The automated soil CO₂/gas flux system (model LI-8100A) was used to continuously observe the soil respiration rate, soil temperature and water content in a semi-arid grassland of Loess Plateau during the growing season (from May to October). The temporal variations of soil respiration were comprehensively analyzed, and the influences of environmental hydrothermal factors on soil respiration were studied. The results are as follows: (1) The diurnal dynamic variations of soil respiration rate were significantly different under diverse weather conditions, and the daily mean value (2.90 µmol·m^-2·s^-1) and variation range (1.73-4.92 µmol·m^-2·s^-1) on sunny days were obviously greater than those on cloudy and overcast days. The average diurnal variations of soil respiration rate in different months all showed a significant unimodal pattern, with the maximum (2.20-4.40 µmol·m^-2·s^-1) appearing at 12:00 BST or 13:00 BST and the minimum (0.71-1.70 µmol·m^-2·s^-1) appearing at 05:00 BST or 06:00 BST. The daily mean values were close to the observed values at 10:00 BST or 19:00 BST. (2) The soil respiration rate in both daytime and nighttime were low in May and June, and increased gradually from June, then reached the peak value (in daytime 3.31±0.98 µmol·m^-2·s^-1, in nighttime 1.80±0.39 µmol·m^-2·s^-1) in August, afterwards gradually decreased. Furthermore, the lowest value occurred in October (in daytime 1.55±0.55 µmol·m^-2·s^-1, in nighttime 0.81±0.12 µmol·m^-2·s^-1), and soil respiration rates in daytime were always higher than those in nighttime. The nocturnal soil respiration contributed 27.2%-32.4% to the total respiration during the whole growing season. Therefore, the effects of nocturnal soil respiration should be considered in the current carbon cycle models of grassland ecosystem. (3) Soil temperature was the main environmental factor affecting soil respiration rate, but the univariate model of soil temperature was not enough to fully explain the diurnal dynamic changes of soil respiration. The bivariate nonlinear model combined with soil temperature and water content could be better fitted the soil respiration rate and accounted for 74.0% of the variation. (4) The temperature sensitivity indices (Q₁₀) of the entire day, daytime, and nighttime throughout the growing season in 2020 varied from 1.38 to 2.14, from 1.22 to 1.96, and from 0.85 to 1.64, respectively, with the corresponding mean values of 1.58±0.23, 1.41±0.19 and 1.20±0.16. This suggested that replacing the daily mean values of Q₁₀ with daytime averages would result in an underestimation of about 10.8%.