Groundwater is the primary source of drinking water for many communities in the United States and elsewhere. This valuable resource is often endangered by the impact of contaminants of variable chemical composition. Much of this impact is a result of leakage from underground storage tanks used by service stations. A key issue considered by hydrogeologists during their site assessment is the subsurface structure of the aquifer and the prevailing conditions that control groundwater flow. In cases where the aquifer system is confined or semi-confined, with an aquitard between a shallow and a deeper aquifer, the aquitard may be a protective layer, preventing the underlying aquifer from being impacted by the contaminants. However, this scenario does not hold true if portions of that aquitard are discontinuous. Such discontinuities, also called pinch-outs or windows, would allow for vertical downward migration of the contaminants either directly (in cases involving DNAPLs) or as a dissolved phase in the presence of a downward hydraulic gradient for either DNAPLs or LNAPLs. For that reason, source areas near aquitard windows pose a greater risk to deeper groundwater contamination and should have a higher priority for cleanup. Identifying and mapping these aquitard windows requires the installation of numerous boreholes in the area of investigation, which increases site characterization costs. The aim of this research was to test the ability and accuracy of two separate and independent approaches, geostatistical analysis and aquifer testing, to estimate the distribution and dimensions of aquitard windows. The data used in this research were obtained from Site Assessment and Groundwater Monitoring Reports prepared by various consultant companies for the Charnock sub-basin of the Santa Monica Groundwater Basin. The Charnock groundwater sub-basin is located in Southern California within the communities of West Los Angeles and Culver City and is an important source for drinking water for the cities of Santa Monica and Culver City. The City of Santa Monica sought assistance from EPA after shutting down a number of its drinking water wells in 1996 due to the presence of increasing levels of the gasoline additive MTBE. In 1997, the EPA and the Los Angeles Regional Water Quality Control Board entered into a partnership to investigate the sources of the MTBE pollution in the sub-basin. This Charnock Project also involves the cleanup of MTBE and other gasoline-related pollution in areas affecting water quality in the Charnock Sub-basin and the restoration of this sub-basin for use as a drinking water supply. Affected residents of Santa Monica and Culver City are currently receiving replacement water. The study focused on two principal objectives: (1) Test the ability of geostatistical tools, such as variograms, kriging, and error estimation to produce accurate contour maps of geologic layers using the least number of data points, and (2) Test the sensitivity of aquifer tests to delineate the sizes of aquitard pinch-outs or windows. A number of software packages were utilized for this study. Mainly, GS+ and Surfer were used for the geostatistical portion of the study, and the integrated MODFLOW component in Groundwater Vistas for the aquifer test simulations. In the geostatistical portion of this study, it was shown that less than 50% of the data avail able in the study area can be used to generate an isopach map of the aquitard with the same level of confidence as with all the sampling data. Any increase observed in the average estimation error values using more than 50% was attributed to the microvariability that existed within closely spaced thickness samples as indicated by the nugget effect values. For the aquifer testing part of this research, a three-layered conceptual model was designed to re plicate the same geologic configuration of the Charnock Site. That is, two aquifers separated by a confining layer. The vertical flow properties of the aquitard were varied to observe their effects on detecting aquitard windows. The aquitard windows were assigned five different sizes and six different radial distances from the pumping well, each of which was individually modeled. A total of 160 2-day aquifer pump tests were simulated. The net drawdown produced by the window was contoured by calculating the difference between the drawdown in the un-pumped aquifer with a window present in the confining bed (observed drawdown) and the drawdown in the same aquifer with the window absent (expected drawdown). The vertical hydraulic conductivity of the aquitard (K'v) has the greatest effect on window detection. Two sources for water level measurement thresholds were used for the window detection distance; those derived from a range of measured K'v from core samples, and those derived from calculated K'v from time-drawdown data. Using K'v from core samples, model results have shown that observation wells installed at 200 ft centers above aquitards with a mean K'v of 0.001 ft/d are sufficient to detect windows with sizes as small as 10 x 10 ft2, at distances up to 700 ft away from the pumping well. A grid of monitoring wells drilled at 40 ft spacing would be effective in locating windows 20 x 20 ft2 or larger in size as close as 140 ft to the pumping well assuming the overall K'v of the aquitard is in the order of 0.01 ft/d. Windows with sizes smaller than 40 ft x 40 ft in aquitards with an average K'v of 0.1 ft/d mare generally non-detectable. Windows smaller than 20 x 20 ft2 were found to be non-detectable using K'v from the aquifer test. Observation wells spaced at 20 ft above aquitards with actual K'v of 0.001 ft/d can detect 20 x 20 ft2 windows located at distances of 70 ft or less from the pumping well, and larger windows up to 700 ft away. A slightly smaller well spacing, 17 ft, would be sufficient in detecting the same windows in aquitards with actual K'v of 0.01 ft/d. Windows 30 x 30 ft2 or larger in size and located up to 700 ft away from the production well are detectable in wells spaced 28 ft apart above leaky aquitards with K'v in the order of 0.1 ft/d. Smaller windows can be detected by the same wells if located within 140 ft or so from the pumping well.