Abstract. We present a global distribution of surface methane (CH₄) emission estimates for 2000–2012 derived using the CarbonTracker Europe-CH₄ (CTE-CH₄) data assimilation system. In CTE-CH₄, anthropogenic and biospheric CH₄ emissions are simultaneously estimated based on constraints of global atmospheric in situ CH₄ observations. The system was configured to either estimate only anthropogenic or biospheric sources per region, or to estimate both categories simultaneously. The latter increased the number of optimizable parameters from 62 to 78. In addition, the differences between two numerical schemes available to perform turbulent vertical mixing in the atmospheric transport model TM5 were examined. Together, the system configurations encompass important axes of uncertainty in inversions and allow us to examine the robustness of the flux estimates. The posterior emission estimates are further evaluated by comparing simulated atmospheric CH₄ to surface in situ observations, vertical profiles of CH₄ made by aircraft, remotely sensed dry-air total column-averaged mole fraction (XCH₄) from the Total Carbon Column Observing Network (TCCON), and XCH₄ from the Greenhouse gases Observing Satellite (GOSAT). The evaluation with non-assimilated observations shows that posterior XCH₄ is better matched with the retrievals when the vertical mixing scheme with faster interhemispheric exchange is used. Estimated posterior mean total global emissions during 2000–2012 are 516±51Tg CH₄yr⁻¹, with an increase of 18Tg CH₄yr⁻¹ from 2000–2006 to 2007–2012. The increase is mainly driven by an increase in emissions from South American temperate, Asian temperate and Asian tropical TransCom regions. In addition, the increase is hardly sensitive to different model configurations ( < 2Tg CH₄yr⁻¹ difference), and much smaller than suggested by EDGAR v4.2 FT2010 inventory (33Tg CH₄yr⁻¹), which was used for prior anthropogenic emission estimates. The result is in good agreement with other published estimates from inverse modelling studies (16–20Tg CH₄yr⁻¹). However, this study could not conclusively separate a small trend in biospheric emissions (−5 to +6.9Tg CH₄yr⁻¹) from the much larger trend in anthropogenic emissions (15–27Tg CH₄yr⁻¹). Finally, we find that the global and North American CH₄ balance could be closed over this time period without the previously suggested need to strongly increase anthropogenic CH₄ emissions in the United States. With further developments, especially on the treatment of the atmospheric CH₄ sink, we expect the data assimilation system presented here will be able to contribute to the ongoing interpretation of changes in this important greenhouse gas budget.

Abstract. Data assimilation systems are used increasingly to constrain the budgets of reactive and long-lived gases measured in the atmosphere. Each trace gas has its own lifetime, dominant sources and sinks, and observational network (from flask sampling and in situ measurements to space-based remote sensing) and therefore comes with its own optimal configuration of the data assimilation. The CarbonTracker Europe data assimilation system for CO₂ estimates global carbon sources and sinks, and updates are released annually and used in carbon cycle studies. CarbonTracker Europe simulations are performed using the new modular implementation of the data assimilation system: the CarbonTracker Data Assimilation Shell (CTDAS). Here, we present and document this redesign of the data assimilation code that forms the heart of CarbonTracker, specifically meant to enable easy extension and modification of the data assimilation system. This paper also presents the setup of the latest version of CarbonTracker Europe (CTE2016), including the use of the gridded state vector, and shows the resulting carbon flux estimates. We present the distribution of the carbon sinks over the hemispheres and between the land biosphere and the oceans. We show that with equal fossil fuel emissions, 2015 has a higher atmospheric CO₂ growth rate compared to 2014, due to reduced net land carbon uptake in later year. The European carbon sink is especially present in the forests, and the average net uptake over 2001–2015 was 0. 17 ± 0. 11PgC yr⁻¹ with reductions to zero during drought years. Finally, we also demonstrate the versatility of CTDAS by presenting an overview of the wide range of applications for which it has been used so far.

Abstract. Climate change projections still suffer from a limited representation of the permafrost–carbon feedback. Predicting the response of permafrost temperature to climate change requires accurate simulations of Arctic snow and soil properties. This study assesses the capacity of the coupled land surface and snow models ISBA-Crocus and ISBA-ES to simulate snow and soil properties at Bylot Island, a high Arctic site. Field measurements complemented with ERA-Interim reanalyses were used to drive the models and to evaluate simulation outputs. Snow height, density, temperature, thermal conductivity and thermal insulance are examined to determine the critical variables involved in the soil and snow thermal regime. Simulated soil properties are compared to measurements of thermal conductivity, temperature and water content. The simulated snow density profiles are unrealistic, which is most likely caused by the lack of representation in snow models of the upward water vapor fluxes generated by the strong temperature gradients within the snowpack. The resulting vertical profiles of thermal conductivity are inverted compared to observations, with high simulated values at the bottom of the snowpack. Still, ISBA-Crocus manages to successfully simulate the soil temperature in winter. Results are satisfactory in summer, but the temperature of the top soil could be better reproduced by adequately representing surface organic layers, i.e., mosses and litter, and in particular their water retention capacity. Transition periods (soil freezing and thawing) are the least well reproduced because the high basal snow thermal conductivity induces an excessively rapid heat transfer between the soil and the snow in simulations. Hence, global climate models should carefully consider Arctic snow thermal properties, and especially the thermal conductivity of the basal snow layer, to perform accurate predictions of the permafrost evolution under climate change.

Abstract. To evaluate the impact of vapor diffusion on isotopic composition variations in snow pits and then in ice cores, we introduced water isotopes in the detailed snowpack model Crocus. At each step and for each snow layer, (1) the initial isotopic composition of vapor is taken at equilibrium with the solid phase, (2) a kinetic fractionation is applied during transport, and (3) vapor is condensed or snow is sublimated to compensate for deviation to vapor pressure at saturation.

We study the different effects of temperature gradient, compaction, wind compaction, and precipitation on the final vertical isotopic profiles. We also run complete simulations of vapor diffusion along isotopic gradients and of vapor diffusion driven by temperature gradients at GRIP, Greenland and at Dome C, Antarctica over periods of 1 or 10 years. The vapor diffusion tends to smooth the original seasonal signal, with an attenuation of 7 to 12% of the original signal over 10 years at GRIP. This is smaller than the observed attenuation in ice cores, indicating that the model attenuation due to diffusion is underestimated or that other processes, such as ventilation, influence attenuation. At Dome C, the attenuation is stronger (18%), probably because of the lower accumulation and stronger δ¹⁸O gradients.