Characterizing climate change sentiments in Alaska on social media

Abstract

The profound impacts of climate change have spurred global concerns. Yet, public perceptions of this issue exhibit significant variations rooted in local contexts. This study investigates public perceptions of climate change in Alaska on Twitter and explores their connections with local socioeconomic and environmental factors. Using geo-located tweets from 2014 to 2017, we identified a collection of climate-related tweets using a deep learning framework. Employing lexicon-based sentiment analysis, we quantified the sentiments with positive and negative scores, further enriched by extracting eight core emotions expressed in each tweet. Furthermore, we applied regression models to assess the influence of regional socioeconomic and environmental attributes on climate-related sentiments at the census tract level. Our findings reveal an overall upward trajectory of Alaska's Twitter-expressed climate change sentiments over time, particularly during the summer months. Insights into the interplay between local demographics and environmental features and climate change perceptions include: (1) Census tracts with higher Native Alaskan or American Indian populations tend to express more negative sentiments, (2) the inclusion of road density stands out as a significant factor, suggesting that climate change is seen/discussed more in areas with more dense-built infrastructure, and (3) the presence of mixed emotions exhibits a profound connection with climate change sentiments—i.e., emotions of disgust and surprise are inversely related, whereas sadness and trust demonstrate positive associations. These outcomes underscore an evolving situation awareness of climate change among individuals, emphasizing the need to consider local factors in understanding public perceptions of this global issue.