Objective The rapid expansion of monoculture rubber plantations in tropical regions has led to severe soil degradation and ecosystem nitrogen limitation. Biological nitrogen fixation plays a crucial role in alleviating nutrient constraints in degraded systems; however, the dynamics and assembly mechanisms of its key functional microbial communities during rubber plantation restoration remain poorly understood. This study aimed to reveal the diversity patterns and assembly mechanisms of soil diazotrophic communities during restoration and to compare the differences between natural restoration and human-assisted restoration pathways.

Method Using a space-for-time substitution approach, four forest types were investigated in Xishuangbanna, China: rubber monoculture (RM), naturally restored forest (NF), artificially restored forest (AF), and mature tropical rainforest (MF) as a reference. Based on high-throughput sequencing of the nifH gene, we classified different ecological groups according to abundance thresholds, and combined β diversity decomposition with community assembly models to systematically analyze the shifts in diversity patterns and assembly mechanisms of soil diazotrophic communities during rubber plantation restoration.

Result (1) The community structures of diazotrophs in restored forests (NF and AF) were significantly different from both RM and MF, with a clear divergence observed between NF and AF; (2) β diversity decomposition revealed that species turnover was the dominant process driving compositional variation among diazotrophic communities, and conditionally rare taxa (CRT, abundance < 1% with partial samples < 0.01%) and absolutely rare taxa (ART, abundance < 0.01% in all samples) were highly consistent with the turnover pattern of the overall community, suggesting that these taxa might be important contributors to the high turnover of the overall community; (3) The assembly of diazotrophic communities is predominantly governed by stochastic processes (ecological drift and dispersal limitation), with variations observed across different restoration strategies and ecological groups. Specifically, the contribution of stochastic processes increased in AF but decreased in NF for overall taxa, CRT, and ART, while it decreased for dominant taxa in both AF and NF. pH, total soil carbon (TC), and total soil nitrogen (TN) were identified as key environmental factors driving the differentiation of different ecological groups, with CRT exhibiting stronger environmental sensitivity than other groups, indicating their potential as environmental indicators.

Conclusion Restoration of rubber monocultures drives the restructuring of soil diazotrophic communities, with assembly mechanisms exhibiting differentiated responses. Our findings suggest that restoration practices should pay attention to the long-term impacts of different restoration strategies on community assembly processes, and we recommend adopting an integrated restoration strategy of “initial artificial facilitation followed by natural succession” to balance recovery efficiency with long-term ecosystem stability and resilience. This study reveals the differential response characteristics of diazotrophic community assembly during rubber plantation restoration, providing a scientific basis for function-oriented restoration of degraded ecosystems.