Abstract. An automated open system for measurement of soil CO2 efflux (Rsc) was developed and calibrated against known fluxes and tested in the field, while measuring soil respiration also by the gradient method (Rsg) at a dry sandy grassland (Bugac, Hungary). Ecosystem respiration (Reco) was measured by the eddy covariance technique. Small chamber size (5 cm in diameter) of the chamber system made it possible to use the chambers also in vegetation gaps, thereby avoiding the necessity of removing shoots, the disturbance of the spatial structure of vegetation and the upper soil layer. Low air flow rates associated with small chamber volume and chamber design allowed the overpressure range to stabilize between 0.05–0.12 Pa. While the correlation between ecosystem and soil CO2 efflux rates as measured by the independent methods was significant, Reco rates were similar or even lower than Rsc in the low flux (up to 2 μmol CO2 m−2 s−1) range, probably due to the larger than assumed storage flux. The gradient method showed both up and downward CO2 fluxes originating from the main rooting zone after rains. Downward fluxes within the soil profile amounted to 15% of the simultaneous upward fluxes and to ~ 7.6% of the total (upward) effluxes during the 3 months study. The upper 5 cm soil layer contributed to ~ 50% of the total soil CO2 efflux. The continuously operated automatic open chamber system and the gradient system makes possible the detection of situations when the eddy system underestimates Reco, gives the lower limit of underestimation (chamber system) and helps in quantifying the downward flux component of soil respiration (gradient method) between the soil layers. These latter (downward) fluxes are expected to seriously affect (1) the Reco vs. temperature response functions and (2) the net ecosystem exchange of CO2 (NEE) vs. photon flux density response functions, therefore potentially affecting also the gap filling procedures and to led to a situation (3) when the measured surface and the real time ecosystem fluxes will necessarily differ in the short term. Simultaneous measurements of Reco and soil CO2 effluxes may reveal the time and degree of the above decoupling, thereby contributing to decrease uncertainty, associated with eddy flux measurements over flat terrains. While the correlation between chamber fluxes and gradient fluxes was strong, gradient fluxes were generally larger than the flux from chambers. Calibration of gradient flux system by chamber effluxes is proposed.