A stochastic-mechanistic modeling approach to reactive solute transport through an integrated soil-groundwater system is used to interpret observed results of subsurface 18O transport through a covered experimental catchment at the Lake G¿rdsj¿n site, Sweden. The model incorporates the effect of variable pathway lengths to the catchment outlet due to the distributed source of 18O over the whole catchment area, the nonuniform mean flow velocity that results from groundwater recharge through the unsaturated zone, the existence of preferential flow and transport pathways and the associated diffusional mass transfer between mobile and relatively immobile water zones, and the random flow heterogeneity that results from spatially variable hydraulic transmissivity. Using available site observations and topographical data from the catchment, most of the required model parameters are quantified by independent estimation. On the basis of this independent estimation and calibration of the remaining few parameter values, a good representation of the 18O transport through the catchment was obtained. The spreading effect of the longitudinal extent of the solute source (due to differences in pathway length to the catchment outlet for different entrance points at the soil surface) does to a large degree explain the observed temporal spreading of 18O at the catchment outlet. Additional solute spreading is obtained by diffusional solute mass transfer between mobile and relatively immobile water regions, associated with the existence of preferential flow paths observed at the site. In comparison with these two spreading mechanisms, flow heterogeneity within the mobile water plays a much less important role. This indicates that natural tracer transport through a whole catchment area does not provide a sufficiently fine tool for distinguishing between and quantifying various flow heterogeneity effects that may be important for smaller solute sources. ¿ 1999 American Geophysical Union