Stephen P. Garabedian

Large-scale natural gradient tracer test in sand and gravel, Cape Cod, Massachusetts: 1. Experimental design and observed tracer movement

A large-scale natural gradient tracer experiment was conducted on Cape Cod, Massachusetts, to examine the transport and dispersion of solutes in a sand and gravel aquifer. The nonreactive tracer, bromide, and the reactive tracers, lithium and molybdate, were injected as a pulse in July 1985 and monitored in three dimensions as they moved as far as 280 m down-gradient through an array of multilevel samplers. The bromide cloud moved horizontally at a rate of 0.42 m per day. It also moved downward about 4 m because of density-induced sinking early in the test and accretion of areal recharge from precipitation. After 200 m of transport, the bromide cloud had spread more than 80 m in the direction of flow, but was only 14 m wide and 4-6 m thick. The lithium and molybdate clouds followed the same path as the bromide cloud, but their rates of movement were retarded about 50% relative to bromide movement because of sorption onto the sediments.

Large-scale natural gradient tracer test in sand and gravel, Cape Cod, Massachusetts: 2. Analysis of spatial moments for a nonreactive tracer

A large-scale natural gradient tracer test was conducted to examine the transport of reactive and nonreactive tracers in a sand and gravel aquifer on Cape Cod, Massachusetts. As part of this test the transport of bromide, a nonreactive tracer, was monitored for about 280 m and quantified using spatial moments. The calculated mass of bromide for each sampling date varied between 85% and 105% of the injected mass using an estimated porosity of 0.39, and the center of mass moved at a nearly constant horizontal velocity of 0.42 m per day. A nonlinear change in the bromide longitudinal variance was observed during the first 26 m of travel distance, but afterward the variance followed a linear trend, indicating the longitudinal dispersivity had reached a constant value of 0.96 m. The transverse dispersivities were much smaller; transverse horizontal dispersivity was 1.8 cm, and transverse vertical dispersivity was about 1.5 mm.

Sampling design for groundwater solute transport: Tests of methods and analysis of Cape Cod tracer test data

Tests of a one-dimensional sampling design methodology on measurements of bromide concentration collected during the natural gradient tracer test conducted by the U.S. Geological Survey on Cape Cod, Massachusetts, demonstrate its efficacy for field studies of solute transport in groundwater and the utility of one-dimensional analysis. The methodology was applied to design of sparse two-dimensional networks of fully screened wells typical of those often used in engineering practice. In one-dimensional analysis, designs consist of the downstream distances to rows of wells oriented perpendicular to the groundwater flow direction and the timing of sampling to be carried out on each row. The power of a sampling design is measured by its effectiveness in simultaneously meeting objectives of model discrimination, parameter estimation, and cost minimization. One-dimensional models of solute transport, differing in processes affecting the solute and assumptions about the structure of the flow field, were considered for description of tracer cloud migration. When fitting each model using nonlinear regression, additive and multiplicative error forms were allowed for the residuals which consist of both random and model errors. The one-dimensional single-layer model of a nonreactive solute with multiplicative error was judged to be the best of those tested. Results show the efficacy of the methodology in designing sparse but powerful sampling networks. Designs that sample five rows of wells at five or fewer times in any given row performed as well for model discrimination as the full set of samples taken up to eight times in a given row from as many as 89 rows. Also, designs for parameter estimation judged to be good by the methodology were as effective in reducing the variance of parameter estimates as arbitrary designs with many more samples. Results further showed that estimates of velocity and longitudinal dispersivity in one-dimensional models based on data from only five rows of fully screened wells each sampled five or fewer times were practically equivalent to values determined from moments analysis of the complete three-dimensional set of 29,285 samples taken during 16 sampling times.

Simulation of variable-density flow and transport of reactive and nonreactive solutes during a tracer test at Cape Cod, Massachusetts

A multispecies numerical code was developed to simulate flow and mass transport with kinetic adsorption in variable-density flow systems. The two-dimensional code simulated the transport of bromide (Br−), a nonreactive tracer, and lithium (Li+), a reactive tracer, in a large-scale tracer test performed in a sand-and-gravel aquifer at Cape Cod, Massachusetts. A two-fraction kinetic adsorption model was implemented to simulate the interaction of Li+ with the aquifer solids. Initial estimates for some of the transport parameters were obtained from a nonlinear least squares curve-fitting procedure, where the breakthrough curves from column experiments were matched with one-dimensional theoretical models. The numerical code successfully simulated the basic characteristics of the two plumes in the tracer test. At early times the centers of mass of Br− and Li+ sank because the two plumes were closely coupled to the density-driven velocity field. At later times the rate of downward movement in the Br− plume due to gravity slowed significantly because of dilution by dispersion. The downward movement of the Li+ plume was negligible because the two plumes moved in locally different velocity regimes, where Li+ transport was retarded relative to Br−. The maximum extent of downward transport of the Li+ plume was less than that of the Br− plume. This study also found that at early times the downward movement of a plume created by a three-dimensional source could be much more extensive than the case with a two-dimensional source having the same cross-sectional area. The observed shape of the Br− plume at Cape Cod was simulated by adding two layers with different hydraulic conductivities at shallow depth across the region. The large dispersion and asymmetrical shape of the Li+ plume were simulated by including kinetic adsorption-desorption reactions.

Using Transport Model Interpretations of Tracer Tests to Study Microbial Processes in Groundwater

It has long been known that microorganisms affect the geochemistry of groundwater. But despite this recognition, little detailed information is available regarding the rates and the factors controlling microbial processes in groundwater. Part of the reason stems from the relatively inaccessible nature of most groundwater and the difficulties encountered in obtaining representative samples of groundwater and subsurface sediments. At the same time, most groundwater systems are nutrient poor or oligotrophic environments in which the resident microorganisms are severely stressed and often nearly inactive. These populations are functioning so slowly that many types of activity measurements designed to assess microbial processes in more productive environments are ineffective for groundwater. However, because groundwater is by far the largest reservoir of freshwater in the world (Freeze and Cherry 1979), our lack of knowledge about groundwater microorganisms and their processes represents a significant void in the study of microbial ecology and in our ability to predict the outcome when these reserves are compromised by contamination.