FUSING MULTIPLE OPEN-SOURCE REMOTE SENSING DATA TO ESTIMATE PM2.5 AND PM10MONTHLY CONCENTRATIONS IN CROATIA

Abstract

The objective of this study is to create a methodology for accurately estimating atmospheric concentrations of PM_2.5 and PM₁₀ using Sentinel-5P and other open-source remote sensing data from the Google Earth Engine (GEE) platform on a monthly basis for June, July and August which are considered as months of non-heating season in Croatia, and December, January and February, which, on the other hand, are considered as months of the heating season. Furthermore, machine learning algorithms were employed in this study to build models that can accurately identify air quality. The proposed method uses open-source remote sensing data accessible on the GEE platform, with in-situ data from Croatian National Network for Continuous Air Quality Monitoring as ground truth data. A common thing for all developed monthly models is that the predicted values slightly underestimate the actual ones and appear slightly lower. However, all models have shown the general ability to estimate PM_2.5 and PM₁₀ levels, even in areas without high pollution. Furthermore, all models can effectively detect all PM_2.5 and PM₁₀  hotspots.