Long-term CO2 production following permafrost thaw

Thawing permafrost represents a poorly understood feedback mechanism of climate change in the Arctic, but with a potential impact owing to stored carbon being mobilized1–5. We have quantified the long-term loss of carbon (C) from thawing permafrost in Northeast Greenland from 1996 to 2008 by combining repeated sediment sampling to assess changes in C stock and >12 years of CO2 production in incubated permafrost samples. Field observations show that the activelayer thickness has increased by >1 cm yr−1 but thawing has not resulted in a detectable decline in C stocks. Laboratory mineralization rates at 5 C resulted in a C loss between 9 and 75%, depending on drainage, highlighting the potential of fast mobilization of permafrost C under aerobic conditions, but also that C at near-saturated conditions may remain largely immobilized over decades. This is confirmed by a three-pool C dynamics model that projects a potential C loss between 13 and 77% for 50 years of incubation at 5 C. The Northern Hemisphere permafrost region contains approximately 1,700 Pg of organic C of which about 90% occurs in permafrost deposits1. This C pool represents about 50% of the estimated global below-ground organic C pool1. With the large amplitude of predicted Arctic climate change, this C pool has been used to imply a critical potential for global scale feedbacks from Arctic climate change if these C reservoirs are destabilized1,3, but actual mobilization rates on decadal to century scales are unknown as well as the consistency on a circumpolar scale of the few studies reporting multiyear incubation results6. Long-term observations indicate that permafrost thawing inNortheastGreenland,measured as increasing active-layer depths, occurs at rates of more than 1 cm yr−1 (refs 4,7). In the present long-term incubation experiment we hypothesized that natural drainage associated with permafrost thawing would considerably increase the potential mineralization and the associated release of CO2. Sites included two different landforms (see Supplementary Figs S1–S3) with two different, but dominating, vegetation types7; a sandy ground moraine with a Cassiope tetragona heath vegetation covered in places by aeolian sediments and the lower part of a nivation site containing a large snow patch with wet grassland vegetation7–10. These two sites differ in geomorphological activity, age, C-substrate quality and the environmental conditions controlling oxygen availability and pHas described in theMethods. Depth and volume-specific samples from the active layer and top permafrost in Zackenberg (Northeast Greenland) were collected down to about 1m depth in three replicate profiles in 1996 and