Verifying Reports of Collapsed Buildings from Twitter Aftermaths of Earthquakes: A Case Study from Turkey

Abstract

On February 6, 2023, two highly severe earthquakes occurred in a wide region encompassing 11 cities in Turkey, resulting in extensive damage and an official death toll exceeding 50,000. In the aftermath of this catastrophic event that affected multiple cities, identifying the locations of debris with potential survivors became a crucial challenge for search and rescue operations. However, another significant obstacle emerged in obtaining accurate and genuine addresses. Individuals who were either trapped themselves or had relatives under the collapsed buildings attempted to report addresses using conventional communication methods. Communication difficulties on lines prompted disaster victims to resort to internet-based communication methods. Consequently, social media platforms emerged as powerful tools for rapidly disseminating information to millions of people. However, alongside the positive impact of social media, the risk of generating significant panic due to the spread of fake news also surfaced. This study analyzes tweets posted on Twitter within the first 24 hours following the earthquakes. Firstly, tweets containing reports of collapsed structures were identified, and text parsing techniques were employed to extract address information. The veracity of destruction at these addresses was confirmed using imagery captured from Unmanned Aerial Vehicles (UAVs) in the aftermath of the earthquakes. As a result, a 90% accuracy rate was observed in confirming the presence of destruction either at the reported addresses or within a 100-meter proximity, based on the top 100 most widely shared reports on social media. Moreover, the presence of numerous unidentifiable addresses highlights the necessity for continued enhancements to the Address Registration System.