Anthropogenic activities emit ~2,000 Mg y ⁻¹ of the toxic pollutant mercury (Hg) into the atmosphere, leading to long-range transport and deposition to remote ecosystems. Global anthropogenic emission inventories report increases in Northern Hemispheric (NH) Hg emissions during the last three decades, in contradiction with the observed decline in atmospheric Hg concentrations at NH measurement stations. Many factors can obscure the link between anthropogenic emissions and atmospheric Hg concentrations, including trends in the reemissions of previously released anthropogenic (“legacy”) Hg, atmospheric sink variability, and spatial heterogeneity of monitoring data. Here, we assess the observed trends in gaseous elemental mercury (Hg ⁰ ) in the NH and apply biogeochemical box modeling and chemical transport modeling to understand the trend drivers. Using linear mixed effects modeling of observational data from 51 stations, we find negative Hg ⁰ trends in most NH regions, with an overall trend for 2005 to 2020 of −0.011 ± 0.006 ng m ⁻³ y ⁻¹ (±2 SD). In contrast to existing emission inventories, our modeling analysis suggests that annual NH anthropogenic emissions must have declined by at least 140 Mg between the years 2005 and 2020 to be consistent with observed trends. Faster declines in 95th percentile Hg ⁰ values than median values in Europe, North America, and East Asian measurement stations corroborate that the likely cause is a decline in nearby anthropogenic emissions rather than background legacy reemissions. Our results are relevant for evaluating the effectiveness of the Minamata Convention on Mercury, demonstrating that existing emission inventories are incompatible with the observed Hg ⁰ declines.

This study systematically examines the global uncertainties and biases in the carbon dioxide (CO₂) mixing ratio provided by the Copernicus Atmosphere Monitoring Service (CAMS). The global greenhouse gas re-analysis (EGG4) data product from the European Centre for Medium-Range Weather Forecasts (ECMWF) was evaluated against ground-based in situ measurements from more than 160 of stations across the world. The evaluation shows that CO₂ re-analysis can capture the general features in the tracer distributions, including the CO₂ seasonal cycle and its strength at different latitudes, as well as the global CO₂ trend. The emissions and natural fluxes of CO₂ at the surface are evaluated on a wide range of scales, from diurnal to interannual. The results highlight re-analysis compliance, reproducing biogenic fluxes as well the observed CO₂ patterns in remote environments. CAMS consistently reproduces observations at marine and remote regions with low CO₂ fluxes and smooth variability. However, the model’s weaknesses were observed in continental areas, regions with complex sources, transport circulations and large CO₂ fluxes. A strong variation in the accuracy and bias are displayed among those stations with different flux profiles, with the largest uncertainties in the continental regions with high CO₂ anthropogenic fluxes. Displaying biased estimation and root-mean-square error (RMSE) ranging from values below one ppmv up to 70 ppmv, the results reveal a poor response from re-analysis to high CO₂ mixing ratio, showing larger uncertainty of the product in the boundaries where the CAMS system misses solving sharp flux variability. The mismatch at regions with high fluxes of anthropogenic emission indicate large uncertainties in inventories and constrained physical parameterizations in the CO₂ at boundary conditions. The current study provides a broad uncertainty assessment for the CAMS CO₂ product worldwide, suggesting deficiencies and methods that can be used in the future to overcome failures and uncertainties in regional CO₂ mixing ratio and flux estimates.

This work was developed on the scope of the national initiative of the Integrated Greenhouse Gas Monitoring System for Germany – ITMS, which focus on reducing and characterizing uncertainties/errors introduced in the full retrieval chain of models, making use of observational data. Determining the magnitude, cause, and agents of carbon fluxes is important to advance our current understanding of the carbon budget and cycle, permitting more accurate predictions of its future behaviour. Sources and sinks of CO2at the Earth's surface can, in principle, be estimated from atmospheric concentrations by inverting atmospheric transport in the atmospheric tracer inversions.Solving for CO2 fluxes in the inversions is highly desirable to better understand the carbon cycle but also to support policies aimed at reducing CO2 emissions. The Jena CarboScope inversion developed based on Bayesian inverse methods is used to obtain data-driven estimates of trace gas exchange, quantifying the large-scale sources and sinks of CO2.To make Jena CarboScope estimates more reliable in understanding sources, sinks and transport of atmospheric CO2from the surface into the troposphere, the reliability of this data product should be evaluated based on independent observations. Quantifying the quality of the inversion estimation by decomposing the inherent uncertainty components is a challenging and key component in product reliability and its use. The overall objective in validating and evaluating uncertainties of the Bayesian data product provided by Jena CarboScope, is to explicitly answer the question: How good is the inversion estimation?The assessment of these products is of special importance for further development and possibly allows for judging the trustworthiness of the inversion outcome.This study explores the potential of CarboScope to reproduce CO2 concentrations recorded during regular flights and aircraft campaigns during the past two decades. The inversion estimations are accomplished by forward runs performed with the inversion having the measurement locations as receptors.This study examines biases and uncertainties in the CarboScope estimations evaluated against flights’ data. It has been found that the CO2 simulated by CarboScope in the forward run (using the transport model TM3 for background concentration and STILT for regional signal) agree reasonably well, into the 10/90th percentile for 3-sigma of the distribution. On the other hand, the inversion exhibit some systematic biases at the edges of the distribution under and overestimating at high and lower mixing ratios, respectively. The CarboScope strength and concerns were enhanced by understanding the differences among observations and the inversion estimation. In comprehensive statistics comparing measurement data from hundreds of flights, we assess the compliance CarboScope`s CO2 estimation. Furthermore, we discuss the estimation and observation mismatches, exploiting the model constraints to reproduce atmospheric transport from the boundary layer to the upper troposphere. Understanding such constraints has the potential to reduce uncertainties of the atmospheric inversion estimates.