Horizon Magazine · Health

AI-Powered Digital Twins Help Cities Tackle Street-Level Air Pollution

Researchers are developing AI-powered 3D "digital twins" of cities to provide real-time, street-level air quality forecasts, aiming to help authorities make better-informed decisions on pollution control.

Urban air pollution can vary significantly between streets, often lingering in areas with tall buildings or busy junctions. Cities frequently lack the detailed forecasts needed to address dangerous pollution levels before they occur. Professor Soledad Le Clainche, leading the ModelFLOWs research group, is part of the EU-funded MODELAIR project, which uses AI and 3D models to create real-time, street-level pollution maps. This initiative aims to provide a detailed understanding of urban air quality, complementing the EU Mission for Climate-Neutral and Smart Cities.

The MODELAIR project is testing its tools in Brussels, Madrid, and Bristol, with each city exploring different applications, such as urban design impacts, monitoring improvements, and practical response testing. The research addresses the significant health risks posed by air pollution, with the World Health Organization (WHO) considering current levels in most European cities unsafe. Air pollution is linked to increased risks of heart disease, stroke, cancers, and respiratory issues, and emerging evidence suggests a possible link to dementia.

The AI agents used in the project continuously integrate data from sensors, weather forecasts, and computer models to generate local forecasts. While some methods are available through the open-source ModelFLOWs-App, the MODELAIR project aims to provide a more detailed picture than current models by combining limited data sources with detailed air movement simulations. This allows for more accurate predictions of pollutant travel within urban areas, assisting cities that struggle with air pollution decision-making.

The complexity of urban environments, with hourly changes in pressure, wind, and heat, complicates pollution behavior. The AI agents collaborate with scientists to predict pollution at specific locations and suggest potential responses, such as rerouting traffic or planning new developments. However, AI suggestions require assessment by scientists and final decisions by policymakers. The insights from these digital twins can inform interventions like tree planting to mitigate heat-island effects and reduce temperatures.

The project is preparing to test a real-time decision-making tool in Brussels and will use models in Madrid to assess traffic routing. Belgian company BuildWind is contributing its BrusAir platform, which maps air pollution in Brussels, and expects the MODELAIR research to enhance its capabilities. Findings from MODELAIR will be made more widely available by the end of the year, potentially offering cities a clearer view of street-level air quality to guide timely interventions.

AI-samenvatting op basis van de bron.