The regional variability in tundra and boreal carbon dioxide (CO2) fluxes can be high, complicating efforts to quantify sink‐source patterns across the entire region. Statistical models are increasingly used to predict (i.e., upscale) CO2 fluxes across large spatial domains, but the reliability of different modeling techniques, each with different specifications and assumptions, has not been assessed in detail. Here, we compile eddy covariance and chamber measurements of annual and growing season CO2 fluxes of gross primary productivity (GPP), ecosystem respiration (ER), and net ecosystem exchange (NEE) during 1990–2015 from 148 terrestrial high‐latitude (i.e., tundra and boreal) sites to analyze the spatial patterns and drivers of CO2 fluxes and test the accuracy and uncertainty of different statistical models. CO2 fluxes were upscaled at relatively high spatial resolution (1 km2) across the high‐latitude region using five commonly used statistical models and their ensemble, that is, the median of all five models, using climatic, vegetation, and soil predictors. We found the performance of machine learning and ensemble predictions to outperform traditional regression methods. We also found the predictive performance of NEE‐focused models to be low, relative to models predicting GPP and ER. Our data compilation and ensemble predictions showed that CO2 sink strength was larger in the boreal biome (observed and predicted average annual NEE −46 and −29 g C m−2 yr−1, respectively) compared to tundra (average annual NEE +10 and −2 g C m−2 yr−1). This pattern was associated with large spatial variability, reflecting local heterogeneity in soil organic carbon stocks, climate, and vegetation productivity. The terrestrial ecosystem CO2 budget, estimated using the annual NEE ensemble prediction, suggests the high‐latitude region was on average an annual CO2 sink during 1990–2015, although uncertainty remains high.

Abstract. Peatlands store substantial amounts of carbon and are vulnerable to climate change. We present a modified version of the Organising Carbon and Hydrology In Dynamic Ecosystems (ORCHIDEE) land surface model for simulating the hydrology, surface energy, and CO 2 fluxes of peatlands on daily to annual timescales. The model includes a separate soil tile in each 0.5 • grid cell, defined from a global peatland map and identified with peat-specific soil hydraulic properties. Runoff from non-peat vegetation within a grid cell containing a fraction of peat is routed to this peat soil tile, which maintains shallow water tables. The water table position separates oxic from anoxic decomposition. The model was evaluated against eddy-covariance (EC) observations from 30 northern peatland sites, with the maximum rate of carboxylation (V cmax ) being optimized at each site. Regarding short-term day-to-day variations, the model performance was good for gross primary production (GPP) (r 2 = 0.76; NashSutcliffe modeling efficiency, MEF = 0.76) and ecosystem respiration (ER, r 2 = 0.78, MEF = 0.75), with lesser accuracy for latent heat fluxes (LE, r 2 = 0.42, MEF = 0.14) and and net ecosystem CO 2 exchange (NEE, r 2 = 0.38, MEF = 0.26). Seasonal variations in GPP, ER, NEE, and energy fluxes on monthly scales showed moderate to high r 2 values (0.57-0.86). For spatial across-site gradients of annual mean GPP, ER, NEE, and LE, r 2 values of 0.93, 0.89, 0.27, and 0.71 were achieved, respectively. Water table (WT) variation was not well predicted (r 2 < 0.1), likely due to the uncertain water input to the peat from surrounding areas. However, the poor performance of WT simulation did not greatly affect predictions of ER and NEE. We found a significant relationship between optimized V cmax and latitude (temperature), which better reflects the spatial gradients of annual NEE than using an average V cmax value.

Abstract. Peatlands store substantial amount of carbon, are vulnerable to climate change. To predict the fate of carbon stored in peatlands, the complex interactions between water, peat and vegetations need more attention. This study describes a modified version of the ORCHIDEE land surface model for simulating the hydrology, surface energy and CO2 fluxes of peatlands on daily to annual time scales. The model, referred to as ORCHIDEE-PEAT, includes a separate soil tile in each 0.5° grid-cell, defined from a global peatland map and identified with peat-specific soil hydraulic properties. Runoff from non-peat vegetation with a grid-cell containing a fraction of peat is routed to this peat soil tile, which maintains shallow water tables. The water table position separates oxic from anoxic decomposition. The model is evaluated against eddy-covariance (EC) observations from 30 northern peatland sites, with the maximum rate of carboxylation (Vcmax) being optimized at each site to match the peak of growing season gross primary productivity (GPP), derived from direct EC measurements. Regarding short-term variations from day to day, the model performance was good for the variations in GPP (r2 = 0.76, Nash-Sutcliff modeling efficiency, MEF = 0.76), with lesser accuracy for latent heat fluxes (LE, r2 = 0.42, MEF = 0.14) and Net ecosystem CO2 exchange (NEE, r2 = 0.38, MEF = 0.26). Seasonal variations in GPP, NEE and energy fluxes on monthly scales showed moderate to high r2 values ranging from 0.57 to 0.86. For spatial across-sites gradients of annual mean GPP, NEE and LE, r2 of 0.93, 0.27, and 0.71, respectively, were achieved. The water table variations are not well predicted (r2 < 0.1), likely due to the uncertain water input to the peat from surrounding areas. However, when using the observed water table in the carbon module to define the fraction of oxic and anoxic decomposition instead of the modeled water table, ORCHIDEE-PEAT shows a small improvement in reproducing NEE. Moreover, we found a significant relationship between optimized Vcmax and the latitude (temperature), which can better reflect the spatial gradients of annual NEE than using an average Vcmax value. In a future version of ORCHIDEE-PEAT, the influences of water table on photosynthesis and depth-dependent influences of soil temperature on respiration may be included.

Abstract. Arctic tundra ecosystems are currently facing amplified rates of climate warming. Since these ecosystems store significant amounts of soil organic carbon, which can be mineralized to carbon dioxide (CO2) and methane (CH4), rising temperatures may cause increasing greenhouse gas fluxes to the atmosphere. To understand how net the ecosystem exchange (NEE) of CO2 will respond to changing climatic and environmental conditions, it is necessary to understand the individual responses of the processes contributing to NEE. Therefore, this study aimed to partition NEE at the soil–plant–atmosphere interface in an arctic tundra ecosystem and to identify the main environmental drivers of these fluxes. NEE was partitioned into gross primary productivity (GPP) and ecosystem respiration (Reco) and further into autotrophic (RA) and heterotrophic respiration (RH). The study examined CO2 flux data collected during the growing season in 2015 using closed-chamber measurements in a polygonal tundra landscape in the Lena River Delta, northeastern Siberia. To capture the influence of soil hydrology on CO2 fluxes, measurements were conducted at a water-saturated polygon center and a well-drained polygon rim. These chamber-measured fluxes were used to model NEE, GPP, Reco, RH, RA, and net primary production (NPP) at the pedon scale (1–10 m) and to determine cumulative growing season fluxes. Here, the response of in situ measured RA and RH fluxes from permafrost-affected soils of the polygonal tundra to hydrological conditions have been examined. Although changes in the water table depth at the polygon center sites did not affect CO2 fluxes from RH, rising water tables were linked to reduced CO2 fluxes from RA. Furthermore, this work found the polygonal tundra in the Lena River Delta to be a net sink for atmospheric CO2 during the growing season. The NEE at the wet, depressed polygon center was more than twice that at the drier polygon rim. These differences between the two sites were caused by higher GPP fluxes due to a higher vascular plant density and lower Reco fluxes due to oxygen limitation under water-saturated conditions at the polygon center in comparison to the rim. Hence, soil hydrological conditions were one of the key drivers for the different CO2 fluxes across this highly heterogeneous tundra landscape.

Methane emissions from natural wetlands tend to increase with temperature and therefore may lead to a positive feedback under future climate change. However, their temperature response includes confounding factors and appears to differ on different time scales. Observed methane emissions depend strongly on temperature on a seasonal basis, but if the annual mean emissions are compared between sites, there is only a small temperature effect. We hypothesize that microbial dynamics are a major driver of the seasonal cycle and that they can explain this apparent discrepancy. We introduce a relatively simple model of methanogenic growth and dormancy into a wetland methane scheme that is used in an Earth system model. We show that this addition is sufficient to reproduce the observed seasonal dynamics of methane emissions in fully saturated wetland sites, at the same time as reproducing the annual mean emissions. We find that a more complex scheme used in recent Earth system models does not add predictive power. The sites used span a range of climatic conditions, with the majority in high latitudes. The difference in apparent temperature sensitivity seasonally versus spatially cannot be recreated by the non-microbial schemes tested. We therefore conclude that microbial dynamics are a strong candidate to be driving the seasonal cycle of wetland methane emissions. We quantify longer-term temperature sensitivity using this scheme and show that it gives approximately a 12% increase in emissions per degree of warming globally. This is in addition to any hydrological changes, which could also impact future methane emissions. Plain Language Summary Wet soils such as bogs, fens, and other wetlands emit methane gas. Methane is a powerful greenhouse gas that adds to climate warming. It is important to understand its net production and also how this might change as the Earth warms. Generally, scientists have found that warmer soils emit more methane. However, there is a discrepancy between comparing warm versus cold sites-where the effect of the temperature difference is relatively small-and comparing warmer and colder seasons of the year, where the effect of temperature is much stronger. Since methane emissions are caused by microbes, we investigated whether their behavior might provide an explanation for this