Drainage networks, consisting of different levels of ditches, play a positive role in removing reactive nitrogen (N) via self-purification before drainage water returns to natural water bodies. However, relatively little is known about the N removal capacity of different levels of drainage ditches in irrigation agricultural systems. In this study, we employed soil core incubation and soil slurry 15N paired tracer techniques to investigate the N removal rate (i.e., N2 flux), denitrification, anaerobic ammonium oxidation (anammox), and dissimilatory nitrate reduction to ammonium (DNRA) rates in the Ningxia Yellow River irrigation district at various ditch levels, including field ditches (FD), a paddy field ditch (PFD), lateral ditches (LD1 and LD2), branch ditches (BD1, BD2, and BD3), and trunk ditches (TD). The results indicated that the N removal rate ranged from 44.7 to 165.22 nmol N g-1 h-1 in the ditches, with a decreasing order of trunk ditches > branch ditches > the paddy field ditch > lateral ditches > field ditches. This result suggested that the N removal rate in drainage ditches is determined by the ditch level. In addition, denitrification and anammox were the primary pathways for N removal in the ditches, contributing 68.40-76.64% and 21.55-30.29%, respectively, to the total N removal. In contrast, DNRA only contributed 0.82-2.15% to the total nitrate reduction. The N removal rates were negatively correlated with soil EC and pH and were also constrained by the abundances of denitrification functional genes. Overall, our findings suggest that the ditch level should be considered when evaluating the N removal capacity of agricultural ditch systems.