The central trick is a well-known chemical companionship. When fossil fuels burn, they release carbon dioxide and nitrogen oxides, collectively known as NOx, in proportions that depend on what is being burned and how. Carbon dioxide is notoriously difficult to measure from orbit because it mixes into a background already saturated with natural carbon-cycle fluxes; fresh anthropogenic emissions represent at most a few percent of the total column of gas above any location. Nitrogen oxides, by contrast, are short-lived, surviving only hours to a day in the atmosphere, so a satellite snapshot of NO2 reveals essentially fresh combustion activity. The researchers exploited this by measuring NO2 from the TROPOspheric Monitoring Instrument, or TROPOMI, aboard the Copernicus Sentinel-5 Precursor satellite, and then translating those observations into CO2 estimates using sector-specific NOx to CO2 emission ratios.

The technical machinery behind the study is an ensemble data assimilation system built around the GEOS-Chem atmospheric transport model. The team assimilated a full year of TROPOMI NO2 observations from 2021 into an Ensemble Kalman Filter, a statistical framework that uses an ensemble of 100 perturbed model states to represent and propagate uncertainty in emissions. Each day, the filter compared simulated NO2 columns, sampled at the time and location of satellite overpasses and convolved with scene-dependent averaging kernels, against the actual retrievals. It then adjusted emission scaling factors to reduce the mismatch, updating the model state before the next day’s simulation began. Because NOx is so short-lived, a one-day assimilation window proved sufficient, since observations are largely insensitive to emissions from previous days.

A key innovation made this computationally feasible. Running full atmospheric chemistry for 100 ensemble members, day after day, would be prohibitively expensive. Instead, the team used a lightweight offline treatment of NOx chemistry, developed in earlier work, in which the chemical rates of change are estimated by scaling a baseline rate according to the relative change in local NOx concentration. This parameterisation reproduces the full-chemistry results with remarkable fidelity, and the team extended its validation to the larger perturbations encountered during the inversion, finding correlation coefficients above 0.94 for chemical loss rates and near-perfect stability in the NO2 to NOx partitioning ratio. The result is a pathway that retains critical chemical feedbacks while enabling large-ensemble inversions that would otherwise be out of reach.

The results show that assimilating TROPOMI data systematically sharpens the picture of European emissions. The mean uncertainty in total European fossil fuel CO2 emissions fell from 5.6 percent in the prior estimate to 3.3 percent after assimilation, a reduction of roughly 40 percent. Model agreement with the satellite observations also improved, with the annual correlation between all daily model-observation pairs rising from 0.42 to 0.54. The largest uncertainty reductions appeared over major source regions such as London, Paris, and Madrid, where dense satellite coverage and strong emission signals combine to give the filter the most information to work with.

Perhaps the most eye-catching finding is the seasonal pattern that emerges when the posterior NOx fluxes are converted into CO2 estimates. The inferred emissions exhibit a far more pronounced seasonal cycle than the prior inventories, with elevated values in autumn and winter, most notably in February, November, and December, and a close correlation with surface temperature variability. During colder months, the posterior adjustments implied widespread increases in total European daily fluxes, ranging from 10 percent to as much as 177 percent, while warmer months showed decreases of up to 22 percent, particularly over northwestern Europe. The team attributes the winter amplification partly to the difficulty that bottom-up inventories face in capturing domestic heating, which is highly seasonal and depends heavily on meteorology, fuel use, and assumed temporal profiles.

The national-level adjustments were substantial. Annual combustion CO2 emissions increased relative to the prior in every one of the twelve high-emitting countries analysed, ranging from 20 percent in the United Kingdom to 91 percent in Turkey, with the changes exceeding the stated prior uncertainties by factors of roughly six to seventy. In Sweden, where February temperatures stayed below minus five degrees Celsius for the entire month, the satellite-derived estimates captured a distinct spike in emissions. Comparisons against the independent EDGAR inventory showed improved agreement for five countries, including Sweden, Romania, Turkey, Spain, and the United Kingdom, but deteriorated for the remaining seven, suggesting that some of the inferred increases may reflect limitations of the inversion framework rather than genuine biases in national reporting.

Independent evaluation against ground-based measurements lent cautious support to the approach. Comparisons with 202 monitoring stations from the European Environment Agency network showed consistent improvements in correlation, mean absolute error, and bias, with the largest gains during the colder months when the posterior flux adjustments were greatest. Evaluations against 47 in situ CO2 sites from the ICOS and UK DECC networks showed more mixed results, with roughly half to three-quarters of sites improving depending on the season, reflecting the weaker sensitivity of rural background stations to localised fossil fuel plumes and the strong influence of biogenic fluxes. Comparisons with NASA’s OCO-2 satellite column measurements of CO2 showed negligible improvement, underscoring just how difficult it remains to isolate the fossil fuel signal from the vast natural carbon cycle using column CO2 observations alone.

The authors are careful to frame the work as a proof of concept rather than a definitive correction to national inventories. A central limitation is the assumption of temporally fixed NOx to CO2 emission ratios, which in reality vary across sectors, regions, and seasons due to differences in combustion conditions, fuel composition, and emissions control technology. Non-combustion and biogenic NOx sources, particularly soil emissions in summer, were held fixed at prior values, meaning their uncertainties may have been compensated for through adjustments to combustion emissions. The reliance on a single daily satellite overpass also requires prescribed diurnal emission profiles. These caveats mean the inferred increases should be interpreted as initial estimates that warrant further validation and multi-species observational constraints.

Even so, the study arrives at a pivotal moment for climate accountability. As the Copernicus CO2M mission and GOSAT-GW prepare to deliver co-located CO2 and NO2 retrievals at high resolution, and geostationary platforms such as TEMPO add constraints on the diurnal cycle, the Edinburgh team’s reduced-complexity chemistry approach suggests that the historical trade-off between chemical accuracy and computational feasibility need not stand in the way of operational monitoring. The next step is a fully coupled NO2-CO2 inversion in which both species are assimilated simultaneously, allowing emission ratios themselves to be dynamically constrained. Such systems could move the world closer to a transparent, satellite-based verification capability, one capable of independently checking the emission pledges that underpin international climate accords and national net-zero targets.

Subject of Research: Satellite-based estimation of European fossil fuel CO2 emissions using TROPOMI nitrogen dioxide observations and sector-based NOx to CO2 emission ratios

Article Title: Inferring European fossil fuel CO2 emissions using TROPOMI NO2 data and sector-based NOx : CO2 emission ratios

Article References: Schooling, C. N., Feng, L., Super, I., & Palmer, P. I. (2026). Inferring European fossil fuel CO 2 emissions using TROPOMI NO 2 data and sector-based NO x : CO 2 emission ratios. Atmospheric Chemistry and Physics, 26(19), 13743-13766. https://doi.org/10.5194/acp-26-13743-2026

Image Credits: AI Generated

DOI: 10.5194/acp-26-13743-2026

Keywords: fossil fuel CO2 emissions, TROPOMI, nitrogen dioxide, NOx, data assimilation, Ensemble Kalman Filter, GEOS-Chem, satellite monitoring, Europe, carbon emissions verification, emission ratios, climate mitigation