In order to analyze the environmental drivers of soil respiration in an extreme arid desert ecosystem, we measured diurnal variation in winter soil respiration at Tazhong, a hinterland of Taklimakan Desert in northwest China. Regression analysis was performed with SPSS 21.0. We observed that:(1) Diurnal variation in winter soil respiration showed a single peak at 12:00 noon (local time), after which soil respiration began to decrease, reaching a minimum value at around 4:00 a.m. (2) Soil respiration and the air temperature at each height tested (0.5 m, 2 m) were significantly and positively correlated. Air temperature at 2 m was able to explain 67.8% of the diurnal variation in soil respiration. (3) Soil temperature at 0 cm, modeled by linear equations, was able to explain 86.3% of the diurnal variation in soil respiration, demonstrating that this process is more sensitive to temperature at 0 cm than at any other soil layer (10 cm, 20 cm, 40 cm). (4) Soil respiration exhibited a positive linear correlation with soil moisture at a depth of 5 cm. When linear regression analysis was used to model the relationship between these variables, the fitted linear model explained 69.5% of the diurnal variation in soil respiration, demonstrating that, in the extreme arid desert ecosystem, this shallow layer of moisture exerts a large effect on soil respiration. (5) The greatest contributors to soil respiration were soil temperature at a depth of 0 cm, followed by soil moisture at 5 cm. (6) Multiple regression analyses showed that a multi-variable model of temperature and soil moisture explains 86.9% of the diurnal variation in soil respiration, which is not significantly better than a single-variable model. (7) For winter soil respiration, the daily average rate of CO₂ absorption was -1.45mg CO₂ m^-2 h^-1.