Arctic ecosystems store vast carbon amounts but also emit greenhouse gases like methane (CH4) and carbon dioxide (CO2). These fluxes vary significantly based on microrelief, water table, and vegetation. Sparse and clustered measurements bias regional budgets, a challenge addressed here by combining pan-Arctic and plot-scale insights across wet-dry gradients. This work introduces a cross-scale framework through three stages: pan-Arctic synthesis, site-scale measurements, and map-based upscaling. Pan-Arctic clustering revealed that CH4 and CO2 (as NEE) are governed by distinct drivers: CH4 clusters are structured by soil moisture and climate, while NEE clusters are primarily governed by climate. Plot-scale evaluation at a heterogeneous tundra site showed that microrelief structures flux contrasts; median CH4 fluxes were six times higher in wet trenches than in other landforms. While all landforms were CH4 sources, only half acted as CO2 sinks. Landform-specific models improved CH4 variability explanation, indicating microrelief must be modeled explicitly. Finally, upscaling July CH4 fluxes using machine learning at 1 m and 10 m resolutions highlighted significant scale sensitivity. At 1 m, algorithms diverged and terrain metrics dominated importance, whereas at 10 m, results converged and soil moisture proxies became more critical. Recommendations include landform-aware sampling and combining 10 m base grids with targeted 1 m calibration in complex zones. This multi-scale approach refines landscape-level carbon estimates and identifies priorities for future Arctic research.
Kseniia Ivanova (Thu,) studied this question.