Direct and continuous longterm measurements of canopyscale carbon dioxide fluxes have only recently become possible due to the improvement of eddy covariance technique. Now it has been widely used in longterm observation studies of CO2 uptake and release by terrestrial ecosystem. A global network of micrometeorological tower sites (FLUXNET) uses eddy covariance methods to measure the exchanges of CO2 between terrestrial ecosystem and atmosphere, but a majority of the site are not in ideal conditions, which can not fully meet the basic assumption of eddy covariance technique for the CO2 flux measurement, thus resulting in some uncertainties in the flux measurement. Chinese Terrestrial Ecosystem Flux Observational Research Network (ChinaFLUX) has been established, which relies on Chinese Ecosystem Research Network (CERN) and applies eddy covariance of microclimatology as main research method to study fluxes of carbon dioxide between the vegetation and soil of typical ecosystem and atmosphere. The objective of this paper is to introduce the basic assumption of the eddy covariance technique, the basic theoretical equations, the types and characteristics of errors in longterm measurements and to elucidate emphatically the uncertainties in the flux measurement due to physical limitation of instrumentation, two and three dimensional air flow motion effects, methods of data processing, and underestimation of nighttime fluxes, etc. Where possible, this paper makes recommendations concerning methodologies and research priorities. Without some understanding of and ability to compensate for these uncertainties, crosssite comparisons and global scale synthesis are difficult and uncertain at best. Thus, dada quality analysis and control (QA/QC) and errors assessments are necessary for crosssite comparisons and global scale synthesis. 