International audience ; An intercomparison is made of the Net Ecosystem Exchange of CO 2 , NEE , for eight Dutch grassland sites; four natural grasslands, two production grasslands and two meteorological stations within a rotational grassland region. At all sites the NEE was determined during at least 10 months per site, using the eddy-covariance (EC) technique, but in different years. The photosynthesis-light response analysis technique is used along with the respiration-temperature response technique to partition NEE among Gross Primary Production ( GPP ) and Ecosystem Respiration ( R e ) and to obtain the eco-physiological characteristics of the sites at the field scale. Annual sums of NEE , GPP and R e are then estimated using the fitted response curves with observed radiation and air temperature from a meteorological site in the centre of The Netherlands as drivers. These calculations are carried out for four years (2002?2005). The estimated annual R e for all individual sites is more or less constant per site and the average for all sites amounts to 1390±30 gC m ?2 a ?1 . The narrow uncertainty band (±2%) reflects the small differences in the mean annual air temperature. The mean annual GPP was estimated to be 1325 g C m ?2 a ?1 , and displays a much higher standard deviation, of ±100 gC m ?2 a ?1 (8%), which reflects the relatively large variation in annual solar radiation. The mean annual NEE amounts to ?65±85 gC m ?2 a ?1 , which implies that on average the grasslands act as a source, with a relatively large standard deviation. From two sites, four-year records of CO 2 flux were available and analyzed (2002?2005). Using the weather record of 2005 with optimizations from the other years, standard deviation of annual GPP was estimated to be 171?206 gC m ?2 a ?1 (8?14%), of annual R e 227?247 gC m ?2 a ?1 (14?16%) and of annual NEE 176?276 gC m ?2 a ?1 . The inter-site standard deviation was higher for GPP and R e , 534 gC m ?2 a ?1 (37.3%) and 486 gC m ?2 a ?1 (34.8%), respectively. However, the ...