Abstract: Soil carbon pool is an important component of terrestrial ecosystem with higher carbon storage and the atmospheric CO₂ pool compensation capacity via CO₂ flux from soil surface, which is of great significance to global carbon cycle and carbon budget. Therefore tiny changes in soil carbon pool have a strong impact on atmospheric CO₂ concentration. This study aimed to assess carbon budget at ecosystem scale by comparing soil carbon efflux, net primary productivity (NPP) and net ecosystem productivity (NEP). The assessment was done through soil-atmosphere interface analysis as influenced by land use change and long-term fertilization in croplands of maize-soybean-wheat rotation system in the black soils of Northeast China. The land use types were grassland (GL) and bareland (BL) and farmlands with three long-term fertilization patterns including non-fertilization (NF), nitrogen, phosphorus and potassium fertilization (NPK), and organic manure amended NPK (NPKOM). Static chambers used to collect gas samples in cropland treatments were of two types - one used to collect gas samples from whole soils and the other to collect gas samples from non-rhizospheric soils. Soil carbon flux was determined by CO₂ concentration detection in a gas chromatograph (Shimadzu GC-2010) equipped with a flame ionization detector (FID). The annual cumulative CO₂ flux was calculated based on soil CO₂ flux rate. The three-year averaged annual carbon flux (R_ann) and annual microbial respiration (R_m) decreased in the order of NPKOM > GL > NPK > NF > BL. In addition, significant difference was noted in Rann and R_m among the five experimental treatments (P < 0.05). However, no significant difference was observed between grassland and NPKOM plots (P > 0.05) in terms of soil CO₂ fluxes. For cropland treatments, R_ann and R_m were significantly (P < 0.05) higher in NPKOM than in NPK and NF. However, no significant difference was found in annual carbon flux in root-free soils (R_rfann) among the three cropland treatments. Total biomass and carbon sequestration in grassland was significantly (P < 0.05) higher than that in other four plots. This was particularly so for grass roots which sequestered 10 20 times of carbon as against cropland treatments. NPP decreased in the order of GL > NPKOM > NPK > NF > BL, with a significant difference among the five treatments (P < 0.05). Grassland NPP increased by 32% 96% compared to cropland treatments. Also NPPs of NPKOM and NPK were respectively 46% and 49% higher than that of NF. NEP was positive for grassland and cropland, indicating net sinks of atmospheric CO₂ at the ecosystem scale. For cropland, carbon sequestration under NPKOM was not significantly (P > 0.05) higher than that under NPK. This was due to the "priming effect" of organic amendment resulting in increased soil respiration and a significant decrease in NEP. In terms of carbon balance through soil-atmosphere interface, grassland soil was the net sink under realistic management practices. In contrast, bareland and cropland soils served as the net sources. To date, carbon storage in arable soils tends to decline. However, increase in organic input could enhance carbon sequestration and thereby attain a new carbon equilibrium.