Seedlabs
Source research

High-resolution modelling of particulate matter chemical composition over Europe: brake wear pollution

Abhishek Upadhyay, Jianhui Jiang, Yun Cheng, Petros Vasilakos, Ying Chen, Daniel Trejo Banos et al. · 2025 · 5 citationsRead the paper

In today’s rapidly evolving society, the sources of atmospheric particulate matter (PM) emissions are shifting significantly. Stringent regulations on vehicle tailpipe emissions, in combination with a lack of control of non-exhaust vehicular emissions, have led to an increase in the relative contribution of non-exhaust PM in Europe. This study analyzes the spatial distribution, temporal trends, and impacts of brake wear PM pollution across Europe by modeling copper (Cu) concentrations at a high spatial resolution of ∼250 m which is a key tracer of brake-wear emissions. We integrated coarse-resolution brake-wear Cu from CAMx chemical transport model and high-resolution land use data into a random forest model to predict Cu concentrations at 250 m over whole of continental Europe. The random forest model was trained using an unprecedented dataset of over 50,000 daily Cu measurements from 152 sites. It corrected CAMx underestimation and downscaled Cu to a higher spatial resolution. In validation, the model showed robust spatial and temporal prediction with good Pearson’s correlation coefficients of 0.6 and 0.7, respectively. We generated 10 years (2010–2019) of daily Cu concentrations over Europe, revealing spatial patterns aligned with urbanization and road networks, with peaks in cities and lower values in rural areas. Temporal trends reveal that Cu concentrations generally peak on weekdays and in winter. Despite a decline in PM across Europe over decades, Cu concentrations showed no decrease in many cities from 2010 to 2019. Cu levels are strongly correlated with population density with more than 12 million Europeans exposed to levels exceeding 40 ng/m 3 , equivalent to around 1 μg/m 3 of total PM 10 from brake wear. Our findings highlight the need for expanded metal measurement for non-exhaust tracers for a better understanding of the health relevance of PM composition including Cu, and more effective regulations of non-exhaust PM emissions as included in EURO 7 vehicles.

What the full text shows

Working prototype
  • More than 12 million Europeans are exposed to copper (Cu) levels exceeding 40 ng/m³. — 40 ng/m³ (Continental Europe, 2010–2019)
  • Brake wear contributes approximately 1 μg/m³ of total PM10 for the population exposed to >40 ng/m³ Cu. — 1 μg/m³ PM10 (Continental Europe, 2010–2019)
  • The Random Forest (RF) model improved prediction accuracy over the CAMx model, increasing the Pearson correlation coefficient for unseen sites. — 0.58 (RF-CAMx) vs 0.43 (CAMx) (Validation on unseen monitoring stations)
  • The CAMx model significantly underestimated Cu concentrations compared to observations. — Average negative bias of a factor of ~7 (Continental Europe, 2010–2019)
  • Cu concentrations showed no decrease in many cities despite overall PM declines. — No decrease (2010–2019)
Setup
Dataset of >50,000 daily Cu measurements from 152 sites across Europe; CAMx simulation domain of 0.25° × 0.125° downscaled to ~250 m.
How
The study integrated coarse-resolution Cu simulations from the CAMx chemical transport model with high-resolution land use data using a Random Forest machine learning model. The RF model predicted the logarithmic ratio of observed to simulated Cu to correct biases and downscale concentrations to a 250 m resolution.
Limitations
Significant data gaps in Eastern Europe and the Balkans · Cu is emitted by sources other than brake wear (e.g., industry, railways), though brake wear is dominant · CAMx bias varies by concentration range (underestimation factor of <2 at low concentrations to >10 at high concentrations)

Summarized by Gemma 4 from the open-access full text · 8 Oct 2026. Read it