MItigating Transport-Related Air Pollution (MI-TRAP)

Duration: January 2024 - December 2027


The EU-funded MI-TRAP aims to use a bottom-up dose-response framework to quantify the impact of pollution from transport on environmental and epidemiological outcomes. The project will be delivered by a large consortium of academic and industry partners, led by the National Centre for Scientific Research (Demokritos) in Greece. Results will be generated across all transport modes for a number of case study cities in Europe including Copenhagen, Milan, and Zurich, among others.

The Challenge

Despite notable progress, urban air quality remains a complex concern with significant public interest. The project addresses critical issues arising from transport emissions, particularly in high-impact zones. The transport sector is one of the most polluting globally, with the latest figures showing that it generates 24% of global CO2 emissions from fossil fuel combustion and consumes 28% of total global energy (International Energy Agency, 2020a,b; United Nations, 2021). Pollution from transport in the form of greenhouse gases and aerosols has two main adverse outcomes: (i) it contributes to climate forcing, which results in a net warming of global temperatures, and (ii) it impacts the health of populations exposed to the pollutants. However, it is currently difficult to link and attribute the impact of transport pollution directly to these adverse outcomes.

Our Role in the Project

In the project, TUM is leading the creation of a real-time multimodal traffic management system that outputs near real-time information on traffic flows, which feeds into connected modules that calculate air pollutant concentrations and noise levels. The key challenge in this part of the project is generating real-time multimodal traffic flows from sparse data. We propose to use a combination of non-conventional data sources (dynamic traffic volumes, link speeds, traffic composition) from commercial providers, open-source data and non-public licensed data from cities (when available). The collected data will be scaled up using state-of-the-art machine learning models to predict current and future traffic volumes and speeds. For data efficiency, we propose to use transfer learning methods to generalise pre-trained models from one case study city to another.

The TUM team will also lead scenario modelling in the project. We propose to apply statistical learning techniques using the baseline air and noise pollution data generated for each case study city as training data. Various scenarios can be tested by altering input feature data such as traffic speeds and flows, vehicle composition, and land use data to generate new air pollution predictions and noise levels for each scenario.


For a detailed list of the 26 partners participating in the project, visit the MI-TRAP website.