Numerical experiments are applied to the problem of modeling base flow water quality originating from nonpoint or spatially distributed sources of groundwater contamination in an idealized stream-aquifer setting. The experiments examine the dependence of base flow concentration on stochastic spatial variability of the nonpoint source, the initial aquifer concentration field, and the hydraulic conductivity field. The deterministic effects of local dispersive mixing, aquifer aspect ratio, and equilibrium sorption are also investigated. For the simulated conditions, base flow concentration is found to be independent of second-spatial moments of the above random parameter fields and independent of the magnitude of local dispersive mixing. The experiments demonstrate the primary importance of defining the mean flow field and the space average concentration when determining “effective” field parameters. Input-output behavior is represented by a nondiffusive integral balance model with exponential or Weibull-type kernel. Parameters of the integral balance model are shown to be adequately represented by space averages.