Abstract: Based on measuring agricultural carbon emission of 31 provinces in China during 1997—2018, the paper analyses the spatio-temporal characteristics of agricultural carbon emission, and uses Kernel density estimate method and Markov chain model to analyze the dynamic evolution characteristics of agricultural carbon emission．It further discusses the relationship between agricultural carbon emission and agricultural economic development based on the theory of decoupling. The results show that:（1） Agricultural carbon emission showed a fluctuating upward trend, agricultural carbon emission intensity presented a gradual decline trend, paddy fields was the largest carbon source of agricultural carbon emission. （2） The regional disparity of agricultural carbon emission is large. The spatial distribution of provinces with lower carbon emission showed a trend of expansion, while the provinces with higher carbon emission emerged downtrend. A spatial distribution pattern of high is in the eastern and central regions and the low is in the western regions, while the spatial distribution of carbon emission intensity is contrary. （3）The agricultural carbon emission of most provinces and regions was growing, the regional gap has expanded, but the phenomenon of “polarization” gradually disappeared. The distribution of agricultural carbon emission has high stability and low internal mobility. （4）The relationship between agricultural carbon emission and agricultural economic development was mainly in the weak and strong decoupling states, and the provinces with the strong decoupling state gradually replaced the weak decoupling state and occupied the absolute dominant position in the spatial pattern.