Abstract: Quantitative researches on the spatial structure and spatial-temporal evolution characteristics of the net carbon sink of regional farmland ecosystem are of great significance to explore the multiple functions of farmland ecosystem and to formulate emission reduction policies appropriate to agricultural regions. Based on the input and output data of major crops at county level units in Jiangsu Province, this paper calculated the net carbon sink of farmland ecosystem in Jiangsu Province from 2005 to 2019. Further by using methods of standard deviation ellipse, spatial autocorrelation, exploratory spatio-temporal data analysis method（ESTDA）, this paper also studied the spatial-temporal evolution characteristics of the net carbon sink of farmland ecosystem in Jiangsu Province. The main results were shown that: (1) The net carbon sink of farmland ecosystem presented a spatial distribution pattern of “from northwest to southeast” in Jiangsu Province, and tended to be concentrated in the shrinking process. The gravity center of net carbon sink had generally moved towards northwest in Jiangsu Province. (2) By observing the Moran's I from 2005 to 2019, this paper found that the net carbon sink of farmland ecosystem in Jiangsu Province had a significant spatial positive correlation, the Global Moran's I value increased from 0.183 in 2005 to 0.528 in 2019. The local spatial agglomeration was stable, which was high in the north and low in the south of Jiangsu Province, while among them, the HH cluster areas and LL cluster areas accounted for more than 70% of the total number of counties. (3) The local spatial pattern of net carbon sink of farmland ecosystem in Jiangsu Province was obviously different in dynamic evolution, while the local spatial structure and spatial dependence direction were both characteristic of strong stability. In comparison, the dynamics of local spatial structure of farmland ecosystem net carbon sink in Central Jiangsu was stronger than that in Southern and Northern Jiangsu. Yancheng, Qidong and Taizhou had greater fluctuation in spatial dependence direction. (4) The occurrence probability of net carbon sink type IV was 68.33% from 2005 to 2019, indicating that the local spatial correlation pattern of net carbon sink of farmland ecosystem was stable, the local spatial linkage weak, as the spatial agglomeration was of high characteristics of path locking.