The density correction, which is widely used in eddy covariance calculations to compensate for density fluctuations arising from heat and water vapor transfer in order to determine the surface (ecosystem) flux of a gaseous constituent such as CO2, was revisited based on its historical background and recent discussions. Firstly, the derivations previously published were categorized into three by their prerequisites: (1) mean vertical wind (the conventional Webb-Pearman-Leuning derivation), (2) expansion/compression of air parcel, and (3) the mass balance and continuity of both the target constituent and dry air. All the categories of derivation implicitly or explicitly assume no sink/source of dry air at the surface and arrive at the same equation for the correction, which relates an eddy covariance flux from mass or molar density measurements to the corresponding mixing ratio flux. Secondly, we examined the underlying assumptions or concepts used in those derivations, especially mean vertical wind to compensate for the turbulent flux of dry air. Thirdly, practical aspects of the density correction were considered for both open- and closed-path eddy covariance, and critical problems such as sensor heating of an open-path IRGA were also discussed. Our conclusions were that the density correction is theoretically sound as long as there is no sink/source of dry air at the surface but difficult to apply correctly because, in disagreement with its theory, different sensors with different time constants, which are usually positioned separately, are used in actual eddy covariance systems to calculate the correction terms. Closed-path systems are advantageous over open-path systems under such a condition that the surface flux is expected to be an order smaller in magnitude than the correction terms, due to their ability to point-by-point convert density measurements to the mixing ratios using the temperature, pressure and water vapor density that are simultaneously measured in the gas analyzer.
