Fertilizer application during sugarcane cultivation is a main source of nitrogen (N) loads to groundwater on small islands in southwestern Japan. The aim of this study was to quantify the effect of reducing the N fertilizer application rate on sugarcane yield, N leaching, and N balance. We conducted a sugarcane cultivation experiment with drainage lysimeters and different N application rates in three cropping seasons (three years). N loads were reduced by reducing the first N application rate in all cropping seasons. The sugarcane yields of the treatment to which the first N application was halved (T2 = 195 kg ha⁻¹ N) were slightly lower than those of the conventional application (T1 = 230 kg ha⁻¹ N) in the first and third seasons (T1 = 91 or 93 tons ha⁻¹, T2 = 89 or 87 tons ha⁻¹). N uptake in T1 and T2 was almost the same in seasons 1 (186–188 kg ha⁻¹) and 3 (147–151 kg ha⁻¹). Based on the responses of sugarcane yield and N uptake to fertilizer reduction in two of the three years, T2 is considered to represent a feasible fertilization practice for farmers. The reduction of the first N fertilizer application reduced the underground amounts of N loads (0–19 kg ha⁻¹). However, application of 0 N in the first fertilization would lead to a substantial reduction in yield in all seasons. Reducing the amount of N in the first application (i.e., replacing T1 with T2) improved N recovery by 9.7–11.9% and reduced N leaching by 13 kg ha⁻¹. These results suggest that halving the amount of N used in the first application can improve N fertilizer use efficiency and reduce N loss to groundwater.