Temporal factors in the high latitude eddy covariance network design

High latitude regions are increasingly affected by climate change, where large stocks of carbon sequestered in permafrost are at risk of being released. Monitoring this change is therefore vital for our understanding of  the Arctic and global climate, but adverse conditions and the heterogeneity of Arctic landscapes make it exceptionally challenging to gain a comprehensive observational view of the carbon cycle. In this work we explore the growth of the high latitude eddy covariance (EC) network, and evaluate ways to improve its design both from a spatial and temporal aspect to better monitor this vital region. We utilise the relative extrapolation index (EI) metric, a method to assess upscaling errors as a factor of a location's distance in predictor variable space to the EC network. Our EC site survey, last updated in 2022, identified 213 site locations in the Arctic, of which 124 are currently active. Of these active sites, 79% intend to remain active for 5 years or longer, although on average these sites only have 3.1 years of funding. We investigated the effect of limited site activity periods on the network, and found that if sites only stayed active for a maximum of 36 month (3 full years) or 18 month (3 full summers), the network’s mean EI would increase by 27.3 and 24.5 percentage points, respectively. This deterioration in network data coverage is similar to setting the network back to 2012 and 2008, i.e. a time when fewer sites by far were active. In addition, we investigated the effect of long time series of data, and the optimal configuration of the network in depth versus breadth. For the former, our results show that even for time series that are already long, adding more data years still contributes to increasing network performance overall. For the latter, we find that with total site-months remaining constant, many sites with fewer site-months results in a lower EI than a few sites with many site-months. Summarising, our findings demonstrate that a top-down network management should ideally combine long-term monitoring sites with observations that are rotated between multiple observation sites, to capture both long-term trends and spatial heterogeneity in exchange flux rates.