SIMULATION OF CONTRIBUTING AREAS IN A FRACTURED CRYSTALLINE AQUIFER

Simulation of all the detail that is revealed by aquifer-test data often is not feasible in contributing area models. In aquifers with continuous, bed-oriented preferred pathways, it is feasible to explicitly simulate aquifer heterogeneity. In aquifers where ground-water flow may not be associated with continuous preferred pathways, such as in fractured crystalline rock, aquifer heterogeneity may have to be generalized. An aquifer test in Old Lyme, Conn., indicates that the aquifer is highly heterogeneous. Although fractures in outcrops and in boreholes have a well-defined dominant orientation, individual fractures are continuous only over a scale of 100s of feet. In a contributing area model, constructed using 1,000-by-1,000-ft grid cells, the observed heterogeneity is lost. The model would be more realistic if the known heterogeneity could be generalized in some way. In order to test whether anisotropy might be detectable in water-level data, a model was used to generate two sets of "perfect" water-level data, with and without anisotropy, for use in a parameter-estimation model. These data were perturbed using normally distributed random numbers. In both cases, a parameter-estimation model correctly returned the degree of anisotropy, meaning that with the correct model structure, it might be possible to detect anisotropy using a ground-water model. To apply the above finding in a more realistic setting, ground-water flow was simulated using parameter estimation in a 24-mi² area near Old Lyme, Conn. Three alternative models were posed-a homogeneous and isotropic aquifer (the basic model), the basic model with horizontal anisotropy, and the basic model with horizontal heterogeneity. Horizontal heterogeneity was simulated by creating two zones for hydraulic conductivity-one zone where land surface is above 200 ft altitude and one zone where it is below. The calibration data consisted of water levels, measured by drillers, in 822 wells. Parameter estimates for each model are realistic, and the residuals of the regression appear unbiased. Particle tracking shows that the basic and heterogeneous models produce similar contributing areas. The anisotropic model produces significantly different contributing areas, particularly where the ground-water flow direction is coincident with the anisotropy.