The permafrost organic carbon (OC) stock is of global significance because of its large pool size and the potential positive feedback to climate warming. However, due to the lack of systematic field observations and appropriate upscaling methodologies, substantial uncertainties exist in the permafrost OC budget, which limits our understanding of the fate of frozen carbon in a warming world. In particular, the lack of comprehensive estimates of OC stocks across alpine permafrost means that current knowledge on this issue remains incomplete. Here, we evaluated the pool size and spatial variations of permafrost OC stock to 3 m depth on the Tibetan Plateau by combining systematic measurements from a substantial number of pedons (i.e. 342 three-metre-deep cores and 177 50-cm-deep pits) with a machine learning technique (i.e. support vector machine, SVM). We also quantified uncertainties in permafrost carbon budget by conducting Monte Carlo simulations. Our results revealed that the combination of systematic measurements with the SVM model allowed spatially explicit estimates to be made. The OC density (OC amount per unit area, OCD) exhibited a decreasing trend from the south-eastern to the north-western plateau, with the exception that OCD in the swamp meadow was substantially higher than that in surrounding regions. Our results also demonstrated that Tibetan permafrost stored a large amount of OC in the top 3 m, with the median OC pool size being 15.31 Pg C (interquartile range: 13.03-17.77 Pg C). 44% of OC occurred in deep layers (i.e. 100-300 cm), close to the proportion observed across the northern circumpolar permafrost region. The large carbon pool size together with significant permafrost thawing suggests a risk of carbon emissions and positive climate feedback across the Tibetan alpine permafrost region.

No abstract

The sign and magnitude of permafrost carbon (C)-climate feedback are highly uncertain due to the limited understanding of the decomposability of thawing permafrost and relevant mechanistic controls over C release. Here, by combining aerobic incubation with biomarker analysis and a three-pool model, we reveal that C quality (represented by a higher amount of fast cycling C but a lower amount of recalcitrant C compounds) and normalized CO 2 -C release in permafrost deposits were similar or even higher than those in the active layer, demonstrating a high vulnerability of C in Tibetan upland permafrost. We also illustrate that C quality exerts the most control over CO 2 -C release from the active layer, whereas soil microbial abundance is more directly associated with CO 2 -C release after permafrost thaw. Taken together, our findings highlight the importance of incorporating microbial properties into Earth System Models when predicting permafrost C dynamics under a changing environment.

Our knowledge of fundamental drivers of the temperature sensitivity (Q10) of soil carbon dioxide (CO2) release is crucial for improving the predictability of soil carbon dynamics in Earth System Models. However, patterns and determinants of Q10 over a broad geographic scale are not fully understood, especially in alpine ecosystems. Here we addressed this issue by incubating surface soils (0–10 cm) obtained from 156 sites across Tibetan alpine grasslands. Q10 was estimated from the dynamics of the soil CO2 release rate under varying temperatures of 5–25°C. Structure equation modeling was performed to evaluate the relative importance of substrate, environmental, and microbial properties in regulating the soil CO2 release rate and Q10. Our results indicated that steppe soils had significantly lower CO2 release rates but higher Q10 than meadow soils. The combination of substrate properties and environmental variables could predict 52% of the variation in soil CO2 release rate across all grassland sites and explained 37% and 58% of the variation in Q10 across the steppe and meadow sites, respectively. Of these, precipitation was the best predictor of soil CO2 release rate. Basal microbial respiration rate (B) was the most important predictor of Q10 in steppe soils, whereas soil pH outweighed B as the major regulator in meadow soils. These results demonstrate that carbon quality and environmental variables coregulate Q10 across alpine ecosystems, implying that modelers can rely on the “carbon‐quality temperature” hypothesis for estimating apparent temperature sensitivities, but relevant environmental factors, especially soil pH, should be considered in higher‐productivity alpine regions.