Assessment of Air Quality Using Satellite Data and Machine Learning Techniques in Uttarakhand, India

The present study assesses the air quality using satellite data and machine learning techniques in Uttarakhand, India. Degrading Air Quality is a major concern for all species on this planet. Over the years, it has been seen that air quality is constantly degrading mainly due to industrialisation, deforestation, and greenhouse effect. Parameters generally considered for measuring Air Quality are Carbon Monoxide (CO), Nitrogen Dioxide (NO2), Sulphur Dioxide (SO2), Ozone (O3) and Aerosols. These are present in the air and changes in the composition of these gases cause major changes in the air that organisms breathe. A study of the change of these parameters over time is necessary to understand the impact on air quality.

Data is collected from the SENTINEL-5P satellite through Google Earth Engine and is processed in Google Collaboratory. In this study, the data of Carbon Monoxide (CO), Nitrogen Dioxide (NO2), Sulphur Dioxide (SO2), Ozone (O3) and Aerosols is taken for the past 5 years and their time series is extracted thereafter a test on stationarity is done so as to know whether these series are stationary or not. Two machine learning models namely Holt Winter’s Smoothing and FbProphet are applied to predict the value adjacent to the original value and an error metric comparison is done to find out which model is best suited for forecasting these Air Quality parameters. The present study concluded that Deep learning and machine learning models are accurate for predicting along Air Quality Components. These models are capable enough to predict daily data of these air quality parameters.