Advancing urban air quality and climate research

Project’s scientific goals

The scientific goal of this project was to advance process-level understanding of urban air quality and greenhouse gas (GHG) emissions in a complex metropolitan environment by jointly investigating emissions, meteorology, and the nonlinear atmospheric chemistry governing secondary pollutant formation, using real-time observations and a range of modeling tools.

Summary of the work carried out

This project studied combined observations, modeling, and data-driven analysis of urban and regional air pollution. These include (i) an assessment of tropospheric nitrogen dioxide (NO₂), ozone (O₃), and fine particulate matter (PM₂.₅) responses to COVID-19 emission reductions over German cities using GEOS-Chem and real-time observations [1] [2]; (ii) developed a machine learning model for modeling spatiotemporal air pollutant concentrations across Germany [3]; (iii) the development and calibration of a low-cost sensor network in Munich [4]. These results highlight the importance of accurate emission characterization, nonlinear chemistry, and advanced modeling for understanding urban air quality.

In addition, two intensive measurement campaigns were conducted during summer 2023 and winter 2024 using a measurement container equipped with online volatile organic compounds (VOC) and particulate matter instruments. Measurements included VOCs, inorganic and organic aerosol components, particle size distributions, black carbon, and key trace gases (NO₂, O₃, NH₃), alongside detailed meteorology. A CHARON–PTR-ToF-MS provided alternating measurements of gas-phase VOCs and semi-volatile organics, while an HR-ToF-AMS quantified non-refractory PM₂.₅ composition at high temporal resolution. These observations and positive matrix factorization (PMF) methods were used to resolve seasonal sources of organic fine particulate matter [5]. The PALM4U LES model was implemented with nested domains over Munich at horizontal resolutions down to 10 m, targeting contrasting pollution episodes to diagnose emission–chemistry–transport interactions.

An international workshop was co-organized at TUM with Jia Chen and Frank N. Keutsch (Harvard University), bringing together researchers from Europe, North America, and Asia. Discussions focused on secondary pollutant formation, advanced modeling, sensor networks, and data-driven approaches.

The Focus Group also addressed gender-related challenges in the daily lives of researchers. Prof. Furth and Prof. List co-organized the Life & Science Career Symposium with the Collaborative Research Center 1371 on 10/10/2024 in Freising, featuring a broad selection of speakers and role models, covering opportunities, challenges, and obstacles in their career paths. The event culminated in a lively exchange with the audience, mostly young researchers who were eager to talk about how family can be fostered during an academic career and also had questions on sometimes overlooked topics such as taking care of elderly family members.

Major findings, outcomes, and their ­significance

Analyses of air quality changes during the COVID-19 lockdown period across Germany demonstrated that large reductions in anthropogenic emissions led to substantial decreases in NO₂ concentrations, while ozone levels showed weak or even opposite responses due to nonlinear photochemical effects and NOx-saturated regimes. These studies further revealed that PM₂.₅ concentrations decreased significantly less than NO₂, highlighting the resilience of secondary particulate matter formation pathways even under strongly reduced primary emissions. Together, these findings underscored the importance of atmospheric chemistry and oxidative capacity in controlling urban air quality, beyond simple emission–concentration relationships.

In parallel, a machine-learning study showed that data-driven approaches can successfully reproduce the spatiotemporal variability of NO₂ and O₃ across Germany at fine spatial resolution. While this work demonstrated strong predictive skill, it also highlighted the limited process-level interpretability of purely statistical models and the need for physically based frameworks to diagnose the underlying drivers of pollutant variability.

The LES framework was shown to realistically reproduce urban boundary-layer structure, wind fields, and pollutant dispersion, capturing strong spatial and temporal gradients that are unresolved in regional-scale models. The simulations revealed the critical role of urban morphology, turbulence, and street-level flow patterns in shaping pollutant concentrations and exposure.

The chemically resolved observations enabled robust source attribution of organic particulate matter and precursor gases. Biomass burning was identified as a dominant contributor during cold seasons, while biogenic VOC emissions and photochemical processing drove secondary organic aerosol formation in summer. Traffic and cooking emissions, although smaller contributors to PM mass, were found to exert a disproportionate influence on ultrafine particle number concentrations.

Overall, the combined findings demonstrate a data-driven analysis toward a mechanistic, urban-scale understanding of air quality. The project shows that only by integrating observational constraints, advanced modeling, and process-level interpretation can effective mitigation strategies for both primary and secondary urban air pollutants be developed.

Use of the final outcomes

The final outcomes of this project provide a comprehensive, process-level understanding of urban air quality and will be used in several ways:

• Machine-learning models of spatiotemporal pollutant variability can be combined with LES outputs to enhance predictive capability and operational air quality forecasting. The approach creates a transferable framework for other European cities, enabling hybrid observation–modeling–data-driven strategies for urban air quality management.
   
• PALM4U large-eddy simulations, constrained by high-time-resolution measurements, offer an improved framework for evaluating turbulence, pollutant dispersion, and chemistry-emission interactions in complex urban environments.
   
• The chemically resolved observations enable robust attribution of particulate matter and precursor sources, distinguishing traffic, cooking, biomass burning, and biogenic contributions.

Future research directions and suggestions

Future work should focus on tighter coupling between LES and detailed chemical mechanisms, improved representation of traffic and non-exhaust emissions, and assimilation of high-time-resolution observational data. Expanding the framework to include machine learning–assisted model correction, multi-seasonal simulations. This provides a scalable blueprint for next-generation urban air quality research. 

Selected publications

• V. Balamurugan et al., “Tropospheric NO₂ and O₃ response to COVID-19 lockdown restrictions at the national and urban scales in Germany,” Journal of Geophysical Research Atmospheres, vol. 126, no. 19, p. e2021JD035440, Sep. 2021, doi: 10.1029/2021jd035440.
   
• V. Balamurugan, J. Chen, Z. Qu, X. Bi, and F. N. Keutsch, “Secondary PM 2.5 decreases significantly less than NO2 emission reductions during COVID lockdown in Germany,” Atmospheric Chemistry and Physics, vol. 22, no. 11, pp. 7105–7129, Jun. 2022, doi: 10.5194/acp-22-7105-2022.
   
• V. Balamurugan, J. Chen, A. Wenzel, and F. N. Keutsch, “Spatiotemporal modeling of air pollutant concentrations in Germany using machine learning,” Atmospheric Chemistry and Physics, vol. 23, no. 17, pp. 10267–10285, Sep. 2023, 
  doi: 10.5194/acp-23-10267-2023.
   
• A. Abu-Hani, J. Chen, V. Balamurugan, A. Wenzel, and A. Bigi, “Transferability of ML-Based Global Calibration Models for NO₂ and NO Low-Cost Sensors,” Atmospheric Measurement Techniques Discussions, pp. 1-23, 2024.
   
• Y. Li et al., “Sources, concentrations, and seasonal variations of VOC and aerosol particles in downtown Munich in 2023/24,” EGUsphere [preprint], Nov. 2025, doi: 10.5194/egusphere-2025-5191. (forthcoming).

Full list of publications