- Email: [email protected]
Modeling and control of saltwater intrusion in a coastal aquifer of Andhra Pradesh, India
Available online at www.sciencedirect.com
Journal of Hydro-environment Research 3 (2009) 148e159 www.elsevier.com/locate/jher
Modeling and control of saltwater intrusion in a coastal aquifer of Andhra Pradesh, India Bithin Datta a,b,*, Harikrishna Vennalakanti b, Anirban Dhar c a
Discipline of Civil and Environmental Engineering, School of Engineering, James Cook University, Townsville QLD 4811, Australia b Department of Civil Engineering, Indian Institute of Technology Kanpur, India c Department of Civil Engineering, Indian Institute of Technology Kharagpur, India Received 14 April 2008; revised 8 April 2009; accepted 3 September 2009
Abstract The saltwater intrusion in a coastal aquifer is a highly complex and nonlinear process. The management of coastal aquifers requires careful planning of withdrawal strategies for control and remediation of saltwater intrusion. Prediction and control of future saltwater distribution in coastal aquifer may be possible by simulating the processes utilizing mathematical models. A finite element based flow and transport simulation model is implemented for a coastal aquifer in Nellore district of Andhra Pradesh, India. The model is calibrated for two years of time period between July 2000 and July 2002, both in terms of hydraulic heads and salt concentration. The calibrated model is validated for next two years in terms of head and salt concentration with the available data for July 2002 and July 2004. The limited data availability in this study area may represent a typical real life situation. The calibrated and partially validated simulation model is used to evaluate the effectiveness of planned pumping strategies to locally control the saltwater intrusion in the study area. These limited evaluations show the potential for using a planned pumping strategy and therefore hydraulic control measures to affect the saltwater intrusion process. These simulations are also useful for evolving planned pumping strategies for control of saltwater intrusion in a coastal aquifer. Ó 2009 Published by Elsevier B.V. on behalf of International Association for Hydro-environment Engineering and Research, Asia Pacific Division. Keywords: Saltwater intrusion; Coastal aquifer; Groundwater modeling; Groundwater management; FEMWATER
1. Introduction Coastal areas are important for human settlement and development. It is estimated that more than half the world’s population lives within 60 km of the shoreline, and this number could increase up to three quarters in over a decade (UNCED, 1992). Over exploitation of groundwater in coastal aquifers may result in intrusion of saltwater. On the whole, contamination of coastal aquifers may lead to serious consequences on environment, ecology and economy of that region. This study is aimed at modeling and prediction of saltwater
* Corresponding author. Tel.: þ61 07 47814983(Office), þ61 07 47283684 (Home). E-mail address: [email protected] (B. Datta).
intrusion in a vulnerable area in the coastal belt of Andhra Pradesh in India, based on very limited hydrogeologic data. This data availability situation may represent a typical study area. The simulated saltwater intrusion process is also utilized to show the impact of planned pumping strategies on the intrusion process. The feasibility of using a spatial and temporal pumping strategy for control of the saltwater intrusion process is also demonstrated. Saltwater intrusion problem has been reported all over the world, e.g., Madras, India (Rouve and Stoessinger, 1980), Nile delta aquifer in Egypt (Sherif et al., 1988), Yun Lin basin in southwestern Taiwan (Willis and Finney, 1988), southwest region of Bangladesh (Nobi and Gupta, 1997), Hernando County, Florida (Guvanasen et al., 2000), Jahe river basin, Shandong province in China (Cheng and Chen, 2001), Savannah Georgia (Kentel et al., 2005).
1570-6443/$ - see front matter Ó 2009 Published by Elsevier B.V. on behalf of International Association for Hydro-environment Engineering and Research, Asia Pacific Division. doi:10.1016/j.jher.2009.09.002
B. Datta et al. / Journal of Hydro-environment Research 3 (2009) 148e159
In most of the cases the main reason for saltwater intrusion in coastal aquifers is due to over exploitation of groundwater and due to lack of proper management measures. Often, the management of coastal aquifers requires careful planning of withdrawal strategies for control and remediation of saltwater intrusion in coastal aquifers. Such strategies can be evolved only if, the physical processes involved in the coastal aquifers are reliably simulated and studied. A number of prescriptive optimal management models for saltwater intrusion is presented in Willis and Finney (1988), Finney et al. (1992), Das and Datta (1999a, 1999b, 2000), Kentel et al. (2005). Mathematical models, once implemented for a specific study area after calibration and validation can be used to simulate the response of the aquifer system to various pumping and recharge strategies. A calibrated and validated simulation model can be used to explore possible management or control strategies, and to predict the response to an implemented strategy. This study also aims at establishing the effectiveness of utilizing hydraulic control in terms of planned pumping to manage the saltwater intrusion process. A 3-D, transient, density dependent, finite element based flow and transport simulation model FEMWATER (Lin et al., 1997) is utilized for modeling of two Mandals (administrative units) in Nellore District, in Andhra Pradesh, India. This area is extensively utilizing pumped water from the underlying aquifers for agricultural, domestic and aqua cultural uses. The simulation model is calibrated using observed head and concentration data, and estimated withdrawal from the aquifer. The calibrated model is then utilized for evaluating the impact of specified pumping strategies for controlling the saltwater intrusion process. 2. Numerical simulation model Lin et al. (1997) developed a three-dimensional finite element computer model (FEMWATER) for simulating density dependent flow and transport in variable saturated media. FEMWATER is designed to solve the flow and transport through saturated e unsaturated porous media. The governing equation for flow is the modified Richards equation (Lin et al., 1997): r vh r r F ¼ V$ K$ Vh þ Vz þ q ð1Þ r0 r0 vt rO In the case of saltwater intrusion, the constitutive relationship between fluid density and concentration takes the form: r c ð2Þ ¼ 1þ3 r0 cmax The governing equations for transport describe the material transport through groundwater systems. These equations are derived based on the laws of continuity of mass and flux. The major processes that are considered in this study are advection, dispersion/diffusion, and injection/ withdrawal. The transport equation is written as
149
vc vh r q þ V$Vc V$ðqD$VcÞ ¼ a0 qc þ qcin qc r vt vt vh r0 r vq c ð3Þ þ F þ V$V vt r r0 vt The flow and transport processes are represented by equations (1) and (3) respectively. These two equations are coupled together by the density coupling coefficient, and by the Darcy velocities which makes the saltwater intrusion problem highly nonlinear. The finite element based simulation model, FEMWATER (Lin et al., 1997) is used in this study to solve these two governing equations simultaneously. 3. General description of the study area The study area is a coastal aquifer (Harikrishna, 2005) located downstream of the Penna river delta (Latitude: 14 350 2400 e 140 480 3600 , Longitude: 79 570 0000 e 80 100 1200 ) of Nellore district in Andhra Pradesh in India (Fig. 1). The coastal aquifer is already affected by saltwater intrusion problem. The study area consists of two Mandals: Allur (15 villages, 197 km2) and Vidavalur (10 villages, 158 km2) with population of 52,990 and 46,793 respectively. (Chief Planning Office, 2004). Pumped water from the underlying aquifers is extensively utilizing for agricultural, domestic, and aquaculture use. Water levels in the villages close to coastal line are already below the mean sea level. This is mainly due to huge requirement of water for aquaculture which is spreading in alarming rate in this area, and for irrigation of paddy the major crop of the area which requires huge amount of water. Adding to this huge requirement of water, there is considerable amount of rainfall deficit (actual rainfall much below the average annual rainfall for Allur: 1133 mm/yr, and Vidavalur: 1141 mm/yr). Annual rainfall is the only source for recharging this coastal aquifer. 3.1. Hydrogeology The geographical location of study area is show in Fig. 1. The study area falls under alluvium soil type. These soils comprise of admixtures of sand, silt and clay in various proportions. The quartz pebbles are invariably encountered at different depths in almost all places in alluvial areas. It is generally light brown to pale gray and sandy in nature. The thickness of coastal alluvium is very large as evident from the exploratory wells drilled by Central Ground Water Board (CGWB). Groundwater occurs in almost all the geological formations of the district. However, its potential depends upon the nature of the geological formation, geographical set up, incidence of rainfall, recharge and other hydrogeological characteristics of the aquifer of the area. Hydrogeologically, the study area comes under unconsolidated formations. These are formed by deltaic alluvium due to the Pennar river, wind blown sands and coastal alluvium along the coast and in inter deltaic areas. These alluvium formations consist of admixtures of sand, silt and clay and form a multilayered aquifer.
150
B. Datta et al. / Journal of Hydro-environment Research 3 (2009) 148e159
Fig. 1. Geographical location of study area.
4. Model development
4.3. Wells and pumping rates
One of the main difficulties in implementing a simulation model is the lack of reliable hydrogeologic and water use data. Data from various sources are utilized for developing and implementing the 3D finite element simulation model utilizing FEMWATER (Harikrishna, 2005).
The study area consists of various types of wells for withdrawing groundwater for various uses. Generally the area covers dug wells, bore wells, hand pumps and filter points. Filter points are dominant in this area. Depth of filter point wells vary from 6 to 30 m. However in most cases depth vary between 6 and 11 m. The only information available is the drafts for the year 2000. The pumping rates are estimated based on the village-wise draft for the time period between July 2000 and July 2001. Available data show that the variation in draft is significant in two Mandals. Allur is having maximum and minimum draft of 100.4 104 m3 and 4.4 104 m3 (for two different villages), whereas the maximum and minimum values for Vidavalur (Mandal) are 652.4 104 m3 and 35 104 m3 respectively. One of the most difficult information to obtain is the withdrawal from each known pumping well. The estimates may be unreliable, and may omit some withdrawals. In this study, the groundwater draft for the year 2000 has been increased by 10% to reflect this. This estimate is used as the pumping rate for the year 2000. As the draft values for the subsequent years are not available, pumping rates for subsequent years are increased by 25% of the previous pumping rates, to reflect the increasing trends of total pumping in this study area. The recharge rate in the form of vertical infiltration is taken as 15% of the normal annual rainfall. The locations of the wells and boundaries of the selected study area are shown in Fig. 2.
4.1. Hydraulic head data In all, 18 piezometers (Observation wells) provide the hydraulic head information for the entire Nellore district. Two of these wells fall in the study area. Hydraulic heads with respect to MSL in these 18 piezometers are used to interpolate the heads for each of the years between 2000 and 2004. Head contours were obtained from these interpolated head values. Initial head values are specified for each node of the 3D finite element model using scattered data points. In the present study, July 2000 is taken as beginning of the first time period. Later, the head contours from the year 2001 and 2002 are used for calibrating the model, and 2003 and 2004 contours are used for validation of the model, in terms of simulated hydraulic heads.
4.2. Chloride concentration data Chloride concentration data in the same locality is very abrupt; such outliers are not considered due to lack of reliability. Year 2000 chloride concentration data is taken as the initial concentration for the simulation model. Concentration data for the year 2001 is used for calibration and year 2002 and 2003 concentration data are used for validation of the model.
4.4. Stratigraphic data The thicknesses and type of the soil layers are specified based on the lithological data of the near by areas. This study
B. Datta et al. / Journal of Hydro-environment Research 3 (2009) 148e159
151
head (Dirichlet boundary) and constant concentration boundary, in which both the head and chloride concentration are constant over time. For this boundary, the specified head is taken as zero (MSL), and specified concentration as 2400 mg/l. The southern boundary is taken as no flow boundary (no change in hydraulic head across the boundary) since the river is almost dry for the past 5 years. The Mandal boundaries are considered as variable flux boundaries (zero Neumann condition). 4.6. Fluid properties The fluid properties needed to be specified are density of water (r0 ¼ 1000 kg/m3), dynamic viscosity of water (m0 ¼ 280985.76 kg/m yr), compressibility of water (b ¼ 4.71e-026 m yr2/kg) and dimensionless density reference ratio (3 ¼ 0.025). 4.7. Hydrogeological parameters
Fig. 2. Location of pumping wells.
area generally falls under alluvium type of soils which are generally admixtures of sand, silt and clay. In the simulation model that was used to simulate the coastal aquifer processes, three layers are considered. First layer below the ground surface is sand, of about 12 m thickness, followed by sandy clay of about 3 m thickness, followed by sand of 15 m thickness. The entire multilayered system is considered as heterogeneous and anisotropic. Stratification details are shown in Fig. 3. 4.5. Boundary conditions The study area is bounded by Bay of Bengal in the east side and by the Penna River in the south side. Administrative (Mandal) boundaries are considered in north and west sides as shown in the Fig. 2. The coastal boundary is defined as constant
Hydrogeological parameters like Hydraulic conductivity, compressibility of the medium are necessary for calibration of the simulation model. For the unsaturated zone, moisture content, relative conductivity and water capacity curves are generated by the model for specified type of material present in each layer. Data adequacy is a major problem in simulating real world systems. Therefore it is necessary to make certain assumptions while constructing the model. Seasonal fluctuations are not considered in this study. Also, tidal effect is not taken into account. The pumping and recharge rates are averaged over the year. The total value is the total pumping or recharge occurring over the year. 5. Model calibration The calibration of the model for the study area is done in two steps. Initially the simulation model is calibrated with respect to the hydraulic heads only, using the flow model. Once, the calibration with respect to the flow model is found to be satisfactory, the calibration of the model is carried out for both the flow and transport processes. Therefore, the first step of calibration is only for approximation of the hydraulic
Fig. 3. Geological stratification of the study area.
152
B. Datta et al. / Journal of Hydro-environment Research 3 (2009) 148e159
parameters and boundary conditions. Using the available data as input, the numerical model is first used for simulating the flow conditions. Subsequently, after approximate calibration of the model for hydraulic parameters, the flow and transport model is used for simulating coupled flow and transport condition. The calibration process is carried out for the time period between July 2000 and July 2002. The simulated and observed head and concentration values area compared for July 2001 and July 2002. The model is then validated for July 2003 and 2004 data, mainly in terms of head values as concentrations are even less reliable for comparison. The calibration model is then utilized for predicting the future scenarios for various pumping patterns. The calibration of flow and coupled flow and transport models are discussed below. 5.1. Simulation model The three-dimensional finite element based flow and transport simulation model FEMWATER is used to simulate the saltwater intrusion process in the coastal aquifer. The total number of nodes and elements of the model are 37,080 and 60,165 respectively. The 3-D model consists of three layers. The different layers of the model are shown in Fig. 3. Prism or wedge elements were used. The calibration is carried out for 2 years, starting from July 2000 with a constant time step of 0.04 yr. (14 days). Recharge is considered as 0.1 m/yr. Longitudinal and transverse dispersivities are taken as 50 m and 15 m respectively. Porosity is assumed to be 0.3. The simulation concentrations were found to be little sensitive to the dispersivities. Poroisity did effect the resulting head distribution. However, porosity value of 0.3 was considered as representative value for the study area based on soil information. The simulation results are certainly affected by specified boundary conditions, which was a major source of uncertainty. With the above input data, the transient flow condition is simulated for the two year period between July 2000 & July 2002. The simulation results are the spatial head distributions
at each time step (0.04 yr interval). The transient flow convergence criterion specified is 0.01. 5.2. Calibration and validation of flow and transport simulation model The three-dimensional finite element model simulates the spatial and temporal values of the hydraulic head. To calibrate the model with respect to heads, the simulated heads are compared with the observed heads at the end of 1st and 2nd years (July 2001 and July 2002). These observation locations are chosen from the set of available observation locations, distributed over the entire study area. 32 observation points are thus specified in the whole study area. For each of these observation points, calibration targets are assigned by considering an acceptable measure of deviation of simulated heads from the observed heads. The observed head are based on interpolation of the spatial head observations at available locations. In Fig. 4, the error bar quantifies the error between the observed head and simulated head. In this study, an interval of 0.5 m is specified as acceptable. This means observed head þ/0.5 m is acceptable in the calibration stage. The simulated hydraulic heads for different specified parameter values (hydraulic conductivity, recharge rate) are compared with the observed hydraulic heads as estimated by interpolation of observation data. The interpolation was necessary as the number of observation points were few in the study area. The spatial interpolation is based on available observation data outside the study area as well. The estimated pumping rates remain unchanged in calibration process as it is based on the observed draft. The hydraulic conductivity and recharge rates are adjusted to obtain simulated water levels which satisfactorily match with observed heads. In addition to high pumping, rainfall was scanty for the four years (2001e2004), and much below the annual average. This had further accelerated the groundwater salinity problem. The
Fig. 4. Comparison between observed heads and simulated heads for trial set 1.
B. Datta et al. / Journal of Hydro-environment Research 3 (2009) 148e159
153
Fig. 5. Comparison between observed heads and simulated heads for trial set 2.
only source of groundwater recharge is the infiltration of excess rainfall. Figs. 4(a) and (b) show the calibration errors in terms of the hydraulic heads at the observation locations, on July 2001 and July 2002. These calibration results are for simulated heads using, Kxx, Kyy ¼ 3650 m/yr and Kzz ¼ 1825 m/yr for sand, and Kxx, Kyy ¼ 150 m/yr and Kzz ¼ 75 m/yr for sandy clay, and infiltration (recharge) rate of 0.1 m/yr (trial set 1). It can be noted that these errors between observed and simulated heads are substantially exceeding the calibration target error of (þ/0.5 m). The calibration errors are smaller when Kxx, Kyy ¼ 7300 m/ yr and Kzz ¼ 3650 m/yr for sand, and Kxx, Kyy ¼ 180 m/yr and Kzz ¼ 90 m/yr for sandy clay, and infiltration (vertical recharge rate) is 0.2 m/yr as seen in Figs. 5(a) and (b) (trail set 2). However, some of the simulated heads are beyond the specified calibration targets. The value of hydraulic conductivity and the infiltration rate were further changed to Kxx, Kyy ¼ 11,000 m/yr and
Kzz ¼ 5500 m/yr for sand, and Kxx, Kyy ¼ 200 m/yr and Kzz ¼ 100 m/yr for sandy clay, and infiltration rate is 0.1 m/yr. The calibration errors for July 2001 and 2002 are shown in Figs. 6(a) and (b) (trial set 3). It is seen that the calibration errors are more or less within the acceptable error limit. Considering the inherent uncertainty in estimating the observed heads, and the reliability of the later, these calibration results are intuitively acceptable. Minimizing the errors further, may not mean a more accurate simulation. The parameter values obtained by the trial and error process of model calibration may not be unique. The calibration process may also leads to an understanding of the factors determining the system’s response and behavior, that is, the parameters and boundary conditions to which the system is most sensitive. Observing the results of these simulations, the best suited combination is Kxx, Kyy ¼ 11,000 m/yr and Kzz ¼ 5500 m/yr for sand, and Kxx, Kyy ¼ 200 m/yr and Kzz ¼ 100 m/yr for sandy clay, and a recharge rate of 0.1 m/yr. The simulated hydraulic heads and estimated observation heads were also
Fig. 6. Comparison between observed heads and simulated heads for trial set 3.
154
B. Datta et al. / Journal of Hydro-environment Research 3 (2009) 148e159
Fig. 7. Simulated versus observed hydraulic heads. July 2001 (Trial Set 3).
compared over time for all selected locations. A comparison of the observed and simulated hydraulic heads for these parameter values are shown in Fig. 7. These parameter values results in the best fit between observed and simulated heads. It was noted that, for smaller hydraulic conductivity values specified, some of the wells tend to dry up after some time. Therefore, for further validation, these parameter values are chosen. Based on these chosen parameter values, and specified boundary conditions, the saltwater concentration is simulated using the coupled 3-D, transient, density dependent flow and transport simulation model. The salt concentration patterns over space and time, as simulated for July 2001, July 2002 and July 2003, and the observed concentration patterns for the same times are shown in Figs. 8e10. All these concentration values are based on the concentrations at the bottom of the top sand layer, in the 3-D model.
Figures (Figs. 8e10) show that the calibration model performs satisfactorily in simulating the salt concentration in the study area over the time horizon for which observation data are available. These comparisons also show, the hydraulic head simulation errors obtained using the coupled flow and transport model are within the acceptable range, considering the inherent uncertainties and errors in the scanty observation data available. Some numerical experiments were conducted with different porosity and hydraulic conductivity values. As expected, dispersivity was not very important. Hydraulic heads were mostly sensitive to porosity and to certain extent to K values. This also affected the spatitemporal concentration distribution. However, the heads and concentrations were certainly sensitive to boundary conditions and to a lesser extent to pumping values. The pumping rates did influence the head and concentration, with localized effects being more prominent. The question is how to normalize these parameters to compare relative sensitivity. This specific quantification, suitable for presenting a meaningful comparison was not undertaken. Basically meaningful parameter values were utilized based on known values for area, and the calibration process was iterative incorporating the sensitivities of these parameters. 6. Results and discussions The calibrated and partially validated simulation model is used to evaluate the impact of few deliberately imposed pumping or withdrawal scenarios on the saltwater intrusion process in the study area. To observe the response of the system, some extra wells are specified in a hypothetical management scenario and the simulated results are compared with those for the existing conditions. The main aim is to establish if a specified or planned pumping strategy can be used to control the intrusion process.
Fig. 8. Salt Concentration Contours for July 2001 (in mg/l).
B. Datta et al. / Journal of Hydro-environment Research 3 (2009) 148e159
Fig. 9. Salt Concentration Contours for July 2002 (in mg/l).
Fig. 10. Salt concentration contours for July 2003 (in mg/l).
Fig. 11. Location of wells for different scenarios.
155
156
B. Datta et al. / Journal of Hydro-environment Research 3 (2009) 148e159
of July 2009, there is a decrement in salt concentration with extra wells in scenario-I. Fig. 14 shows concentration variation with time for both the pumping patterns at specified locations 1 in Fig. 12. Although only marginally, by inducing additional pumping at selected and carefully chosen locations, it is possible to reduce the salt concentration levels at some other locations. Therefore it is possible to lower salt concentrations by careful planning of pumping patterns. This may improve groundwater condition in areas away from the coast line by locally controlling the saltwater intrusion. 6.2. Scenario-II: additional pumping wells near coast line in Iskapalli village of Allur Mandal
Fig. 12. Concentration contours for existing pumping pattern (in mg/l).
6.1. Scenario-I: additional pumping wells near coast in Utukuru village of Vidavalur Mandal Additional pumping is assumed from five extra wells near the coast in the Utukuru village of Vidavalur Mandal. Fig. 2 and Fig. 11 show the old and new pattern (with extra wells) of the well locations, respectively. The pumping rates are kept at 3,50,0000 m3/yr for all additional wells specified in scenario-I. The pumping in the new wells starts from July 2005. Fig. 12 and Fig. 13 (a) show the simulated concentration contours after five years from July 2004, i.e., in July 2009. It is observed from the concentration contours
Five extra pumping wells are hypothetically specified near the east coast line in Iskapalli village of Allur Mandal. Figs. 2 and 11 show the old and new pattern of the well locations, respectively. The pumping rates in the additional wells are 3,50,0000 m3/yr. The pumping in these new wells starts from the fifth year i.e., in July 2005. Fig. 14 shows concentration variation with time for the existing and new pumping patterns at specified locations 2, as shown in Fig. 12. Actually, the increasing trend in concentration is arrested in locations 2 by inducing these additional pumping. 6.3. Scenario-III: additional pumping wells at Dampuru village of Vidavalur Mandal In scenario-III, five additional pumping wells are induced at Dampuru village of Vidavalur Mandal. Figs. 2 and 11 shows the old and new pattern of the well locations respectively. The pumping rates in the additional wells are 3,50,0000 m3/yr. The pumping in these new wells starts from the fifth year i.e., in July 2005. Figs. 12 and 13(b) show the concentration contours after five years from July 2004, i.e. in July 2009. It
Fig. 13. Concentration contours for scenarios I and III (in mg/l).
B. Datta et al. / Journal of Hydro-environment Research 3 (2009) 148e159
157
Fig. 14. Concentration at different locations for different scenarios.
can be observed from the resulting concentration contours of July 2009, that there is a decrement in concentration, with additional pumping at new wells in some locations, compared to the existing concentration pattern. In both the cases, concentration is decreasing with time, but the decrement rate is more in the new pattern compared to the old one. Fig. 14 shows concentration variations with time in the both pumping patterns, at specified location 3 shown in Fig. 12. Evaluating results also show that very close to the new pumping wells, the salt concentration may increase as well. Similarly, on the coastal side of a series of pumping wells near the sea front, the salt concentration may also increase. 6.4. Scenario-IV: pumping rates increased by 10% over the previous years from July 2004 to July 2009 in all the existing wells
evident near the coastline compared to other parts of the study area. These evaluation results obtained for hypothetical pumping scenarios, selected with some specific consideration for possible control of saltwater intrusion, no doubt establish the fact that it is possible to alter the salt concentration in the aquifer by modifying the pumping patterns. These results, although very limited in scope, do show that spatially varied pumping as a planned strategy may be utilized to control salt concentration patterns and therefore, to manage the saltwater intrusion process. The aim was to only establish the feasibility of using a planned pumping strategy for effective control and management of the saltwater intrusion process. These simulation results are also useful in predicting the salt concentration scenario in the future, if the existing pumping patterns continue in the future. 7. Conclusions
The pumping rates in all the existing wells are increased by 10% over the previous year from July 2004 to July 2009. The pumping rates for July 2004, are as described before. It is observed from the concentration contours of July 2009, there is some marginal increment in salt concentration with increase in pumping rates. Fig. 14 shows concentration variation for both the pumping rates at specified location 4 (Fig. 12). Fig. 14 shows that the concentration values at location 4 marginally increase with increase in pumping rates. This trend is more
A 3D, transient, density dependent, finite element based flow and transport simulation model, FEMWATER, is implemented for simulating the coupled flow and transport processes of saltwater intrusion in a coastal aquifer in Nellore district of Andhra Pradesh, India. Available data for a selected study area of around 355 km2 was obtained from different sources to be used as input data for implementing the numerical simulation model for the study area. Due to the
158
B. Datta et al. / Journal of Hydro-environment Research 3 (2009) 148e159
scanty nature of available data and questionable reliability of all available data, the best but subjective judgment was used in selecting the data for implementing the model. The numerical model was calibrated for two years time period, between July 2000 and July 2002, both in terms of hydraulic heads and salt concentration. The calibrated model was further validated for the chosen hydrogeological parameters with the data available between for 2002 and July 2004. The aquifer was considered heterogeneous in terms of vertical stratification. Both flow and transport are considered transient. Withdrawal from aquifer is estimated based on available data, and assuming an increasing trend over the period of calibration and validation. The calibrated and partially validated simulation model was used to predict the saltwater transport scenario in the study area at future time periods. It is worth mentioning that one can easily perform linked optimization parameter estimation, based on repeated call to the simulation model. In present situation the single run of the model takes huge computational time. Thus, it is not realistic to perform such optimization based calibration. The simulation was carried out till July 2009. The predicted head and concentration values show the future saltwater intrusion patterns if the present trend of pumping continues. The calibrated model was also used to evaluate the effect of modifying the pumping patterns on the future saltwater intrusion process. Effect of increase in withdrawal rates uniformly spread over the study area was also evaluated. These evaluation results show that a carefully planned pumping strategy is capable of modifying the saltwater intrusion process and the spatial and temporal concentration patterns. These results show that a well planned pumping strategy can help in controlling the saltwater intrusion process in space and time. However, due to a limited period of analysis, these impacts are evident only locally. These simulation results also show, if the withdrawal rate continues to increase over time it may have detrimental effect on the salt concentration in the study area. This study for a real life study area does demonstrate the feasibility of using planned pumping strategy to control the saltwater intrusion process. The main aim was to demonstrate that in a realistic study area, hydraulic control measures can influence the spatial and temporal distribution of saline concentration. Theoretically, it should be possible to identify a management strategy for coastal aquifers, using such planned control measures. This study does not try to identify an optimal control strategy. It only tries to establish the plausibility of actually obtaining such a strategy. In order to demonstrate the feasibility of using hydraulic control as a management measure, the study area was selected. It was felt that the sparsity of data available for this study area may represent a typical real life situation. Therefore, an attempt was made to calibrate and validate a numerical simulation model for this study area with very limited data. These evaluation results are based on very limited data and are limited in scope. Actual implementation of planned pumping strategy for economic management of the coastal aquifer will require more rigorous calibration and validation with additional reliable data. Only then the impact of any planned pumping strategy can be accurately evaluated.
Acknowledgement Authors are thankful to different government organizations of the Nellore district: State Groundwater Department (Andhra Pradesh, INDIA), Chief Planning office, Mandal Revenue Offices of both Mandals, and the Water Supply Departments for various types of data. References Cheng, J.M., Chen, C.X., 2001. Three-dimensional modeling of density dependent saltwater intrusion in multilayered coastal aquifers in Jahe River basin, Shandong province, China. Ground Water 39 (1), 137e143. Chief Planning Office 2004 ‘‘Hand book of statistics 2003e04, Nellore District’’, Nellore. Das, A., Datta, B., 1999a. Development of multiobjective management models for coastal aquifers. Journal of Water Resources Planning and Management, ASCE 125 (2), 76e87. Das, A., Datta, B., 1999b. Development of management models for sustainable use of coastal aquifers. Journal of Irrigation and Drainage Engineering, ASCE 125 (3), 112e121. Das, A., Datta, B., 2000. Optimization based solution of density dependent seawater intrusion in coastal aquifers. Journal of Hydrologic Engineering, ASCE 5 (1), 82e89. Finney, B.A., Samsuhadi, S., Willis, R., 1992. Quasi-three-dimensional optimization model of Jakarta basin. Journal of Water Resources Planning and Management, ASCE 118, 18e31. Guvanasen, V., Wade, S.C., Barcelo, M.D., 2000. ‘Simulation of regional groundwater flow and saltwater intrusion in Hernando county, Florida’. Ground Water 38 (5), 772e783. Harikrishna, V., 2005. ‘‘Modeling prediction and control of saltwater intrusion in a coastal aquifer of Andhra Pradesh India’’ M.Tech. dissertation, I.I.T. Kanpur, India. Kentel, E., Gill, H., Aral, M.M., 2005. ‘‘Evaluation of groundwater resources potential of Savannah Georgia region,’’ Multimedia Environmental Simulations Laboratory, Report No. MESL-01-05. Lin, H.J., Rechards, D.R., Talbot, C.A., Yeh, G.T., Cheng, J.R., Cheng, H.P., Jones, N.L., 1997. A Three e Dimensional Finite Element Computer Model for Simulating Density -Dependent Flow and Transport in Variable Saturated Media. Version 3.1. U.S Army Engg. Research and Development Center, Vicksburg, M.S. Nobi, N., Gupta, A.D., 1997. Simulation of regional flow and salinity intrusion in an integrated stream aquifer system in a coastal region: southwest region of Bangladesh. Groundwater 35 (5), 786e796. Rouve, G., and Stoessinger, W. (1980), Simulation of the Transient Position of the Saltwater Intrusion in the Coastal Aquifer near Madras Coast, Proc. 3rd Internat. Conf. Finite Elements in Water Resources, Univ. of Miss., Oxford, Miss. Sherif, M.M., Singh, V.P., Amer, A.M., 1988. ‘ A two e dimensional finite element model for dispersion (2D-FED) in coastal aquifer’. Journal of Hydrology 103, 11e36. UNCED, 1992. Agenda 21: Chapter 17.3. http://www.un.org/esa/sustdev/ documents/agenda21/english/agenda21chapter17.htm [August, 2007]. Willis, R., Finney, B.A., 1988. Planning model for control of saltwater intrusion. Journal of Water Resources Planning and Management, ASCE 114 (3), 333e347.
Notations The following symbols are used in this paper: c: material concentration in aqueous phase; cin: material concentration in the source; cmax: maximum material concentration; D: dispersion tensor; F: storage coefficient; h: pressure head;
B. Datta et al. / Journal of Hydro-environment Research 3 (2009) 148e159 K: hydraulic conductivity tensor; Kxx: hydraulic conductivity in x direction; Kyy: hydraulic conductivity in y direction; Kzz: hydraulic conductivity in z direction; q: volumetric flow rate of source and/or sink; t: time; V: discharge; z: potential head; a0 : compressibility of the medium;
b: compressibility of water; 3: dimensionless density reference ratio; q: moisture concentration; m0: dynamic viscosity of water; r: water density at chemical concentration c; r0: referenced water density at zero chemical concentration; r*: density of either the injection fluid or the withdrawn water; V: del operator; (.): vector dot product.
159
Journal of Hydro-environment Research 3 (2009) 148e159 www.elsevier.com/locate/jher
Modeling and control of saltwater intrusion in a coastal aquifer of Andhra Pradesh, India Bithin Datta a,b,*, Harikrishna Vennalakanti b, Anirban Dhar c a
Discipline of Civil and Environmental Engineering, School of Engineering, James Cook University, Townsville QLD 4811, Australia b Department of Civil Engineering, Indian Institute of Technology Kanpur, India c Department of Civil Engineering, Indian Institute of Technology Kharagpur, India Received 14 April 2008; revised 8 April 2009; accepted 3 September 2009
Abstract The saltwater intrusion in a coastal aquifer is a highly complex and nonlinear process. The management of coastal aquifers requires careful planning of withdrawal strategies for control and remediation of saltwater intrusion. Prediction and control of future saltwater distribution in coastal aquifer may be possible by simulating the processes utilizing mathematical models. A finite element based flow and transport simulation model is implemented for a coastal aquifer in Nellore district of Andhra Pradesh, India. The model is calibrated for two years of time period between July 2000 and July 2002, both in terms of hydraulic heads and salt concentration. The calibrated model is validated for next two years in terms of head and salt concentration with the available data for July 2002 and July 2004. The limited data availability in this study area may represent a typical real life situation. The calibrated and partially validated simulation model is used to evaluate the effectiveness of planned pumping strategies to locally control the saltwater intrusion in the study area. These limited evaluations show the potential for using a planned pumping strategy and therefore hydraulic control measures to affect the saltwater intrusion process. These simulations are also useful for evolving planned pumping strategies for control of saltwater intrusion in a coastal aquifer. Ó 2009 Published by Elsevier B.V. on behalf of International Association for Hydro-environment Engineering and Research, Asia Pacific Division. Keywords: Saltwater intrusion; Coastal aquifer; Groundwater modeling; Groundwater management; FEMWATER
1. Introduction Coastal areas are important for human settlement and development. It is estimated that more than half the world’s population lives within 60 km of the shoreline, and this number could increase up to three quarters in over a decade (UNCED, 1992). Over exploitation of groundwater in coastal aquifers may result in intrusion of saltwater. On the whole, contamination of coastal aquifers may lead to serious consequences on environment, ecology and economy of that region. This study is aimed at modeling and prediction of saltwater
* Corresponding author. Tel.: þ61 07 47814983(Office), þ61 07 47283684 (Home). E-mail address: [email protected] (B. Datta).
intrusion in a vulnerable area in the coastal belt of Andhra Pradesh in India, based on very limited hydrogeologic data. This data availability situation may represent a typical study area. The simulated saltwater intrusion process is also utilized to show the impact of planned pumping strategies on the intrusion process. The feasibility of using a spatial and temporal pumping strategy for control of the saltwater intrusion process is also demonstrated. Saltwater intrusion problem has been reported all over the world, e.g., Madras, India (Rouve and Stoessinger, 1980), Nile delta aquifer in Egypt (Sherif et al., 1988), Yun Lin basin in southwestern Taiwan (Willis and Finney, 1988), southwest region of Bangladesh (Nobi and Gupta, 1997), Hernando County, Florida (Guvanasen et al., 2000), Jahe river basin, Shandong province in China (Cheng and Chen, 2001), Savannah Georgia (Kentel et al., 2005).
1570-6443/$ - see front matter Ó 2009 Published by Elsevier B.V. on behalf of International Association for Hydro-environment Engineering and Research, Asia Pacific Division. doi:10.1016/j.jher.2009.09.002
B. Datta et al. / Journal of Hydro-environment Research 3 (2009) 148e159
In most of the cases the main reason for saltwater intrusion in coastal aquifers is due to over exploitation of groundwater and due to lack of proper management measures. Often, the management of coastal aquifers requires careful planning of withdrawal strategies for control and remediation of saltwater intrusion in coastal aquifers. Such strategies can be evolved only if, the physical processes involved in the coastal aquifers are reliably simulated and studied. A number of prescriptive optimal management models for saltwater intrusion is presented in Willis and Finney (1988), Finney et al. (1992), Das and Datta (1999a, 1999b, 2000), Kentel et al. (2005). Mathematical models, once implemented for a specific study area after calibration and validation can be used to simulate the response of the aquifer system to various pumping and recharge strategies. A calibrated and validated simulation model can be used to explore possible management or control strategies, and to predict the response to an implemented strategy. This study also aims at establishing the effectiveness of utilizing hydraulic control in terms of planned pumping to manage the saltwater intrusion process. A 3-D, transient, density dependent, finite element based flow and transport simulation model FEMWATER (Lin et al., 1997) is utilized for modeling of two Mandals (administrative units) in Nellore District, in Andhra Pradesh, India. This area is extensively utilizing pumped water from the underlying aquifers for agricultural, domestic and aqua cultural uses. The simulation model is calibrated using observed head and concentration data, and estimated withdrawal from the aquifer. The calibrated model is then utilized for evaluating the impact of specified pumping strategies for controlling the saltwater intrusion process. 2. Numerical simulation model Lin et al. (1997) developed a three-dimensional finite element computer model (FEMWATER) for simulating density dependent flow and transport in variable saturated media. FEMWATER is designed to solve the flow and transport through saturated e unsaturated porous media. The governing equation for flow is the modified Richards equation (Lin et al., 1997): r vh r r F ¼ V$ K$ Vh þ Vz þ q ð1Þ r0 r0 vt rO In the case of saltwater intrusion, the constitutive relationship between fluid density and concentration takes the form: r c ð2Þ ¼ 1þ3 r0 cmax The governing equations for transport describe the material transport through groundwater systems. These equations are derived based on the laws of continuity of mass and flux. The major processes that are considered in this study are advection, dispersion/diffusion, and injection/ withdrawal. The transport equation is written as
149
vc vh r q þ V$Vc V$ðqD$VcÞ ¼ a0 qc þ qcin qc r vt vt vh r0 r vq c ð3Þ þ F þ V$V vt r r0 vt The flow and transport processes are represented by equations (1) and (3) respectively. These two equations are coupled together by the density coupling coefficient, and by the Darcy velocities which makes the saltwater intrusion problem highly nonlinear. The finite element based simulation model, FEMWATER (Lin et al., 1997) is used in this study to solve these two governing equations simultaneously. 3. General description of the study area The study area is a coastal aquifer (Harikrishna, 2005) located downstream of the Penna river delta (Latitude: 14 350 2400 e 140 480 3600 , Longitude: 79 570 0000 e 80 100 1200 ) of Nellore district in Andhra Pradesh in India (Fig. 1). The coastal aquifer is already affected by saltwater intrusion problem. The study area consists of two Mandals: Allur (15 villages, 197 km2) and Vidavalur (10 villages, 158 km2) with population of 52,990 and 46,793 respectively. (Chief Planning Office, 2004). Pumped water from the underlying aquifers is extensively utilizing for agricultural, domestic, and aquaculture use. Water levels in the villages close to coastal line are already below the mean sea level. This is mainly due to huge requirement of water for aquaculture which is spreading in alarming rate in this area, and for irrigation of paddy the major crop of the area which requires huge amount of water. Adding to this huge requirement of water, there is considerable amount of rainfall deficit (actual rainfall much below the average annual rainfall for Allur: 1133 mm/yr, and Vidavalur: 1141 mm/yr). Annual rainfall is the only source for recharging this coastal aquifer. 3.1. Hydrogeology The geographical location of study area is show in Fig. 1. The study area falls under alluvium soil type. These soils comprise of admixtures of sand, silt and clay in various proportions. The quartz pebbles are invariably encountered at different depths in almost all places in alluvial areas. It is generally light brown to pale gray and sandy in nature. The thickness of coastal alluvium is very large as evident from the exploratory wells drilled by Central Ground Water Board (CGWB). Groundwater occurs in almost all the geological formations of the district. However, its potential depends upon the nature of the geological formation, geographical set up, incidence of rainfall, recharge and other hydrogeological characteristics of the aquifer of the area. Hydrogeologically, the study area comes under unconsolidated formations. These are formed by deltaic alluvium due to the Pennar river, wind blown sands and coastal alluvium along the coast and in inter deltaic areas. These alluvium formations consist of admixtures of sand, silt and clay and form a multilayered aquifer.
150
B. Datta et al. / Journal of Hydro-environment Research 3 (2009) 148e159
Fig. 1. Geographical location of study area.
4. Model development
4.3. Wells and pumping rates
One of the main difficulties in implementing a simulation model is the lack of reliable hydrogeologic and water use data. Data from various sources are utilized for developing and implementing the 3D finite element simulation model utilizing FEMWATER (Harikrishna, 2005).
The study area consists of various types of wells for withdrawing groundwater for various uses. Generally the area covers dug wells, bore wells, hand pumps and filter points. Filter points are dominant in this area. Depth of filter point wells vary from 6 to 30 m. However in most cases depth vary between 6 and 11 m. The only information available is the drafts for the year 2000. The pumping rates are estimated based on the village-wise draft for the time period between July 2000 and July 2001. Available data show that the variation in draft is significant in two Mandals. Allur is having maximum and minimum draft of 100.4 104 m3 and 4.4 104 m3 (for two different villages), whereas the maximum and minimum values for Vidavalur (Mandal) are 652.4 104 m3 and 35 104 m3 respectively. One of the most difficult information to obtain is the withdrawal from each known pumping well. The estimates may be unreliable, and may omit some withdrawals. In this study, the groundwater draft for the year 2000 has been increased by 10% to reflect this. This estimate is used as the pumping rate for the year 2000. As the draft values for the subsequent years are not available, pumping rates for subsequent years are increased by 25% of the previous pumping rates, to reflect the increasing trends of total pumping in this study area. The recharge rate in the form of vertical infiltration is taken as 15% of the normal annual rainfall. The locations of the wells and boundaries of the selected study area are shown in Fig. 2.
4.1. Hydraulic head data In all, 18 piezometers (Observation wells) provide the hydraulic head information for the entire Nellore district. Two of these wells fall in the study area. Hydraulic heads with respect to MSL in these 18 piezometers are used to interpolate the heads for each of the years between 2000 and 2004. Head contours were obtained from these interpolated head values. Initial head values are specified for each node of the 3D finite element model using scattered data points. In the present study, July 2000 is taken as beginning of the first time period. Later, the head contours from the year 2001 and 2002 are used for calibrating the model, and 2003 and 2004 contours are used for validation of the model, in terms of simulated hydraulic heads.
4.2. Chloride concentration data Chloride concentration data in the same locality is very abrupt; such outliers are not considered due to lack of reliability. Year 2000 chloride concentration data is taken as the initial concentration for the simulation model. Concentration data for the year 2001 is used for calibration and year 2002 and 2003 concentration data are used for validation of the model.
4.4. Stratigraphic data The thicknesses and type of the soil layers are specified based on the lithological data of the near by areas. This study
B. Datta et al. / Journal of Hydro-environment Research 3 (2009) 148e159
151
head (Dirichlet boundary) and constant concentration boundary, in which both the head and chloride concentration are constant over time. For this boundary, the specified head is taken as zero (MSL), and specified concentration as 2400 mg/l. The southern boundary is taken as no flow boundary (no change in hydraulic head across the boundary) since the river is almost dry for the past 5 years. The Mandal boundaries are considered as variable flux boundaries (zero Neumann condition). 4.6. Fluid properties The fluid properties needed to be specified are density of water (r0 ¼ 1000 kg/m3), dynamic viscosity of water (m0 ¼ 280985.76 kg/m yr), compressibility of water (b ¼ 4.71e-026 m yr2/kg) and dimensionless density reference ratio (3 ¼ 0.025). 4.7. Hydrogeological parameters
Fig. 2. Location of pumping wells.
area generally falls under alluvium type of soils which are generally admixtures of sand, silt and clay. In the simulation model that was used to simulate the coastal aquifer processes, three layers are considered. First layer below the ground surface is sand, of about 12 m thickness, followed by sandy clay of about 3 m thickness, followed by sand of 15 m thickness. The entire multilayered system is considered as heterogeneous and anisotropic. Stratification details are shown in Fig. 3. 4.5. Boundary conditions The study area is bounded by Bay of Bengal in the east side and by the Penna River in the south side. Administrative (Mandal) boundaries are considered in north and west sides as shown in the Fig. 2. The coastal boundary is defined as constant
Hydrogeological parameters like Hydraulic conductivity, compressibility of the medium are necessary for calibration of the simulation model. For the unsaturated zone, moisture content, relative conductivity and water capacity curves are generated by the model for specified type of material present in each layer. Data adequacy is a major problem in simulating real world systems. Therefore it is necessary to make certain assumptions while constructing the model. Seasonal fluctuations are not considered in this study. Also, tidal effect is not taken into account. The pumping and recharge rates are averaged over the year. The total value is the total pumping or recharge occurring over the year. 5. Model calibration The calibration of the model for the study area is done in two steps. Initially the simulation model is calibrated with respect to the hydraulic heads only, using the flow model. Once, the calibration with respect to the flow model is found to be satisfactory, the calibration of the model is carried out for both the flow and transport processes. Therefore, the first step of calibration is only for approximation of the hydraulic
Fig. 3. Geological stratification of the study area.
152
B. Datta et al. / Journal of Hydro-environment Research 3 (2009) 148e159
parameters and boundary conditions. Using the available data as input, the numerical model is first used for simulating the flow conditions. Subsequently, after approximate calibration of the model for hydraulic parameters, the flow and transport model is used for simulating coupled flow and transport condition. The calibration process is carried out for the time period between July 2000 and July 2002. The simulated and observed head and concentration values area compared for July 2001 and July 2002. The model is then validated for July 2003 and 2004 data, mainly in terms of head values as concentrations are even less reliable for comparison. The calibration model is then utilized for predicting the future scenarios for various pumping patterns. The calibration of flow and coupled flow and transport models are discussed below. 5.1. Simulation model The three-dimensional finite element based flow and transport simulation model FEMWATER is used to simulate the saltwater intrusion process in the coastal aquifer. The total number of nodes and elements of the model are 37,080 and 60,165 respectively. The 3-D model consists of three layers. The different layers of the model are shown in Fig. 3. Prism or wedge elements were used. The calibration is carried out for 2 years, starting from July 2000 with a constant time step of 0.04 yr. (14 days). Recharge is considered as 0.1 m/yr. Longitudinal and transverse dispersivities are taken as 50 m and 15 m respectively. Porosity is assumed to be 0.3. The simulation concentrations were found to be little sensitive to the dispersivities. Poroisity did effect the resulting head distribution. However, porosity value of 0.3 was considered as representative value for the study area based on soil information. The simulation results are certainly affected by specified boundary conditions, which was a major source of uncertainty. With the above input data, the transient flow condition is simulated for the two year period between July 2000 & July 2002. The simulation results are the spatial head distributions
at each time step (0.04 yr interval). The transient flow convergence criterion specified is 0.01. 5.2. Calibration and validation of flow and transport simulation model The three-dimensional finite element model simulates the spatial and temporal values of the hydraulic head. To calibrate the model with respect to heads, the simulated heads are compared with the observed heads at the end of 1st and 2nd years (July 2001 and July 2002). These observation locations are chosen from the set of available observation locations, distributed over the entire study area. 32 observation points are thus specified in the whole study area. For each of these observation points, calibration targets are assigned by considering an acceptable measure of deviation of simulated heads from the observed heads. The observed head are based on interpolation of the spatial head observations at available locations. In Fig. 4, the error bar quantifies the error between the observed head and simulated head. In this study, an interval of 0.5 m is specified as acceptable. This means observed head þ/0.5 m is acceptable in the calibration stage. The simulated hydraulic heads for different specified parameter values (hydraulic conductivity, recharge rate) are compared with the observed hydraulic heads as estimated by interpolation of observation data. The interpolation was necessary as the number of observation points were few in the study area. The spatial interpolation is based on available observation data outside the study area as well. The estimated pumping rates remain unchanged in calibration process as it is based on the observed draft. The hydraulic conductivity and recharge rates are adjusted to obtain simulated water levels which satisfactorily match with observed heads. In addition to high pumping, rainfall was scanty for the four years (2001e2004), and much below the annual average. This had further accelerated the groundwater salinity problem. The
Fig. 4. Comparison between observed heads and simulated heads for trial set 1.
B. Datta et al. / Journal of Hydro-environment Research 3 (2009) 148e159
153
Fig. 5. Comparison between observed heads and simulated heads for trial set 2.
only source of groundwater recharge is the infiltration of excess rainfall. Figs. 4(a) and (b) show the calibration errors in terms of the hydraulic heads at the observation locations, on July 2001 and July 2002. These calibration results are for simulated heads using, Kxx, Kyy ¼ 3650 m/yr and Kzz ¼ 1825 m/yr for sand, and Kxx, Kyy ¼ 150 m/yr and Kzz ¼ 75 m/yr for sandy clay, and infiltration (recharge) rate of 0.1 m/yr (trial set 1). It can be noted that these errors between observed and simulated heads are substantially exceeding the calibration target error of (þ/0.5 m). The calibration errors are smaller when Kxx, Kyy ¼ 7300 m/ yr and Kzz ¼ 3650 m/yr for sand, and Kxx, Kyy ¼ 180 m/yr and Kzz ¼ 90 m/yr for sandy clay, and infiltration (vertical recharge rate) is 0.2 m/yr as seen in Figs. 5(a) and (b) (trail set 2). However, some of the simulated heads are beyond the specified calibration targets. The value of hydraulic conductivity and the infiltration rate were further changed to Kxx, Kyy ¼ 11,000 m/yr and
Kzz ¼ 5500 m/yr for sand, and Kxx, Kyy ¼ 200 m/yr and Kzz ¼ 100 m/yr for sandy clay, and infiltration rate is 0.1 m/yr. The calibration errors for July 2001 and 2002 are shown in Figs. 6(a) and (b) (trial set 3). It is seen that the calibration errors are more or less within the acceptable error limit. Considering the inherent uncertainty in estimating the observed heads, and the reliability of the later, these calibration results are intuitively acceptable. Minimizing the errors further, may not mean a more accurate simulation. The parameter values obtained by the trial and error process of model calibration may not be unique. The calibration process may also leads to an understanding of the factors determining the system’s response and behavior, that is, the parameters and boundary conditions to which the system is most sensitive. Observing the results of these simulations, the best suited combination is Kxx, Kyy ¼ 11,000 m/yr and Kzz ¼ 5500 m/yr for sand, and Kxx, Kyy ¼ 200 m/yr and Kzz ¼ 100 m/yr for sandy clay, and a recharge rate of 0.1 m/yr. The simulated hydraulic heads and estimated observation heads were also
Fig. 6. Comparison between observed heads and simulated heads for trial set 3.
154
B. Datta et al. / Journal of Hydro-environment Research 3 (2009) 148e159
Fig. 7. Simulated versus observed hydraulic heads. July 2001 (Trial Set 3).
compared over time for all selected locations. A comparison of the observed and simulated hydraulic heads for these parameter values are shown in Fig. 7. These parameter values results in the best fit between observed and simulated heads. It was noted that, for smaller hydraulic conductivity values specified, some of the wells tend to dry up after some time. Therefore, for further validation, these parameter values are chosen. Based on these chosen parameter values, and specified boundary conditions, the saltwater concentration is simulated using the coupled 3-D, transient, density dependent flow and transport simulation model. The salt concentration patterns over space and time, as simulated for July 2001, July 2002 and July 2003, and the observed concentration patterns for the same times are shown in Figs. 8e10. All these concentration values are based on the concentrations at the bottom of the top sand layer, in the 3-D model.
Figures (Figs. 8e10) show that the calibration model performs satisfactorily in simulating the salt concentration in the study area over the time horizon for which observation data are available. These comparisons also show, the hydraulic head simulation errors obtained using the coupled flow and transport model are within the acceptable range, considering the inherent uncertainties and errors in the scanty observation data available. Some numerical experiments were conducted with different porosity and hydraulic conductivity values. As expected, dispersivity was not very important. Hydraulic heads were mostly sensitive to porosity and to certain extent to K values. This also affected the spatitemporal concentration distribution. However, the heads and concentrations were certainly sensitive to boundary conditions and to a lesser extent to pumping values. The pumping rates did influence the head and concentration, with localized effects being more prominent. The question is how to normalize these parameters to compare relative sensitivity. This specific quantification, suitable for presenting a meaningful comparison was not undertaken. Basically meaningful parameter values were utilized based on known values for area, and the calibration process was iterative incorporating the sensitivities of these parameters. 6. Results and discussions The calibrated and partially validated simulation model is used to evaluate the impact of few deliberately imposed pumping or withdrawal scenarios on the saltwater intrusion process in the study area. To observe the response of the system, some extra wells are specified in a hypothetical management scenario and the simulated results are compared with those for the existing conditions. The main aim is to establish if a specified or planned pumping strategy can be used to control the intrusion process.
Fig. 8. Salt Concentration Contours for July 2001 (in mg/l).
B. Datta et al. / Journal of Hydro-environment Research 3 (2009) 148e159
Fig. 9. Salt Concentration Contours for July 2002 (in mg/l).
Fig. 10. Salt concentration contours for July 2003 (in mg/l).
Fig. 11. Location of wells for different scenarios.
155
156
B. Datta et al. / Journal of Hydro-environment Research 3 (2009) 148e159
of July 2009, there is a decrement in salt concentration with extra wells in scenario-I. Fig. 14 shows concentration variation with time for both the pumping patterns at specified locations 1 in Fig. 12. Although only marginally, by inducing additional pumping at selected and carefully chosen locations, it is possible to reduce the salt concentration levels at some other locations. Therefore it is possible to lower salt concentrations by careful planning of pumping patterns. This may improve groundwater condition in areas away from the coast line by locally controlling the saltwater intrusion. 6.2. Scenario-II: additional pumping wells near coast line in Iskapalli village of Allur Mandal
Fig. 12. Concentration contours for existing pumping pattern (in mg/l).
6.1. Scenario-I: additional pumping wells near coast in Utukuru village of Vidavalur Mandal Additional pumping is assumed from five extra wells near the coast in the Utukuru village of Vidavalur Mandal. Fig. 2 and Fig. 11 show the old and new pattern (with extra wells) of the well locations, respectively. The pumping rates are kept at 3,50,0000 m3/yr for all additional wells specified in scenario-I. The pumping in the new wells starts from July 2005. Fig. 12 and Fig. 13 (a) show the simulated concentration contours after five years from July 2004, i.e., in July 2009. It is observed from the concentration contours
Five extra pumping wells are hypothetically specified near the east coast line in Iskapalli village of Allur Mandal. Figs. 2 and 11 show the old and new pattern of the well locations, respectively. The pumping rates in the additional wells are 3,50,0000 m3/yr. The pumping in these new wells starts from the fifth year i.e., in July 2005. Fig. 14 shows concentration variation with time for the existing and new pumping patterns at specified locations 2, as shown in Fig. 12. Actually, the increasing trend in concentration is arrested in locations 2 by inducing these additional pumping. 6.3. Scenario-III: additional pumping wells at Dampuru village of Vidavalur Mandal In scenario-III, five additional pumping wells are induced at Dampuru village of Vidavalur Mandal. Figs. 2 and 11 shows the old and new pattern of the well locations respectively. The pumping rates in the additional wells are 3,50,0000 m3/yr. The pumping in these new wells starts from the fifth year i.e., in July 2005. Figs. 12 and 13(b) show the concentration contours after five years from July 2004, i.e. in July 2009. It
Fig. 13. Concentration contours for scenarios I and III (in mg/l).
B. Datta et al. / Journal of Hydro-environment Research 3 (2009) 148e159
157
Fig. 14. Concentration at different locations for different scenarios.
can be observed from the resulting concentration contours of July 2009, that there is a decrement in concentration, with additional pumping at new wells in some locations, compared to the existing concentration pattern. In both the cases, concentration is decreasing with time, but the decrement rate is more in the new pattern compared to the old one. Fig. 14 shows concentration variations with time in the both pumping patterns, at specified location 3 shown in Fig. 12. Evaluating results also show that very close to the new pumping wells, the salt concentration may increase as well. Similarly, on the coastal side of a series of pumping wells near the sea front, the salt concentration may also increase. 6.4. Scenario-IV: pumping rates increased by 10% over the previous years from July 2004 to July 2009 in all the existing wells
evident near the coastline compared to other parts of the study area. These evaluation results obtained for hypothetical pumping scenarios, selected with some specific consideration for possible control of saltwater intrusion, no doubt establish the fact that it is possible to alter the salt concentration in the aquifer by modifying the pumping patterns. These results, although very limited in scope, do show that spatially varied pumping as a planned strategy may be utilized to control salt concentration patterns and therefore, to manage the saltwater intrusion process. The aim was to only establish the feasibility of using a planned pumping strategy for effective control and management of the saltwater intrusion process. These simulation results are also useful in predicting the salt concentration scenario in the future, if the existing pumping patterns continue in the future. 7. Conclusions
The pumping rates in all the existing wells are increased by 10% over the previous year from July 2004 to July 2009. The pumping rates for July 2004, are as described before. It is observed from the concentration contours of July 2009, there is some marginal increment in salt concentration with increase in pumping rates. Fig. 14 shows concentration variation for both the pumping rates at specified location 4 (Fig. 12). Fig. 14 shows that the concentration values at location 4 marginally increase with increase in pumping rates. This trend is more
A 3D, transient, density dependent, finite element based flow and transport simulation model, FEMWATER, is implemented for simulating the coupled flow and transport processes of saltwater intrusion in a coastal aquifer in Nellore district of Andhra Pradesh, India. Available data for a selected study area of around 355 km2 was obtained from different sources to be used as input data for implementing the numerical simulation model for the study area. Due to the
158
B. Datta et al. / Journal of Hydro-environment Research 3 (2009) 148e159
scanty nature of available data and questionable reliability of all available data, the best but subjective judgment was used in selecting the data for implementing the model. The numerical model was calibrated for two years time period, between July 2000 and July 2002, both in terms of hydraulic heads and salt concentration. The calibrated model was further validated for the chosen hydrogeological parameters with the data available between for 2002 and July 2004. The aquifer was considered heterogeneous in terms of vertical stratification. Both flow and transport are considered transient. Withdrawal from aquifer is estimated based on available data, and assuming an increasing trend over the period of calibration and validation. The calibrated and partially validated simulation model was used to predict the saltwater transport scenario in the study area at future time periods. It is worth mentioning that one can easily perform linked optimization parameter estimation, based on repeated call to the simulation model. In present situation the single run of the model takes huge computational time. Thus, it is not realistic to perform such optimization based calibration. The simulation was carried out till July 2009. The predicted head and concentration values show the future saltwater intrusion patterns if the present trend of pumping continues. The calibrated model was also used to evaluate the effect of modifying the pumping patterns on the future saltwater intrusion process. Effect of increase in withdrawal rates uniformly spread over the study area was also evaluated. These evaluation results show that a carefully planned pumping strategy is capable of modifying the saltwater intrusion process and the spatial and temporal concentration patterns. These results show that a well planned pumping strategy can help in controlling the saltwater intrusion process in space and time. However, due to a limited period of analysis, these impacts are evident only locally. These simulation results also show, if the withdrawal rate continues to increase over time it may have detrimental effect on the salt concentration in the study area. This study for a real life study area does demonstrate the feasibility of using planned pumping strategy to control the saltwater intrusion process. The main aim was to demonstrate that in a realistic study area, hydraulic control measures can influence the spatial and temporal distribution of saline concentration. Theoretically, it should be possible to identify a management strategy for coastal aquifers, using such planned control measures. This study does not try to identify an optimal control strategy. It only tries to establish the plausibility of actually obtaining such a strategy. In order to demonstrate the feasibility of using hydraulic control as a management measure, the study area was selected. It was felt that the sparsity of data available for this study area may represent a typical real life situation. Therefore, an attempt was made to calibrate and validate a numerical simulation model for this study area with very limited data. These evaluation results are based on very limited data and are limited in scope. Actual implementation of planned pumping strategy for economic management of the coastal aquifer will require more rigorous calibration and validation with additional reliable data. Only then the impact of any planned pumping strategy can be accurately evaluated.
Acknowledgement Authors are thankful to different government organizations of the Nellore district: State Groundwater Department (Andhra Pradesh, INDIA), Chief Planning office, Mandal Revenue Offices of both Mandals, and the Water Supply Departments for various types of data. References Cheng, J.M., Chen, C.X., 2001. Three-dimensional modeling of density dependent saltwater intrusion in multilayered coastal aquifers in Jahe River basin, Shandong province, China. Ground Water 39 (1), 137e143. Chief Planning Office 2004 ‘‘Hand book of statistics 2003e04, Nellore District’’, Nellore. Das, A., Datta, B., 1999a. Development of multiobjective management models for coastal aquifers. Journal of Water Resources Planning and Management, ASCE 125 (2), 76e87. Das, A., Datta, B., 1999b. Development of management models for sustainable use of coastal aquifers. Journal of Irrigation and Drainage Engineering, ASCE 125 (3), 112e121. Das, A., Datta, B., 2000. Optimization based solution of density dependent seawater intrusion in coastal aquifers. Journal of Hydrologic Engineering, ASCE 5 (1), 82e89. Finney, B.A., Samsuhadi, S., Willis, R., 1992. Quasi-three-dimensional optimization model of Jakarta basin. Journal of Water Resources Planning and Management, ASCE 118, 18e31. Guvanasen, V., Wade, S.C., Barcelo, M.D., 2000. ‘Simulation of regional groundwater flow and saltwater intrusion in Hernando county, Florida’. Ground Water 38 (5), 772e783. Harikrishna, V., 2005. ‘‘Modeling prediction and control of saltwater intrusion in a coastal aquifer of Andhra Pradesh India’’ M.Tech. dissertation, I.I.T. Kanpur, India. Kentel, E., Gill, H., Aral, M.M., 2005. ‘‘Evaluation of groundwater resources potential of Savannah Georgia region,’’ Multimedia Environmental Simulations Laboratory, Report No. MESL-01-05. Lin, H.J., Rechards, D.R., Talbot, C.A., Yeh, G.T., Cheng, J.R., Cheng, H.P., Jones, N.L., 1997. A Three e Dimensional Finite Element Computer Model for Simulating Density -Dependent Flow and Transport in Variable Saturated Media. Version 3.1. U.S Army Engg. Research and Development Center, Vicksburg, M.S. Nobi, N., Gupta, A.D., 1997. Simulation of regional flow and salinity intrusion in an integrated stream aquifer system in a coastal region: southwest region of Bangladesh. Groundwater 35 (5), 786e796. Rouve, G., and Stoessinger, W. (1980), Simulation of the Transient Position of the Saltwater Intrusion in the Coastal Aquifer near Madras Coast, Proc. 3rd Internat. Conf. Finite Elements in Water Resources, Univ. of Miss., Oxford, Miss. Sherif, M.M., Singh, V.P., Amer, A.M., 1988. ‘ A two e dimensional finite element model for dispersion (2D-FED) in coastal aquifer’. Journal of Hydrology 103, 11e36. UNCED, 1992. Agenda 21: Chapter 17.3. http://www.un.org/esa/sustdev/ documents/agenda21/english/agenda21chapter17.htm [August, 2007]. Willis, R., Finney, B.A., 1988. Planning model for control of saltwater intrusion. Journal of Water Resources Planning and Management, ASCE 114 (3), 333e347.
Notations The following symbols are used in this paper: c: material concentration in aqueous phase; cin: material concentration in the source; cmax: maximum material concentration; D: dispersion tensor; F: storage coefficient; h: pressure head;
B. Datta et al. / Journal of Hydro-environment Research 3 (2009) 148e159 K: hydraulic conductivity tensor; Kxx: hydraulic conductivity in x direction; Kyy: hydraulic conductivity in y direction; Kzz: hydraulic conductivity in z direction; q: volumetric flow rate of source and/or sink; t: time; V: discharge; z: potential head; a0 : compressibility of the medium;
b: compressibility of water; 3: dimensionless density reference ratio; q: moisture concentration; m0: dynamic viscosity of water; r: water density at chemical concentration c; r0: referenced water density at zero chemical concentration; r*: density of either the injection fluid or the withdrawn water; V: del operator; (.): vector dot product.
159