The northern high latitudes (NHL, >50°N) are experiencing significant changes in carbon cycling, showing increased annual terrestrial net CO2 uptake and amplified seasonal atmospheric CO2 cycles over the past five decades. However, the mechanisms driving these changes remain unclear. Net CO2 uptake is the difference between larger gross fluxes of photosynthesis and ecosystem respiration, which respond asynchronously to seasonal changes. Increased photosynthesis might be offset by enhanced respiration in the fall or winter, complicating the detection of climate-carbon feedbacks over longer timescales. Seasonal compensation in CO2 uptake might vary among biomes due to differing sensitivities of above and belowground carbon cycle processes to climate and environmental controls under changing permafrost conditions. Understanding the magnitude, trends, and spatial patterns of seasonal net CO2 uptake is crucial to determining whether net CO2 exchange has changed in the NHL, particularly in rapidly changing permafrost regions. This study investigates the trends and mechanisms of seasonal CO2 exchange in the NHL by analyzing long-term atmospheric CO2 inversions (ACIs) and eddy covariance (EC) observations to address three key questions: (1) the trends in net CO2 uptake and their relationship to climate, vegetation, and environmental gradients; (2) the mechanisms underlying different net CO2 uptake trends; and (3) how well the latest land surface models replicate these dynamics.

Previous research has highlighted the enhanced seasonal exchange of CO2 by northern ecosystems since 1960, with amplified plant productivity identified as a major contributor to this enhanced exchange. Studies have also shown the effects of non-uniform climate warming on terrestrial carbon cycles and the sensitivity of peatland CO2 sinks to seasonal warming trends. The greening of the Arctic and its drivers have been extensively investigated, along with the characteristics, drivers, and feedbacks of global greening. There’s a body of work on the large CO2 losses in winter across the northern permafrost region and the carbon dioxide sources from Alaska, driven by increasing early winter respiration from Arctic tundra. Studies have also explored widespread seasonal compensation effects of spring warming on northern plant productivity and the increasing summer net CO2 uptake in high northern ecosystems, which are inferred from atmospheric inversions and comparisons to remote-sensing NDVI. Significant research addresses the potential impact of climate change and permafrost carbon feedback, highlighting the dependence of carbon dynamics evolution in the northern permafrost region on the climate change trajectory and the assessment of the carbon balance of Arctic tundra.

This study used an ensemble mean of six long-term atmospheric CO2 inversions (ACIs) from 1980 to 2017 and a network of 48 eddy covariance (EC) sites with at least three years of continuous measurements (1990-2017) to analyze trends in net CO2 uptake across the NHL. The ACIs showed high correlation with independent estimates of net CO2 exchange at a global scale, providing confidence in detecting regional trends. To understand the mechanisms driving these trends, the researchers tested two hypotheses: (H1) different temperature sensitivities of vegetation primary productivity, and (H2) the degree to which seasonal net CO2 uptake is compensated by seasonal respiration losses. Structural equation modeling was used to explore climatic, environmental, and vegetation controls on seasonal CO2 dynamics. The observationally constrained estimates were compared with an ensemble of ten Dynamic Global Vegetation Models (DGVMs) from the TRENDY intercomparison project to assess model performance. Multiple datasets were used: gridded satellite-based estimates of vegetation productivity (GPP from a light use efficiency approach and GIMMS NDVI), permafrost extent data from the ESA Climate Change Initiative, and ESA CCI soil moisture data. A robustness analysis was conducted, considering uncertainties in ACI estimates and using individual ACI analyses, random years and length analyses, site-level EC comparisons, and GLMM to account for uncertainties from inversion spread. The study also compared its findings with the Global Carbon Budget 2020. Path analysis, using structural equation models (SEMs), was conducted for both the early-growing season (EGS) and late-growing season (LGS) to investigate the relative influences of air temperature, PAR, soil moisture, tree cover, permafrost extent, and preseason GPP on CO2 fluxes. Lastly, the direct and legacy effects of spring temperature on seasonal net CO2 uptake were analyzed using regression against spring air temperature for EC observations, regional-level ACI ensemble, and the TRENDY NBP ensemble.

The study found that positive trends of annual net CO2 uptake are smaller with increasing tree cover. This pattern is attributed to increased seasonal compensation due to larger respiratory CO2 losses during the late-growing season at higher tree cover levels, rather than differences in temperature sensitivities of vegetation productivity. Analysis of atmospheric inversions (ACIs) revealed that approximately half of the NHL showed significant increases in annual net CO2 uptake, mostly in permafrost regions dominated by tundra. Conversely, only a small percentage showed significant decreases, primarily in non-permafrost forested regions. The trend of net CO2 uptake most strongly correlated with tree cover, followed by mean annual temperature and permafrost extent. While net CO2 uptake in forested regions increased at a slower rate, these regions remained stronger CO2 sinks than non-forest tundra in colder permafrost regions. Regional aggregation showed significantly faster net CO2 uptake increases in tundra permafrost regions compared to tree-dominated non-permafrost regions. Permafrost regions shifted from being CO2 neutral before 2000 to being a CO2 sink after 2000. The study ruled out several confounding factors affecting the increase in net CO2 uptake in short-vegetated permafrost regions through various robustness checks. Analysis of vegetation productivity showed a positive correlation with net CO2 uptake, suggesting warming-induced increases in productivity. However, trends in productivity and net CO2 uptake did not correlate along the tree cover gradient. Seasonal compensation was identified as the key mechanism, with accelerating net CO2 uptake in the early-growing season (EGS) being partially offset by accelerating net CO2 release in the late-growing season (LGS). While EGS uptake trends were similar regardless of tree cover, LGS release trends increased significantly with greater tree cover, leading to a slower rate of annual net CO2 uptake and larger seasonal compensation with greater tree cover. Short-vegetated permafrost regions showed positive sensitivity of annual net CO2 uptake to mean annual temperature, while tree-dominated non-permafrost regions had non-significant negative sensitivity. DGVMs failed to reproduce the observed trends of net CO2 uptake along permafrost-vegetation gradients, particularly due to underestimation of increased CO2 release in the LGS. Structural equation modeling (SEM) revealed that in the EGS, net CO2 uptake was strongly influenced by ecosystem productivity, primarily controlled by spring air temperature and tree cover. In the LGS, respiration strongly controlled net CO2 uptake, driven by increased labile carbon from enhanced early-season productivity and temperature. The study also analyzed the direct and legacy effects of spring temperature on seasonal net CO2 uptake.

The findings challenge the notion that warming-induced acceleration of soil decomposition in permafrost regions will cause a regime shift to a net CO2 source. Instead, the study suggests that net CO2 uptake has increased faster in permafrost regions than in tree-dominated regions, with most permafrost tundra regions shifting from sources to sinks. These increases are primarily driven by warming-enhanced early-growing season productivity and increased shrub/tree cover. While the coarse resolution of ACI data prevents identification of specific ecosystem types, the study suggests that ecosystems with intermediate permafrost extent might be particularly strong contributors to enhanced net C uptake. The slower net CO2 uptake increase in forest-dominated regions is attributed to enhanced late-season CO2 release due to warming-induced respiration and emerging moisture limitations on productivity. Under future warmer and drier conditions, enhanced late-season respiration could offset or exceed photosynthetic gains, potentially turning these regions into net CO2 sources. The study highlights the importance of both productivity and respiration in regulating seasonal and annual CO2 dynamics, but with varying magnitudes and mechanisms across gradients. The discrepancy between observations and DGVM simulations underscores the need to improve model representations of late-growing season respiration, especially considering the strong influence of respiration on the net CO2 balance and its sensitivity to early-season productivity and vegetation changes induced by warming.

This study provides strong evidence that northern high-latitude permafrost regions are not currently net sources of CO2, but rather are significant contributors to the increased northern terrestrial carbon sink. Increased early-growing season CO2 uptake surpasses late-growing season emissions, driving this trend. However, greater seasonal compensation in net CO2 uptake with increased tree cover is an important driver of the increasing seasonal atmospheric CO2 amplitude. Improvements are needed in DGVMs to accurately simulate respiration processes in response to changing landscape conditions for more accurate predictions of climate-carbon feedbacks in high latitudes. While positive carbon-climate feedbacks are not yet evident, thresholds may occur as these ecosystems respond to changing permafrost conditions.

The coarse resolution of ACI data limits the identification of specific ecosystem types contributing to enhanced net CO2 uptake. The study acknowledges that some episodic ecosystem processes, such as fire disturbance, were not fully accounted for in the analysis. The reliance on satellite-based estimates of vegetation productivity introduces uncertainties due to the inherent limitations of remote sensing techniques.