Fig. 5: Impacts of thermokarst landforms on CO₂ and CH₄ fluxes in alpine grasslands on the Tibetan Plateau.

a Conceptual representation of CO₂ and CH₄ fluxes associated with three major thermokarst landforms: retrogressive thaw slumps (RTS), thermo-erosion gullies (TEG) and thermokarst lakes (TL). Upward arrows indicate emission and downward arrows indicate uptake. Arrow lengths illustrate direction of change only and not flux magnitude. White stars mark statistically significant changes in gas flux. Flux patterns for TEG and TL are adapted from previous studies^35,50,52,53. b Conceptual trajectory for RTS effects on the carbon budget of alpine grasslands. RTS development comprises two stages: expansion and stabilization. During the expansion stage, total slump area increases rapidly together with the areal fraction of exposed, vegetation-free surfaces. These newly exposed patches are defined by bare ground lacking established vegetation communities, a relatively high proportion of particulate organic carbon (POC) in soil organic carbon (SOC), and a net carbon-source state. As exposed area expands, the site-level carbon budget shifts from a sink to a source. During the subsequent stabilization stage, the fraction of exposed area ceases to increase. Vegetation progressively colonizes bare ground and forms stable communities. Due to limited carbon replenishment before vegetation is fully established, the fraction of POC in SOC decreases significantly. With vegetation recovery and depletion of labile carbon (POC), emissions from exposed patches decrease and the site-level carbon budget trends back toward a sink. c Conceptual pattern illustrating the impact of TEG on the carbon budget of alpine grasslands. d Reported CO₂ and CH₄ fluxes datasets for major thermokarst landforms. Datasets for TEG and TL are from previous studies. For TEG, CH₄ fluxes are derived from the dataset obtained from July to August, 2016³⁵; CO₂ fluxes are derived from the dataset obtained from May to October, 2018⁵⁰. For TL, CO₂ diffusion fluxes and CH₄ diffusion fluxes are sourced from the dataset obtained from May to August, and September, 2020⁵³; CH₄ ebullition fluxes are sourced from the dataset obtained from June to October, 2020⁵².