Determining regional and global carbon cycles hinges on investigating the dynamic characteristics and influencing factors of soil respiration in various types of natural grasslands located in arid regions, and these characteristics are important indicators for assessing the structural and functional health of grassland ecosystems. Such investigations also provide theoretical support for carbon sink monitoring, energy conservation, emission reduction and low-carbon development in the western arid zone and are important for obtaining an in-depth understanding of the carbon cycle, as well as for ecosystem management, restoration and the reconstruction of arid areas. In this study, during the growing season (from May to October) of 2022, the LI-8100A automated soil CO2 flux system was used to measure the soil respiration rate (Rs), temperature from 1.5 m above the surface to depths of 5–25 cm (T, T5, T10, T15, T20 and T25) and the soil moisture content (SM) at a depth of 20 cm in four types of grasslands: lowland meadow, alpine meadow, temperate desert steppe and temperate steppe desert. Five replicates were established for each plot, and the responses of Rs to T and SM were fitted to construct the optimal regression model. The results revealed that (1) the daily average soil respiration was highest in the lowland meadow (0.07 to 5.76 μmol·m−2·s−1), followed by the alpine meadow (−0.57 to 0.95 μmol·m−2·s−1), the temperate desert steppe (−0.45 to 3.0 μmol·m−2·s−1) and the temperate steppe desert (−1.29 to 1.61 μmol·m−2·s−1); (2) the soil respiration rates of the four grassland types were significantly correlated with the temperature in the 5–15 cm soil layer, and the best model was an exponential function; the peak values generally appeared between 13:00 and 17:00 (h), with the minimum values at 2:00 or 8:00 (h); the maximum value was observed in July–August, and the minimum value was observed in October; and the soil respiration in the lowland meadow was higher than that in the other three types of grassland during the same period. The average variation intensities of the soil respiration from May to October were as follows: temperate steppe desert (91.78%) > temperate desert steppe (76%) > alpine meadow (58.77%) > lowland meadow (43.93%). (3) The partial correlation analysis revealed that when soil temperature was used as a control, the correlation between SM and soil respiration in the four types of grasslands changed, and the coefficient of determination (R2) increased to varying degrees, explaining up to 80% of the variation in the soil respiration in the lowland meadows. The correlation between soil respiration and the SM normalized to 10 °C explained up to 93.8% of the variation in soil respiration; the two-factor fitting equations revealed that the model with soil temperature and SM was superior to the single-factor model with either soil temperature or SM.
