Multi-year observations reveal a larger than expected autumn respiration signal across northeast Eurasia

Abstract

Abstract. Site-level observations have shown pervasive cold season CO₂ release across Arctic and boreal ecosystems, impacting annual carbon budgets. Still, the seasonality of CO₂ emissions are poorly quantified across much of the high latitudes due to the sparse coverage of site-level observations. Space-based observations provide the opportunity to fill some observational gaps for studying these high-latitude ecosystems, particularly across poorly sampled regions of Eurasia. Here, we show that data-driven net ecosystem exchange (NEE) from atmospheric CO₂ observations implies strong summer uptake followed by strong autumn release of CO₂ over the entire cold northeastern region of Eurasia during the 2015–2019 study period. Combining data-driven NEE with satellite-based estimates of gross primary production (GPP), we show that this seasonality implies less summer heterotrophic respiration (Rₕ) and greater autumn Rₕ than would be expected given an exponential relationship between respiration and surface temperature. Furthermore, we show that this seasonality of NEE and Rₕ over northeastern Eurasia is not captured by the TRENDY v8 ensemble of dynamic global vegetation models (DGVMs), which estimate that 47%–57% (interquartile range) of annual Rₕ occurs during August–April, while the data-driven estimates suggest 59%–76% of annual Rₕ occurs over this period. We explain this seasonal shift in Rₕ by respiration from soils at depth during the zero-curtain period, when sub-surface soils remain unfrozen up to several months after the surface has frozen. Additional impacts of physical processes related to freeze–thaw dynamics may contribute to the seasonality of Rₕ. This study confirms a significant and spatially extensive early cold season CO₂ efflux in the permafrost-rich region of northeast Eurasia and suggests that autumn Rₕ from subsurface soils in the northern high latitudes is not well captured by current DGVMs.