Three-Dimensional Modelling of Heterogeneous Coastal Aquifer: Upscaling from Local Scale

The aquifer heterogeneity is often simplified while conceptualizing numerical model due to lack of field data. Conducting field measurements to estimate all the parameters at the aquifer scale may not be feasible. Therefore, it is essential to determine the most significant parameters which require field characterization. For this purpose, the sensitivity analysis is performed on aquifer parameters, viz., anisotropic hydraulic conductivity, effective porosity and longitudinal dispersivity. The results of the sensitivity index and root mean square deviation indicated, that the longitudinal dispersivity and anisotropic hydraulic conductivity are the sensitive aquifer parameters to evaluate seawater intrusion in the study area. The sensitive parameters are further characterized at discrete points or at local scale by using regression analysis. The longitudinal dispersivity is estimated at discrete well points based on Xu and Eckstein regression formula. The anisotropic hydraulic conductivity is estimated based on established regression relationship between hydraulic conductivity and electrical resistivity with R2 of 0.924. The estimated hydraulic conductivity in x and y-direction are upscaled by considering the heterogeneous medium as statistically homogeneous at each layer. The upscaled model output is compared with the transversely isotropic model output. The bias error and root mean square error indicated that the upscaled model performed better than the transversely isotropic model. Thus, this investigation demonstrates the necessity of considering spatial heterogeneous parameters for effective modelling of the seawater intrusion in a layered coastal aquifer.

Modeling Spatial Distribution of 2D Anisotropic Hydraulic Conductivity Tensor of Fractured Groundwater Flow Media using Neural Network Case Study Grasberg Open Pit of PTFI and Surrounding

Highly fractured rocks in Grasberg open pit and surrounding of PT Freeport Indonesia (PTFI) result in fractured groundwater flow media. It is due to the complex geological structure and lithological condition. Accordingly, it leads to anisotropic distribution of hydraulic conductivity. The paper will be devoted tothe modeling of two dimensional (2D) spatial distribution of hydraulic conductivity using neural network. Surface fracture mapping database will be used to estimate 2D equivalent anisotropic hydraulic conductivity tensor based on the Oda et al (1996) approach. Modeled anisotropic hydraulic conductivity is then checked at some points where the slug tests for isotropic conductivity are observed. Co-relation, validation and training between modeled and observed hydraulic conductivity is then carried out using transformation of vector anisotropic hydraulic conductivity into the scalar isotropic hydraulic conductivity. Following training step, neural network will then generate two dimensional model of anisotropic hydraulic conductivity distribution. It is beneficial for modeling of shallow anisotropic flow of groundwater distribution