作者：Zhuang, X., Liu, X., Xu, S., Wang, X., Shah, A., Wang, L., Wu, S., Jiang, C., Ouyang, Z. & Piao, S.

Human activities amplify climate-induced greenhouse gas emissions from small water bodies (SWBs), creating critical but unquantified feedback in the global carbon cycle. Here, by training machine learning models on 470 field observations and upscaling to a global database of 3.28 million water bodies, we quantify this human amplification, which drives SWBs to emit 84.5 Tg CO2 y-1 and 11.0 Tg CH4 y-1, a disproportionate share of total inland water emissions (15% of CO2 and 28% of CH4) from only 6% of Earth's surface area. This amplification is primarily fueled by agricultural nutrient loading and land use intensity, which elevate CH4 fluxes in agricultural catchments five times higher than those in forested systems. Future projections show this synergy will increase emissions by up to 30% (CO2) and 14% (CH4) by 2100 under SSP5-8.5, whereas sustainable pathways (SSP1-2.6) could mitigate this emission acceleration through nutrient mitigation efforts, a largely neglected feedback process in current climate change assessments.

人类活动放大了小型水体（SWBs）中气候诱导的温室气体排放，在全球碳循环中形成了关键但尚未量化的反馈。本研究基于470个野外观测样本训练机器学习模型，并将其上推至包含328万个水体的全球数据库，定量揭示了这种人类放大效应：SWBs排放CO₂ 84.5 Tg·y⁻¹、CH₄ 11.0 Tg·y⁻¹，仅占地球表面积的6%，却贡献了内陆水体温室气体总排放的“超比例份额”（CO₂的15%与CH₄的28%）。这一放大效应主要由农业营养盐负荷与土地利用强度驱动：农业集水区的CH₄通量达森林系统的5倍。未来预估显示，在SSP5-8.5情景下，这种协同效应将使排放量至2100年最多增加30%（CO₂）与14%（CH₄）；而在可持续路径（SSP1-2.6）下，通过营养盐削减措施有望减缓这一排放加速，这是当前气候变化评估中很大程度上被忽视的反馈过程。

（来源：PNAS 2026 Issue 22  DOI: 10.1073/pnas.2537678123）