草原
表土
环境科学
草原
土壤碳
自然地理学
气候变化
土壤科学
土壤水分
农学
生态学
地理
生物
作者
Shangshi Liu,Yuanhe Yang,Haihua Shen,Huifeng Hu,Xia Zhao,Li He,Taoyu Liu,Jingyun Fang
标识
DOI:10.1016/j.scitotenv.2017.12.254
摘要
The grasslands of northern China store a large amount of soil organic carbon (SOC), and the small changes in SOC stock could significantly affect the regional C cycle. However, recent estimates of SOC changes in this region are highly controversial. In this study, we examined the changes in the SOC density (SOCD) in the upper 30cm of the grasslands of northern China between the 1980s and 2000s, using an improved approach that integrates field-based measurements into machine learning algorithms (artificial neural network (ANN) and random forest (RF)). The RF-generated SOCD averaged 5.55kgCm-2 in the 1980s and 5.53kgCm-2 in the 2000s, and the change ranged from -0.17 to 0.22kgCm-2 at the 95% confidence level, suggesting that the overall SOCD did not vary significantly during the study period. However, the change in SOCD exhibited large regional variability; the topsoil of the Inner Mongolian grasslands experienced significant C loss (4.86 vs. 4.33kgCm-2), while that of the Xinjiang grasslands exhibited an accumulation of C (5.55 vs. 6.46kgCm-2). Furthermore, the topsoil C in the Tibetan alpine grasslands remained relatively stable (6.12 vs. 6.06kgCm-2). A comparison of the different grassland types indicated that SOCD significantly decreased in typical steppe, whereas it increased in mountain meadow, and remained stable in the other grasslands (alpine meadow, alpine steppe, mountain steppe and desert steppe). Climate change could partly explain the changes in the SOCD of the different grassland types. Increases in precipitation could lead to SOC accumulation in temperate grasslands and SOC loss in alpine grasslands, while climate warming is likely to cause SOC loss in temperate grasslands. Overall, our study suggests that the grasslands of northern China remained a neutral SOC sink between the 1980s and 2000s.
科研通智能强力驱动
Strongly Powered by AbleSci AI