QUALARIA Project: An AI System to Monitor and Predict Metropolitan Area of São Paulo Street-Level Air Quality

The German-Brazilian Project “QUALARIA: Artificial Intelligence based system for sub-urban scale air quality prediction” aims to create an operational artificial intelligence-based system to monitor, simulate and predict air quality in the Metropolitan Area of São Paulo (MASP), with high spatial resolution (https://meteoia.com/qualaria/). Advanced global and regional chemical-meteorological models, such as CAMS global composition forecast, and WRF-Chem simulations are applied to derive the climatological state of air composition, mainly the average levels of air pollutant based on existing local emission inventories. Measurements of PM10, PM2.5 NO2, and O3 concentrations from the São Paulo State Environmental Agency (CETESB) Air Quality Network are used to train the downscaling to capture the pollutant concentration sub-grid spatial variations. The spatial and temporal disaggregated local vehicular emission inventory and the building height from the Global Human Settlement Layer dataset are also used as input. Preliminary results produced air pollution concentration maps at 100 m and showed an increase in Pearson correlation and a reduction in the mean absolute error compared to CAMS forecast. Other high spatial resolution datasets and measurement from other states air quality networks are being explored to increase the input datasets. In the following steps of the project, low-cost sensors are going to be deployed to increase the spatial coverage of MASP and its surroundings, complementing the CETESB stations. Then, from their predicted downscaled pollutant concentration fields, the system will provide an online dashboard to display relevant air quality indicators, and to inform the impacts of air pollution on human health. To improve the dashboard design, stakeholders from the public and private sector are being engaged and consulted for the development of its user interface and features.