| dc.description.abstract | In this study Twitter data have been analysed to gain information about public response to climate hazards by using artificial intelligence techniques. An increasing body of research has suggested that the processing of information, decision-making and action-taking, and behaviour with regards to climate change, hazards and risks are largely determined by emotions. This opens up the possibility to study the relation between climate hazards and emotional expressions, responses and effects. Understanding tweet activity and its associated emotions before, during and after climate hazards can have significant implications for interventions towards preparedness, safety, regulations and recovery during future climate hazards. In addition to global response, information about tweet activity and emotional response in different regions can add to these implications and interventions. Twitter is chosen as social network forum due to its global usage, the current purpose of Twitter with regards to sharing opinions and expressions, the coverage of a very wide range of topics and the fact that the analysis of a large amount of data is facilitated for scientific purposes.
The main question of this thesis is as follows: How does the public respond (emotionally) to climate hazards on Twitter?
First a literature study was conducted that discusses climate change and natural hazards, psychology of emotions and behaviour and its link to climate change and the expression and recognition of emotions in linguistics. This is followed by an outline of previous research regarding climate hazards, social media and emotions.
Concepts from literature regarding emotions and emotion recognition were integrated to collect and analyse Tweet datasets from Twitter for three different climate hazards. Tweets were collected before, during and after these climate hazards. Tweet datasets were collected by selection on specific words, start and end date and locations. This was followed by the extraction of emotions from the tweet texts in these datasets. The resulting tweet and emotion data were analysed to identify patterns regarding tweet activity and emotional response. Three climate hazards were selected: A sudden temperature rise on Antarctica, a Tornado outbreak in South West USA and Hurricane Ida in South West USA. All three events were short-lasting events that lasted between 1-3 days to be able to analyse the tweet and emotion trend before, during and after the event on a 4-week timescale. Tweet activity peak during the core event and show a gradual decrease during the recovery phase. The results show that the emotion anger is substantially larger for the hurricane and tornado cases in comparison with the Antarctica case, which can be explained by the visible impacts during the hurricane and tornado hazards and the association of hurricanes and tornadoes to climate hazards. The tornado event shows substantially more fearful reactions than the hurricane event. It is suggested that this difference in fear and anger proportions lies in the predictability or controllability of the event. The found tweet activity and emotional response patterns in this thesis can be useful for governments and organisations to communicate more effectively with the public by aiming for better preparation and recovery of climate hazards and also enhancing positive attitude towards climate change topics. | |
