Abstract
Thermokarst lakes, serving as significant sources of methane (CH(4)), play a crucial role in affecting the feedback of permafrost carbon cycle to global warming. However, accurately assessing CH(4) emissions from these lakes remains challenging due to limited observations during lake ice melting periods. In this study, by integrating field surveys with machine learning modeling, we offer a comprehensive assessment of present and future CH(4) emissions from thermokarst lakes on the Tibetan Plateau. Our results reveal that the previously underestimated CH(4) release from lake ice bubble and water storage during ice melting periods is 11.2 ± 1.6 Gg C of CH(4), accounting for 17 ± 4% of the annual total release from lakes. Despite thermokarst lakes cover only 0.2% of the permafrost area, they annually emit 65.5 ± 10.0 Gg C of CH(4), which offsets 6.4% of the net carbon sink in alpine grasslands on the plateau. Considering the loss of lake ice, the expansion of thermokarst lakes is projected to lead to 1.1-1.2 folds increase in CH(4) emissions by 2100. Our study allows foreseeing future CH(4) emissions from the rapid expanding thermokarst lakes and sheds new lights on processes controlling the carbon-climate feedback in alpine permafrost ecosystems.