Complex behavior, associated with soil respiration of an oak-grass savanna ecosystem in California, was quantified with continuous measurements of CO2 exchange at two scales (soil and canopy) and with three methods (overstory and understory eddy covariance systems, soil respiration chambers, and a below-ground CO2 flux gradient system). To partition soil respiration into its autotrophic and heterotrophic components, we exploited spatial gradients in the landscape and seasonal variations in rainfall. During the dry summer, heterotrophic respiration was dominant in the senesced grassland area, whereas autotrophic respiration by roots and the feeding of microbes by root exudates was dominant under the trees. A temporal switch in soil respiration occurred in the spring. But the stimulation of root respiration lagged the timing of leaf-out by the trees. Another temporal switch in soil respiration occurred at the start of autumn rains. This switch was induced by the rapid germination of grass seed and new grass growth. Isolated summer rain storms caused a pulse in soil respiration. Such rain events stimulated microbial respiration only; the rain was not sufficient to replenish soil moisture in the root zone or to germinate grass seed. Soil respiration lagged photosynthetic activity on hourly scales. The likely mechanism is the slow translocation of photosynthate to the roots and associated microbes. Another lag occurred on daily scales because of modulations in photosynthesis and stomatal conductance by the passage of dry and humid air masses.