Characterizing and visualizing air pollution mixtures for air quality management decision-making in Canada

While air quality policy has traditionally focused on individual pollutants, in a real-world context, we are exposed to a mixture of pollutants simultaneously. Therefore, current air quality management approaches can fail to capture the synergistic effects of real-world exposures, which often disproportionately impact marginalized communities. This work examines existing approaches used for assessing air pollution mixtures in Canadian decision-making for air quality management and air quality research, to inform the pilot design of fit-to-purpose multi-pollutant analytical and visualization tools.

We collected data from a variety of sources to inform the design of more action-oriented mixture visualizations. These sources included decision-maker interviews and a scoping literature review. Through interviews we investigated the decision context of various decision-informing actors, focusing on data formats, availability and presentation. Interview participants have included municipal, provincial and federal environmental policy-makers and researchers, representatives of health authorities and (frontline) community advocates and researchers. The study collected the questions that these actors are seeking to answer, as well as their perceptions of the limitations of available data, including regarding data format or potential insights. As a next step, this work then combined the insights gained with technical methods for interpreting complex multipollutant data sets from the literature review. Many analysis techniques for multipollutant data sets are effective in reducing the complexity of high-dimensional data sets and produce informative multivariate patterns (e.g. dimensionality reduction techniques, clustering, correlation, and cumulative indicators). However, these methods often generate abstract, technical outputs, which can be difficult for non-expert audiences to interpret, and lack context or place-specific variables.

In a pilot demonstration, multivariate data analyses combined with situated knowledge of the decision context were translated into understandable measures of cumulative exposures to air pollutants. This synthesis is visually represented in a user-friendly, interactive dashboard and will be evaluated through user testing. The objective of the dashboard is to provide insights into the data to help answer user questions given their respective decision context. Examples of questions include: ‘What are the multipollutant profiles in the area? What are the major emission sources contributing to these profiles? What are the characteristics of the population most impacted by these multipollutant exposures?' The pilot dashboard will be tested with users to determine how the multivariate insight can best be communicated and to identify possible limitations and uncertainties. We hypothesize that the dashboard will be positively received if it supports users in answering their relevant questions, while clearly communicating how data insights are generated.
These findings contribute to ongoing efforts to refine analytical tools for multipollutant datasets and situate them within the complexities of local decision-making contexts. These approaches aim to better reflect real-world exposure patterns while balancing analytical complexity with interpretability.