Integrating Data-Driven Techniques for Assessing Urban Air Quality Dynamics

As a strategy to improve environmental quality, air pollution, and its impact on health, monitoring systems for emissions and environmental variables have been proposed. These systems provide information on air pollution levels in areas of interest and serve as support for analyzing the state of the environment. This enables the assessment of effects on human health and quality of life. A model is designed and implemented to estimate pollutant gases, air quality, and meteorological variables in a city using scattered local measurements and information from remote measurements provided by various sources. These are associated and processed to determine relevant information that helps generate accurate estimation models and their visualization through pollution maps. Meteorological data, gas, particulate matter, local energy consumption, and on-site measurements across the city, involving pedestrians and cyclists, contribute to environmental characterization studies, they can make use of a low-power, wide-range LoRa communication network. Using inference from a lower-resolution grid, maps are interpolated to resolutions of 50m or 100m. The kriging method is then applied to enhance the resolution to 5.5m, utilizing a geostatistical model based on the variance and covariance of the samples to model the variance at a lower resolution. Finally, machine learning algorithms are employed to model and visualize data, generating pollution maps and environmental variables to understand pollution dynamics.