关键词: 机器学习, 氮输入, 碳氮化学计量, 植物生产力, 植物茎, 湿地生态系统

Abstract: Nitrogen (N) input is one of the key global change contributors that has profound effects on the carbon (C) and N cycling of wetland ecosystems. However, the information is very limited on the response patterns of wetland productivity to N input-induced changes in elemental stoichiometric composition. Here we investigated the effects of global N input on the stoichiometry and productivity of wetland ecosystems using 103 individual data points from 74 studies. The results showed that global N input significantly enhanced wetland aboveground and belowground productivity by 45.5% and 21.7%, respectively. N input significantly altered C and N content of plant, soil and microbes: C content increased significantly in plant roots (4.9%), soil (4.8%), and microbes (29.3%), but decreased significantly in plant stems (0.8%); also, C content in plant leaf tissue showed no significant change. While N content increased significantly in all components. N input generally enhanced wetland productivity and reduced C:N ratios across ecosystem components, but its effect intensity was modulated by multiple environmental factors. More critically, statistical analysis revealed that changes in C/N stoichiometry in plant stems—rather than in plant leaves or roots—constituted the core mechanism linking N input to wetland productivity responses. This mechanism explained 31.0% and 56.3% of the variation in above-ground and below-ground productivity, respectively. Our meta-analysis shows that plant stems stoichiometry under N input is key factor to wetland productivity. These responding processes contribute to a better understanding of the N input induced changes in wetland productivity, and improve ecosystem modeling.

Key words: machine learning, nitrogen input, C/N stoichiometry, plant productivity, plant stem, wetland ecosystem