Past and Current State of the Arctic Eddy Covariance Network

The Arctic stores enormous pools of sequestered carbon currently locked away in permafrost soils. Due to projected climate change and specifically Arctic amplification, the future sustainability of these stocks is unclear. There are strong indicators that the Arctic could turn from a sink into a source of carbon to the atmosphere. Depending on local environmental conditions, the largest portion of this carbon could either be released as CO₂ or CH₄.

One of the most reliable methods to measure long-term surface-atmosphere exchange of carbon is through the use of eddy covariance (EC) towers. However, due to logistical difficulties, in the Arctic the existing EC network is still comparatively sparse, with apparent gaps regarding certain regions and ecosystem types. To facilitate an objective evaluation of this network, in this study we first assessed the availability of Arctic EC datasets for past and current timeframes, and made this information available to the research community through an online mapping tool. Subsequently, we applied two different methods to quantify the representativeness of the EC network regarding the quantification of pan-Arctic carbon budgets, and identified locations where new flux towers would ideally complement the existing infrastructure.

Based on our first method, we indicate the networks representativeness with regards to its coverage of the ecoregions of the Arctic. Here ecoregions are computed based on combinations of key ecosystem variables, such as e.g. land cover type or prevailing temperature regimes. The representativeness of different subsets of eddy covariance sites from our database was subsequently evaluated based on how well the conditions found at the chosen EC sites correspond to the means of ecosystem variables within each ecoregion. Through our second network evaluation we focus more on the carbon fluxes. Here, key variables used for the upscaling of CO₂ fluxes identified by the Fluxcom project were utilized with a K nearest neighbors method to compare EC site coverage to the drivers of these flues. Both methods show clear gaps in data acquisition: Even though the network has clearly expanded over the last two decades, coverage is sparse in mountainous regions and large parts of Russia. More specifically, wintertime fluxes and CH₄ fluxes are clearly lagging behind.