Potassium (K) imbalances are of growing concern in southern China, where rice (Oryza sativa L.) is the primary food resource for a growing population. This study examined rice yield, K uptake and apparent balance under long-term fertilization in rice-based systems at four experimental sites, including both rice-rice as well as rice-wheat rotations. The experiments consist of four treatments: control (no fertilizer), nitrogen and phosphorus (NP), nitrogen, phosphorus and potassium (NPK), and NPK plus manure (NPKM). Across all sites, rice yields increased by 3-20% due to K fertilization (NPK vs. NP) and 4-20% due to manure application (NPKM vs. NPK). The mean internal K use efficiency (IE) was lower (32-56 kg kg -1 ) in treatments receiving K (NPK and NPKM) than in those without K application (36-91 kg kg -1 -control and NP). Estimated from the logarithmic model, a total K uptake of 38-212 kg ha -1 was needed to produce 3-7 Mg ha -1 of rice grain. The annual apparent K balances were negative (17-245 kg ha -1 year -1 ), irrespective of mineral K application and site. But the negative K balance reduced by 27-88 kg ha -1 year -1 through application of mineral K in combination with manure. The higher negative apparent K balances under ricewheat cropping system were related to the lower K application rate and the soils rich in K-bearing minerals, while the lower negative apparent K balances under rice-rice cropping system were related to the higher K application rate and the soils low in K-bearing minerals. We conclude that a re-adjustment of the current K application rate is needed to improve the long-term rice production in southern China.