An Introduction to the Effect of Heterogeneities on the Characterization and Remediation of Alluvial Geosystems

Heterogeneous granular sediments—sands and gravels —form the most important water-supply aquifers in many areas of North America. In the past 60 years many such aquifers have become contaminated with nonaqueous phase liquids (NAPLs) and the dissolved phases associated with them. These include glacial outwash deposits in New England, Ohio, and Ontario, the floodplain alluvium along the Columbia and Fraser rivers, the buried channel aquifers beneath the U.S. Midwest and the Canadian prairies, and the huge coalescing alluvial-fan aquifers of the U.S. Southwest basin-andrange province. These heterogeneous granular sediments are referred to in this article as ‘alluvial geosystems’ when contaminated by NAPLs. The term geosystem comprises that system composed of alluvial aquifer materials, contaminated groundwater, NAPLs, and any capillary barrier that traps or redirects migrating NAPLs. This term might be compared with the oilfield term reservoir but reflects the smaller scale, absence of lithification, and shallower depth of these aquifers, the irregular distribution of the NAPLs, as well as its much lower average NAPL saturations than are displayed by crude oil in petroleum reservoirs. Furthermore, the immiscible fluids in these pages of this journal refer to not just lighter-than-water, refined fuel hydrocarbons (LNAPLs) but also to denser-than-water, chlorinated solvents, coal tar, and creosote (DNAPLs). Hydrogeologists working before World War II, such as Meinzer (1923) and Tolman (1937), sought to quantify the relationship between textural heterogeneities and groundwater flow mainly by reference to the experimental studies of Allen Hazen in the 1890s. Hazen had shown the importance of the ‘effective diameter’ of grain size on hydraulic conductivity, although it fell short when applied to nonuniform mixtures of sands (Tolman, 1937, p. 202). Krumbein and Monk (1943) and Masch and Denny (1966) later addressed this relationship between nonuniformity in grain size (i.e., the variation in particle diameter) and permeability and hydraulic conductivity. Once hydraulic testing of aquifers had become commonplace, Tolman (1937, p. 214) advised hydrogeologists that ‘‘the heterogeneity of alluvial deposits may render this method of permeability determination [the aquifer pumping test] more reliable than the laboratory method [the permeameter test] which is limited to one type of material in each test and results are then applied to assumed field conditions.’’ However, it was the use of hydrogeological tracers— in particular, radionuclides and dyes—that showed hydrogeologists how important heterogeneous sedimentary units would be at the field scale in the transport of contaminants through groundwater flow systems. It was the U.S. Geological Survey (USGS) that laid the groundwork for our present understanding of the heterogeneity of alluvial systems by undertaking research on behalf of the U.S. Atomic Energy Commission, now the U.S. Department of Energy. Looking back at research published in the 1950s on bomb-tritium and radioactivewaste migration, the USGS’ Theis (1963) pointed out the very large difference in the measured values of tracer dispersion in the field when compared with the much smaller values measured in laboratory soil columns. He concluded, ‘‘the reason for the increased longitudinal dispersion must lie in the wide distribution of permeabilities in any suite of sedimentary beds.’’ Theis’ interest in this matter appears to have led to the series of fundamental studies of the dispersion process by the USGS (e.g., Ogata and Banks, 1961; Simpson, 1962; Skibitzke and Robinson, 1963; and Ogata, 1970). These studies in turn led to further studies that sought to establish a clear link between aquifer heterogeneity and field-scale dispersion as measured by the length-scale parameter, the dispersivity (see Pickens and Grisak, 1981; Gelhar et al., 1985; and Neuman, 1990). However, interconnecting high-permeability units that preferentially channel contamination make ‘‘it impossible to use the standard advection-dispersion equation with effective dispersivity values’’ (Anderson, 1990). More recent attempts to address aquifer heterogeneity using stochastic processes cannot be said to have had a profound effect on the present practice of contaminant hydrogeology, a result most likely attributed to the high cost of acquiring the necessary data on the spatial