Achieving global climate goals requires more than scientific insight, it needs trusted, operational evidence on greenhouse gas (GHG) emissions and how they change as mitigation is put in place. That evidence increasingly comes from combining measurements from many sites and networks, often together with models and inventories. The challenge is not only measuring well, but delivering data that stay consistent over time, are comparable between sites, and are ready for routine use by different communities. Climate action therefore needs an operational level of data: regular releases with clear metadata and uncertainty information.A key part of this is traceability. Traceability means being able to answer simple questions about every value in a dataset: How was it measured? How was it calibrated? What corrections and quality checks were applied? Which software produced it? What does the uncertainty mean? This becomes especially important over time, because instruments, calibrations, and processing methods evolve, and users need to understand what changed and why.A practical blueprint will be presented for running traceable GHG and related tracer datasets at scale, based on the day-to-day experience of a large team of measurement scientists, data specialists, and modellers. The blueprint is built around tiered data releases, where products are published at different data levels (raw → quality controlled → derived products), each with uncertainty information appropriate to that level and clear links between levels. A recorded history of processing and version changes is maintained for every release, together with harmonised metadata and uncertainty fields so both people and machines can interpret the data in the same way. Practical operational tools are discussed, such as automated checks, written decision rules, routine reprocessing, and release practices that support stable identifiers and proper credit.Examples using tracer-based diagnostics, with radon as one example, show how good traceability enables routine, reproducible products that can be used directly in modelling and emissions workflows. The contribution closes with lessons learned on how to keep this working in practice, including coordination, shared standards, and training across teams.

Abstract. A large discrepancy of at least 10 Gg yr-1 exists between reported emissions of the potent greenhouse gas HFC-23 (CHF3, trifluoromethane) and emissions derived from atmospheric measurements. In-atmosphere production of HFC-23 from the breakdown of fluorinated source gases such as hydrofluorocarbons and hydrofluoroolefins contributes to this gap, but the magnitude of this source is weakly constrained. This uncertainty is due, in part, to limited experimental measurements of the photolysis quantum yield of trifluoroacetaldehyde (CF3CHO), a key degradation product which forms HFC-23 via photolysis. The parameters governing CF3CHO deposition are also poorly understood. Previous work reported an upper limit of the contribution of the in-atmosphere source to the global HFC-23 burden. Here, we use a 3D chemistry and transport model to further constrain this contribution, using recent estimates of source gas emissions, kinetic rate constants, photolysis rates and deposition parameters, as well as considering the uncertainties in these values. We find that in-atmosphere production of HFC-23 is in the range 0.013–0.035 Gg yr−1, significantly lower than previous estimates. This accounts for <0.5 % of the discrepancy between reported emissions and those derived from atmospheric observations, suggesting that this source makes a negligible contribution to the overall HFC-23 budget. As part of this work, we also calculate indirect global warming potentials for the HFC-23 source gases HFO-1234ze(E), HFO-1336mzz(Z) and HCFO-1233zd(E) and find that their impact on climate is up to ten times higher than previously reported.

Inverse modelling systems relying on Lagrangian Particle Dispersion Models (LPDMs) are a popular way to quantify greenhouse gas emissions using atmospheric observations, providing independent evaluation of countries' self-reported emissions. For each GHG measurement, the LPDM performs backward-running simulations of particle transport in the atmosphere, calculating source-receptor relationships (“footprints”). These reflect the upwind areas where emissions would contribute to the measurement. However, the increased volume of satellite measurements from high-resolution instruments like TROPOMI cause computational bottlenecks, limiting the amount of data that can be processed for inference. Previous approaches to speed up footprint generation revolve around interpolation, therefore still requiring expensive new runs. In this work, we present the first machine learning-driven LPDM emulator that once trained, can approximate satellite footprints using only meteorology and topography. The emulator uses Graph Neural Networks in an Encode-Process-Decode structure, similar to Google’s Graphcast [1], representing latitude-longitude coordinates as nodes in a graph. We apply the model for GOSAT measurements over Brazil to emulate footprints produced by the UK Met Office’s NAME LPDM, training on data for 2014 and 2015 on a domain of size approximately 1600x1200km at a resolution of 0.352x0.234 degrees. Once trained, the emulator can produce footprints for a domain of up to approximately 6500x5000km, leveraging the flexibility of GNNs. We evaluate the emulator for footprints produced across 2016 on the 6500x5000km domain size, achieving intersection-over-union scores of over 40% and normalised mean absolute errors of under 30% for simulated CH4 concentrations. As well as demonstrating the emulator as a standalone AI application, we show how to integrate it with the full GHG emissions pipeline to quantify Brazil’s emissions. This method demonstrates the potential of GNNs for atmospheric dispersion applications and paves the way for large-scale near-real time emissions emulation. [1] Remi Lam et al.,Learning skillful medium-range global weather forecasting. Science 382,1416-1421 (2023). DOI:10.1126/science.adi233

,2-Dibromotetrafluoroethane (C2Br2F4, Halon-2402, H-2402) is used as a fire suppressant due to its stability because it maintains a liquid state at room temperature with a relatively high boiling point. However, H-2402, containing bromine (Br), was identified as a potent ozone-depleting substance, with a destructive capacity six times higher than that of CFC-11. Therefore, under the Montreal Protocol, its production and consumption were phased out globally in 2010, with developed countries starting their phase out in 1994. Russia, the primary producer of H-2402, reportedly ceased production after 2000. For essential uses where no alternatives are available (e.g., military fire extinguishers, oil and gas pipelines), existing or recycled supplies of H-2402 are permitted to be used. Despite H-2402 being under strict international regulation, accurate reporting and statistical information on essential production and consumption by countries remain limited. This study analyzes the atmospheric mole fractions records of H-2402 measured from 2008 to 2020 at Gosan station, South Korea. While the background mole fractions of H-2402 at Gosan station are gradually decreasing at -0.01 ppt/yr, similar to the global decreasing trend, high pollution cases were continuously observed throughout the entire period. Also, the frequency of occurrence also increased by more than three times in 2020 compared to 2008. This increase in pollution signals, not observed at major Northern Hemisphere background monitoring stations (such as Mace Head, Trinidad Head, and Jungfraujoch), suggests potential regional emissions in eastern Asia. Based on long-term atmospheric observations, and a combined analysis using the Lagrangian particle dispersion model (FLEXPART) and the Bayesian inverse framework (FLEXINVERT+), we have estimated the annual regional emissions in eastern Asia. We present observation-based results on the long-term, regional-scale variability of H-2402 emissions over eastern Asia and their significant contributions from a global perspective.

Data-driven emulators are increasingly being used to learn and emulate physics-based simulations, reducing computational expense and run time. Here, we present a structured way to improve the quality of these high-dimensional emulated outputs, through the use of prototypes: an approximation of the emulator's output passed as an input, which informs the model and leads to better predictions. We demonstrate our approach to emulate atmospheric dispersion, key for greenhouse gas emissions monitoring, by comparing a baseline model to models trained using prototypes as an additional input. The prototype models achieve better performance, even with few prototypes and even if they are chosen at random, but we show that choosing the prototypes through data-driven methods (k-means) can lead to almost 10% increased performance in some metrics.

Monitoring greenhouse gas emissions and evaluating national inventories require efficient, scalable, and reliable inference methods. Top-down approaches, combined with recent advances in satellite observations, provide new opportunities to evaluate emissions at continental and global scales. However, transport models used in these methods remain a key source of uncertainty: they are computationally expensive to run at scale, and their uncertainty is difficult to characterise. Artificial intelligence offers a dual opportunity to accelerate transport simulations and to quantify their associated uncertainty. We present an ensemble-based pipeline for estimating atmospheric transport "footprints", greenhouse gas mole fraction measurements, and their uncertainties using a graph neural network emulator of a Lagrangian Particle Dispersion Model (LPDM). The approach is demonstrated with GOSAT (Greenhouse Gases Observing Satellite) observations for Brazil in 2016. The emulator achieved a ~1000x speed-up over the NAME LPDM, while reproducing large-scale footprint structures. Ensembles were calculated to quantify absolute and relative uncertainty, revealing spatial correlations with prediction error. The results show that ensemble spread highlights low-confidence spatial and temporal predictions for both atmospheric transport footprints and methane mole fractions. While demonstrated here for an LPDM emulator, the approach could be applied more generally to atmospheric transport models, supporting uncertainty-aware greenhouse gas inversion systems and improving the robustness of satellite-based emissions monitoring. With further development, ensemble-based emulators could also help explore systematic LPDM errors, offering a computationally efficient pathway towards a more comprehensive uncertainty budget in greenhouse gas flux estimates.

Atmospheric trace gas measurements can be used to independently assess national greenhouse gas inventories through inverse modelling. Here, atmospheric nitrous oxide (N2O) measurements are used to derive monthly U.K. N2O emissions for 2013-2022 – using the InTEM and RHIME inverse methods – and Swiss N2O emissions for 2017-2022 – using the ELRIS inverse method. We find mean U.K. emissions of 90.5±23.0 and 111.7±32.1 Gg N2O yr-1 for 2013-2022 and corresponding trends of -0.68±0.48 and -2.10±0.72 Gg N2O yr-2, respectively, derived using InTEM and RHIME. The 2013-2022 mean U.K. N2O emissions as reported by the U.K. National Atmospheric Emissions Inventory were relatively constant at 74 Gg N2O yr-1 across this period, which is 14-33% smaller than the U.K. emissions derived from atmospheric data. Top-down Swiss emissions of 10.8±3.8 Gg N2O yr-1 derived using atmospheric measurements were very comparable to those reported in the Swiss National Inventory: 11.5 (8.3 to 14.9) Gg N2O yr-1 over 2017-2021. Pronounced seasonal N2O emissions cycles are inferred in the U.K. and Swiss data with similar seasonal magnitudes observed in both countries. In the U.K., the primary seasonal peak occurs in the spring with a second smaller peak occurring in the late summer for certain years. The springtime peak has a long seasonal decline that contrasts with the sharp rise and fall of N2O emissions estimated from the bottom-up U.K. Emissions Model (UKEM). Similarly, Swiss seasonal N2O emissions peak during the summer with a second smaller peak also occurring in the late summer/early autumn for certain years. Bayesian inference is used to minimize the U.K. seasonal cycle mismatch between the average top-down (atmospheric data-based) and UKEM bottom-up (process model and inventory-based) seasonal emissions at a sub-sector level. Increasing agricultural manure management and decreasing synthetic fertiliser N2O emissions reduces some of the discrepancy between the average U.K. top-down and bottom-up seasonal cycles. Other possibilities could also explain these discrepancies, such as missing emissions from NH3 deposition, but these require further investigation.

Hydrochlorofluorocarbons (HCFCs) are ozone-depleting substances whose production and consumption have been phased out under the Montreal Protocol in non-Article 5 (mainly developed) countries and are currently being phased out in the rest of the world. Here, we focus on two HCFCs, HCFC-123 and HCFC-124, whose emissions are not decreasing globally in line with their phase-out. We present the first measurement-derived estimates of global HCFC-123 emissions (1993–2023) and updated HCFC-124 emissions for 1978–2023. Around 5 Gg yr−1 of HCFC-123 and 3 Gg yr−1 of HCFC-124 were emitted in 2023. Both HCFC-123 and HCFC-124 are intermediates in the production of HFC-125, a non-ozone-depleting hydrofluorocarbon (HFC) that has replaced ozone-depleting substances in many applications. We show that it is possible that the observed global increase in HCFC-124 emissions could be entirely due to leakage from the production of HFC-125, provided that its leakage rate is around 1 % by mass of HFC-125 production. Global emissions of HCFC-123 have not decreased despite its phase-out for production under the Montreal Protocol, and its use in HFC-125 production may be a contributing factor to this. Emissions of HCFC-124 from western Europe, the USA and East Asia have either fallen or not increased since 2015 and together cannot explain the entire increase in the derived global emissions of HCFC-124. These findings add to the growing evidence that emissions of some ozone-depleting substances are increasing due to leakage and improper destruction during fluorochemical production.

Atmospheric transport model (ATM) uncertainty continues to be a significant constraining factor in making confident top-down (inverse model based) GHG emission estimates. Despite its importance, accurately gauging model uncertainty and capturing its temporal fluctuations remains a challenge. Inversion frameworks typically involve an empirical selection of data to be assimilated whereby only the data from periods where the ATM has the lowest uncertainties are used for the inversion. There are numerous data filtering methods, that often depend on modelled parameters (mixing height, wind speed, potential temperature), which could result in data selection bias.To address this, we present analysis of radon measurements, a natural radioactive noble gas with simple and well-constrained source and sink. Radon’s unique characteristics make it an ideal tracer to study the transport and mixing of air and thus has potential to act as an independent metric to evaluate ATM performance. A new approach involves utilising measured and modelled radon (calculated using the Met Office Numerical Atmospheric Modelling Environment (NAME) dispersion model and radon flux map) to classify the ATM output uncertainty as either high (poor performance) or low (the best performance). This approach could be universally applied to any location measuring radon from a single inlet height and in conjunction with any other dispersion modelling scenarios. To evaluate the effectiveness of the radon selection method, we assess the methane (CH4) emissions across the UK using four tall tower sites (part of the Deriving Emissions linked to Climate Change - DECC network): Heathfield, Ridge Hill, Tacolneston and Weybourne. The CH4 emissions are estimated by the Met Office’s inversion modelling system – Inversion Technique for Emission Modelling (InTEM). We will compare how emissions sensitivity varies between our radon-based approach and the current selection method, which relies on model parameters and the vertical gradient of CH4 measurements. This comparative analysis aims to demonstrate the potential advantages of using radon as a tool for improving the accuracy of ATM performance assessments in GHG emission estimates.

HFC-23 (trifluoromethane) is a potent greenhouse gas, believed to be emitted to the atmosphere primarily as a by-product during the production of the refrigerant and feedstock HCFC-22 (chlorodifluoromethane). Due to the high global warming potential of HFC-23 (GWP100 ~ 14,700), the Kigali Amendment to the Montreal Protocol requires countries to limit their emissions of HFC-23 as much as possible and report these emissions to the United Nations Environment Programme. Global reported emissions have been in the range 2-3 Gg yr-1 since 2019 and reflect the near-total destruction of emissions from HCFC-22 production reported by the countries with major HCFC-22 manufacturers, such as China and India. However, atmospheric observations show that, whilst emissions fell from their maximum in 2019 of 17.3 ± 0.8 Gg yr-1 to 14.0 ± 0.9 Gg yr-1 in 2023, they remain many times higher than reported. In addition, regional inverse modelling was performed based on measurements from the AGAGE site at Gosan, South Korea, using three different Bayesian inverse models (FLEXINVERT+, InTEM and RHIME) to estimate emissions from eastern China. These inversions use the same observational data, but different transport models, baselines, priors and uncertainties. Results are compared to better quantify regional emissions and their uncertainties. The results suggest that emissions from eastern China are four to six times higher than reported for the whole of China. In addition, we examine the emission of HFC-23 as a by-product during the production of other hydrofluorocarbons and fluorochemicals. In-atmosphere HFC-23 production (from the breakdown of certain hydrofluoroolefins used as replacements for HFCs) is also investigated further using a 3D chemical transport model incorporating photolysis and ozonolysis reactions. Our results indicate that, based on currently available information, these potential alternative sources contribute less than 2.0 Gg yr-1 to global emissions. This suggests that HFC-23 emissions from HCFC-22 production have been consistently under-reported since the implementation of the Kigali Amendment. It therefore appears likely that abatement of HFC-23 emissions has not occurred to the extent reported in this period. Improved monitoring and verification of HFC-23 emissions from industrial sources is essential to the continued success and efficacy of the Kigali Amendment.

Hydrofluorocarbons (HFCs) are potent greenhouse gases whose global abundance continues to rise and subsequently warm the Earth. Southern China is a rapidly developing region that has experienced a sharp increase in its HFC consumption. Here, we present the first high-frequency HFC observations in Southern China from 2022 to 2023, analyzing the atmospheric mole fractions of four HFCs (HFC-134a, HFC-32, HFC-125, and HFC-143a) and using inverse modeling to estimate their emissions in Southern China. We find that HFC emissions in Southern China are primarily concentrated in Jiangsu, Zhejiang, and Guangdong, with Jiangsu having the highest HFC-134a emissions (4.1 ± 0.5 Gg yr-1, ± 1 standard deviation). HFC-125 and HFC-32 emissions are elevated in Anhui, Jiangsu, and Guangdong, while HFC-143a emissions are predominantly in Jiangsu and Zhejiang. From 2022-2023, HFC emissions in the Pearl River Delta are expected to increase, while in the Yangtze River Delta, HFC-134a, HFC-125, and HFC-32 emissions are 94.2% ± 54.6%, 200.9% ± 28.7%, and 187.5% ± 24.2% higher than 2012-2016 levels, respectively. The rise in HFC consumption and the delayed emissions from HFC banks in Southern China highlight the necessity of estimating HFC emissions. Our findings will support local emission reduction policies and contribute to global climate change efforts.

U.K. top-down nitrous oxide (N2O) emissions estimates for 2013-2022 Top-down emissions were derived using the InTEM (Manning et al., 2021) and RHIME (Ganesan et al., 2014) inverse models with atmospheric N2O mole fraction measurements from the U.K. DECC network (O'Doherty et al., 2020). This dataset provides monthly top-down emissions estimates for the U.K. (land-only, sea-only, land+sea components) with 68% confidence interval ranges. A priori total monthly N2O emissions created by the U.K. Emissions Model are also provided. This dataset was used in the publication "Combining top-down and bottom-up approaches to evaluate recent trends and seasonal patterns in U.K. N2O emissions" that has been submitted to JGR: Atmospheres by Saboya et al. Further information about the generation of this dataset and its implications can be found in Saboya et al. (2024). Please note InTEM emissions data hold a Crown Copyright.

Nitrogen trifluoride (NF3) is a potent and long-lived greenhouse gas that is widely used in the manufacture of semiconductors, photovoltaic cells, and flat panel displays. Using atmospheric observations from eight monitoring stations from the Advanced Global Atmospheric Gases Experiment (AGAGE) and inverse modeling with a global 3-D atmospheric chemical transport model (GEOS-Chem), we quantify global and regional NF3 emission from 2015 to 2021. We find that global emissions have grown from 1.93 +/- 0.58 Gg yr(-1) (+/- one standard deviation) in 2015 to 3.38 +/- 0.61 Gg yr(-1) in 2021, with an average annual increase of 10% yr(-1). The available observations allow us to attribute significant emissions to China (0.93 +/- 0.15 Gg yr(-1) in 2015 and 1.53 +/- 0.20 Gg yr(-1) in 2021) and South Korea (0.38 +/- 0.07 Gg yr-1 to 0.65 +/- 0.10 Gg yr(-1)). East Asia contributes around 73% of the global NF3 emission increase from 2015 to 2021: approximately 41% of the increase is from emissions from China (with Taiwan included), 19% from South Korea, and 13% from Japan. For Japan, which is the only one of these three countries to submit annual NF3 emissions to UNFCCC, our bottom-up and top-down estimates are higher than reported. With increasing demand for electronics, especially flat panel displays, emissions are expected to further increase in the future.

Sulfur hexafluoride (SF 6 ) is a potent greenhouse gas. Here we use long-term atmospheric observations to determine SF 6 emissions from China between 2011 and 2021, which are used to evaluate the Chinese national SF 6 emission inventory and to better understand the global SF 6 budget. SF 6 emissions in China substantially increased from 2.6 (2.3-2.7, 68% uncertainty) Gg yr −1 in 2011 to 5.1 (4.8-5.4) Gg yr −1 in 2021. The increase from China is larger than the global total emissions rise, implying that it has offset falling emissions from other countries. Emissions in the less-populated western regions of China, which have potentially not been well quantified in previous measurement-based estimates, contribute significantly to the national SF 6 emissions, likely due to substantial power generation and transmission in that area. The CO 2 -eq emissions of SF 6 in China in 2021 were 125 (117-132) million tonnes (Mt), comparable to the national total CO 2 emissions of several countries such as the Netherlands or Nigeria. The increasing SF 6 emissions offset some of the CO 2 reductions achieved through transitioning to renewable energy in the power industry, and might hinder progress towards achieving China’s goal of carbon neutrality by 2060 if no concrete control measures are implemented.

Top-down evaluation of the UK’s methane, nitrous oxide and halogenated greenhouse gas emissions has been possible since the 1990s, firstly due to measurements from the Advanced Global Atmospheric Gases Experiment (AGAGE), combined later with the UK national-scale Deriving Emissions linked to Climate Change (UK DECC) network. Here, we show that carbon dioxide-equivalent emissions inferred from observations of these gases have declined more slowly than stated in the UK’s National Inventory Report (NIR); a decline of approximately 50 Tg/yr CO2-e is found between 1990 and 2021, compared to approximately 100 Tg/yr CO2-e in the inventory. This difference, which is driven largely by a smaller-than-reported reduction in methane emissions, suggests that the UK may be approximately 3 years behind its stated progress toward net-zero. This paper will describe the evolution of greenhouse gas monitoring in the UK, including an overview of the first decade of national-scale emissions estimation from the UK DECC network. It will show how top-down emissions are calculated, and how atmospheric observation-based estimates are used, in close collaboration with inventory teams, to improve the national inventory. Finally, it will discuss the UK’s plans for a prototype “operational” emissions evaluation system, the Greenhouse gas Emissions Measurement and Modelling Advancement (GEMMA).

Rapid growth in the emissions of nitrogen trifluoride (NF3), a potent greenhouse gas, poses a threat to the environment and the climate system. This study estimated NF3 emissions and their spatial distribution in China from 2017 to 2021 based on atmospheric observations from nine background stations in China, by employing a Lagrangian-dispersion-model-based Bayesian inversion technique. We found that NF3 emissions in China increased from 0.95 (0.82-1.07) Gg yr(-1) in 2017 to 1.41 (1.28-1.55) Gg yr(-1 )in 2021, representing a substantial growth of 48% over this period. The absolute increase in NF3 emissions in China over 2017-2020, 0.65 (0.57-0.74) Gg yr(-1) is comparable to the increase in global emissions (0.63 (0.50-0.75) Gg yr(-1)) over the same period. We identified substantial NF3 emissions in the Pearl and Yangtze River Delta regions and Hubei Province, where well-established semiconductor industries could have contributed to NF3 emissions. Moreover, large NF3 emissions were identified in northern China, including Hebei, Henan, and Shandong Provinces. If control measures are not implemented, increasing NF3 emissions may delay China's progress toward achieving its carbon neutrality target by 2060.

The perfluorocarbons tetrafluoromethane (CF4, PFC-14) and hexafluoroethane (C2F6, PFC-116) are potent greenhouse gases with near-permanent atmospheric lifetimes relative to human timescales and global warming potentials thousands of times that of CO2. Using long-term atmospheric observations from a Chinese network and an inverse modeling approach (top-down method), we determined that CF4 emissions in China increased from 4.7 (4.2-5.0, 68% uncertainty interval) Gg y(-1) in 2012 to 8.3 (7.7-8.9) Gg y(-1) in 2021, and C2F6 emissions in China increased from 0.74 (0.66-0.80) Gg y(-1) in 2011 to 1.32 (1.24-1.40) Gg y(-1) in 2021, both increasing by approximately 78%. Combined emissions of CF4 and C2F6 in China reached 78 Mt CO2-eq in 2021. The absolute increase in emissions of each substance in China between 2011-2012 and 2017-2020 was similar to (for CF4), or greater than (for C2F6), the respective absolute increase in global emissions over the same period. Substantial CF4 and C2F6 emissions were identified in the less-populated western regions of China, probably due to emissions from the expanding aluminum industry in these resource-intensive regions. It is likely that the aluminum industry dominates CF4 emissions in China, while the aluminum and semiconductor industries both contribute to C2F6 emissions. Based on atmospheric observations, this study validates the emission magnitudes reported in national bottom-up inventories and provides insights into detailed spatial distributions and emission sources beyond what is reported in national bottom-up inventories.

Top-down evaluation of the UK’s methane, nitrous oxide and halogenated greenhouse gas emissions has been possible since the 1990s, firstly due to measurements from the Advanced Global Atmospheric Gases Experiment (AGAGE), combined later with the UK national-scale Deriving Emissions linked to Climate Change (UK DECC) network. Here, we show that carbon dioxide-equivalent emissions inferred from observations of these gases have declined more slowly than stated in the UK’s National Inventory Report (NIR); a decline of approximately 50 Tg/yr CO2-e is found between 1990 and 2021, compared to approximately 100 Tg/yr CO2-e in the inventory. This difference, which is driven largely by a smaller-than-reported reduction in methane emissions, suggests that the UK may be approximately 3 years behind its stated progress toward net-zero. This paper will describe the evolution of greenhouse gas monitoring in the UK, including an overview of the first decade of national-scale emissions estimation from the UK DECC network. It will show how top-down emissions are calculated, and how atmospheric observation-based estimates are used, in close collaboration with inventory teams, to improve the national inventory. Finally, it will discuss the UK’s plans for a prototype “operational” emissions evaluation system, the Greenhouse gas Emissions Measurement and Modelling Advancement (GEMMA).

We investigate the use of atmospheric oxygen (O2) and carbon dioxide (CO2) measurements for the estimation of the fossil fuel component of atmospheric CO2 in the UK. Atmospheric potential oxygen (APO) – a tracer that combines O2 and CO2, minimizing the influence of terrestrial biosphere fluxes – is simulated at three sites in the UK, two of which make APO measurements. We present a set of model experiments that estimate the sensitivity of APO simulations to key inputs: fluxes from the ocean, fossil fuel flux magnitude and distribution, the APO baseline, and the exchange ratio of O2 to CO2 fluxes from fossil fuel combustion and the terrestrial biosphere. To estimate the influence of uncertainties in ocean fluxes, we compare three ocean O2 flux estimates from the NEMO–ERSEM, the ECCO–Darwin ocean model, and the Jena CarboScope (JC) APO inversion. The sensitivity of APO to fossil fuel emission magnitudes and to terrestrial biosphere and fossil fuel exchange ratios is investigated through Monte Carlo sampling within literature uncertainty ranges and by comparing different inventory estimates. We focus our model–data analysis on the year 2015 as ocean fluxes are not available for later years. As APO measurements are only available for one UK site at this time, our analysis focuses on the Weybourne station. Model–data comparisons for two additional UK sites (Heathfield and Ridge Hill) in 2021, using ocean flux climatologies, are presented in the Supplement. Of the factors that could potentially compromise simulated APO-derived fossil fuel CO2 (ffCO2) estimates, we find that the ocean O2 flux estimate has the largest overall influence at the three sites in the UK. At times, this influence is comparable in magnitude to the contribution of simulated fossil fuel CO2 to simulated APO. We find that simulations using different ocean fluxes differ from each other substantially. No single model estimate, or a model estimate that assumed zero ocean flux, provided a significantly closer fit than any other. Furthermore, the uncertainty in the ocean contribution to APO could lead to uncertainty in defining an appropriate regional background from the data. Our findings suggest that the contribution of non-terrestrial sources needs to be better accounted for in model simulations of APO in the UK to reduce the potential influence on inferred fossil fuel CO2 using APO.