Surface mining can have a significant impact on various aspects of the environment, including human health. One of the most harmful effects on the human body comes from particulate pollution, which is dispersed into the air through mining processes. These pollutants, known as particulate matter (PM10), are particles smaller than 10 micrometers that can be inhaled and are associated with respiratory and cardiovascular illnesses. Monitoring this type of pollution is challenging due to its spatial distribution. The atmosphere is a three-dimensional space, and it is impossible to cover an entire locality with in-situ monitoring stations. However, remote sensing satellite data offers an effective alternative for monitoring airborne dust. Although satellite-based data may have limitations, they provide stable, spatial, and temporal coverage that is valuable for long-term monitoring and offers advantages in monitoring over large, inaccessible areas. This study has been carried out in Chalkidiki peninsula, Greece encompassing three active surface mines—Olympias, Stratoni, and Skouries—operated by Hellas Gold, a subsidiary of Eldorado Gold Corporation. Three Earth Observation (EO) datasets were used: the PM10 variable from the CAMS European air quality analysis dataset provided by the Copernicus Atmosphere Monitoring Service (CAMS), Aerosol Optical Depth (AOD) data at 0.55 micrometers from NASA’s MODIS Terra and Aqua satellites, and data from NASA’s SEVIRI sensor, where the Dust product was used for visual evaluation and the creation of a dust index. In-situ measurements (2015-2023) from Hellas-Gold dust monitoring stations (https://environmental.hellas-gold.com/) were used for building and validating spatial and temporal models. CAMS and MODIS data were aggregated into daily values for further evaluation. The CAMS data showed strong correlations with in-situ data, making them a valuable resource for monitoring, with correlation coefficients (R) around 0.8, depending on the locality. In contrast, MODIS data presented a different scenario, with correlations peaking at 0.5 and in some cases turning negative. This incompatibility may be attributed to long periods of missing data due to cloud coverage, so different MODIS preprocessing levels will be further examined. CAMS data was also aggregated into monthly and yearly values to observe trends, which indicated a decreasing tendency in concentrations, suggesting an improvement in air quality over the years. Additionally, machine learning methods, including Random Forest, LSTM, and XGBoost, were applied for data modeling and prediction. The XGBoost model produced the best results, with R values above 0,65 slightly varying within the sites of the interest. Future steps in this research will include further data modeling and prediction methods, with an emphasis on integrating artificial intelligence to develop an optimal workflow for this use case. Additionally, this methodology will be tested on another locality in Finland. ACKNOWLEDGEMENTS The presented analysis was conducted under the support of the EC grant MultiMiner. The MultiMiner project is funded by the European Union’s Horizon Europe research and innovation actions programme under Grant Agreement No. 10109137474.
Sedláčková et al. (Thu,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: