| dc.description.abstract | Seawater Intrusion (SWI) into the fresh groundwater aquifers is one of the major hydrological problems in coastal regions. This global issue is aggravated by increasing demands for fresh groundwater due to on-going urbanization, population growth, and economic development in these regions. Freshwater and seawater mixing is by mechanical dispersion and hydrodynamic diffusion, which results in the formation of the transition zone. The position and width of the transition zone change with hydrogeological, hydrological circumstances, and anthropogenic activities such as excessive groundwater extraction, climate variations/and sea-level fluctuations, etc. The groundwater models provide a scientific and predictive tool for determining the extent of SWI at field-scale. But, the development of numerical models to determine the groundwater flow and SWI in the heterogeneous aquifers requires several hydrogeological field data. Due to incomplete knowledge of geological structures such as heterogeneity, anisotropy, and layering, the development of conceptual numerical models is often simplified. This indicates that there is a wide gap in understanding the hydrogeology characterization in three-dimensional (3D) modeling of the groundwater flow and SWI. With this motivation, the work presented in this thesis is aimed at estimating the anisotropic heterogeneous parameters in a layered coastal aquifer for the simulation of hydraulic head and SWI.
The coastal plain in between River Swarna and River Netravati of Karnataka in India is selected as the study area. As most of the West flowing rivers of Karnataka are seasonal and tidal, due to this saline water intrudes into rivers during the non-monsoon period up to several kilometers contaminating the adjacent fresh groundwater aquifers. And due to the permeable geological formation, thousand million cubic feet of recharged rainwater is designated as subsurface groundwater discharge, thus, making it an interesting and challenging case study to pursue. The study area between the rivers extends over 1191 km2 (large-scale) with the Arabian Sea in the West. Applying the different parameter estimation approaches to this large-scale area requires an efficient
iv
machine and adequate field measurements. Due to insufficient field data and to reduce the computational cost, the parameter estimation approaches are applied on the small-scale coastal aquifer of 8 km2 areal extent, which lies within the large-scale area.
The work reported in this thesis contributes towards developing a 3D, variable-density conceptual model integrated with the Geographic Information System (GIS). The developed basic conceptual model is constrained with real-field data such as layering, aquifer bottom topography, and appropriate initial conditions. The initial aquifer parameters are layered heterogeneous but spatially homogeneous. The developed basic conceptual model showed poor correlation with observed state variables (hydraulic head and solute concentration), signifying the importance of spatial heterogeneity of aquifer parameters in all the layers. A sensitivity analysis is carried out to determine the most critical aquifer parameters that affect SWI modeling. From the analysis results, it is inferred that the anisotropic heterogeneous hydraulic conductivity and longitudinal dispersivity are the most significant parameters to be estimated spatially. This investigation demonstrates the necessity of considering anisotropy and heterogeneity for effective modeling of the SWI in a layered coastal aquifer.
An inverse modeling approach is used to estimate heterogeneous aquifer parameters. The solution for the inverse SWI problem has not been studied as extensively as forward modeling. Inverse modeling aims to estimate layerwise anisotropic heterogeneous hydraulic conductivity and longitudinal dispersivity. The inverse code minimizes the least square error of state variables and errors induced by regularization to estimate heterogeneous parameters at pilot points. It is found that the inverse model output for longitudinal dispersivity is mostly the same as that of the empirical equation given by Xu and Eckstein (1995). Thus, for further investigation, Xu and Eckstein equation (1995) is used to obtain heterogeneous longitudinal dispersivity. The calibrated inverse model showed substantial improvement in the simulated state variables compared to the basic conceptual model output. The validation for the simulated hydraulic head is performed at discrete validation wells and a novel methodology is developed to validate simulated solute concentration for Electrical Resistivity Tomography (ERT) data, which shows good results against field
v
measurements. The estimated heterogeneous hydraulic conductivity (K) is used to verify the layering and geological formation; by verifying the cross-sectional profiles with ERT profiles. To conclude, the results obtained from the calibrated inverse model are better, but the approach is computationally demanding, which motivated to estimate anisotropic heterogeneous K by a geostatistical approach.
Estimation of aquifer parameters in coastal aquifers by conducting field experiments is a challenging task due to its complex hydrochemical conditions. Several studies have addressed 1D or 2D parameters estimation in coastal aquifers, based on geophysical methods but not many studies have addressed 3D parameter estimation with clay lenses being present. A methodology is proposed to estimate K and porosity in an aquifer containing clayey formation. The regression equations between K and bulk resistivity is established using resistivity and fluid conductivity data in petrophysical equations. The established equations are validated for the small-scale area as wells as for the extended area in between River Gurupura and River Pavanje which showed R2 of 0.83, indicating good reliability. The established equations enable the estimation of anisotropic K at 2D resistivity profiles. The horizontal K i.e., in both x and y-direction, varies from 0.37–50 m/d, which agrees with the inverse model results. At each K profile, equivalent vertical K and anisotropy ratio are also estimated. This significantly enriches the 3D K database for numerical modeling, which the earlier studies have failed to address. The geophysical data is not only used to estimate anisotropic K but also to characterize the aquifer.
The locally estimated K data at ERT profiles are used for upscaling to aquifer-scale intrinsically, where boundary conditions and size of the domain are not considered. The intrinsic upscaling is used to estimate heterogeneous K field at an aquifer-scale where the field (resistivity) data is lacking. Earlier studies focused on generating the K fields by neglecting the anisotropy but here based on the major principal axes, the layerwise anisotropic heterogeneous K random fields are generated. The estimated K in x and y-direction at ERT profiles are upscaled by considering Gaussian covariance function with known layerwise mean, variance, and correlation length. The upscaled K is used as input to the numerical model for transient
vi
simulation of the state variables. The upscaled model output is compared with the calibrated inverse model and basic conceptual model output. The performance of the upscaled model is better than the basic conceptual model but showed a relatively high error when compared with the calibrated inverse model output and with observed data. The computational time and computational effort taken by intrinsic upscaling are less than that of the inverse modeling. Thus, this method which also accounts for parameter randomness is applied to estimate anisotropic heterogeneous K for the large-scale aquifer.
The 3D large-scale conceptual model is constructed with real field data about layering, aquifer bottom topography along with appropriate initial and boundary conditions. The intrinsic upscaling method is used to generate multiple realizations spatially correlated K random fields at each layer in x and y-direction. And the other flow and solute transport parameters are considered as spatially homogeneous and values are taken from the literature. Spatio-temporal varying recharge rates are estimated from the water-table fluctuation method and pumping rates at wells are taken from the literature. The transient simulation of a 3D variable-density large-scale conceptual model is carried out for 1647 days. The model results showed a high error to observed state variables. This may be due to unknown locations of groundwater draft, neglecting tidal influence, land-use land cover change, evapotranspiration, etc. The performance can be improved by constraining the model with the above-mentioned inputs. This study is a building block towards large-scale 3D modeling in a coastal anisotropic heterogeneous aquifer. The improved conceptual model can be further used for the predictive analysis of groundwater flow and SWI. | en_US |
