Co-designing an interactive tool to communicate the uncertainty of urban air quality models: uncertAIR

According to the World Health Organization, air pollution is the main environmental threat to public health. Urban environments are particularly critical due to their high-density population centers with often poor air quality. To characterize the exposure of citizens, the use of numerical models corrected with observational data has become a fundamental tool. Despite recent efforts, bias-corrected air quality models at the street scale exhibit significant uncertainty, partly due to the limited number of traffic and air quality observations.

Model uncertainty can critically increase far from measurement points and in regions with characteristics different from those used for calibration. In such locations, modeled data should be interpreted with caution. When the street-scale air quality models are intended to inform policy makers, estimating uncertainty is highly valuable to support decision-making protocols. A simpler air quality model with an estimation of the spatial uncertainty distribution may be preferred over a very sophisticated model that does not give any notion of uncertainty.

Within this context, we aimed at co-designing and co-developing an interactive tool to report the uncertainty of urban air quality simulations, disseminating the results tailored to the users’ needs.

The methodology consists of a geostatistical post-processing of the raw simulations of NO₂ concentrations of the CALIOPE-urban air quality model in the city of Barcelona. The methodology is replicable to other cities and pollutants. The uncertainty estimation is based on the error variance of the Universal Kriging technique, which can be subsequently used to produce hourly maps of the probability of exceeding a certain threshold. Additionally, relevant social-ecological-technological variables were identified to explore the interconnections among different types of data, as well as broadening the social impact of this project. For instance, locations associated with vulnerable citizens (e.g., schools and nursing homes), or other variables potentially linked with air quality (e.g., public parks and green spaces). 

A user-centric approach was adopted, involving policymakers from local administrations, urban planners from private companies, environmental social agents and scientific personnel from research institutions and universities. To get a deep understanding of how uncertainty maps can add value to users’ objectives, we conducted a series of individual interviews and a co-design workshop based on design thinking, which allowed for the co-design of the interactive platform. The prototype of the interactive platform was presented in a second workshop, where the users tested the prototype and provided input to further developing the final tool.

The final product is the uncertAIR platform, an open-source interactive tool that integrates modeled NO₂ concentrations, their uncertainty and probability of exceedances of legal thresholds, together with  social-ecological-technological variables at different scales of time and spatial resolution. Data can be visualized and downloaded with a temporal resolution of annual or daily averages, and a spatial resolution of 20 m or aggregated at census areas. This integrated dataset serves as the foundational step to integrate uncertainty information on future air quality policy making in Barcelona, such as health impact assessments, official communications, campaign planning, and location optimization of new monitoring stations.