This study uses time series analysis to examine long-term stream water nitrate concentration records from a pair of forested catchments at the Coweeta Hydrologic Laboratory, North Carolina, USA. Monthly average concentrations were available from 1970 through 1997 for two forested catchments, one of which was clear-felled in 1977 and the other maintained as a control. The time series were decomposed into their trend and annual cycle before modeling as an autoregressive (AR) process. AR models were calculated for both an expanding and a shifting window so that prefelling could be directly compared with the effects of tree clearance. In comparison with flow records for both of the catchments, transfer function-noise models were calculated on a moving window basis, and the impulse functions were derived. Analysis shows that both catchments show an annual memory effect but that the clear-felled catchment shows, in addition, a 6-month memory effect. The annual effect in the control catchment responds to drought conditions while in the felled catchment, it reflects the change in vegetation. The 6-month effect in the felled catchment responds to drought conditions independent of both the annual effect and of logging operations. The control catchment shows no significant impulse function with respect to flow, while for the felled catchment a distinct impulsivity develops over time subsequent to logging and coincident with the onset of nitrogen saturation. By examining the nature of the nitrate export, rather than solely the levels of export, a systems approach can be taken to understanding catchment behavior. Such an approach shows that the catchment is in metastable equilibrium with respect to its hydrological pathways and nitrogen reserves but in dynamic equilibrium with respect to long-term temperature change. The onset of nitrogen saturation may represent irreversible changes in the catchment behavior, and impulsivity, with respect to streamflow, represents a new indicator of N-saturated conditions.