BackgroundNon-optimum ambient temperature has recently been acknowledged as an independent risk factor for disease burden, but its impact on atrial fibrillation (AF) episodes was rarely evaluated. ObjectivesTo examine the associations between ambient non-optimum temperature and symptom onset of AF episodes and calculate the corresponding disease burden. MethodsWe conducted an individual-level, time-stratified, case-crossover analysis based on a nationwide registry, which comprises of 94,711 eligible AF patients from 1993 hospitals in 322 Chinese cities from January 2015 to December 2021. Multiple moving 24 h average temperatures prior to the symptom onset of AF episodes were calculated as lag days. The associations were analyzed using conditional logistic regression combined with distributed lag non-linear models with a duration of lag 0–7 days, after controlling for criteria air pollutants. Stratification analyses were performed to explore possible effect modifiers. ResultsThere was a monotonically increasing relationship of AF onset risk with decreasing temperature. The excess AF risk occurred at lag 1 d and lasted for 5 days. Nationally, the cumulative relative risk of AF episode onset associated with extremely low temperature (−9.3 °C) over lag 0–7 d was 1.25 (95 % confidence interval: 1.08, 1.45), compared with the reference temperature (31.5 °C). The exposure-response curve was steeper in the south than in the north where there was levelling-off at lower temperature. Nationally, 7.59 % of acute AF episodes could be attributable to non-optimum temperatures. The attributable fraction was larger for southern residents, males and patients <65 years. ConclusionThis nationwide study provides novel and robust evidence that declining ambient temperature could increase the risk of AF episode onset. We also provide the first-hand evidence that a considerable proportion of acute AF episodes could be attributable to non-optimum temperatures.