The degradation of alpine cushion plants is known to disrupt soil microbial networks. To explore the mechanisms behind this, we collected soils across a degradation sequence on the Qinghai-Xizang Plateau and, using a microcosm approach, simulated the effects of three global change factors: warming (T), nitrogen addition (N), and dry-wet cycling (D). Our findings indicated that fungal and bacterial networks exhibited contrasting structures and response patterns. Fungal networks displayed hub-dependent structure: they maintained complexity and enhanced robustness under T, N, and D, yet showed vulnerability when keystone nodes were preferentially removed. Bacterial networks exhibited connector-mediated redundancy that conferred high baseline robustness but limited adaptive capacity; their robustness did not increase under T or N, though assortativity rose under N. Network complexity showed nonlinear shifts across degradation stages. At two thresholds—individual-level Stage 3–4 and community-level balanced-to-stable transition—fungal networks reorganized while bacterial networks simplified. Both abiotic and biotic factors predicted network dynamics. Abiotic predictors included microbial biomass, nutrients, pH, and polyphenol oxidase (PPO) activity. Beyond these, keystone taxa abundance emerged as a biotic driver strongly correlated with bacterial network complexity, though this relationship was attenuated under warming. We propose a management framework that prioritizes protecting bacterial network integrity through mitigating warming and nitrogen deposition, targets pre-threshold stages as intervention windows, and integrates soil-microbial indicators for early warning. This study provides a practical basis for predicting and managing microbial network dynamics in alpine tundra under global change. • Fungal networks remained stable; bacterial networks were vulnerable. • Nonlinear thresholds occurred at key degradation stages. • Initial soil properties predicted microbial responses. • A vulnerability-focused framework is proposed for ecosystem management.
Yin et al. (Fri,) studied this question.