OBJECTIVE: This study evaluated the effects of extreme temperatures on hospital admissions for respiratory diseases (RD) in a semi-arid region in the Northwest of China. METHODS: Distributed lag non-linear model was constructed and stratified analysis by gender and age was performed. RESULTS: The exposure-response curve between temperature and RD hospital admissions was almost W-shaped. Either extremely cold temperatures or moderately cold temperatures presented a short-term acute harmful effect, and the relative risks were higher among males (1.976, 95%CI: 1.773-2.203; 1.242, 95% CI: 1.203-1.282) and the elderly (2.363, 95% CI: 1.724-3.240; 1.267, 95% CI: 1.154-1.390). Both extreme and moderately hot temperatures had higher risks among females (2.040, 95% CI: 1.815-2.292; 1.328, 95% CI: 1.276-1.381). CONCLUSIONS: The relationship between air temperature and RD hospital admissions was non-linear. Vulnerable populations varied according to extreme temperature conditions.