The city of Columbus, Ohio, operates four radial collector wells, designed to yield 42 Mgal/d (million gallons per day), in southern Franklin County, Ohio, as part of their municipal supply of water. The collector wells are adjacent to, and designed to induce infiltration from, Big Walnut Creek and Scioto River. A previously constructed, three-dimensional, steady-state and transient ground-water-flow model of this river-aquifer system was used to estimate contributing recharge areas (CRA's) and calculate particle flowpaths in southern Franklin County. The simulations were of two steady-state periods (October 1979 and March 1986) and one 5-year transient period (March 1986---June 1991). The first simulation (1979) was of conditions before construction of the collector wells. The second simulation (1986) was of conditions when the collector wells were producing 8 Mgal/d. During the 5 years covered in the transient simulation, production at the well field averaged 18.5 Mgal/d. Under the 1979 conditions, the largest ground-water contributing areas were of the quarries and Scioto River (41 and 47 percent of the study area, respectively). During 1986, when 8 Mgal/d was withdrawn, the primary contributing areas were of the quarries (40 percent), collector wells (34 percent), and rivers (8 percent). Travel times associated with simulated particles of water tracked from cells along Big Walnut Creek to their discharge points in cells along Scioto River were about 5 to 60 years in the 1979 simulation and about 7 to 41 years in the 1986 simulation. The endpoints of these particles varied as simulated pumping rates were increased to 22 Mgal/d. The 1986, 10-year CRA's of the collector wells under 8 Mgal/d-conditions totalled about 4.5 mi2. As the pumping rate was increased to 22 Mgal/d in a predictive simulation, 10-year CRA's of the collector wells increased to 6.7mi2. Because the transient simulation encompassed only 5 years, the 10-year CRA's could not be estimated from the transient simulation. However, the size of the 1- to 5-year CRA's for the transient simulation was similar to the size of the 1- to 5-year CRA's for a steady-state predictive simulation if well-field production were 16 Mgal/d. The transient simulations predicted discontinuous CRA's, especially adjacent to the rivers, due to changes in hydrologic stresses. Analyses of the steady-state and transient models showed that sizes of CRA's were most sensitive to changes in porosity, pumping rate, riverbed conductance, and horizontal hydraulic conductivity.