摘要：There is a significant knowledge gap in the current stateof the terrestrial carbon (C) budget. Recent studies have highlighted a poorunderstanding particularly of C pool transit times and of whether productivityor biomass dominate these biases. The Arctic, accounting for approximately50% of the global soil organic C stocks, has an important role in theglobal C cycle. Here, we use the CARbon DAta MOdel (CARDAMOM) data-assimilation system toproduce pan-Arctic terrestrial C cycle analyses for 2000–2015. This approachavoids using traditional plant functional type or steady-state assumptions.We integrate a range of data (soil organic C, leaf area index, biomass, andclimate) to determine the most likely state of the high-latitude C cycle ata 1∘×1∘ resolution and also to provide generalguidance about the controlling biases in transit times. On average, CARDAMOMestimates regional mean rates of photosynthesis of 565gCm−2yr−1(90% confidence interval between the 5th and 95th percentiles: 428, 741),autotrophic respiration of 270gCm−2yr−1 (182, 397) andheterotrophic respiration of 219gCm−2yr−1 (31, 1458),suggesting a pan-Arctic sink of −67 (−287, 1160)gCm−2yr−1, weaker in tundra and stronger in taiga. However, ourconfidence intervals remain large (and so the region could be a source of C),reflecting uncertainty assigned to the regional data products. We show aclear spatial and temporal agreement between CARDAMOM analyses and differentsources of assimilated and independent data at both pan-Arctic and localscales but also identify consistent biases between CARDAMOM and validationdata. The assimilation process requires clearer error quantification for leaf area index (LAI) and biomass products to resolve these biases. Mapping of vegetation C stocksand change over time and soil C ages linked to soil C stocks is requiredfor better analytical constraint. Comparing CARDAMOM analyses to globalvegetation models (GVMs) for the same period, we conclude that transit timesof vegetation C are inconsistently simulated in GVMs due to a combination ofuncertainties from productivity and biomass calculations. Our findingshighlight that GVMs need to focus on constraining both current vegetation Cstocks and net primary production to improve a process-based understanding ofC cycle dynamics in the Arctic.
