Marine Remaud

The Arctic has experienced the largest increase in atmospheric CO ₂ seasonal cycle amplitude (SCA) since the 1960s. Previous studies primarily attributed this to enhanced regional warming. Here, we show that at numerous locations in the Arctic, atmospheric CO ₂ SCAs have increased at a relative rate similar to atmospheric CO ₂ abundance over the past five decades, suggesting a likely strong role of CO ₂ fertilization. Using a tagged‐tracer transport experiment and long‐term atmospheric observations of CO ₂ , carbonyl sulfide (COS), and the ¹³ C: ¹² C ratio of CO ₂ (), we quantified high northern latitude (HNL, 53°–90°N) ecosystem process changes. Results suggest that HNL terrestrial ecosystems drive Arctic CO ₂ SCA increases, with gross primary productivity (GPP) exerting a more dominant impact than ecosystem respiration. Using atmospheric COS measurements and mass balance calculation, we estimated 2000–2020 growing‐season (May–August) average HNL stomatal conductance at 0.024 ± 0.004 mol m ⁻² s ⁻¹ . This exhibited a slight negative trend of −4.3 ± 3.8 (1) %, consistent with CO ₂ –induced reductions in stomata observed previously at smaller scales. However, accounting for regional leaf area expansion, leaf‐area‐integrated ecosystem conductance was estimated to be 0.044 ± 0.006 mol m ⁻² s ⁻¹ with a small positive trend of (1) %. Furthermore, observations suggest the growing‐season intercellular‐to‐ambient CO ₂ ratio in HNL ecosystems remained constant. Integrating these estimates, we estimate growing‐season GPP in HNL ecosystems averaged 33 ± 3 (1 ) PgC yr ⁻¹ from 2000 to 2020 and increased by 12 ± 2 (1) % over this period, driven primarily by CO ₂ fertilization.

Forest structure is shaped by forest management practices, land‐use changes, and natural disturbances, including droughts, fires, storms, and insect outbreaks that drive species‐specific and size‐specific mortality. By modifying the carbon‐water‐energy exchanges with the atmosphere, it influences a stand's capacity to buffer against or succumb to extreme weather events, which in turn determines the long‐term stability of the terrestrial carbon stocks. Given the importance of forest structure for forest land sink, land surface models are moving toward explicit representations of forest structure and management strategies. We present a new procedure to initialize forest diameters over Europe and document its implications for simulations of future forest carbon sinks. The simulated diameters for each grid cell covered by forests are initialized toward the diameter from a forest inventory. To this end, a 300‐year semi‐analytical spinup was carried out to bring the soil carbon pools into equilibrium. A lookup table with the simulated diameter and plant functional type as its entries was built by clearcutting all forests followed by a 200‐year simulation over Europe. For each grid point, the year associated with the simulated diameter that is the closest to the observation is selected, enabling the production of new initial state files over Europe. The new initialization procedure makes the initial state of forest more realistic and therefore significantly modifies the evolution of the forest carbon sink. The method could be further extended to initialize other forest state variables such as height or aboveground biomass.

L'Oscillation de Madden-Julian (MJO) est le principal mode de variabilité intrasaisonnière (20-90 jours) dans les tropiques. La MJO se caractérise par une perturbation de la dynamique de grande échelle et des précipitations se propageant principalement vers l'Est à une vitesse 5 m/s à l'équateur, et aussi vers le Nord en été boréal. Malgré son importance, cette perturbation est mal reproduite dans les modèles de circulation générale (MCG). Ce défaut des MCGs a été attribué en partie à un déclenchement trop fréquent du schéma de convection profonde du modèle qui préserverait trop la stabilité statique dans les tropiques et empêcherait la formation de fortes perturbations organisées de la convection et le développement de la réponse dynamique de grande échelle qui en résulte. Cette thèse a donc pour objectif d'étudier l'impact de l'inhibition de la convection sur l'état moyen et la variabilité tropicale dans le MCG LMDZ. Pour cela, deux paramètres - la fermeture et l'entraînement - d'un même schéma convectif, qui étaient initialement basés sur le profil vertical de la convergence d'humidité, ont été modifiés. Le taux entraînement du schéma modifié est proportionnel à l'humidité relative de l'environnement. Des études de sensibilité dans un modèle unidimensionnel montrent que, par rapport au schéma initial, cette modification a pour effet d'inhiber la convection dans une atmosphère sèche et de la favoriser en atmosphère humide, les précipitations deviennent donc plus rares et plus intenses. Avec ce nouvel entraînement, les simulations globales du modèle atmosphérique LMDZ présentent effectivement une plus forte variabilité tropicale de la dynamique et des précipitations à toutes les échelles de temps. Le cycle saisonnier des précipitations est aussi mieux reproduit ainsi que la propagation vers l'Est et vers le Nord des perturbations intrasaisonnières. L'ajout d'une fermeture en CAPE modifie peu ce comportement, mais accélère légèrement la propagation vers le Nord et vers l'Est de la MJO, et donne un état moyen et une variabilité un peu plus réaliste dans les tropiques.

Forest structure is shaped by forest management practices, land-use changes and disturbances including droughts, fires, storms and insect outbreak. It plays an important role in climate by modifying the carbon-water-energy exchanges with the atmosphere and affects the capability of forests to undergo future disturbances in a changing climate. Given the importance of forest structure for the climate, land surface models are moving towards explicit representations of forest structure and management strategies. We present a new procedure to initialize forest diameters over Europe and document its implications for simulations of future forest carbon sinks. The simulated diameters for each grid cell covered by forests are initialized toward the diameter from a forest inventory. To this end, a 300-years semi-analytical spinup was carried out to bring the soil carbon pools into equilibrium. A look-table with the simulated diameter as its sole entry was built by clearcutting all forests followed by a 200-years biosphere simulation over Europe. For each grid point, the year associated with the simulated diameter that is the closest to the observation is selected, enabling the production of new initial state files over Europe. The new initialization procedure makes the initial state of forest more realistic and therefore significantly modifies the evolution of the forest carbon sink. The method could be further extended to initialize other forest state variables such as height or aboveground biomass.

Carbonyl sulfide (COS), a gas used as a tracer in carbon cycle studies, has a potential missing sink in Northern high latitudes. Boreal COS budgets typically account for the contribution by forests, but ignore any uptake that wetland ecosystems, widespread in Northern latitudes, may contribute. The first direct measurements of the ecosystem‐atmosphere COS exchange of a boreal wetland, presented here, demonstrate their likely importance for Northern latitude COS budgets. The investigated wetland (Siikaneva, Finland) took up on average 11 pmol m ⁻² s ⁻¹ COS, which was c. 72% of the nearby boreal forest COS uptake. During nighttime, the COS uptake rates were similar at both sites. Upscaling our measurements to the boreal region using the Organizing Carbon and Hydrology In Dynamic Ecosystems model revealed a Northern wetland sink of c. 13 Gg S/y, changing the simulated budget from a small source to a COS sink impacting Northern latitudes carbon uptake estimates based on COS.

Record breaking atmospheric methane growth rates were observed in 2020 and 2021 (15.2±0.5 and 17.8±0.5 parts per billion per year), the highest since the early 1980s. Here we use an ensemble of atmospheric inversions informed by surface or satellite methane observations to infer emission changes during these two years relative to 2019. Results show global methane emissions increased by 20.3±9.9 and 24.8±3.1 teragrams per year in 2020 and 2021, dominated by heightened emissions from tropical and boreal inundated areas, aligning with rising groundwater storage and regional warming. Current process-based wetland models fail to capture the tropical emission surges revealed by atmospheric inversions, likely due to inaccurate representation of wetland extents and associated methane emissions. Our findings underscore the critical role of tropical inundated areas in the recent methane emission surges and highlight the need to integrate multiple data streams and modeling tools for better constraining tropical wetland emissions.

Not-yet-known not-yet-known not-yet-known unknown Carbonyl sulfide (COS), an atmospheric gas used as a tracer in carbon cycle studies, has an inferred missing sink in high Northern latitudes. Boreal COS budgets typically account for the contribution by forests and ignore any uptake that wetland ecosystems, widespread in Northern latitudes, may contribute. The first direct measurements of the ecosystem-atmosphere COS exchange of a boreal wetland, presented here, demonstrate their likely importance in that Northern latitude COS budgets. The investigated wetland (Siikaneva, Finland) took up on average 11 pmol m−2s−1 COS, which was ~72 % of the nearby boreal forest COS uptake. During nighttime, the COS uptake rates were similar at both sites. Upscaling our measurements to the boreal region using the ORCHIDEE model revealed in a Northern wetland sink of ~13 Gg S/y, changing the budget model output from a small source to a COS sink impacting Northern latitudes carbon uptake estimates based on COS.

We present a comparison of atmospheric transport models that simulate carbonyl sulfide (COS). This is part II of the ongoing Atmospheric Transport Model Inter‐comparison Project (TransCom–COS). Differently from part I, we focus on seven model intercomparison by transporting two recent COS inversions of NOAA surface data within TM5‐4DVAR and LMDz models. The main goals of TransCom‐COS part II are (a) to compare the COS simulations using the two sets of optimized fluxes with simulations that use a control scenario (part I) and (b) to evaluate the simulated tropospheric COS abundance with aircraft‐based observations from various sources. The output of the seven transport models are grouped in terms of their vertical mixing strength: strong and weak mixing. The results indicate that all transport models capture the meridional distribution of COS at the surface well. Model simulations generally match the aircraft campaigns HIAPER Pole‐To‐Pole Observations (HIPPO) and Atmospheric Tomography Mission (ATom). Comparisons to HIPPO and ATom demonstrate a gap between observed and modeled COS over the Pacific Ocean at 0–40°N, indicating a potential missing source in the free troposphere. The effects of seasonal continental COS uptake by the biosphere, observed on HIPPO and ATom over oceans, is well reproduced by the simulations. We found that the strength of the vertical mixing within the column as represented in the various atmospheric transport models explains much of the model to model differences. We also found that weak‐mixing models transporting the optimized flux derived from the strong‐mixing TM5 model show a too strong seasonal cycle at high latitudes.

Boreal forests absorb a significant amount of atmospheric CO2 through gross primary production (GPP), representing about 20% of the global GPP. However, direct observations of GPP over the whole boreal region are not available as plant photosynthetic rate cannot be measured at scales larger than the leaf scale. At large scales, Land Surface Models (LSMs) can simulate GPP but the lack of direct GPP measurements makes it challenging to evaluate and improve the GPP representation in LSMs. In addition, boreal forests are highly sensitive to environmental changes, impacting gas exchanges and leading to high uncertainties in GPP estimates simulated by LSMs or obtained from data driven methods. Carbonyl sulfide (COS) has emerged as a promising proxy to infer GPP estimates or to better constrain GPP representation in LSMs. Because COS is absorbed by vegetation following the same diffusion pathway as CO2 during photosynthesis and not emitted back to the atmosphere, implementing a mechanistic representation of vegetation COS uptake in LSMs allows using COS data to constrain GPP representation. In this study, we performed ecosystem COS flux and GPP assimilations to constrain the COS and GPP related parameters in the ORCHIDEE LSM. We focused on Hyytiälä forest, where the longest time-series of ecosystem COS flux measurements was reported. We found that assimilating ecosystem COS fluxes increases the estimated net ecosystem COS uptake by 14%. However, a persistent underestimation of the ecosystem COS flux seasonal amplitude after data assimilation points towards structural errors in the COS model, possibly related to COS internal conductance representation. In comparison with an assimilation of GPP only, adding ecosystem COS flux assimilation leads to a stronger reduction in the stomatal conductance, highlighting the potential of COS to inform stomatal diffusion. Consequently, including COS data in the assimilations also impacts the resulting latent heat flux and water use efficiency. Finally, we scaled up this assimilation framework to the boreal region and found that the joint assimilation of COS and GPP fluxes increased the modeled vegetation COS uptake up to 18%, but not the GPP budget. This contrasts with previous inversion studies that simultaneously increase vegetation COS uptake and GPP budgets based on a linear relationship relating the two.

We present a comparison of atmospheric transport models that simulate carbonyl sulfide (COS). This is part II of the ongoing Atmospheric Transport Model (ATM) Inter-comparison Project (TransCom–COS). Differently from part I, we focus on seven model intercomparison by transporting two recent COS inversions of NOAA surface data within TM5-4DVAR and LMDz models. The main goals of TransCom-COS part II are (a) to compare the COS simulations using the two sets of optimized fluxes with simulations that use a control scenario (part I) and (b) to evaluate the simulated tropospheric COS abundance with aircraft-based observations from various sources. The output of the seven transport models are grouped in terms of their vertical mixing strength: strong and weak mixing. The results indicate that all transport models capture the meridional distribution of COS at the surface well. Model simulations generally match the aircraft campaigns HIPPO and ATom. Comparisons to HIPPO and ATom demonstrate a gap between observed and modelled COS over the Pacific Ocean at 0–40$\degree$N, indicating a potential missing source in the free troposphere. The effects of seasonal continental COS uptake by the biosphere, observed on HIPPO and ATom over oceans, is well reproduced by the simulations. We found that the strength of the vertical mixing within the column as represented in the various atmospheric transport models explains much of the model to model differences. We also found that weak-mixing models transporting the optimized flux derived from the strong-mixing TM5 model show a too strong seasonal cycle at high latitudes.