Soil erosion can dramatically alter soil nitrogen (N) cycling, but it has rarely been examined under in-situ conditions due to the complicated interaction between erosion and vegetation. Comparing the effects of vegetation presence at eroding and depositional sites would provide a direct understanding of such effects. Herein, we investigated how soil N dynamics (mineral N pools and net N mineralization rates) responded to vegetation presence (plot with vs. without vegetation), landscape position (eroding vs. depositional sites) and vegetation type (grassland vs. woodland) during a two-year manipulation experiment using an in-situ incubation method in the hilly-gully region of China’s Loess Plateau. The results showed that soils on eroding site had higher concentrations of nitrate (3.820 ± 0.302 vs. 2.642 ± 0.280 mg kg^-1) and mineral N (5.853 ± 0.320 vs. 4.308 ± 0.287 mg kg^-1) but similar rates of net nitrification (0.057 ± 0.013 vs. 0.051 ± 0.011 mg kg^-1 d^-1) and mineralization (0.056 ± 0.014 vs. 0.056 ± 0.013 mg kg^-1 d^-1) in comparison to depositional site in all grassland plots. The concentrations of soil nitrate and mineral N and rates of net nitrification and mineralization in legume woodland were 26%, 19%, 88% and 64% higher than that in grassland in all slope plots, respectively. Although vegetation presence significantly decreased soil nitrate (−57%) and total mineral N (−27%) concentrations, it did not affect the rates of net nitrification and mineralization. The effects of vegetation presence on soil net N mineralization occurred irrespective of landscape position and vegetation type(P > 0.05), but varied with season or year, with greater effects in 2017 growing season than in the non-growing season and 2018 growing season. Redundancy analysis and hierarchical partitioning indicated that the above- and blow-ground biomass accounted for most of the variation in soil N mineralization. These results suggested that the variations in soil mineral N and net N mineralization in relation to vegetation presence in this eroding watershed were consistent across different landscape positions and vegetation types, but showed intra- and inter-annual variability.