There are few studies analyzing tracer experiments conducted in rock fractures at large scales (>50 m) and fewer still which have succeeded in simulating experimental data without the use of parameters which increase/decrease with increasing scale of observation (i.e. scaling or effective parameters). To explore the use of effective parameters at large scales, a numerical model is used to examine the results of a divergent tracer experiment conducted in sparsely fractured rock where five observation points ranged over distances from 30 to 125 m. The experiment was conducted in what was first interpreted as a single horizontal fracture, traceable using geological and hydraulic evidence over that range in distance. Initial simulations were conducted using measured values of porosity, fracture aperture, and regional hydraulic gradient but were not successful in matching the field data at all measured locations. Multiple fits of equal quality at each observation point however were obtained with models using effective parameters (such as increasing porosity with scale) to represent heterogeneity. Additional fits were achieved by physically representing the same heterogeneity with different conceptual models, such as added horizontal and vertical fractures. This illustrates the non-uniqueness that arises in the interpretation of a divergent tracer experiment with multiple observation well distances. Using the models, predictions at distances beyond the measured domain were then generated. These showed that the choice of conceptual model significantly impacted simulated arrival times and concentrations at distances of only 75-175 m larger than the experimental scale. In particular, predictions made with effective parameters provided estimates that were less conservative (i.e. under predicting concentrations) than those made by directly adding heterogeneity to the numerical model.Copyright © 2021 Elsevier B.V. All rights reserved.
