Aim The temperature sensitivity of labile (Q(10L)) and recalcitrant (Q(10R)) pools of soil organic carbon (SOC) decomposition is a critical for predicting soil carbon (C) fluxes. Methods Soils under six grass covers, namely, Cenchrus ciliaris, Chrysopogon fulvus, Panicum maximum, Sehima nervosa, Heteropogon contortus, and Vetiveria zizanioides from semi-arid India were evaluated for Q(10L) and Q(10R) of bulk soil, macroaggregates, microaggregates, and silt + clay associated SOC. Soil fractions (from 0-20 and 21-40 cm depths) were incubated at 25, 32, and 37 degrees C for 100 days, and cumulative C mineralization was measured. The Q(10L) and Q(10R) were calculated using a two-pool decay model. The quality of root litter C and SOC was assessed through FTIR spectroscopy. Results Q(10L) and Q(10R) of microaggregate-C was significantly higher (22-64% and 22-24%, respectively) than macroaggregate-C and silt + clay-C. Among grasses, Q(10L) and Q(10R) values under C.ciliaris, H.contortus, and S. nervosa were lower (by 6-35%) than other grasses, indicating their capability to store SOC under global warming scenarios. Q(10L) at topsoil layer was correlated with root litter C quality (r = -0.641 to 0.633) and at the sub-surface soil layer, it was influenced by labile C concentration (r > 0.637). The Q(10R) was correlated with the recalcitrant C concentration (r > 0.721) and SOC quality in both soil layers, indicating that quality and availability of recalcitrant SOC had pivotal roles in governing Q(10R) in restored land. Conclusions Soil C and litter C quality should be potentially incorporated into the biogeochemical models to better predict SOC dynamics in managed ecosystems in the context of global warming and land use changes.