A long-term (33 years) experiments were conducted to investigate the effect of chemical fertilization on rice yield, yield trends, soil properties, agronomic efficiency of applied nutrients and nutrient balance for the double rice cropping systems in subtropical China. The treatments were different combinations of N, P and K fertilizers (N, NP, NK and NPK), double dose of recommended NPK (2NPK) and no fertilizer control (control). Compared with no fertilizer control, all fertilization treatments had no significant effects on soil pH and SOC contents (P > 0.05), but generally increased nutrients content when corresponding elements were applied. The impact of fertilizers on grain yields was 2NPK > NPK > NP > NK > N, and application of P fertilizer not only increased the rice yield, but improved yield stability. The trend of agronomic use efficiency of applied P was significantly positive (P < 0.05) only for the first rice crop, suggesting that P fertilizer played a less important role in the second rice season than in the first rice season. The study indicated that the current local fertilizer recommendations should be optimized for the consideration of differences in indigenous nutrient supplies in different rice seasons.