Terrestrial ecosystems are major sources of N pollution to aquatic ecosystems. Predicting N export to streams is a critical goal of nonpoint-source modeling. This study was conducted to assess the effect of terrestrial N cycling on stream N export using long-term monitoring data from Hubbard Brook Experimental Forest (HBEF) in New Hampshire. The field scale DAYCENT model was used to quantify N pools and long-term annual streamflow and mineral N export for six subwatersheds at the HBEF By combining DAYCENT with the Soil and Water Assessment Tool (SWAT) watershed model, mineral N export simulations were extended to the watershed scale. Our study indicated that only 13% of external N input was exported to streams during 1951-2000 at HBEF As much as 4763 kg/ha of N was stored in forest litter, soil organic matter (SOM), and living plant biomass. Net N mineralization of SOM and forest litter contributed 93% of total available N for export within the HBEF ecosystem. The Nash-Sutcliffe coefficient (E-ns) evaluating model performance of DAYCENT at six subwatersheds ranged from 0.72 to 0.82 for simulating annual streamflow (1964-2000) and from 0.48 to 0.67 for annual mineral N export (1971-1995), indicating reasonable simulated values. DAYCENT successfully predicted the effect of ecosystem disturbance such as forest cut and insect invasion on stream mineral N export. The watershed-scale simulation suggested that soil spatial variability affects stream N export in addition to the accepted controls of land cover, external N input, climate, and ecosystem disturbance.