Pederzh/Anomaly-detection-of-pollution-emissions-using-TROPOMI-satellite-data

Monitoring established in-land human activities using pollution satellite data. Copernicus Sentinel-5P (TROPOMI) is the satellite data imagery source for remote pollution sensing on which the entire project is based. The image processing phase has taken a significant amount of effort since it was crucial to extract useful and correct information for pollution source identification and time-series analysis. Starting from the assumption that we do not know where human activities are in advance, we have developed a method for top-down detection of pollution sources in areas of interest. During our work, we have developed a Gaussian reconstruction of the emissions (GROTE) method to estimate the emissions by analyzing pollution. Once the data has been processed, we use the processed data to train a time-series machine learning method and generate data on expected pollution emissions for each identified location. Finally, our service can be integrated into the ARCOS project and raise an alert if the difference between the forecast value and the actual value exceeds the reference baseline for determining whether the pollution emissions value falls into the category of "usual" or "anomalous" behavior.