Aristotle University Of Thessaloniki (AUTH)
Development of an inverse model simulation methodology for localizing and assessing the strength of unspecified atmospheric pollution sources
Abstract
dc:descriptionThe rapid increase in urbanization over recent decades has significantly heightened human exposure to airborne pollutants. Air pollutants such as particulate matter (PM), nitrogen oxides (NOx), sulphur dioxide (SO2), carbon monoxide (CO), and ozone (O3) contribute to both short- and long-term health issues, including respiratory and cardiovascular diseases, as well as cancer. Additionally, accidental or deliberate releases of highly toxic chemical, biological, radiological, or nuclear (CBRN) substances could have catastrophic effects, particularly in densely populated urban areas. To mitigate air pollution exposure and prepare for toxic airborne releases, authorities and policymakers, such as the European Commission, must implement effective measures. In this regard, accurate information on air pollutant concentration levels and emission sources can be obtained using in-situ measurements and remote sensing, as well as Atmospheric Transport and Dispersion (ATD) models. At the same time, various approaches and methodologies analyse different aspects of air pollution and its negative impacts on the population and the environment. Inverse dispersion modelling is a key research area focused on identifying and quantifying unknown air pollutant source characteristics. By combining ATD model calculations with sensor observations and appropriate algorithms, these techniques can determine the location and release rate of unknown sources, either related to CBRN emissions or established source categories that can significantly impact Air Quality (AQ) levels. Additionally, they can quantify emissions from multiple known sources. These methodologies operate across different scales, from global and continental to the national and urban levels, utilizing the appropriate ATD model for each application. The primary objective of this thesis is to develop, apply, and evaluate inverse dispersion modelling methodologies for both emergency response to unknown CBRN releases and urban AQ assessment. To achieve this, numerical simulations from Computational Fluid Dynamics (CFD) dispersion models, sensor network measurements, and inverse algorithms are utilized. CFD models, known for their high spatial resolution, provide accurate street-scale predictions of airflow and pollutant dispersion by resolving turbulence caused by wind interactions with urban structures. These turbulence effects play a crucial role in determining pollutant concentration levels within complex urban environments. Various approaches, methods and algorithms have been evaluated for developing a robust urban inverse modelling system. The present thesis is structured into two main parts, each based on a specific case study. The first one, the Michelstadt wind tunnel experiment, which is used to assess and analyse inverse modelling applications and approaches based on its geometry and measurement dataset. The second one, the Augsburg city centre, serves as the case for developing the inverse modelling system in a real urban setting. In the Michelstadt case study, various CFD approaches and inverse algorithms were tested against reliable observational data. A comparison between Reynolds Averaged Navier-Stokes (RANS) and Large Eddy Simulation (LES) (used in a RANS-like method) approaches, both using the same inverse algorithm, showed similar accuracy in estimating source location and release rate of an unknown source. However, LES required over an order of magnitude more computational time than RANS. In a second application, RANS simulations for wind flow and pollutant dispersion were validated against wind tunnel measurements. Furthermore, the Bayesian-based Metropolis-Hastings Markov Chain Monte Carlo (MCMC) algorithm was tested in three independent source release scenarios, achieving high accuracy in estimating the unknown source term. A third application examined different sensor configurations in Source Term Estimation (STE) results for an unknown source, analysing the impact of sensor number and ATD model accuracy. Results showed that increasing the number of sensors and improving ATD model accuracy generally enhanced algorithm predictions. However, beyond a certain number of sensors in each release scenario, additional sensors or further improvements in ATD model performance had negligible impact on STE results. The RANS approach and the Metropolis-Hastings MCMC algorithm were utilized in the Augsburg city centre case study, based on findings from the Michelstadt experiment. The first application examined the capability of the previously used methodology for STE of individual unknown sources. However, no measurement datasets from individual sources were available for the Augsburg case study. To address this, a synthetic observational dataset was generated by extracting ATD model concentrations at hypothetical sensor positions and adding Gaussian noise. The algorithm was tested for different source positions under varying wind directions using synthetic measurement datasets, both with and without additional Gaussian noise. Results showed that when more than 10 sensors detected pollutant plumes (measuring concentrations above a specified threshold), the algorithm provided accurate predictions for both source location and release rate. However, when only 3 or 4 sensors captured concentrations above the threshold, the algorithm’s performance declined significantly. To further investigate the impact of monitoring network configuration, five sensors affected by the pollutant plume were added in each release scenario. The outcomes indicated higher accuracy when the number of affected sensors increased from 3 or 4 to 8 or 9, respectively. However, in the cases that more than 10 sensors detected concentrations above the threshold, further increases in sensor number had little impact on the algorithm’s performance. The last application introduced a novel methodology that combines the Metropolis-Hastings MCMC algorithm with the RANS-based CFD dispersion model to estimate emission rates from multiple traffic sources, using only prior knowledge of each source’s range of release rates. This approach also used synthetic measurements with Gaussian noise. The algorithm was tested on different hypothetical sensor subgroups, under various wind directions. Results indicated that the methodology is effective, achieving higher accuracy when denser monitoring networks and narrower emission rate ranges were used. However, further evaluation with real monitoring data is required. In conclusion, this study provides the necessary information, methods, and tools for developing inverse modelling systems and deploying sensor networks to estimate distinct unknown sources parameters in urban areas. Additionally, the findings to traffic-related multiple sources quantification highlights that the proposed inverse modelling methodology is capable of identifying the contributions of different sources to air quality levels. This system is highly valuable, as it can support authorities in rapid responses to emergency CBRN releases and provide policymakers with crucial information to implement measures for mitigating urban air pollution.
Degree
thesis:*- Grantor dc:publisher
- Aristotle University Of Thessaloniki (AUTH)
- Year dc:date
- 2025
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Gkirmpas, Panagiotis
- Γκίρμπας, Παναγιώτης
Subjects
dc:subject × 14- Μοντελοποίηση αντίστροφης διασποράς
- Εκτίμηση όρου πηγής
- Ποσοτικοποίηση πολλαπλών πηγών ατμοσφαιρικής ρύπανσης
- Μοντέλα υπολογιστικής ρευστομηχανικής
- Inverse dispersion modelling
- Source term estimation
- Multiple atmospheric pollution sources quantification
- Computational fluid dynamics models
- Επιστήμη Μηχανολόγου Μηχανικού
- Επιστήμες Μηχανικού και Τεχνολογία
- Μηχανολογία
- Mechanical Engineering
- Engineering and Technology
- Mechanics
Rights
- Language dc:language
- eng
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/58752