Introduction We address the problem of characterizing heterogene- ity in regional aquifer systems using databases derived from borehole logs in the public domain (predominantly water well logs of varying quality). We also address the problem of representing that heterogeneity within models for ground water flow and transport. We focus on sedimen- tary aquifer systems in which sedimentary unit types have different permeability modes (Ritzi et al. 2003; Gaud et al. 2004). We bring a number of methods involving indicator geostatistics to bear on these problems (Johnson 1995; Ritzi et al. 1995; Carle and Fogg 1996, 1997; Deutsch and Jour- nel 1998; Carle 1998; Ritzi et al. 2000; Ritzi 2000). We apply the geostatistical methods in ways that reduce bias in statistics derived from borehole log data. We consider sed- imentary units defined at two different scales within a hier- archical framework. Our focus is on characterizing and modeling the three-dimensional interconnection of higher- permeability units. The three-dimensional architecture of sedimentary unit types can strongly influence mass transport (Anderson 1990; Desbarats 1990; Ritzi et al. 2000; Weissmann et al. 2002). Residence-time distributions can differ significantly from unimodal, normal distributions. Shorter residence time is influenced by transport in interconnected unit types of higher permeability. Longer residence time is influ- enced, in part, by transport through lower-permeability unit types. In some cases, the relevant aspects of transport have been shown to be sensitive to the architecture of units defined at one scale, but insensitive to the heterogeneity at smaller scales within those units (Desbarats 1990). How- ever, this may not always be true. In some cases, the archi- tecture at multiple scales may be important. Methods in indicator geostatistics are particularly well suited to characterizing and modeling sedimentary archi- tecture. In briefly reviewing them, we can divide them into two groups. In one group are the methods that are used to Abstract A number of methods involving indicator geostatistics were combined in a methodology for characterizing and modeling multiscale heterogeneity. The methodology circumvents sources of bias common in data from borehole logs. We applied this methodology to the complex heterogeneity within a regional system of buried valley aquifers, which occurs in the western glaciated plains of North America and includes the Spiritwood Aquifer. The region is conceptualized as having a hierarchical organization with three facies assemblage types (large-scale heterogeneity) and two facies types within each assemblage (small-scale heterogeneity). We statistically characterized the sedimen- tary architecture at both scales, formulated indicator correlation models from those characterizations, and used the models to simulate the architecture in a multiscale realization. We focused on the interconnectivity of units creating higher-permeability pathways. Higher-permeability pathways span the realization even though the proportion of higher-permeability facies is less than the percolation threshold. Thus, geologic structures as represented in the indi- cator correlation models create interconnectivity above that which would occur if the higher-permeability facies were randomly placed. This amount of interconnection among higher-permeability facies within the multiscale realization is consistent with that suggested in prior hydraulic and geochemical studies of the regional system. 658 M odeling M ultiscale Heterogeneity and Aquifer Interconnectivity by Christopher J. Proce 1 , Robert W. Ritzi 1,2 , David F. Dominic 1 , and Zhenxue Dai 1 1 Department of Geological Sciences, Wright State University, Dayton, OH 45435 2 Corresponding author: (937) 775–2460; rritzi@wright.edu Received July 2002, accepted September 2003. Copyright © 2004 by the National Ground Water Association. Vol. 42, No. 5—GROUND WATER—September–October 2004 (pages 658–670)