Soil respiration is one of the main processes of loss of organic C, but few studies so far have evaluated losses from arid and semiarid soils as compared with other ecosystems. Spatial variation in CO 2 efflux is especially important in semiarid areas, where the distribution of ecological factors and organisms is markedly patchy. We have evaluated the small-scale spatial variation of soil respiration in a semiarid Stipa tenacissima steppe from late winter to early summer. We selected the five most frequent soil cover types in the study site: upslope and downslope of S. tenacissima tussocks (SU and SD surface units, respectively), patches of the perennial grass Brachypodium retusum (BR surface unit), biological crusts (BC surface unit), surface earthworm casts (EC surface unit) and bare ground (BG surface unit). The dynamics of soil CO 2 efflux clearly reflected the changes in soil moisture. We observed significant differences in CO 2 efflux between surface units, with BR and BG showing maximum and minimum CO 2 efflux rates, respectively. Soil temperature increased during the course of the study period, showing significant differences between the most shaded (SU and SD) and the other surface units (BC, BG, BR and EC), with higher temperatures in the latter group. Soil CO 2 efflux was poorly correlated with soil temperature, but significant non-linear relationships were found for SU, BR and BC surface units. Rainfall simulation significantly enhanced CO 2 efflux in SU, BR and EC surface units. Our results highlight the importance of the spatial variation of both vegetation and surface soil features for the estimation of CO 2 efflux in semiarid ecosystems. They also suggest that estimations based only on vegetated and bare-ground areas can be rather conservative if the sampling effort is not high enough to capture the spatial variability in CO 2 efflux in the latter areas.