Previous studies have reported that daily average temperature is connected with respiratory diseases (RD), but proof is limited for the influence of the extreme low temperature on RD in Lanzhou, a northwestern China of temperate area. Generalized additive model (GAM) was built in this work to describe the relationship between daily mean temperature and RD in Lanzhou, China from 2012 to 2017. The results indicated that the exposure-response curve was inverse J-shaped, showing the lower the temperature, the larger the relative risk (RR). The RR of daily emergency room (ER) admissions in P5 extreme low temperature (the temperature below the fifth percentile, etc.) was larger than that in P10. The P5 extreme low temperature has the strongest effect at lag 0, and the RRs were 1.043 (95% CI: 1.030, 1.055) for the total, 1.031 (95% CI: 1.015, 1.046) for males and 1.058 (95% CI: 1.039, 1.077) for females. For different age groups, the largest RRs were 1.026 (95% CI: 1.013, 1.039) for the children (age < 16 years) at lag 5, 1.057 (95% CI: 1.030, 1.085) for the young adults (aged 16-45 years), 1.060 (95% CI: 1.023, 1.099) for the middle-aged (aged 46-60 years) and 1.121 (95% CI: 1.077, 1.166) for the elderly group of age > 60 years. Meanwhile, females and the elderly were more vulnerable to extreme temperature. The results could strengthen the scientific evidence of effects of extreme low temperature on RD in temperate areas.