We developed a Bayesian framework that permitted quantitative integration of borehole flow meter data and cross hole geophysical information for estimating the hydraulic conductivity distribution at each pixel along the tomographic cross sections. Figure 2 illustrates the estimates of hydraulic conductivity along two vertical cross sections at the Oyster site, obtained using the Bayesian estimation framework and radar velocity tomograms. Hydraulic conductivity data from an electromagnetic borehole flow meter log are superimposed on top of the estimates obtained using radar tomographic data within the Bayesian framework. Comparison of the estimated hydraulic conductivity field and tracer breakthrough data suggested that the tomographic estimates were extremely useful for helping to reduce the ambiguity associated with interpreting bacterial and chemical transport data collected during tracer tests. Even though this site was fairly homogeneous and it had extensive borehole control ,planting blueberries in a pot it was difficult to capture the variability of hydraulic conductivity using borehole data alone sufficiently to ensure reliable transport predictions.
By comparing numerical model predictions with tracer test measurements at the Oyster Site, Scheibe and Chien found that conditioning the models to the geophysical estimates of hydraulic conductivity significantly improved the accuracy and precision of the model predictions relative to those obtained using borehole data alone. This study suggested that the geophysical based methods provided information at a reasonable scale and resolution for understanding field-scale processes. This is an important point, as it is often difficult to scale the information gained at the laboratory scale or even from discrete well bore samples for use at the remediation field scale.In addition to the importance of hydrogeological heterogeneity in controlling contaminant and bacterial transport, geochemical heterogeneity plays a large role in reactive flow and transport. In the same general area as the study discussed above, we used crosshole ground penetrating radar amplitude data to estimate lithology and sediment Fe and Fe. In this study, we developed a sampling-based Bayesian approach to fuse the diverse geochemical, lithological, and geophysical datasets into an integrated interpretation. The geophysical data do not sense the geochemical parameters directly, but the geophysical attribute is sensitive to lithology, which is in turn sensitive to sediment geochemistry.
Our developed estimation approach exploited this mutual dependence to estimate lithology and sediment geochemical parameters using geophysical data. Figure 3a illustrates the 2-D geophysical attribute field obtained from inversion of GPR amplitude data. Figures 3b-3d illustrate the mean values of the estimated Fe and Fe distributions; variances associated with these estimates are available but are not shown in the figure. Cross-validation exercises revealed that the estimates obtained using the geophysical data were accurate and greatly improved the 2-D identification of the geological and geochemical properties. This study represents perhaps the highest resolution field-scale characterization of geochemical properties performed todate. Although geophysical methods have been used to estimate hydrological parameter probability density functions with very high resolution , this level of detail may not always be necessary to adequately describe the controls on transport. Many studies suggest that spatial variations of lithological changes tend to be closely correlated with hydrological parameters and with geochemical parameters. As such, if an understanding of the relationships between lithofacies, hydrological and geochemical parameters is available, field-scale mapping of lithofacies may provide sufficient information about the controls on transport over large spatial scales. Figure 4 illustrates how geophysical methods have been used with sparse hydrological data and with stochastic estimation techniques to estimate the probability of being within hydraulically important units.
The top image illustrates the probability of being within a transmissive fracture zone at the NABIR Field Research Center at the Oak Ridge National Laboratory in Tennessee obtained using seismic velocity tomographic and flow meter data within a Monte Carlo markov Chain approach. This fracture zone is the ongoing focus of a Uranium bio-stimulation experiment at this site . The middle image shows the probably of being within a sandy lense at the Hanford 100H site in WA, and was obtained using seismic and radar tomographic data and a discriminant analysis technique. Finally, the lower figure illustrates the probability of being within a sandy lithofacies at the DOE Bacterial Transport Site in Oyster, VA , which was estimated using radar velocity data within a Bayesian estimation approach.Remediation approaches, such as in situ chemical manipulation and bio-stimulation, often induce dynamic spatiotemporal transformations in subsurface systems, such as the dissolution and precipitation of minerals, gas evolution, redox variations, biofilm generation, and changes in permeability and porosity. The limited understanding of bio-geochemical-hydrological processes and the inadequacy of conventional approaches for characterizing or monitoring these processes hinders our ability to guide contamination remediation. We have investigated the capability of time-lapse geophysical datasets for remotely detecting changes in hydrological-biogeochemical properties during bio-stimulation at both the laboratory and the field scale. Using column-scale experiments, we have tested the sensitivity of different geophysical methods to reaction products that occur during bio-stimulation, such as gasses, precipitates and biofilms during nitrate and sulfate reduction. As a consequence of the greater energy available from nitrate reduction, a majority of nitrate in a groundwater system is reduced prior to reduction of the less energetic manganese, iron, and sulfate. Thus, the onset of denitrification is an important indicator for determining the redox state of the aquifer that is being perturbed, and advanced techniques to detect this onset and monitor associated processes in a remote and accurate manner are needed. We conducted laboratory based experiments to investigate the sensitivity of radar attributes for detecting N2 gas generated during denitrification as the gas replaces the original pore water. The key steps in the experimental design included packing and inoculating sand columns with Acidovorax sp. , providing nutrients to the microbes to stimulate denitrification, and measuring the hydraulic and geophysical responses during the bio-stimulation experiment.
TDR measurements were made along the length of the column and over time to estimate the spatiotemporal variations in dielectric constant during the stimulation experiment. We used a volumetric averaging/mixing model to represent the effective dielectric measurement as a function the individual components that contribute to the measurement,raspberries in pots including the dielectric constants of N2 , of water, and of the sand grains. Figure 5 illustrates the estimated percent of the pore space that was filled by N2 gas as a function of time after the bio-stimulation experiment began for three different TDR measurement locations. This figure illustrates how the pore water started to be significantly replaced by gas after about thirteen days. Gas was first detected by probe #1, or the top probe in the column. The apparent ‘jumps’ in the estimated evolved N2 gas are associated with the pressure release procedure. With time, all of the probes sensed the presence of a significant volume of gas. At the end of the experiment, we estimated that approximately 22% of the pore spaces were filled with N2 gas for the bottom third of the column, about 21% of the pore spaces were filled with gas for the middle third of the column, and that 31% of the pore spaces were filled with N2 gas at the top third of the column, yielding an average estimated gas saturation over the entire column of 24.6%. The experiment suggested that the radar velocity may be a good indicator of the onset and extent of denitrification that occurs during bio-stimulation. Our laboratory and field-scale experiments have also indicated that the geophysical methods hold good potential for detecting some amendments as they are introduced into an aquifer and as they distribute over time. We have conducted geophysical imaging in support of a Cr bio-remediation effort that is being performed at the 100H Site at the Hanford Reservation in Hanford. Here, HRC, which is a slow-release polylactate was injected into a sandy portion of a saturated aquifer to reduce and immobilize Cr. Field tomographic data were collected before and after HRC injection. Figure 6 illustrates some of the field-estimates obtained using radar and seismictomographic data at the site. The top image in Figure 6 illustrates the hydraulic conductivity zonation in the stimulation zone, obtained using a statistical approach with radar and seismic tomographic data together with borehole flow meter data. Although the entire section is characterized as being ‘sandy’ using conventional data, there are areas within the section that are more permeable . The subsequent images in Figure 6 show the changes in electrical conductivity as a function of time after HRC was injected into the aquifer, obtained using radar amplitude and velocity information following Peterson.
We find that the HRC, which is electrically conductivity upon injection due to the release of lactic acid when mixed with the groundwater, originally distributes near the injection area. However, very soon after injection , the lactic acid redistributes itself into the higher hydraulic conductivity section, and migrates down gradient. After 9 days, the change in the electrical response due to the lactate injection is minimal compared to the change in the conductivity associated with the metals precipitation around the HRC reaction front . The bottom image in Figure 6 shows that the seismic energy is completely attenuated in the vicinity of the HRC. These images illustrate the utility of high-resolution geophysical methods for imaging the amendment location and redistribution as a function of heterogeneity. The geophysical responses to HRC that we observed at the field scale were similar to the responses observed in the laboratory.In 2007, 311 million kg of pesticide active ingredients were used in agriculture in the United States . California, the state with the largest agricultural output, uses 25% of all U.S. agricultural pesticides, or 84.5 million kg annually. Although recent studies have demonstrated widespread organophosphate pesticide exposures in the general U.S. population, including pregnant women and children, exposures are often higher in agricultural populations . Among pregnant women in the Center for the Health Assessment of Mothers and Children of Salinas study, all of whom lived in an agricultural region, and many of whom either worked in agriculture or lived with people who did, OP urinary metabolites or dialkylphosphate levels were ∼40% higher than those in a representative sample of U.S. women of childbearing age . We observed adverse associations in the CHAMACOS study between prenatal maternal DAP concentrations and children’s performance on the Bayley Scales of Infant Development at 2 y , measures of attention at 5 y , and on the Wechsler Intelligence Scale for Children at 7 y . Several studies in other populations have similarly reported adverse associations of prenatal exposure to OP pesticides and child neurodevelopment , but few studies have examined the effects of other potentially neurotoxic pesticides on child cognitive development. Populations residing in agricultural areas may be exposed to a complex mixture of neurotoxic pesticides through pesticide drift and para-occupational exposures . Some of these pesticide classes—including OPs as well as carbamates— share atleast one mode of action, depression of acetylcholinesterase , and there isin vitro evidence that there may be additive inhibitory effects from exposure to certain pesticide mixtures . Furthermore, although biomarkers such as DAP metabolites are an important tool for assessing exposures to pesticides, several challenges limit the utility of pesticide biomarkers in epidemiologic analyses. For example, many pesticides have a short halflife in the body, and biomarkers reflect only very recent exposures . In addition, no biomarkers are available for some pesticides, leaving only environmental concentrations or modeling to characterize exposure. Since 1990, all agricultural pesticide applications in California have been compiled in the uniquely comprehensive Pesticide Use Reporting database. In several studies, PUR data have been shown to correlate with pesticide levels in various media. For example, we have shown significant associations between nearby use of specific pesticides based on the PUR data and levels in house dust , and moderate to strong associations between agricultural use of malathion, chlorpyrifos, and diazinon with community air samples . In addition, several epidemiologic studies conducted in California have shown that higher nearby agricultural pesticide use is associated with various adverse health outcomes, including OP and fungicide use with Parkinson disease , OP and pyrethroid use with autism and birthdeffects , organochlorine pesticide use with autism , and the use of carbamates and the neonicotinoid imidacloprid with neural tube deffects in children .