Studies using eddy covariance have shown grasslands to be both sinks and sources of carbon dioxide (CO2). However, such studies do not take into account the exports of carbon (C), such as in meat and milk and imports of C, such as off-farm derived C in cattle feed supplement. By coupling eddy covariance results with farm management data we quantified the farm scale C balance during 2004 for two dairy farms in South West Ireland. The system boundary for inputs and outputs of C is the farm perimeter. Carbon sequestration is determined as the difference between all C inputs and C outputs. Carbon inputs are similar in both farms with net ecosystem exchange (NEE) (2.9 +/- 0.5 t C ha(-1) year(-1)) accounting for 88 and 81\% of C inputs in Farms A and B, respectively. Carbon in concentrate feed accounts for 12 and 19\% of C inputs in Farms A and B, respectively. Respiration by cattle during the winter housing period, and respiration by cows during milking throughout the grazing season, are the largest C outputs and account for approximately half of C outputs on both farms. The other major sources of C output are milk, CH4 produced by enteric fermentation and emitted during slurry spreading and dissolved organic carbon (DOC) in streamflow. Carbon in meat and CH4 emissions from dung (both in the farmyard and fields) and animal slurry in farmyard storage are minor sources of C output. The annual total C inputs are 3.30 and 3.58 t C ha(-1) and the total C outputs are 1.25 and 1.43 t C ha(-1) in Farms A and B, respectively. The net difference is 2.05 and 2.15 t C ha(-1) in Farms A and B, respectively. This suggests that both farms were net C sinks for 2004. Further work on below ground process and soil C turnover is required to determine if this C sink estimate is reflected in changes in soil C stocks. Furthermore, we estimate the global warming potential (GWP) of this grassland to be a sink for similar to 1 t CO2 equiv. ha(-1) year(-1). (c) 2006 Elsevier B.V. All rights reserved.