Abstract: Based on a long-term experiment on maize-soybean relay intercropping system in Huanjiang Observation and Research Station for Karst Ecosystem, Chinese Academy of Sciences, the effects of straw in place of chemical potassium on spring maize yield and soil potassium (K) balance were investigated, aiming to provide scientific basis for the rational use of straw and fertilizer in karst peak-cluster depression. The experiment was started in 2006, and the data were collected during 2010-2014. Four treatments with four repetitions were set in this study, including CK (no fertilizer), NPK (chemical fertilizer), C7S3 (30% K from straws and 70% K from chemical fertilizer), C4S6 (60% K from straws and 40% from the chemical fertilizer). The results showed that:1) the annual yields of spring maize in fertilization treatments from 2010 to 2014 were 4.12-4.17 times that in CK treatment. The spring maize yields in C7S3 and C4S6 were 98.3% and 98.7% of that in NPK, respectively. The yields of grain and straw increased with time under fertilization treatments. 2) Compared with CK, fertilization treatments significantly increased K amount in maize straw, with 2.21-2.53 times that in CK. N, P and K absorption amounts in maize grain under fertilization treatments significantly increased (P < 0.01), but there was no significant difference among fertilizer treatments (P > 0.05). 3) There were no significant differences in recovery efficiency, absorption efficiency, agronomic efficiency and partial factor productivity of K among treatments of NPK, C7S3 and C4S6. 4) The annual surplus amount of K in NPK treatment was 3.00 kg·hm^-2, which was significantly higher than that in C7S3 (-1.90 kg·hm^-2). The most annual surplus amount of K was in C4S6, 8.22 kg·hm^-2, with the surplus rate of 7.4%. 5) Compared to the beginning of the experiment in 2006, soil rapidly available K declined by 15.9% in 2014. The fertilization treatments significantly increased amount of soil rapidly available K (P < 0.01). The order of annual increase of soil rapidly available K from 2010 to 2014 in fertilization treatments were NPK > C7S3 > C4S6, but no significant difference among treatments was observed (P > 0.05). 6) After 8 years of cultivation, soil slowly available K declined by 10.9% and 4.9% in CK and C4S6, while increased by 1.3% and 22.4% in NPK and C7S3, respectively. Overall, the straw returning to replace some K fertilizer was suitable for the soil in karst peak-cluster depression, which not only maintained the yield of spring maize, but also sustained soil K dynamic balance. And 60% of the straw K substitution was more suitable for sustaining K balance, while 30% of the straw K substitution was more beneficial to improve soil slowly available potassium.