Thermokarst landforms, particularly retrogressive thaw slumps (RTS), are proliferating across the Qinghai-Tibet Plateau (QTP), yet their ecosystem consequences remain poorly constrained. We integrated vegetation surveys, soil physicochemical analyses, and static-chamber CO2 flux measurements across four representative RTS and four surface types (control check (CK), disturbed ground (DG), vegetated raft (VR), and exposed (EX)) to test how slump-induced microtopography governs ecosystem development. RTS increased soil bulk density and re-sorted particles, altering near-surface moisture and redistributing substrates (soil organic carbon, SOC; alkali-hydrolyzable nitrogen). At 0–15 cm, fine particles increased in disturbed ground (+23.4 %) but decreased in exposed areas (−58.7 %), where gravel increased by approximately 347.1 % (p ≤ 0.05); soil bulk density rose most in exposed areas (+27.6 %); soil moisture increased in vegetated rafts (+44.4 %) but declined in exposed areas (−8%). These edaphic shifts favored hygrophilous assemblages (Kobresia royleana), redirecting communities from alpine meadow toward swamp meadow on disturbed ground. However, functional recovery lagged structural change: gross primary productivity (GPP) declined by ∼80 % in exposed and remained ∼25 % lower in disturbed ground and vegetated raft despite substantial vegetation cover. SOC and alkali-hydrolyzable nitrogen densities were tightly coupled (R2 = 0.92, p ≤ 0.001) and concentrated in surface soils. Partial least squares path modeling identified a dominant pathway from texture through moisture, substrate, and vegetation to GPP, in which vegetation exerted the strongest direct control on GPP and slump age provided a secondary direct effect. Overall, permafrost collapse accelerates ecosystem development via altered microtopography, expediting shifts toward hygrophilous communities while temporarily depressing carbon fixation until soils and vegetation stabilize. Incorporating patch-scale heterogeneity and slump age is therefore essential for forecasting QTP carbon dynamics.

https://doi.org/10.1016/j.catena.2025.109692