Soil CO2 flux (Fs) can be measured either manually or automatically. While manual measurements are well suited to capture the spatial variability of Fs, automated measurements are able to capture its temporal variability at fine time scales. The manual method is the most commonly used method to estimate annual carbon budgets. However, such estimates can be biased depending on the measurement frequency, the time at which measurements are made, and the method used to interpolate Fs between two measurements. In this study, we investigated the effects of within-day measurement frequency and of the time of measurement on the estimation of daily Fs. We also investigated the effects on cumulative Fs estimates of weekly and fortnightly sampling frequencies over several months and of the interpolation method used to cumulate Fs. We based our analyses on two complete datasets of automated measurements (one 12-month and one 4-month) recorded in two contrasting ecosystems (a tropical eucalypt plantation and a temperate poplar plantation). Low-frequency time step within a day (every 360 min for the eucalypt and every 180 min for the poplar plantations) was sufficient to capture mean daily Fs accurately. Furthermore, in the tropical site, measurements averaged over any 6h period provided good estimates of the daily flux. By contrast, biases were observed in the temperate site. With one measurement per week, linear interpolation methods provided accurate cumulative fluxes at both sites. However, all interpolation methods failed to produce robust estimates of cumulative Fs in the temperate plantation with one measurement every two weeks. Automated measurements will help to select the best time slot for manual measurements or to correct manual measurements from the apparent deviation between measurements collected during the sampling period and the 24 h-mean CO2 flux. It will also be useful to elaborate empirical equations used to cumulate Fs obtained manually. Combining manual and