A better understanding of residual N fate is important for N management in agricultural production. The N application rate and time may dramatically influence N recovery in the plant–soil system, in turn affecting residual N uptake by subsequent crops. A 3-year field experiment was conducted in plastic-mulched maize in semiarid farmland. 15N-labeled urea was applied to microplots with a single application (100% before sowing, N1), two splits (4:6 at sowing and eight-leaf stages, N2), and three splits (4:3:3 at sowing, eight-leaf, and silking stages, N3), and the fate of residual fertilizer N in soils over the following two cropping seasons was examined. Approximately 14.6–18.7% and 5.4–5.8% of labelled fertilizer N were recovered by maize in the second and the third seasons, respectively, with the cumulative recovery efficiency reaching 47.6–60.8% over 3 years. Applying N with three splits significantly increased residual fertilizer N recovery by 24.6% in the second cropping season compared to N1 and N2. About 22.7–32.4% and 15–21% of total labelled N applied was residual in the 0–2 m soil layer after the second- and the third-season harvest, respectively, with the higher residual amount from split N applications. In conclusion, split N applications significantly increased the cumulative fertilizer N recovery in the plant–soil system while decrease the potential losses over 3 years, due to the higher recovery efficiency and the lower N losses from topdressed N.