Abstract
Air pollution is a big environmental issue across the world because of the vast number of industries, automobiles, and machines that emit gases and particles. This harms the health of living beings, the ecosystem, and the climate in general. As per WHO, 90% of the world population is exposed to poor air quality. PM2.5 is a fine particulate matter that is 2.5 micrometers or less in diameter and is hazardous to human health when levels of it in the air are very high. PM2.5 also reduces visibility by making the air appear hazy when its concentration in the air increases. Traditionally, PM2.5 is measured using ground monitors and often represents point measurements. The data from many places is missing because they do not have ground monitor availability. Advancements in satellite remote sensing and machine learning algorithms provide new opportunities to resolve the drawback and to estimate PM2.5 at higher spatial scales (1 km). In this project, satellite data on aerosols, local meteorology, and land use parameters will be used in machine learning algorithms to generate high spatiotemporal maps of PM2.5 over selected urban centers in California, US.
The proposed architecture in this paper is a combination of bagging and boosting, which estimates the daily PM2.5 concentration. This estimation is in-turn plotted on maps by using the Environmental Protection Agency (EPA) color coding standard. This research thereby improves the temporal resolution of the existing research.