Prediction of ecosystem responses to multifactor global changes in a future world strongly relies on our understanding of their interactions. Our study showed that among the three factors that we examined in our experiments, warming and doubled precipitation had significant main effects on soil CO2 efflux, whereas the main effect of clipping was significant only in the transient study. The interactive effects of the three factors were not significant except for warming × clipping in the transient study. The temperature sensitivity of soil CO2 efflux significantly decreased under the warming treatment in Experiment 2 and under the clipping treatment in the transient study. Below we discuss magnitude of soil CO2 efflux, main effects of single factors, and interactive effects of multiple factors.
4.1. Magnitude of Soil CO2 Efflux
 Soil CO2 efflux measured in the control plots ranged from 0.52 to 7.64 μmol m−2 s−1, which is comparable to previous measurements in grasslands [Bremer et al., 1998; Wan and Luo, 2003]. Although annual soil CO2 efflux is not the main focus of this study, our estimates are consistent with the studies on Konza Prairie [Bremer et al., 1998] but greater than estimates of 340 to 480 g C m−2 yr−1 from less productive grasslands in California [Luo et al., 1996]. Overall, our estimates fall within the upper limits of the estimates which range from 160 to 1060 g C m−2 yr−1 in North America and Europe [Hanson et al., 1993]. The difference in annual precipitation (890 mm in 2002 and 647 mm in 2003) likely contributed to the significant difference in annual soil CO2 efflux between 2002 and 2003 in the control plots of Experiment 2 (Table 4).
4.2. Main Effects of Single Factors on Soil CO2 Efflux
 The increase in soil CO2 efflux in response to warming has been observed in various ecosystems [Rustad et al., 2001]. The short-term response to warming in Experiment 2 is similar to those observed in a temperate forest [McHale et al., 1998] and a boreal pine forest [Niinistö et al., 2004]. The observed increase of soil CO2 efflux in our study is 0.74 μmol m−2 s−1, which is slightly lower than the mean increase of 1.20 μmol m−2 s−1 in the first-year warming from a meta-analysis of 17 ecosystem warming experiments [Rustad et al., 2001]. The increased respiration likely resulted from enhanced oxidation of labile soil carbon compounds on warmed plots [Peterjohn et al., 1993; Lin et al., 2001].
 The long-term response of soil CO2 efflux to warming is regulated by acclimatization [Luo et al., 2001], physiological adjustments to pool size changes by plants and microbes [Melillo et al., 2002], extension of growing seasons [Dunne et al., 2002; Wan et al., 2005], and stimulated C4 plant productivity [Wan et al., 2005]. In Experiment 1, soil CO2 efflux increased by 9.9% in the fourth year (Figure 2), by 8.0% and 15.6% in the third and second year, respectively [Wan et al., 2005], and decreased by 5% in the first year [Luo et al., 2001]. The increases in soil CO2 efflux observed in this study are lower than the 20% mean increase reported from a meta-analysis [Rustad et al., 2001]. The meta-analysis synthesized studies mainly from high latitude regions. The year-to-year variation in warming-induced changes in soil CO2 efflux observed in Experiment 1 likely resulted from changes in productivity [Wan et al., 2005] and other abiotic factors such as drought. The lower response of soil CO2 efflux to warming observed in our experiments is likely related to the fact that our grassland has lower soil organic C content than other ecosystems [Luo et al., 2001].
 This study demonstrated that warming significantly increased basal respiration rate (coefficients a) and decreased temperature sensitivity of soil CO2 efflux (coefficient b) in Experiment 2, whereas neither of the parameters was significantly altered by warming in Experiment 1 (Table 3). The different responses of the two parameters to warming between the experiments may be due to a few reasons. First, the temperature increase was ∼ 2°C in Experiment 1 and 4.4°C in Experiment 2. Thus the experimental forcing was stronger in Experiment 2 than in Experiment 1. Second, Experiment 1 was in the fourth year. Ecosystem processes may adjust to warming treatment over time [Melillo et al., 2002]. After 3-year warming in Experiment 1, labile carbon could be in a steady state between supply and depletion (A. Tedla and Y. Luo, unpublished data, 2003). In addition, the shift in soil microbial community structure toward more fungi [Zhang et al., 2005] likely resulted in lower sensitivity of soil CO2 efflux to temperature because fungi are more tolerant of higher soil temperature and drying owing to their filamentous nature. The opposite responses of coefficients a and b to warming could result from changes in root phenology and acclimation of roots and microbes to climate [Janssens and Pilegaard, 2003].
 Doubled precipitation significantly increased soil CO2 efflux in Experiment 2 (Table 2), greatly owing to stimulation of soil CO2 efflux in the dry growing season of 2003 (Figure 3). Similar effects of additional water on soil CO2 efflux have been observed in other experiments [Laporte et al., 2002; Liu et al., 2002]. During the period of the transient study, CO2 efflux from soils was not significantly affected by doubled precipitation owing to the absence of water stress. Although the basal respiration rate and temperature sensitivity were not affected by doubled precipitation (Table 3), the apparent Q10 value in the control was significantly higher in 2003 than 2002 (p < 0.05), largely resulting from differences in precipitation. Dörr and Münnich  found that the apparent Q10 values were low in the wet years and high in the dry years in a multiyear study of a grassland and a beech-spruce forest in Germany. However, others found that the apparent Q10 values were lower in the well-drained sites than the wetter sites [Davidson et al., 1998; Xu and Qi, 2001]. Complex interactions between soil water and temperature, which influence CO2/O2 diffusion, root and microbial activities, could result in these diverse responses of the temperature sensitivity of soil CO2 efflux to water availability.
 A large portion of soil CO2 efflux is derived from recently fixed carbon, thus making it responsive to changes in carbon supply due to clipping, girdling, and shading [Craine et al., 1999; Högberg et al., 2001; Wan and Luo, 2003]. Clipping reduces soil CO2 efflux by 19% to 49% in grassland ecosystems [Bremer et al., 1998; Craine et al., 1999; Wan and Luo, 2003]. Our study showed that yearly clipping had no significant effects on soil CO2 efflux in the fourth year of Experiment 1 and clipping significantly reduced soil CO2 efflux in the transient study within two months (Figures 1 and 5 and Table 2). In Experiment 1, we evaluated the effect of yearly clipping against monthly measurements of soil CO2 efflux over a whole year. The treatment of yearly clipping in our study likely has less impact on soil CO2 efflux than mowing several times per year. However, the transient effects of clipping were examined within 2 months in the transient study. In addition, Wan and Luo  kept clipping aboveground biomass to maintain bare ground in the clipped plots during the whole study period of one year, leading to a 33% decrease in mean soil CO2 efflux. Thus frequency of clipping and durations of study can be sources of variable results. Our transient study showed that clipping significantly reduced respiratory sensitivity to temperature (Table 3), similar to the results in other studies both from the laboratory [Townsend et al., 1997] and field experiments [Boone et al., 1998; Wan and Luo, 2003].
4.3. Interactive Effects of Warming, Precipitation, and Clipping
 Global climate change in the real world involves changes in multiple factors [Shaw et al., 2002; Norby and Luo, 2004]. Therefore the effects of warming on terrestrial ecosystems must be evaluated in combination with other factors. In this study, we found that interactive effects of warming, precipitation, and clipping on soil CO2 efflux were minor except for the warming × clipping interaction in the transient study. Minor interactive effects among multiple global change factors on soil CO2 efflux have been reported in the literature. For example, Edwards and Norby  and Niinistö et al.  did not find interactive effects of elevated CO2 and temperature on soil CO2 efflux statistically significant. Similarly, there were no significant interactions among elevated CO2, nitrogen supply, and plant diversity on soil CO2 efflux [Craineet al., 2001] and between elevated CO2 and O3 [Kasurinen et al., 2004]. However, significant interactive effects of elevated CO2 and warming were found on “old” pool C decomposition in a warming-CO2-N experiment in tunnels with ryegrass swards [Loiseau and Soussana, 1999]. The interaction was largely regulated by N supply.
 The lack of significant interactive effects in Experiment 1 suggest that soil CO2 efflux was determined by warming and yearly clipping treatments in a statistically independent manner. Warming increased soil CO2 efflux while yearly clipping decreased it. The effect size of the warming plus yearly clipping treatment was between that of the warming treatment and the one of the yearly clipping treatment. The insignificant interaction between warming and doubled precipitation in Experiment 2 resulted largely from the anomalously low precipitation in 2003. Precipitation was 647 mm, which was 29.3% less than the average (915 mm). The long period of drought in June and July (34 days without rain) negated the doubled precipitation treatment. A heavy rain of 108.0 mm in two days on 30–31 August 2003 resulted in substantial water loss through surface runoff. Although doubled precipitation increased soil water content by 10.6% and soil CO2 efflux by 9.0% relative to those without extra precipitation treatments, high variability in rainfall events in our ecosystem did not generate statistically significant interaction. In addition, our monthly measurements may not detect fast transient responses of soil CO2 efflux to individual rainfall events [Liu et al., 2002]. Thus we do expect that soil water content and temperature interactively regulate soil CO2 efflux under different circumstances in spite of the fact that we did not detect significant interactions between them in this particular study.
 An interactive response to warming and clipping was observed on soil CO2 efflux and its temperature sensitivity in the transient study (Tables 2 and 3). Clipping immediately reallocated assimilate to regrowth of shoots [Bremer et al., 1998; Craine et al., 1999] and reduced supply of current photosynthates to roots and their mycorrhizal fungi [Högberg et al., 2001]. As a consequence, soil respiration decreases. However, experimental warming accelerated plant regrowth in comparison with that in unwarmed plots after clipping either with or without doubled precipitation. Thus warming made soil CO2 efflux more responsive to clipping, contributing to the observed significant interaction during the transient period. In addition, complex and unpredictable interactions do occur in regulating soil CO2 efflux in other ecosystems [Loiseau and Soussana, 1999] or other ecosystem attributes such as biomass growth [Shaw et al., 2002]. A mechanistic understanding of interactions of warming and other global change factors on soil CO2 efflux also requires study of root and microbial processes, which may have different sensitivities to temperature and other factors in complex soil physical and chemical environments.