Observation-based estimation of urban CO2 emissions can help cities track their pathway to net zero emissions, a goal many cities worldwide have adopted. While mesoscale atmospheric transport models are an effective component in inversion systems estimating country-level emissions, their use in urban-scale inversions presents a significant challenge. Here, we present one-year flux inversion results with the mesoscale ICON-ART atmospheric transport model for two cities with contrasting size and topographic complexity: Zurich and Paris. Inversions were performed with an ensemble square root filter, assimilating observations from a dense rooftop CO2 sensor network in Zurich and from a tall tower network in Paris. The inversion framework optimized gridded anthropogenic and biospheric fluxes, along with background mole fractions from eight inflow regions. Prior anthropogenic emissions were based on detailed inventories provided by local authorities. In Zurich, the inversion resulted in a posterior annual anthropogenic emission of 1012.3 +/- 38.8 ktyr-1, representing approximately a 30 % reduction compared to the prior, with the most significant decreases during winter periods of elevated ambient temperatures. In contrast, the posterior fluxes in Paris remained close to the prior, with an annual emission of 3580.0 +/- 101.9 ktyr-1, which is 7 % higher than the prior. This comparison highlights the influence of city-specific factors - such as topography, city size, and observational network - on the inversion system performance. Furthermore, our findings demonstrate the potential of mesoscale models to refine urban emission estimates, offering valuable insights for policymakers and researchers working to improve emission inventories and advance urban climate strategies.

Accurate emission tracking (e.g., locating and quantifying hot spots) using satellite images requires a good signal-to-noise ratio (SNR) of total column images. Achieving this SNR is challenging for satellite-based trace gas imagers, especially when enhancements are small relative to the background or small relative to retrieval uncertainty. Therefore, some satellites carry additional trace gas imagers with high SNR, such as NO2, which is co-emitted with the trace gas of interest. While NO2 is frequently used qualitatively for plume detection or plume fitting, its potential for quantitative noise reduction remains largely untapped. This paper presents two methods to enhance the SNR of total column images using co-registered NO2 images through minimum mean square error (MMSE) Bayesian denoising, which are a simple form of a Kalman filter or maximum a posteriori estimate. The first "joint MMSE" method relies on the presence of plumes in both the low- and co-registered high-SNR NO2 images. The second "self-similar MMSE" method utilizes image self-similarity and is based on an existing technique called BM3D. The methods are evaluated using a synthetic dataset (SMARTCARB) of atmospheric CO2 and NO2 concentrations, achieving over +40 dB improvement in peak SNR (i.e., an over 10(40/10) increase in SNR). Additionally, the methods are applied to TROPOMI SO2 and NO2 data over South Africa and used to compute a divergence image, demonstrating that an estimated 40 %-80 % noise reduction is possible. By enhancing the SNR of total column images, these techniques improve the detectability of subtle emission signals, which could benefit atmospheric monitoring applications.

Abstract Estimating biogenic CO2 fluxes is essential to quantify urban anthropogenic emissions, yet urban vegetation heterogeneity presents a significant challenge to making accurate estimations. We have developed an hourly temporal, 10‐m spatial resolution biogenic CO2 flux estimation framework based on the Vegetation Photosynthesis and Respiration Model (VPRM) and its variants (UrbanVPRM and VPRM‐modified). Unlike lower‐resolution models, our approach captures finer‐scale variability, particularly in fragmented urban green spaces like street trees and lawns. Results show that vegetation in Munich offsets 2.0%–2.8% of annual anthropogenic CO2 emissions in the study domain, with tree‐covered areas as primary sinks and grasslands as net sources. During summer, daytime CO2 uptake can match or exceed anthropogenic emissions. Evaluations employing city park field measurements and eddy covariance towers confirm strong performance of our models, while highlighting VPRM‐modified's advantage in grasslands and croplands, and UrbanVPRM's improvements in urban areas via impervious surface correction. These findings highlight the value of high‐resolution modeling in improving urban carbon flux assessments.

As part of the ICOS Cities project, a network of low-cost NDIR (non-dispersive infrared) CO2 sensors was set up across the city of Zurich (Switzerland), known as ZiCOS-L. The network consists of 56 sites with paired low-cost sensors spread out over the urban area of Zurich. This publication focuses on the period from August 2022 to July 2024. The sensors require in-field training for model calibration before deployment and further post-processing steps to account for drift and outliers. After data processing, the hourly mean root mean squared error (RMSE) was 13.6 +/- 1.4 ppm and the mean bias 0.75 +/- 1.67 ppm when validated against parallel reference measurements from the mid-cost sensor network ZiCOS-M. CO2 concentrations were highly variable with site means in Zurich ranging from 438 to 465 ppm. These differences can largely be explained by the nearby surroundings, with vegetation, traffic density and human activity being dominant factors while altitude and distance from the city centre had a minor effect. Vegetation (mainly grassland) amplified the morning concentration in summer by up to 20 ppm due to ecosystem respiration, while heavy traffic increased the morning rush hour concentration by 15 ppm. Human activity was shown to locally enhance CO2 concentrations during two public events. Despite its lower measurement accuracy, the ZiCOS-L network enables the study of concentration dynamics at a spatial and temporal scale that could only be achieved at much higher cost with mid-cost or high-precision instrumentation. The observations generated by ZiCOS-L will be further used in ICOS Cities activities to validate CO2 emission inventories with inversion modelling systems.

Climate change is primarily driven by anthropogenic emissions of greenhouse gases. Reducing these emissions globally requires a massive effort at the individual, city, and national scales. Urban areas are hotspots of anthropogenic emissions, given the high density of human-based activities, and clarity on the variations of emissions in these areas will enable effective targeted reduction plans. However, efforts to do so are hampered by a lack of direct measurements of temporal and spatial trends of the emissions. This study combines emission inventories with flux observations and footprint modeling in three pilot sites for urban emission studies (Zurich, Munich, and Paris) with a focus on carbon dioxide (CO2), carbon monoxide (CO), and methane (CH4). Results indicate that the sectors contributing most significantly to CO2 fluxes, stationary combustion and road transport, are consistent across the cities and require future reduction plans to target winter months and daytime hours (05:00-17:00 UTC). The sectors contributing to CO and CH4 fluxes vary by city and do not always have consistent seasonal or diurnal patterns. Results also provide a basis for improving emission inventories and temporal scaling factors across sites and species in order to achieve better agreement with observations.

Abstract. Accurate methane emission quantification is critical for climate mitigation efforts in the oil and gas industry. This study evaluates the performance of five academic methane measurement systems through single-blind controlled release testing at the TotalEnergies Anomaly Detection Initiatives (TADI) facility in France during June and September 2024. Vehicle-based teams from Technical University of Denmark, Heidelberg University, and a collaborative team from Utrecht University/LSCE/Cyprus Institute/Royal Holloway deployed mobile in situ measurement systems, while aircraft-based solutions from Empa/UZH and FAAM BAe-146 utilized hyperspectral imaging and airborne in situ measurements, respectively. Vehicle-based systems demonstrated strong detection capabilities with true positive rates of 93–100 % and minimum detection thresholds below 1 kg CH₄ h⁻¹. Quantification accuracy varied significantly, with slopes ranging from 0.38 to 1.04 when comparing estimated versus true emission rates. Aircraft systems showed more variable performance due to operational constraints and limited data availability. Post-unblinding analysis revealed critical insights into systematic errors, including background concentration calculation issues and wind measurement limitations. Low wind conditions (<2 m s⁻¹) particularly challenged quantification accuracy across all platforms. These findings highlight the importance of robust validation procedures and high-quality meteorological data for reliable methane emission quantification in real-world applications.

Atmospheric inversions are widely used to evaluate and improve inventories of methane (CH 4 ) emissions across scales from global to local, combining observations with atmospheric transport models. This study uses the dense network of in situ stations of the Integrated Carbon Observation System (ICOS) to explore how well in situ data can constrain European CH 4 emissions. Following the concept of inter-comparison studies of the atmospheric tracer transport model inter-comparison Project (TransCom), a CH 4 inverse inter-comparison modeling study has been performed, focusing on Europe for the period 2006–2018. The aim is to investigate the capability of inverse models to deliver consistent flux estimates at the national scale and evaluate trends in emission inventories, using a detailed dataset of CH 4 emissions described and presented here for first time. Study participants were asked to perform inverse modelling computations using a common database of a priori CH 4 emissions and in-situ observations as specified in a protocol. The participants submitted their best estimates of CH 4 emissions for the 27 European Union (EU-27) member states, the United Kingdom (UK), Switzerland, and Norway. Results were collected from 9 different inverse modelling systems, using 7 different global and regional transport models. The range of outcomes allows us to assess posterior emission uncertainty, accounting for transport model uncertainty and inversion design decisions, including a priori emission and model-data mismatch uncertainty. This paper presents inversion results covering 15 years, that are used to investigate the seasonality and trends of CH 4 emissions. The different inversion systems show a range of a posteriori emission adjustments, pointing to factors that should receive further attention in the design of inversions such as optimising background mole fractions. Most inverse models increase the seasonal cycle amplitude, by up to 400 Gg month −1 , with the largest adjustments to the a priori emissions in Western and Eastern Europe. This might be due to underestimation of emissions from wetlands during summer or the importance of seasonality in other microbial sources, such as landfills and waste water treatment plants. In Northern Europe, absolute flux adjustments are comparatively small, which could imply that the emission magnitude is relatively well captured by the a priori, though the lower station density could contribute also. Across Europe, the inverse models yield a similar decreasing trend in CH 4 emissions compared to the a priori emissions (−12.3 % instead of −9.1 %) from 2006 to 2018. While both the a priori and the a posteriori trend for the EU-27 are statistically significant from zero, their difference is not. On a subregional scale, the differences between a posteriori and a priori trends are more statistically significant over regions with more in-situ measurement sites, such as over Western and Southern Europe. Uncertainties in the a priori anthropogenic emissions, such as in the agriculture sector (cows, manure), or waste sector (microbial CH 4 emissions), but also in the a priori natural emissions, e.g. wetlands, might be responsible for the discrepancies between the a priori and a posteriori emission shift in the trends in Western, Eastern and Southern Europe. Our results highlight the importance of improving the inversion setup, such as the treatment of lateral boundary conditions and the model representation of measurement sites, to narrow the uncertainty ranges further. The referenced dataset related to the analysis and figures are available at the ICOS portal: https://doi.org/10.18160/KZ63-2NDJ (Ioannidis et al., 2025).

Inverse modelling is employed to reconcile greenhouse gas (GHG) emission inventories, based on bottom-up methods, with the observed atmospheric GHG concentrations. The Community Inversion Framework (CIF) was created to unify inverse-model developments and simplify the generation of inversions. It makes atmospheric transport models and inversion algorithms easily interchangeable and facilitates the comparison of inversion results obtained using such diverse components.After several years of development and the coupling of CIF with a wide range of transport models used by the inversion community, we present the first intercomparison study conducted with CIF. This exercise focuses on Europe and aims to refine CO₂ natural emissions for the year 2019, following a strict protocol. It involves five transport models (CHIMERE, ICON-ART, LMDz, STILT, and WRF-CHEM) and two inversion algorithms (variational and ensemble-based). Two additional transport models, TM5 and FLEXPART, will be incorporated in the near future.The results show a good agreement, both across transport models, and inversion algorithms. It paves the way towards using CIF as an operational tool for intercomparison studies. It also highlights its strong potential to support the systematic derivation of GHG budgets with multiple transport models, enable a proper and easy quantification of the modelling uncertainty, and improve the robustness of emission estimates, for any relevant atmospheric species, at any scale.

Abstract. Radiocarbon (14C) is a valuable tracer to determine the relative fossil fractions of emitted carbonaceous greenhouse gases, such as CO2 and CH4. While atmospheric Δ14CO2 measurements have been conducted at multiple sites for several decades, Δ14CH4 measurements remain more limited, mainly due to measurement challenges. In addition, 14CH4 emissions from nuclear power plants (NPPs) can complicate data interpretation. In this study, fortnightly Δ14CH4 and Δ14CO2 measurements at the Swiss High-Altitude Research Station Jungfraujoch (JFJ, about 3500 m a.s.l.) between 2019 and 2024 are presented. Over this period, Δ14CH4 values showed an increase from 350 ± 19 ‰ to 381 ± 13 ‰, while Δ14CO2 values decreased from −2.0 ± 3.8 ‰ to −12.7 ± 2.0 ‰, respectively. The former is related to the slight increase of 14CH4 emissions from the nuclear industry over the last years, while the latter is linked to the continued dilution of the 14CO2 signal due to the release of 14C-devoid CO2 from combustion of fossil fuels. Despite its high elevation, JFJ is still influenced by NPPs operating in Europe. To assess the nuclear 14C contribution to our individual measurements, we use a combination of in situ 222Radon measurements and Lagrangian particle dispersion model convolved with bottom-up inventory of 14C emissions from NPPs. Furthermore, our Δ14CH4 measurements reasonably agree with simulated atmospheric values of Δ14CH4 estimated by a global atmospheric one-box model and an estimation of global nuclear 14CH4 emissions.

Atmospheric concentration of methane (CH4), a critical greenhouse gas, increased significantly since pre-industrial times, with anthropogenic emissions originating primarily from agriculture, fossil fuel use and waste management. However, considerable uncertainties persist in the detection and quantification of anthropogenic CH4 emissions. In this study, we present first CH4 observations, plume detections and emission estimates from the new state-of-the-art Airborne Visible InfraRed Imaging Spectrometer 4 (AVIRIS-4), which participated in a blind controlled release experiment in September 2024 in southern France. We used an albedo-corrected matched filter to retrieve CH4 maps from the spectral images and estimated CH4 emission with the Integrated Mass Enhancement (IME) and Cross-Sectional Flux (CSF) methods. Our results demonstrate that AVIRIS-4 can reliably detect emissions as low as 5.5 kg CH4 h(-1) under good weather conditions at low flight altitudes (<1500 m) and 1.45 kg CH4 h(-1) under ideal conditions. While AVIRIS-4 provides highly accurate CH4 maps at <0.5 m resolution, emission estimation is limited by the accuracy of the effective wind speed, whose uncertainty and natural variability contribute substantially to the overall uncertainty. Using wind speed at source height performs well for small releases (below 20 kg CH4 h(-1)) (rRMSE = 1.065; rMBE = 0.361) and overall (rRMSE = 0.702; rMBE = -0.204). Using literature-derived effective wind speeds improves the apparent fit between estimated and reported CH4 emissions, but degrades performance both in overall agreement (rRMSE = 2.098; rMBE = 0.964) and for low-emission events (rRMSE = 2.367; rMBE = 1.711). Interestingly, the high spatial resolution makes it possible to retrieve the cast shadow of the CH4 plume, which can be used to estimate source and plume height, and could provide an approach for better constraining the height-dependency of the effective wind speed. On the bottom line, the controlled release experiment provides critical insights into the sensor's capabilities and guides further improvements to detect and quantify low intensity sources in the fossil fuel and waste management sectors, with implications for more accurate global greenhouse gas monitoring.

We assess future air quality in Europe, expected from the transition to a net-zero greenhouse gas society, using the comprehensive chemistry transport model ICON-ART. Simulations are conducted based on both present-day emissions and a Maximum Feasible Reduction (MFR) anthropogenic emission scenario for 2050. For current conditions, we provide the first evaluation of the ICON-ART model with MOZART-T1 chemistry against EMEP measurements across Europe, focusing on ground-level ozone (O _3 ) and nitrogen oxides (NO _x ). The model closely matches observed mean levels and the temporal and spatial variability of O _3 . However, it slightly underestimates daytime NO _x levels, as well as the morning peak in summer, though overall agreement with ground-based observations remains strong. For a one-month summer period under current conditions, we found that the maximum daily 8 h mean ozone threshold of 100 µg/m3 was exceeded on at least five days at 45

Zurich aims for net-zero direct greenhouse gas emissions by 2040, a target supported by 75 % of voters. Progress is tracked through a detailed CO2 inventory covering energy, transport, industry, and waste. Under the European ICOS Cities project, a monitoring program was launched using two approaches: (i) a network of mid- and low-cost CO2 sensors combined with atmospheric inverse modeling, and (ii) CO2 flux measurements from an eddy-covariance system on a city-center high-rise building, paired with footprint modeling.Here, we focus on the mid-cost (ZiCOS-M) and low-cost (ZiCOS-L) NDIR (nondispersive infrared) CO2 networks, which were both operational for at least 3 years since 2022.ZiCOS-M consists of 26 monitoring sites, 21 in the city and 5 outside the urban area. Daily calibrations using two reference gas cylinders, and corrections of the sensors’ spectroscopic response to water vapour were performed. The hourly mean root mean squared error (RMSE) was 0.98 ppm (0.46 - 1.5 ppm) and the mean bias ranged between 0.72 and 0.66 ppm compared to parallel measurements with a high-precision reference gas analyser for a period of 2 weeks or more. CO2 concentrations in the city were highly variable with site means ranging from 434 to 460 ppm, and Zurich’s mean urban CO2 increment was 15.4 ppm above the regional background.ZiCOS-L consists of 56 sites with paired sensors. The sensors require in-field training for model calibration before deployment and further post-processing steps to account for drift and outliers. After data processing, the hourly RMSE was 13.6±1.4 ppm, and the mean bias 0.75±1.67 ppm when validated against parallel reference measurements from ZiCOS-M. CO2 concentrations were highly variable with site means in Zurich ranging from 438 to 465 ppm, reflecting mainly the influence of sources in the nearby surroundings. Vegetation (mainly grassland) amplified the morning concentration on average in summer by up to 20 ppm due to ecosystem respiration, while heavy traffic increased the morning rush hour concentration by 15 ppm. Despite its lower measurement accuracy, the ZiCOS-L network enables the study of concentration dynamics at a spatial and temporal scale that is not accessible by any other means.The ZiCOS-M data was extensively used to derive top-down CO2 emissions. Similar modelling activities are currently ongoing with the ZiCOS-L data, and both are compared to emissions derived from the eddy covariance system and to the city's emission inventory. Grange SK, … Emmenegger L, The ZiCOS-M CO2 sensor network: measurement performance and CO2 variability across Zurich. https://doi.org/10.5194/acp-25-2781-2025.Creman L, … Bernet L, The Zurich Low-cost CO2 sensor network (ZiCOS-L): data processing, performance assessment and analysis of spatial and temporal CO2 dynamics. https://doi.org/10.5194/egusphere-2025-3425Brunner D, … Emmenegger L, Building-resolving simulations of anthropogenic and biospheric CO2 in the city of Zurich with GRAMM/GRAL. https://doi.org/10.5194/acp-25-14279-2025.Hilland R, … Christen A, Sectoral attribution of greenhouse gas and pollutant emissions using multi-species eddy covariance on a tall tower in Zurich, Switzerland. https://doi.org/10.5194/acp-25-14279-2025.Ponomarev N, … Brunner D, Estimation of CO2 fluxes in the cities of Zurich and Paris using the ICON-ART CTDAS inverse modelling framework. https://doi.org/10.5194/egusphere-2025-3668.

Accurate and efficient modeling of atmospheric composition, including aerosols and trace gases and their interactions with radiation, clouds, and dynamics is essential for improving predictions and understanding of air quality, weather, climate, and related health impacts. The Aerosols and Reactive Trace gases (ART) component extends the ICOsahedral Nonhydrostatic (ICON) modeling framework by enabling online, fully coupled simulations of atmospheric composition processes across scales. ICON-ART includes modules for emissions, transport, gas-phase chemistry, and aerosol microphysics in both the troposphere and stratosphere, allowing for the investigation of feedbacks between atmospheric composition and physical processes from the large-eddy to global scale.This paper presents an updated overview of the ICON-ART framework as implemented in version 2025.10, highlighting recent developments in emission parameterizations, chemical mechanisms, aerosol processes, and coupling to the physical core of ICON via aerosol-radiation and aerosol-cloud interactions. We summarize the structure of the code infrastructure and demonstrate the model's flexibility and scalability across a wide range of applications. ICON-ART provides a unified and modular platform for research and operational use in atmospheric composition, bridging the gap between regional air quality modeling and global Earth system simulations.

Atmospheric inversions are widely used to evaluate and improve inventories of methane (CH4) emissions across scales from global to local, combining observations with atmospheric transport models. This study uses the dense network of in situ stations of the Integrated Carbon Observation System (ICOS) to explore how well in situ data can constrain European CH4 emissions. Following the concept of inter-comparison studies of the atmospheric tracer transport model inter-comparison Project (TransCom), a CH4 inverse inter-comparison modeling study has been performed, focusing on Europe for the period 2006-2018. The aim is to investigate the capability of inverse models to deliver consistent flux estimates at the national scale and evaluate trends in emission inventories, using a detailed dataset of CH4 emissions described and presented here for first time.Study participants were asked to perform inverse modelling computations using a common database of a priori CH4 emissions and in-situ observations as specified in a protocol. The participants submitted their best estimates of CH4 emissions for the 27 European Union (EU-27) member states, the United Kingdom (UK), Switzerland, and Norway. Results were collected from 9 different inverse modelling systems, using 7 different global and regional transport models. The range of outcomes allows us to assess posterior emission uncertainty, accounting for transport model uncertainty and inversion design decisions, including a priori emission and model-data mismatch uncertainty.This paper presents inversion results covering 15 years, that are used to investigate the seasonality and trends of CH4 emissions. The different inversion systems show a range of a posteriori emission adjustments, pointing to factors that should receive further attention in the design of inversions such as optimising background mole fractions. Most inverse models increase the seasonal cycle amplitude, by up to 400 Gg month-1, with the largest adjustments to the a priori emissions in Western and Eastern Europe. This might be due to underestimation of emissions from wetlands during summer or the importance of seasonality in other microbial sources, such as landfills and waste water treatment plants. In Northern Europe, absolute flux adjustments are comparatively small, which could imply that the emission magnitude is relatively well captured by the a priori, though the lower station density could contribute also.Across Europe, the inverse models yield a similar decreasing trend in CH4 emissions compared to the a priori emissions (-12.3 % instead of -9.1 %) from 2006 to 2018. While both the a priori and the a posteriori trend for the EU-27 are statistically significant from zero, their difference is not. On a subregional scale, the differences between a posteriori and a priori trends are more statistically significant over regions with more in-situ measurement sites, such as over Western and Southern Europe.Uncertainties in the a priori anthropogenic emissions, such as in the agriculture sector (cows, manure), or waste sector (microbial CH4 emissions), but also in the a priori natural emissions, e.g. wetlands, might be responsible for the discrepancies between the a priori and a posteriori emission shift in the trends in Western, Eastern and Southern Europe.Our results highlight the importance of improving the inversion setup, such as the treatment of lateral boundary conditions and the model representation of measurement sites, to narrow the uncertainty ranges further. The referenced dataset related to the analysis and figures are available at the ICOS portal: 10.18160/KZ63-2NDJ (Ioannidis et al., 2025).

Eddy-covariance measurements allow us to directly monitor the vertical turbulent CO2 flux at a specific point in the urban atmosphere. Under some assumptions such as stationarity and sufficient turbulence, this flux corresponds to the net emissions in a variable footprint area. Combined with a footprint model and a biospheric CO2 flux model, this method has a high potential for validating and optimizing urban emission inventories. However, the reliability of EC measurements depends on a careful site selection, data processing and quality control. Often, sensor heights below z=50 m a.g.l. are chosen to mitigate issues associated with horizontal heterogeneity, storage flux, and horizontal and vertical advection. The storage flux describes the temporal change of the CO2 amount in the control volume between the surface and sensor height. Tall-tower sites (z>50 m a.g.l.) would be beneficial to capture emissions from a larger part of the city but require careful consideration of these issues. While a few studies have reported plausible EC measurements for urban tall-tower sites, little is known about the impact of the storage flux and advection terms. In the ICOS-Cities project, tall-tower EC systems and networks of mid-cost and low-cost CO2 concentration sensors were installed in three cities. Here, we aim to better quantify the storage flux and identify periods with horizontal advection by leveraging data from the sensor networks in Zurich, Switzerland, and Munich, Germany, and thus improve the reliability of the observed net CO2 emissions. The low-cost sensors were deployed in the urban canopy layer while the mid-cost sensors were mostly located at the rooftop level and collocated with wind and temperature sensors. We estimate the storage flux by dividing the control volume into three to four layers and averaging data from different sensors in the same layer. The storage flux is then added to the turbulent flux to estimate net surface emissions. To filter out periods in which this estimate is biased by horizontal advection, we consider horizontal CO2 gradients determined using mid-cost sensors at rooftop sites. This approach is compared to the often-used filtering with a friction velocity threshold.As expected, the storage flux is most important on days with a pronounced diurnal cycle in atmospheric stability. It reduces the net CO2 emission estimates in the morning hours after sunrise and generally increases these estimates at night. From 1.5 to 5 h after sunrise, this effect amounts on average to -7.3 and -8.0 µmol m-2 s-1 in Zurich and Munich, respectively, while in the first 3.5 hours after sunset, it amounts to +4.7 and +3.0 µmol m-2 s-1 (46% and 24% of the turbulent flux) in Zurich and Munich, respectively. On days with a small diurnal cycle in stability, the storage flux plays a smaller role, especially in winter. We will also present insights in the frequency of horizontal advection and favorable conditions for it. Finally, we will discuss the plausibility of median diurnal cycles of the derived net CO2 emissions, considering the directional dependence on land cover and associated sources and sinks.

Urban canopy schemes are essential for urban climate modeling, yet their performance depends on the level of detail in Urban Canopy Parameters (UCPs). In this study, we evaluate the ICON TERRA_URB urban scheme against dense urban observations (near-surface sites, higher-level sites, and eddy-covariance towers) and satellite surface-temperature products, focusing on air and surface temperature, sensible and latent heat fluxes, and wind speed over Zurich and Basel during summer 2023 using four experimental configurations at 500 m resolution. These include simulations without an urban canopy scheme (No TU), TERRA_URB with spatially uniform UCPs (Constant TU), local-climate-zone-based spatially varying parameters (LCZ TU), and city-specific urban canopy parameters (Real TU). Results show that activating TERRA_URB, regardless of parameter detail, provides the largest improvement by reducing the nocturnal cold bias, increasing urban surface temperatures, reproducing slightly higher nocturnal wind speeds, and improving surface energy flux partitioning. Spatially varying UCPs further refine the simulations. In particular, Real TU best captures the nocturnal urban signal, improving agreement with observed nighttime temperatures and sensible heat fluxes, although some biases remain in city centers. Their added benefit is modest compared to the main gain from activating the urban scheme itself. A trade-off emerges aloft, as TERRA_URB tends to overestimate warming, with No TU often performing better at most higher-level sites, although this may also reflect coupled boundary-layer processes beyond the urban scheme itself. These results suggest that while bulk urban canopy schemes effectively capture surface exchanges, improved boundary-layer representation likely requires multi-layer canopy approaches.

As part of the ICOS Cities project, a network of low-cost NDIR (non-dispersive infrared) CO 2 sensors was set up across the city of Zurich (Switzerland), known as ZiCOS-L. The network consists of 56 sites with paired low-cost sensors spread out over the urban area of Zurich. This publication focuses on the period from August 2022 to July 2024. The sensors require in-field training for model calibration before deployment and further post-processing steps to account for drift and outliers. After data processing, the hourly mean root mean squared error (RMSE) was 13.6 ± 1.4 ppm and the mean bias 0.75 ± 1.67 ppm when validated against parallel reference measurements from the mid-cost sensor network ZiCOS-M. CO 2 concentrations were highly variable with site means in Zurich ranging from 438 to 465 ppm. These differences can largely be explained by the nearby surroundings, with vegetation, traffic density and human activity being dominant factors while altitude and distance from the city centre had a minor effect. Vegetation (mainly grassland) amplified the morning concentration in summer by up to 20 ppm due to ecosystem respiration, while heavy traffic increased the morning rush hour concentration by 15 ppm. Human activity was shown to locally enhance CO 2 concentrations during two public events. Despite its lower measurement accuracy, the ZiCOS-L network enables the study of concentration dynamics at a spatial and temporal scale that could only be achieved at much higher cost with mid-cost or high-precision instrumentation. The observations generated by ZiCOS-L will be further used in ICOS Cities activities to validate CO 2 emission inventories with inversion modelling systems.

To support the European Green Deal and to assist cities in reaching net-zero emissions, we have developed an urban CO2 emission monitoring system combining a mesoscale atmospheric transport and inversion model with measurements from dense sensor networks. We have set up such a system for the city of Zurich, which includes a comprehensive measurement network and intensive campaigns conducted in the framework of the ICOS Cities project to provide a rich dataset for data assimilation and model validation. The network includes low- and mid-cost CO2 sensors and a tall flux tower. For CO2 data assimilation, we primarily use observations from the 21 mid-cost sensors, in particular the 14 sensors installed on rooftops, as they are easier for the model to reproduce. Additionally, we used measurements from three background sites located outside the city, as well as wind speed and temperature observations from meteorological sensors installed at most of the rooftop sensor locations. The atmospheric transport model ICON-ART was set up at a high resolution of about 600 m to resolve the complex topography of the area. The model domain extends about 60 km in the north-south and east-west directions, encompassing the city and all background stations. CO2 background concentrations at the domain boundaries were taken from a separate European-wide simulation, which itself was nested into global inversion-optimized CO2 simulations from the Copernicus Atmospheric Monitoring Service (CAMS). Prior anthropogenic emissions were based on the TNO-GHGco inventory for the European domain and on a composite of three inventories of increasing detail for the high-resolution domain, TNO-GHGco, a Swiss national inventory, and a Zurich city inventory. Another important source and sink of CO2 is the exchange with vegetation, which was calculated online in ICON-ART using the Vegetation Photosynthesis and Respiration Model (VPRM). Based on comparisons with observations, we continuously improved the forward simulations by introducing high-quality land-cover data, emissions from human respiration, and temporal profiles for the heating sector accounting for daily temperatures. Anthropogenic emissions and biospheric fluxes (respiration and gross photosynthetic production separately) are inversely estimated by coupling ICON-ART with the “CarbonTracker Data Assimilation Shell” (CTDAS), which employs an ensemble Kalman smoother to optimize a large number of flux scaling factors. Here we present our initial inversion experiments with both synthetic and real observations. The idealized setup with synthetically generated observations was used to optimize the system before applying it to real observations. Fluxes were estimated on a weekly scale at a grid cell level for multiple months between July 2022 and July 2023. The simulations show generally good agreement with the observations, but estimating anthropogenic emissions is challenging due to uncertainties in the biospheric fluxes and background CO2 concentrations. In its current state of development, the combination of measurements and the model allow reliable emission estimations mainly in winter when the regional anthropogenic CO2 signal is at its highest (20 – 50 ppm) and the biospheric signal is at its lowest.Acknowledgements: ICOS-Cities/PAUL, has received funding from the European Union's H2020 Programme under grant agreement No. 101037319