Monitoring Air Pollution through Twitter Data

The final session at Social Media & Society 2018 starts with Supraja Gurajala, whose interest is in using Twitter data for responding to air quality issues. Air quality is a major health issue in population and industrial centres around the world, and metrics like the Air Quality Index (AQI) and Particulate Matter index (PM) are key to its assessment.

Air quality monitoring stations exist around the world, but are unevenly distributed. How might the gaps in monitoring be addressed by increasing the number of monitoring stations, given the costs involved in setup and maintenance, then? One option may be to draw on social media data: some existing studies have drawn especially on the Chinese platform Weibo, and have shown a correlation between air quality-related posts and actual air quality, and the present study extends this to Twitter.

Key to this is an analysis of tweet frequencies and tweet content in tweets related to air quality, and the present project collected some 25 million such tweets. The present paper focusses especially on tweets from New Delhi, Paris, and London. For Delhi, tweet frequency and poor air quality are broadly correlated, but tweets lag behind air quality events by about 12 hours on average. This differs somewhat across different hashtags relating to air quality, however, and hashtags tend to be different depending on the city being studied. (These patterns are different from other crisis events, such as earthquakes or bushfires.)

It is also worth exploring the different tweeting topics that are sparked by air pollution, from health through politics to climate; these move differently depending on the event and on the city being examined. New, often location-specific topics may also emerge from the analysis.