A novel regression method for pollutant prediction using Albedo, normalized difference vegetation index, and surface kinetic temperature metrics derived from ASTER satellite image: Pollutants prediction by Albedo, NDVI, and Surface Kinetic Temperature metrics derived from ASTER

Atmospheric pollution is largely influenced by human and industrial activities, especially in densely populated urban areas, although natural sources and regional factors also contribute significantly. Air pollution affects human health, ecosystems, materials and climate. Pollutants released into the air can be responsible for various diseases of the respiratory and cardiovascular system. Predicting air pollution can be considered a worthwhile investment for individual and community. An accurate forecast helps people to reduce health effects, severity of local pollution levels and associated costs. In this paper, a regression system is proposed to forecast air pollutants (NO, NO2, NOx and PM10), using the values of Albedo, Normalized Difference Vegetation Index, and Surface Kinetic Temperature derived from ASTER satellite image. The implemented system showed the best prediction results for NO and PM10, where RMSE values <5μg/m3 have been reported. These results overcome the values published in recent works using similar approaches but with different combinations of indices and/or methods. A thorough analysis of our results suggests that PM10 appears more strongly associated with temperature related surface conditions, therefore changes in thermal patterns over time could influence PM10 behavior more than the other pollutants.