (boundary between Soc Trang and Tra Vinh province on the west).On the east of Tra Vinh province is the East Sea with 65km costal line (Thu 2007).The province is a lowland costal area, the elevation of this area is only 0.5-1m above sea level (Nguyen 2000).
The area of Tra Vinh is about 2215.1km2,there are about 1.1 million inhabitants live there.Population density is about 460/km2 (Nguyễn 2011).It comprises one provincial capital city (Tra Vinh) and seven districts (Cang long, Chau Thanh, Cau Ke, Tieu Can, Cau Ngang, Tra Cu and Duyen Hai).The topography is very complex, as there are hundreds of mounds and sand caves, alluvial deposits, coastal plain and complex network of rivers and cannels in Tra Vinh province show in gure1.
Sea transgression and regression formed and developed Mekong Delta in Quaternary period (Chiem 1993).The soil is deposited by oods from the rivers and sea in this delta, meanwhile different particles of sediment deposited in different area which cause distribution of different soil textures in different areas (Ve 1990).The type of soil is an important factor which determine the use of land in Vietnam (Postma 2007).
There are three major soil types in Tra Vinh (1) sandy soil; (2) alluvial land; and (3) acidic soil (Nguyen 2020).Figure 2 shows the map of soil types in Tra Vinh province.
Nowadays, people abstract groundwater to supply daily life, industries production and irrigate agriculture in Tra Vinh (Ha 2015).Meanwhile the demands of fresh groundwater are increasing due to climate change, urbanization, social-economic development and population density growth in Tra Vinh (Danh 2008).So excessive groundwater extraction for different usages is a serious problem in Mekong Delta, especially in Tra Vinh province which is a typically coastal province of Vietnam in Mekong Delta (Minh 2014).The abstraction of groundwater volume is about 347,793m3/d in 2016 in Tra Vinh, it could increase in the future (Van 2017, Van Hiep 2018).
As time goes by, the balance between abstraction and recharge in coastal aquifer is disturbed due to overexploitation.This leads to groundwater level decrease and salinity intrusion in Tra Vinh (Vermeulen 2013).On the other hand, coastal aquifers are vulnerable to seawater intrusion.This problem is also caused by sea level rise which is triggered by global warming.The quality effects on valuable groundwater resources results in a shortage of water availability and other serious damages (Nguyen 2014).Seawater intrusion threatens nearly 38000 hectares of cultivated rice eld every year in Tra Vinh (Herrera-Pantoja and Hiscock 2008), it has also caused fresh water shortages in people daily life and economic development.

Construction of conceptual model
The domain of the model is for the whole Tra Vinh province which is about 2215.1 km2.Groundwater abstraction data, recharge and hydraulic data from 2007 to 2016 was put into GMS model to simulate groundwater ow process under the GIS map in Tra Vinh province.A hydrogeological conceptual model has been built to describe basic characteristics of groundwater system.It is the basic of for creating numerical models such as MODEFLOW, MT3DMS and SEAWAT.The conceptual model include: Aquifer structures; Boundary conditions; Hydrogeological parameters; Hydrological stresses and observations.the situation of observation wells show in table1.

numerical model
Groundwater ow models are normally used to simulation the hydraulic ow patterns process with the use of three-dimensional model.MODFLOW is the most commonly used software in numerical groundwater ow models with the use of nite different method.MODFLOW is applied in groundwater simulation process, aiming to solve unsustainable problems and then optimize system process.The partial differential equation is the base of MODFLOW for con ned aquifer (McDonald and Harbaugh 1988).It is: Where K xx K yy K zz are values of hydraulic conductivity along x, y and z coordinate axes, all of these are assumed to be parallel to the major axes of hydraulic conductivity (Lt-1); h is the groundwater head (L); W is a volumetric ux per unit volume and represents sources and sinks of water (t-1); Ss is the special storage of the porous material (L-1); t is the time (t).
MT3DMS is multi-species and three-dimensional transport model for simulation of groundwater systems which include advection, dispersion and chemical reaction of dissolved constituents process, it is evolved from MT3D (Zheng and Wang 1999).
MT3DMS linked with MODFLOW, it uses hydrologic and discretization features of MODFLOW which support the ow solution which is required through the transport simulation process.The governing equations in MT3DMS is (Zheng and Wang 1999) Where Ck is the dissolved concentration of species k, ML-3; θ is the porosity of the subsurface medium, dimensionless;t is the time, T; xi is the distance along the respective Cartesian co-ordinate axis, L; Dij is the hydrodynamic dispersion coe cient tensor, L2T-1; vj is the seepage or linear pore water velocity, LT-1; It is related to the speci c discharge or Darcy ux through the relationship v= qi/θ; qs is the volumetric ow rate per unit volume of aquifer representing uid sources (positive) and sink (negative), T-1; Csk is the concentration of the source or sink ux for species k, ML-3; ∑Rn is the chemical reaction term, ML-3T-1.

Darcy's law connect transport equation with ow equation, it is:
Where K i is principal component of the hydraulic conductivity tensor, LT-1; h is hydraulic head which can be calculated from the 3D groundwater ow equation, L.

Set-up of numerical models
A numerical groundwater ow model was built.The model consisted of 135 rows and 151 columns with a constant cell size of 500m x 500m grids, and included 13 layers that consist of 7 aquifers and 6 aquitards.
In the new model, each of 6 con ned aquifers were divided into two model layers.The top uncon ned aquifer and 6 aquitards were still simulated with one mode layer since they are very thin.In total, 19 model layers were included in the model.Stress period is de ned one month length and 121 stress periods were set up from 2006 to 2017 which was similar with old model.
The main step to rebuild the numerical model are in follow: 1. a grid frame was created to cover the same model area which has a dimension of length is x as 75000, length in y as 67000.
2. the old model grid data was saved and converted to 2D scatter data sets to be used to provide top and bottom elevations and starting heads for the new model.4. the conceptual model was revised according to 19 model layers.Nineteen coverages of hydrogeological parameters were created, abstraction wells were redistributed to new aquifer layers, observation wells were assigned to new aquifer layers.The lateral boundaries in the north, east, and west of the aquifer layers were de ned as head-dependent ow boundaries and were simulated with MODFLOW General Head Boundary (GHB) package.The coastal line was simulated by two different boundary conditions: head-dependent ow boundary (GHB) and a constant head boundary.The results were compared in sensitivity analysis.
5. the data de ned in the conceptual model was transferred into MODFLOW model by automatic mapping and conversion tool in GMS.Table 2 provides the link between GMS conceptual model coverage and MODFLOW packages.

Numerical model calibration
The transient groundwater ow and transport model were calibrated through comparing the observed values and computed values, hydrogeological parameters were adjust to achieve the best simulation between calculated and observed groundwater head values.Calibration aimed at layers 1, 3, 13 and 19, because there are a big error between observed and computed groundwater head and salt concentration.
Manual calibrated was used in this research, the accuracy of the calibration was checked with statistic indicators: mean error (ME), root mean squared error (RMSE) and Standard Deviation.

Sensitivity analysis
Advection package play an important role in MT3DMS simulation process, there are some solution schemes that can be divided into three major techniques which can used to solve transport equation in this package.i.e., standard nite different method, higher-order nite-volume TVD method and partial tracking based Eulerian-Lagrangian methods.
Dispersion and sink mixing package are also main packages to solve the concentration change due to dispersion with explicit nite different method and uid sink mixing with the explicit nite different method separately.
Sensitive analysis was carried through adjusting the hydraulic conductivity, speci c storage, boundary conductance and with the use of GHB boundaries.
The hydraulic conductivity values from the calibrated model were used as benchmark values.The hydraulic conductive values of all aquifer layers were increased by 50% and decreased by 50% systematically, The output data were different by increasing 50% and decreasing 50% of hydraulic conductivities values (changing K h and K v at same time).The computed groundwater levels and water budgets were computed with values from the benchmark values to show sensitivity of groundwater levels and budget to hydraulic conductivity.
In the same way, sensitivity of speci c storage was checked by increasing 50% and decreasing 50% of the benchmark values.Where K is the average hydraulic conductivity; W is the thickness of the saturated aquifer perpendicular to the ow direction; L is the boundary length perpendicular to the ow direction; D is the distance from the general head boundary to the model boundary The conductance values were also increased and decreased by 50% of the benchmark values to check the in uences of boundaries.When the coastal line was simulated by general head boundary, the conductance was also changed accordingly.When the coastal lime was simulated by constant head boundary, only conductance of other boundaries were changed.

Model calibration
The transient MODFLOW model was calibrated by try and error method.Hydraulic conductivities and abstraction rates were adjusted in order to get the better t of calculated and observed groundwater head.The plot of calculated groundwater head and observed groundwater head in wells Q07701H (layer 1and Q021050 (layer 19) are shown in Figure 3.
The water budget for the whole model are shown in Figure 5.The groundwater abstraction as the main out ow of groundwater has increased, and changes of the groundwater storage have decreased from 2006 to 2017.Meanwhile changes of the groundwater storage and recharge from top uncon ned aquifer which as the main in ow source of groundwater have the same trend in each stress periods.The boundary in ow slowly rises and is an important source of the groundwater storage.River leakage has a little effect on the groundwater storage.
SEAWAT was run to check TDS variations in space and on time.Currently, the SEAWAT model could not be calibrated due to lack of long-term continuous measurements of the concentration.The plot of calculated TDS and observed TDS in the observation wells Q07701H (layer 1) are shown in Figure 6.

in uences of Hydraulic conductivities
Groundwater head results in each stress periods in wells Q07701H (layer 1), Q404020 (layer 4), Q40403T (layer 7), Q40403Z (layer 13), Q2017040 (layer 16) and Q021050 (layer 19) are shown in Figure 8.The mean computed groundwater head and difference between previous and new results of K values of 0.5K 0 and 1.5K 0 , are shown in Table 3.
Groundwater levels in different aquifers react differently on the changes of hydraulic conductivities as shown in Figure 8.In the top uncon ned aquifer, groundwater level decreases when the hydraulic conductivity increases since more water will ow downwards to lower aquifers through leakage (Q07701H (layer 1)).Groundwater levels in lower con ned aquifers increase when the hydraulic conductivity increases since con ned aquifers receive the leakage from the top aquifer.The magnitude of groundwater levels changes comparing to the benchmark values (k 0 ) is presented in Table 3.The relative changes of groundwater levels are higher when the hydraulic conductivity values are decreased by 50%.
The water budget has a strong relation with the hydraulic conductivities, especially the boundary in ow and the change of storage.The boundary in ow results with the use of different K values and the difference between each situations in Tra Vinh are shown in Figure 9 and Table 4, respectively.
Changes of the hydraulic conductivities have slight in uence on the boundary in ow as shown in gure 9.The boundary in ow increases when the hydraulic conductivity increases since more water will in ow the model domain through boundaries.The changes of the boundary in ow compared to the benchmark values (k 0 ) are shown in Table 4.The comparative changes of the boundary in ow are higher when the hydraulic conductivity values increase with 50%.ues are decreased by 50%.
The change of storage is also impacted by changes of the hydraulic conductivities as shown in Figure 10.Change of storage increases when the hydraulic conductivity increases since more water will be stored in the aquifers.The change of storage of 0.5k 0 and 1.5k 0 comparing to the one of benchmark values (k 0 ) is shown in Table 5.The relative change of storage is higher when hydraulic conductivity values are increased by 50% (1.5k 0 ).

3..2.2 in uences of Storage coe cients
The storage coe cient is also one of the important factors which has signi cant impact on the model output values.The output data were different by increasing 50 percent and decreasing 50% of the speci c storage values (Ss).The groundwater head results in each stress periods in wells Q07701H (layer 1), Q404020 (layer 4), Q40403T (layer 7), Q40403Z (layer 13), Q2017040 (layer 16) and Q021050 (layer 19) are shown in Figure 11.It is shown in Table 6 with the mean computed groundwater head and difference between new results of Ss values changes into 0.5 Ss0, 1.5 Ss 0 and previous results in 11 wells.
The groundwater levels in different aquifers react differently on the changes of speci c storage coe cients as shown in Figure 11.Groundwater level in aquifers increases when the speci c storage coe cients increases since more water will storage in aquifers.The magnitude of changes of groundwater levels comparing to the benchmark values (Ss 0 ) is presented in Table 6.The relative changes of groundwater levels are small in each aquifers when the hydraulic conductivity values are decreased by 50% or increased by 50%.
Changing of speci c storage have obvious in uences on boundary in ow is shown in Figure 12.
Boundary in ow increase when the speci c storage change 50% since more water will in ow the model domain through boundaries.The change of boundary in ow comparing to the benchmark values is shown in Table 7 The comparative changes of boundary in ow are higher when hydraulic conductivity values are decreased by 50%.
Change of storage is also impacted by changes of speci c storage shown in Figure 13.Change of storage increase when the speci c storage changed by increased 50% and decreased 50% since more water will storage in aquifers.The varying of change of storage comparing to the benchmark values (Ss 0 ) is shown in Table 8.The relative change of change of storage is higher when speci c storage values are decreased by 50%.

in uences of Boundary conductance
Output data were different by increasing 50% and decreasing 50% of conductance values of general head boundary (C ghb0 ).Groundwater head result in each stress periods in wells Q07701H (layer 1), Q404020 (layer 4), Q40403T (layer 7), Q40403Z (layer 13), Q2017040 (layer 16) and Q021050 (layer 19) were shown in Figure 14.The mean compute groundwater head and different between new results which C ghb0 values are changed into 0.5 C ghb0 and 1.5 C ghb0 and previous results in 11 wells can be seen in Table 9.
Groundwater levels in different aquifers react on the changes of boundary conductance as shown in Figure 14.Groundwater level increases when the boundary conductance increases since more water will ow downwards to aquifers through recharge and leakage from different boundaries.The magnitude of changes of groundwater levels comparing to the benchmark values (C ghb0 ) is presented in Table 9.The relative changes of groundwater levels are higher when the boundary conductance values are decreased by 50%.11.The relative change of change of storage is higher when boundary conductance values are decreased by 50%.boundary in ow are higher when boundary conductance values are decreased by 50%.

Groundwater head sensitive to different parameters
Groundwater levels were affected by adjusting hydraulic conductivity, speci c storage coe cient, and conductance of boundaries.Hydraulic conductivity as a measurement for the di culty of water ows through sediments is the most sensitive parameter for groundwater levels.Groundwater levels in different aquifers changes from 0.3 to 4.5m by decreasing or increasing 50% of hydraulic conductivity, and it just changes from 0.06 to 0.5m and 0.1 to 0.6m for adjusting 50% of speci c storage values and boundaries conductance values separately.But groundwater levels in different aquifers have different sensitivity for hydraulic conductivity.Groundwater levels in lower con ned aquifers are more sensitive than the top uncon ned aquifer and the changes of top uncon ned aquifer are below 0.5m and it is over 1m in lower aquifers.The major reason is that recharge happened in uncon ned aquifer which caused that groundwater resilience in top uncon ned aquifer is higher than the one in lower con ned aquifers.Accuracy of hydraulic conductivity is exact signi cant to obtain accurate groundwater levels.

Boundary in ow sensitive to different parameters
Boundary in ow is also a signi cant observed values to sensitive analysis for hydraulic conductivity, speci c storage and boundary conductance.The most sensitive parameter for boundary in ows in the whole model domain is boundary conductance.The change of boundary in ow by increasing 50% of boundary conductance is about 10000 m 3 /day, but there are only nearly 500 m 3 /day and 1600 m 3 /day of change by increasing 50% for hydraulic conductivity and speci c storage, separately.Comparing with change of about 24 m 3 /day and 6800 m 3 /day by decreasing 50% hydraulic conductivity and speci c storage values respectively, there is a change of more than 15000m 3 /day by decreasing 50% of boundary conductance.The reason might be that conductance is more directly factor than hydraulic conductivity and speci c storage to impact the boundary in ow.In order to get reasonable boundary in ow values, it is suitable to adjust boundary conductance values closer to the real.

Change of storage sensitive to different parameters
The changes of hydraulic conductivity, speci c storage and boundary conductance also have different in uences towards change of storage values.Change of storage is more sensitive to speci c storage, sensitive to increase of hydraulic conductivity when increase 50% of hydraulic conductivity the different between changed change of storage and benchmark change of storage values is trend 5000 m 3 /day, but it trend to zero when decrease hydraulic conductivity by 50% .And it is less sensitive to boundary conductance, the different between changed boundary conductance values and benchmark values are trend to zero.Both hydraulic conductivity and speci c storage coe cient are important for further precise storage values change.

Conclusions
A groundwater ow and salinity transport model is developed to simulating seawater intrusion in a coastal area with complex hydraulic network in this study.Parameter sensitivity analyses are applied to study the importance of different hydrogeology conditions in seawater intrusion costal aquifers processes.
Form the results of groundwater ow and salinity intrusion model, The abstraction of groundwater as the main water supply source is limited in the fresh lenses.Overexploitation of groundwater in recent years caused groundwater levels continuously to decline and saline groundwater is distributed in many of the areas of Tra Vinh province.The freshwater supply is in shortage for domestic and industries because of salinity intrusion.
In the other hand, abstraction of groundwater is continuously increasing from 2006 to 2017 as the groundwater recharge rate is less than the abstraction rate in the same period.It cause groundwater levels and groundwater storage to decrease rapidly.Seawater intrusion in aquifers happens in many areas of Tra Vinh province.The aquifers most effected happen in the top Holocene aquifer as an important source to provide fresh groundwater by abstraction wells.Seawater intrusion caused total dissolved solid to exceed the standards for domestic and industries water in groundwater.So control of groundwater abstraction has to be considered.
Overall, groundwater head and tot dissolved solid are important observation parameters to seawater intrusion simulation.Hydrological parameters of the groundwater ow and saline transport system are also important to simulations in the costal aquifer and groundwater abstraction systems.The parameter sensitives analysis indicates that simulations of salinity intrusion in costal aquifers can be advantageous to understand groundwater hydrogeological conditions, and play an important role in model calibrations.
Hydraulic conductivity, speci c storage and boundary conductance are important hydrological parameters for this model.The output of groundwater levels, change of storage and boundary in ows in the model were used to conduct a sensitivity analysis of the hydraulic conductivity, speci c storage and boundary conductance.Some signi cant conclusions for parameter sensitives are summarized, Different outputs have different sensitivities with the same changes of the values of the parameters.With the use of a head-dependent boundary, groundwater levels are most sensitive in case of a decrease of 50% of hydraulic conductivity, boundary in ow is most sensitive in case of a decrease of 50% of boundary conductance and speci c storage coe cient is more sensitive parameters for change of storage.
The constructed groundwater ow and salinity transport model are useful to simulate groundwater hydraulic processes.But there are also a few limitations in the model, which includes (1) insu cient data (2) uncertain parameters Overexploitation of groundwater is the major reason of salinity intrusion.Reduced abstraction of groundwater is an effective method to mitigate seawater ow into costal aquifers in Tra Vinh province.It is suitable alternative to store rainwater in containers for living and irrigation instead of groundwater abstraction, which can be carried out in Tra Vinh.Since recharge in sandy areas is more effective than clay areas, it is a good strategy to increase recharge in sandy areas through in ltration trenches, road side in ltration trenches in cities and contour farming and trench in farmlands.Groundwater TDS concentration value in duration observed wells Q07701H (layer 1)

Tables
Page 28/37 meanwhile the study aims to conduct an in-depth analysis of factors affecting model accuracy in Tra Vinh province.This research include follow steps.1. model setup, conceptual model adjusting, numerical model construction, and model calibration.2. Sensitive analysis for parameters.3. Discussing the results of parameters sensitivity analysis.4. Conclusion through analysing results.
3. a new 3D grid was created with the model re nement which consisted of 750 rows, 670 columns and 19 layers.Then, a new MODFLOW model was constructed with the new model grid.The top and bottom elevations and starting heads of the new model were interpolated from the 2D scatter data sets of the old model.
6. MODFLOW model was run to compute transient groundwater levels and ow budgets.The computed levels were checked with observed values.7. a new MT3DMS model was constructed with the new grid, and cell-by-cell ow component generated by MODFLOW.Starting concentrations distribution in space were de ned the same with old model.Meanwhile third order TVD scheme technique was choose to solve the advection terms under MT3DMS Advection Package.8. the data de ned in the conceptual model was transferred into MT3DMS model by mapping, then running MT3DMS model to compute transient total dissolved solid concentration, the computed values were checked with observed values.9. new SEAWAT model was constructed based on MODFLOW and MT3DMS model, then transient salinity intrusion results were generated.
The sensitivity of the boundary conditions were checked under two different coastal line boundary: headdependent ow boundary and constant head boundary.The lateral boundary of north, west and east were simulated with: head-dependent ow boundaries.The amount of water ow through the boundaries is controlled by the conductance values as de ned as(McDonald and Harbaugh 1988): Figure 15.Boundary in ow increases when the boundary conductance increases since more water will in ow the model domain through boundaries.The change of boundary in ow comparing to the benchmark values (C ghb0 ) is shown in Table 10.The comparative change of Change of storage is also impacted by changes of boundary conductance shown in Figure 16.Change of storage increase when the boundary conductance increase since more water will storage in aquifers through boundary.The varying of change of storage comparing to the benchmark values (C ghb0 ) is shown in Table

Figures Figure 1
Figures

Figure 10 Change
Figure 10

Figure 13 Change
Figure 13

Table 1
Name and location of observation wellsTable 11 Change of storage results with different Cghb0 values