Ground-water recharge estimates for selected locations in the eastern half of the United States were obtained by Darcian and chloride-tracer methods and compared using statistical analyses. Recharge estimates derived from unsaturated-zone (RUZC) and saturated-zone (RSZC) chloride mass balance methods are less variable (interquartile ranges or IQRs are 9.5 and 16.1 cm/yr, respectively) and more strongly correlated with climatic, hydrologic, land use, and sediment variables than Darcian estimates (IQR = 22.8 cm/yr). The unit-gradient Darcian estimates are a nonlinear function of moisture content and also reflect the uncertainty of pedotransfer functions used to estimate hydraulic parameters. Significance level is <0.001 for nearly all explanatory variables having correlations with RUZC of <-0.3 or >0.3. Estimates of RSZC were evaluated using analysis of variance, multiple comparison tests, and an exploratory nonlinear regression (NLR) model. Recharge generally is greater in coastal plain surficial aquifers, fractured crystalline rocks, and carbonate rocks, or in areas with high sand content. Westernmost portions of the study area have low recharge, receive somewhat less precipitation, and contain fine-grained sediment. The NLR model simulates water input to the land surface followed by transport to ground water, depending on factors that either promote or inhibit water infiltration. The model explains a moderate amount of variation in the data set (coefficient of determination = 0.61). Model sensitivity analysis indicates that mean annual runoff, air temperature, and precipitation, and an index of ground-water exfiltration potential most influence estimates of recharge at sampled sites in the region. Soil characteristics and land use have less influence on the recharge estimates, but nonetheless are significant in the NLR model. ?? 2006 Elsevier B.V. All rights reserved.