MINING SOCIAL MEDIA TO ASSESS PUBLIC PERCEPTION OF WATER QUALITY

Abstract

Social media provides potentially new sources of information for the detection and management of undesirable water quality events such as harmful algae blooms (HABs) in surface waters. Current methods for identifying HAB include field sampling and laboratory tests which are time-intensive and can cause a delay in the issuance of warning advisories, resulting in public health consequences. The potential strengths of social media as water quality indicators are that social media data can be collected in real-time using Application Programming Interfaces (APIs) and is less expensive compared to traditional water quality sampling methods. But the challenge lies in understanding what water quality parameters the public perceives and responds to. To address this challenge, we explored tweets (2016 – 2020) expressing negative sentiment related to the water quality of Utah lake which is well-known for its algae blooms. We used sentiment analysis, natural language processing, spatial interpolation, and count regression modeling to evaluate temporal correlations of social media posts obtained using Twitter API and water quality data collected by the Utah Department of Environmental Quality. We found that the negative tweet counts were significantly and positively associated with many of the perceivable water quality parameters studied such as turbidity, chlorophyll-a, phytoplankton cell count, phytoplankton biovolume, cyanobacteria cell count, and cyanobacteria biovolume. Surface samples for algae concentration and population were also significantly related to the negative tweet counts while the composite samples were not significant, thereby supporting the idea that the public perceives and responds to the toxic water quality near the water surface. Our work serves as a preliminary study that highlights the potential of using social media for identifying water quality events in lakes. To achieve the ultimate goal of developing a real-time public warning system, further studies should be conducted to develop metrics that can translate social media sentiment and activity to a quantitative measurement of water quality health.