No abstract

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

Multi-hazard mapping in urban areas is relevant for preventing and mitigating the impact of nature- and human-induced disasters while being a challenging task as different competencies have to be put together. Artificial intelligence models are being increasingly exploited for single-hazard susceptibility mapping, from which multi-hazard maps are ultimately derived. Despite the remarkable performance of these models, their application requires the identification of a list of conditioning factors as well as the collection of relevant data and historical inventories, which may be non-trivial tasks. The objective of this study is twofold. First, based on a review of recent publications, it identifies conditioning factors to be used as an input to machine and deep learning techniques for singlehazard susceptibility mapping. Second, it investigates open datasets describing those factors for two European cities, namely Milan (Italy) and Sofia (Bulgaria) by exploiting local authorities’ databases. Identification of the conditioning factors was carried out through the review of recent publications aiming at hazard mapping with artificial intelligence models. Two indicators were conceived to define the relevance of each factor. A first research result consists of a relevance-sorted list of conditioning factors per hazard as well as a set of open and free access data describing several factors for Milan and Sofia. Based on data availability, a feasibility analysis was carried out to investigate the possibility to model hazard susceptibility for the two case studies as well as for the limit case of a city with no local data available. Results show major differences between Milan and Sofia while pointing out Copernicus services’ datasets as a valuable resource for susceptibility mapping in case of limited local data availability. Achieved outcomes have to be intended as preliminary results, as further details shall be disclosed after the discussion with domain experts.