Spatial variability of CO₂ uptake in polygonal tundra: Assessing low-frequency disturbances in eddy covariance flux estimates

The large spatial variability in Arctic tundra complicates the representative assessment of CO₂ budgets. Accurate measurements of these heterogeneous landscapes are, however, essential to understanding their vulnerability to climate change. We surveyed a polygonal tundra lowland on Svalbard with an unmanned aerial vehicle (UAV) that mapped ice-wedge morphology to complement eddy covariance (EC) flux measurements of CO₂. The analysis of spectral distributions showed that conventional EC methods do not accurately capture the turbulent CO₂ exchange with a spatially heterogeneous surface that typically features small flux magnitudes. Nonlocal (low-frequency) flux contributions were especially pronounced during snowmelt and introduced a large bias of -46 gCm^-2 to the annual CO₂2 budget in conventional methods (the minus sign indicates a higher uptake by the ecosystem). Our improved flux calculations with the ogive optimization method indicated that the site was a strong sink for CO₂ in 2015 (82 gCm2). Due to differences in light-use efficiency, wetter areas with lowcentered polygons sequestered 47% more CO₂ than drier areas with flat-centered polygons. While Svalbard has experienced a strong increase in mean annual air temperature of more than 2K in the last few decades, historical aerial photographs from the site indicated stable ice-wedge morphology over the last 7 decades. Apparently, warming has thus far not been sufficient to initiate strong ice-wedge degradation, possibly due to the absence of extreme heat episodes in the maritime climate on Svalbard. However, in Arctic regions where ice-wedge degradation has already initiated the associated drying of landscapes, our results suggest a weakening of the CO2 sink in polygonal tundra.