Twitter characteristics of an extreme weather event in The Netherlands
For this reason, the dynamics of social media activity during a severe weather event in The Netherlands (snow accumulation in a densely populated area, mostly affecting commuters) was investigated.
A historical Twitter dataset of almost seven million tweets was examined for characteristics like frequencies of retweets of the severe weather warnings, tweets with relevant keywords, speed of distribution of the weather warning, spread and geographic properties, etc.
Results show that a significant increase in the number of tweets with the extreme weather event associated keywords (like snow, extreme weather, KNMI) is noticeable during the extreme weather event. The number of georeferenced messages with associated keywords was found as high as 20% of all georeferenced messages during the peak of the event. Tweets with geolocation and with extreme weather associated keywords tend to reflect the geographical area of the severe weather. It was further noted that the extreme weather event seems to be very short-lived in terms of social media attention. Also, the communication of the weather event by traditional media (TV) influences the activity on social media, making it necessary for communication experts to communicate about the weather event at the right moment.
The presentation will show the results of the research and some of the pitfalls of research with social media data, like incomplete data and lack of geo reference.