Much stronger tundra methane emissions during autumn freeze than spring thaw.

Warming in the Arctic has been more apparent in the non-growing season than in the typical growing season. In this context, methane (CH4 ) emissions in the non-growing season, particularly in the shoulder seasons, account for a substantial proportion of the annual budget. However, CH4 emissions in spring and autumn shoulders are often underestimated by land models and measurements due to limited data availability and unknown mechanisms. This study investigates CH4 emissions during spring thaw and autumn freeze using eddy covariance CH4 measurements from three Arctic sites with multi-year observations. We find that the shoulder seasons contribute to about a quarter (25.6 ± 2.3%, mean ± SD) of annual total CH4 emissions. Our study highlights the three to four times higher contribution of autumn freeze CH4 emission to total annual emission than that of spring thaw. Autumn freeze exhibits significantly higher CH4 flux (0.88 ± 0.03 mg m-2  hr-1 ) than spring thaw (0.48 ± 0.04 mg m-2  hr-1 ). The mean duration of autumn freeze (58.94 ± 26.39 days) is significantly longer than that of spring thaw (20.94 ± 7.79 days), which predominates the much higher cumulative CH4 emission during autumn freeze (1,212.31 ± 280.39 mg m-2  year-1 ) than that during spring thaw (307.39 ± 46.11 mg m-2  year-1 ). Near-surface soil temperatures cannot completely reflect the freeze-thaw processes in deeper soil layers and appears to have a hysteresis effect on CH4 emissions from early spring thaw to late autumn freeze. Therefore, it is necessary to consider commonalities and differences in CH4 emissions during spring thaw versus autumn freeze to accurately estimate CH4 source from tundra ecosystems for evaluating carbon-climate feedback in Arctic.