ANNOUNCEMENTS
Air Pollution has become a critical environmental issue as well as health and economy. With 42 out of 50 most polluted cities located in India, it has become a serious issue and require strict rules and regulation to combate air pollution. Gurugram ranked 7 most polluted city in the world and its continuously increasing due to economic growth and its geographical location. Efforts are made by local and state-level authority to reduce the air quality. However, due to lack of knowledge of area wise contribution of air quality and sources of air pollution are difficult to be identified. In this current study, efforts are made to create multispectral empirical model to calculate particulate matter concentration in Gurugram with the assistance of Landsat OLI imagery and ground level concentration of particulate matter from Central Pollution Control Board. The accuracy of the model for both the pollutants was calculated using R-square, showing an considerable level of accuracy ranging between 0.76 to 0.84. Furthermore, an LSTM model to predict PM2.5 concentration was developed. Three years' worth of pollutant level data and meteorological conditions were used to build the LSTM model. To maximize the model's performance, different hidden layer, neuron, and learning rate configurations were tried. With a low Mean Square Error (MSE) of 0.00698 at 250 epochs and a high correlation coefficient (R) of 0.856 between the predicted and actual PM2.5 levels, the final model showed significant predictive performance.
KEYWORDS: Particulate Matter, Hotspot, Air Quality, Prediction, LSTM.
