Estimating Daily NO2 Ground Level Concentrations Using Sentinel-5P and Ground Sensor Meteorological Measurements

Environmental and health deterioration due to the increasing presence of air pollutants is a pressing topic for governments organizations. Institutions such as European Environment Agency have determined that more than 350,000 premature deaths can be attributed atmospheric pollutants. The measurement trace gas concentrations key environmental agencies fight against decreased quality. NO2, which one most harmful pollutants, has potential cause diseases Chronic Obstructive Pulmonary Disease (COPD). Unfortunately, not all countries local pollutant monitoring networks perform ground measurements (especially Low- Middle-Income Countries). Although some alternatives, satellite technologies, provide good approximation tropospheric these do measure at level. In this work, we aim an alternative sensor measurements. We used combination meteorological with Sentinel-5P observations estimate NO2. For task, state-of-the-art Machine Learning models, linear regression feature selection algorithms. From results obtained, found Multi-layer Perceptron Regressor Kriging in Random Forest algorithm achieved lowest RMSE (2.89 µg/m3). This result, comparison real data standard deviation models using only data, represented decrease 55%. Future work will focus on replacing use sensors satellite-based data.