An analytic element ground-water-flow model was constructed to help understand the ground-water-flow system in the vicinity of the Ho-Chunk Nation communities of Indian Mission and Sand Pillow in Jackson County, Wisconsin. Data from interpretive reports, well-drillers' construction reports, and an exploratory augering program in 2003 indicate that sand and gravel of varying thickness (0-150 feet[ft]) and porous sandstone make up a composite aquifer that overlies Precambrian crystalline rock. The geometric mean values for horizontal hydraulic conductivity were estimated from specific-capacity data to be 61.3 feet per day (ft/d) for sand and gravel, 6.6 ft/d for sandstone, and 12.0 ft/d for the composite aquifer. A ground-water flow model was constructed, the near field of which encompassed the Levis and Morrison Creeks Watershed. The flow model was coupled to the parameter-estimation program UCODE to obtain a best fit between simulated and measured values of ground-water levels and estimated Q50 flow duration (base flow). Calibration of the model with UCODE provided a ground-water recharge rate of 9 inches per year and a horizontal hydraulic conductivity of 13 ft/d for the composite aquifer. Using these calibrated parameter values, simulated heads from the model were on average within 5 ft of the measured water levels. In addition, these parameter values provided an acceptable base-flow calibration for Hay, Dickey, and Levis Creeks; the calibration was particularly close for Levis Creek, which was the most frequently measured stream in the study area. The calibrated model was used to simulate ground-water levels and to determine the direction of ground-water flow in the vicinity of Indian Mission and Sand Pillow communities. Backward particle tracking was conducted for Sand Pillow production wells under two pumping simulations to determine their 20-year contributing areas. In the first simulation, new production wells 6, 7, and 8 were each pumped at 50 gallons per minute (gal/min). In the second simulation, new production wells 6, 7, and 8 and existing production well 5 were each pumped at 50 gal/min. The second simulation demonstrated interference between the existing production well 5 and the new production wells when all were pumping at 50 gal/min.