This paper describes a new method to monitor continuously soil CO2 profiles using small solid-state CO2 sensors buried at different depths of the soil. Based on the measurement of soil CO2 profile and a gaseous diffusivity model, we estimated soil CO2 efflux, which was mainly from heterotrophic respiration, and its temporal variation in a dry season in a Mediterranean savanna ecosystem in California. The daily mean values of CO2concentrations in soils had small variation, but the diurnal variation was significant and correlated well with soil temperature. The daily mean CO2 concentration remained steady at 396 μmol mol−1 at 2 cm depth during the dry summer from days 200 to 235 in 2002. Over the same period, CO2 concentration decreased from 721 to 611 μmol mol−1at 8 cm depth, and from 1044 to 871 μmol mol−1 at 16 cm. The vertical soil CO2concentrations changed almost linearly with depth up to 16 cm, but the gradient varied over time. Based on the soil CO2 gradient and the diffusion coefficient estimated from the Millington–Quirk model, continuous soil CO2 efflux was calculated. The daily mean values of CO2 efflux slightly decreased from 0.43 to 0.33 μmol m−2 s−1 with a mean of 0.37 μmol m−2 s−1. The mean diurnal range of CO2 efflux was greater than the range of daily mean CO2 efflux within the study period. The diurnal variation of soil CO2 efflux ranged from 0.32 to 0.45 μmol m−2 s−1 with the peak value reached between 14:30 and 16:30 h. This pattern corresponded well with the increase in soil temperatures during this time. By plotting CO2 efflux vs. soil temperature, we found that CO2 efflux correlated exponentially with soil temperature at the depth of 8 cm, with R2 of 0.86 and Q10 of 1.27 in the summer dry season. The Q10 value increased with the depth of soil temperature measurements. The high correlation between CO2 efflux and temperature explains the diurnal pattern of CO2 efflux, but moisture may become another factor driving the seasonal pattern when moisture changes over seasons. The estimated CO2 efflux using this method was very close to chamber measurements, suggesting that this method can be used for long-term continuous measurements of soil CO2 efflux.