Tracking Climate Change Opinions from Twitter Data

The Fifth Assessment Report (AR5) of the Intergovernmental Panel on Climate Change (IPCC) released in 2014 states that warming of the climate system is “unequivocal,” and it is “extremely likely” that human influence has been the dominant cause. However, public perceptions of anthropogenic climate change have varied widely, and indeed may have been significantly influenced by a disproportionate set of non-scientific opinion makers. While the statistics of extremes such as heat waves and heavy rainfall have been scientifically attributed to climate change, such attributions are not possible for single extreme events. Nevertheless, articles in social science and climate journals, including Nature Climate Change, have suggested that exposure to extreme weather events can directly influence opinions about climate change. Greenhouse-gas reduction policies, resilience to natural hazards, and adaptation decisions ultimately rely significantly on having adequate public support, but conducting real-time surveys of public perceptions can be difficult, expensive, and occasionally even impossible. The role of the micro-blogging site Twitter (http://twitter.com) has turned the Web into a major repository of topical comments, and hence a potential source of information for social science research. This paper attempts to understand whether Twitter data mining can complement and supplement insights about climate change perceptions, especially how such perceptions may change over time upon exposure to climate related hazards. A combination of techniques drawn from text mining, hierarchical sentiment analysis and time series methods is employed for this purpose. Future research is motivated in these areas, while potential pitfalls are discussed.