SiteTitle | • | High-throughput computing for the analysis of tracer tests in fractured aquifers. | [X] |
Organization | • | Illinois State Water Survey | [X] |
| | 1: | | Title: | | | | | Volume/Number: | 2006 | | | | Issuing Agency: | | | | | Description: | Traditional approaches to characterization and modeling of fractured dolomite aquifers face many conceptual and technical challenges. One alternative strategy begins with the Generalized Radial Flow interpretation of hydraulic tests, which infers an additional parameter, the flow dimension, to describe the geometry of groundwater flow. This study examines the behavior and variability of the apparent flow dimension, n*, and advective transport for four stochastic models of heterogeneous transmissivity, T(x). This is accomplished through Monte Carlo analysis of numerical models simulating aquifer tests and converging flow tracer tests (CFTTs) in two-dimensional systems. For ln T(x) distributed as a multivariate Gaussian (mvG) variable of variance less than one, the apparent flow dimension of an aquifer test converges to n* = 2 if the scale of the test is large relative to the scale of correlation. The variability of the apparent flow dimension depends on the variance and integral scale of the transmissivity, suggesting that it may be possible to identify the variance and integral scale from a set of aquifer tests. For variances greater than one, the results suggest that the average of the apparent flow dimension is less than two initially, then converges to n* = 2, similar in some respects to a percolation network. The simulation of an uncorrelated log-Gaussian model suggests that the flow dimension of an aquifer test converges to n* = 2 even for large variances. For ln T(x) distributed as fractional Brownian motion (fBm), the apparent flow dimension averages to n* = 2 and its variability increases with time. An approximation of a percolation network model showed an average apparent flow dimension stabilizing between n* = 1.4 to 1.6, followed by an increasing trend. These characteristics apparently are functions of the transmissivity contrast between the percolating and nonpercolating fractions. In the low-variance mvG, uncorrelated log-Gaussian, and fBm models, CFTTs influenced by matrix diffusion showed late-time breakthrough curves (BTCs) with log-log slopes of -3/2, the characteristic behavior of matrix diffusion. In the percolation network model, a simulated CFTT influenced by matrix diffusion had late-time BTC with log-log slopes of -5/4, attributed to slow advection through low transmissivity regions. This indicates that some heterogeneity models can systematically affect the late-time behavior of a BTC for a CFTT. These results suggest that the flow dimension may be a useful diagnostic for selecting models of heterogeneity, and that flow dimensions n ? 2 may be associated with unique tracer behavior. Additional research is advocated to infer the general behavior of the flow dimension at various field sites, to assess a broader range of parameters, to examine other stochastic models, and to conduct a more detailed examination of transport behavior versus the flow dimension. | | | | Date Created: | 4 13 2006 | | | | Agency ID: | CR-2006-04 | | | | ISL ID: | 000000000958 Original UID: 999999994479 FIRST WORD: High | |
|
RSS