Mapping climate discourse to climate opinion: An approach for augmenting surveys with social media to enhance understandings of climate opinion in the United States

locals who fear the projects will negatively impact the environmental quality of their communities. Another result is the importance of the trustworthiness of climate science across all regions, either as Climate Denial, Global Warming, or Trust in Science. Though the framing and lan- guage of these topics is distinct, they all include tweets that question or promote trust in the scientific community. Additionally, there are topics which are focused on a specific event or policy. Event-related topics include Activism which primarily discusses London’s Extinction Rally and Swedish activist Greta Thunberg, and Earth Day which largely focuses on activities and events that occurred on April 22nd, 2019 celebrating environmental protection. Policy-related topics include Green New Deal which largely focuses on the eponymous House bill, and Carbon Tax, a discussion centered on Canada’s recently proposed carbon tax legislation. Because these cate- gories deal with ephemeral subjects, we do not believe they are as relevant to discussions of long term trends in the climate debate. We do include them in this study, however, to demon- strate the rapidity and flexibility of social media mining for developing a regional analysis of climate change opinions. Finally, the Weather Reports category, which is composed nearly exclusively of daily weather updates delivered via Twitter, can be used as a loose proxy for an area’s adoption of technology but is not particularly informative in framing climate policy.

Because relevant issues and topics are dynamic and change over time, an important consid- eration with this work is which topics in this model speaks to long-term concerns, and which are in response to specific events. As a means of sensitivity analysis, a separate Twitter corpus was collected from the 15th to the 21st of March, 2019, which precedes the corpus used in this analysis by almost exactly one month (S1 File). By comparing both time periods, we demon- strate that the only topics that differ between the two corpora are those which are specific to an individual event or policy—such as Green New Deal, Carbon Tax, and Earth Day. Fig 2. Regional Twitter topic distributions. Each region of the US used in this analysis is shown along with its corresponding most important LDA topics. Topics are listed in order of their importance (top to bottom), and colored based on how frequently they occur in regional discussion: darker is more unique to the region while lighter is more common nationally. https://doi.org/10.1371/journal.pone.0245319.g002 PLOS ONE An approach for augmenting surveys with social media to understand climate opinion in the United States PLOS ONE | https://doi.org/10.1371/journal.pone.0245319 January 14, 2021 10 / 16