Winter carbon (C) flux is important for annual C balance over an ecosystem level, but the field measurements are still lacking, especially in alpine grasslands which occupy approximately one third of the Tibetan Plateau area. Consequently, a determination of winter carbon flux is essential for the assessment of alpine grassland carbon budget. Previous studies showed that temperature coefficient (Q₁₀), an index of temperature sensitivity for respirations, increased with decreasing temperature in other ecosystem. This research therefore hypothesized that Q₁₀ of soil and ecosystem respiration in alpine meadows would be higher in winters than in summers and higher in cold winters than in warm winters. To test the hypothesis above, ecosystem and soil CO₂ fluxes were measured by a static chamber-chromatography method in an alpine Kobresia humilis meadow in two continues winters from 2003 to 2005 at Haibei Alpine Meadow Ecosystem Research Station of Chinese Academy of Sciences. The objectives of this study were to: 1) clarify the characteristics of winter soil and ecosystem CO₂ fluxes and their control factors, and 2) compare temperature sensitivity of CO₂ fluxes under different thermal regimes. Results showed that both ecosystem and soil respiration showed clear diurnal and seasonal patterns. Temperature mainly controlled ecosystem and soil respiration, explaining more than 44% of the respiration variance. Ecosystem respiration rates averaged 88.43 mgCO₂ · m^-2 · h^-1 (ranging from 51.63 to 206.07 mgCO₂ · m^-2 · h^-1) in the first winter and 89.50 mgCO₂ · m^-2 · h^-1 (ranging from 35.12 to 145.17 mgCO₂ · m^-2 · h^-1) in the second winter. By comparison, soil respiration rates varied from 47.41 to 152.94 mgCO₂ · m^-2 · h^-1 (with mean of 77.01 mgCO₂ · m^-2 · h^-1) in the first winter and ranged from 28.21 to 107.89 mgCO₂ · m^-2 · h^-1 (with mean of 68.64 mgCO₂ · m^-2 · h^-1) in the second winter. Because the aboveground parts were dead in winters, no significant difference was observed between ecosystem and soil respirations. Soil respiration accounted for more than 85% of ecosystem respiration, indicating that soil respiration composed the main C loss from alpine meadow ecosystems in winters. Our estimate showed soil respiration could release 86.9 gC/m² in one winter, which approximately counteracted 15% carbon fixed by alpine plants in the growing season. Higher ecosystem respiration than soil respiration in winters could be ascribed to additional decomposition of dead litter above the ground. The Q₁₀ values of ecosystem respiration in the two winters were 1.86 and 1.53, respectively. By comparison, the Q₁₀ values of soil respiration were 1.68 and 1.38, respectively. The Q₁₀ values of ecosystem respiration were higher than those of soil respiration in both winters. This is mainly due to that aboveground standing or falling litter comprised more labile carbon, such as sugars, starch, etc, resulting in higher substrate-availability for ecosystem respiratory activities than soil. Additionally, the Q₁₀ values of respirations were higher in the growing season than in the winter and higher in warm winter than that in cold winter, not supporting our hypothesis. This indicates that the Q₁₀ was influenced not only by temperature, but also by substrate availability and microbial activities.