The Sub-Arctic Carbon Cycle: Assimilating Multi-Scale Chamber, Tower and Aircraft Flux Observations into Ecological Models

The Arctic has already warmed significantly, and warming of 4-7 °C is expected over the next century. However, linkages between climate, the carbon cycle, the energy balance, and hydrology mean that the response of arctic ecosystems to these changes remains poorly understood. The release by warming of considerable but poorly quantified carbon stores from high latitude soils could accelerate the build-up of atmospheric CO2. The Arctic Biosphere Atmosphere Coupling at Multiple Scales (ABACUS) project, part of International Polar Year, was designed to improve predictions of the response of the Arctic terrestrial biosphere to climate change. The project operated at two sites (Abisko, Sweden and Kevo, Finland) over multiple years, utilising roving flux chambers (CO2/CH4), five flux towers (CO2/CH4/H2O) and a research aircraft equipped for fluxes (CO2/H2O) to directly measure multi-scale exchanges in-conjunction with other observations (both plot level and satellite). We show how these data can be combined using data assimilation approaches to address the question “what controls the temporal and spatial variability of carbon exchange by sub-Arctic ecosystems?” Eddy covariance measurements of mire methane exchanges agreed with chamber estimates, indicating that mires were strong summer sources, while birch woodland was a weak sink. However, remote sensing of mire extent was limited at resolutions > 30 m, and variations in sink/source activity suggested that upscaling CH4 exchanges (from chamber, to tower, to landscape) required higher resolution (ideally <10 m) landcover data in heterogeneous Arctic landscapes. Chamber and eddy covariance measurements of CO2 exchange recorded similar seasonal timing over a range of vegetation types. Birch woodlands had the greatest range of CO2 exchanges compared to tundra and mires. The challenge of measuring continuous fluxes across the full annual cycle, and inherent uncertainties in the methods, complicates the determination of source/sink status of the landscapes, but cold season data were successfully collected. Aircraft flux measurements during the peak growing season provided an estimate of landscape variability alongside the temporal sampling from fixed tower systems, and a means to constrain upscaling via models. Assimilation of these multi-scale data into an ecosystem carbon model yielded improved constraints on processes, particularly the turnover rates of soil carbon. Our work shows that these improvements cannot be attained with a single source of data (chamber, tower or aircraft).