1/10/2016 Assessment of Social Media Usage during Hurricane Sandy and the Development of a Twitter‐based, Web Application for Real‐time Communication of Storm‐related Information 32nd Conference on Environmental Information Processing Technologies New Orleans, Louisiana January 11, 2016 John F. Edwards, Ph.D. Somya D. Mohanty, Ph.D. Patrick Fitzpatrick, Ph.D. David Martin, B.S. Introduction Funding • National Oceanic and Atmospheric Administration Coastal Storm Awareness Program Connecticut Sea Grant Objectives • Survey affected population regarding storm‐related communication • Identify and codify storm‐related messages from Twitter • Identify major “Influencers” disseminating weather‐related info • Develop a Twitter‐based, data collection and bidirectional communication system for weather‐related emergencies 1 1/10/2016 Survey Data Telephone and Twitter‐based Survey Sampling and Methods Cell Phone Sampling Frame Landline Sampling Frame Phone Modality (n=514) Geo‐Coded Tweets (n=15,000) Twitter‐Distributed Modality (n=179) Survey Data (n=693) Survey Data Catchment Area for both Telephone and Twitter‐based Samples 2 1/10/2016 Survey Data Loss of Power and Communication Mediums Power 62% Television 58% Wired Internet 53% Landline 45% Cellphone 27% 0% 10% 20% 30% 40% 50% 60% 70% Survey Data Communication Mediums Relationship to Evacuation 60% 50% 53% 50% 40% 39% 30% 20% 25% 21% 19% 14% 10% 8% 0% Evacuated (12%) Face‐to‐Face Television 14% 13% 7% Prepared in Place (69%) 9% Did Nothing (19%) Telephone (Voice & Text) Social Media 3 1/10/2016 Survey Data Weather‐related Entities Followed on Twitter Red Cross 4.3% NOAA 6.5% Local Radio 6.6% Meterorologist 6.8% FEMA 6.8% TWC 10.3% NWS 11.6% Local News 13.0% 0.0% 2.0% 4.0% 6.0% 8.0% 10.0% 12.0% 14.0% Survey Data Type of Information Shared on Twitter Photographs 62% Personal Experiences 56% Storm Damage 53% Storm Warnings 47% General Messages 47% 0% 10% 20% 30% 40% 50% 60% 70% 4 1/10/2016 Social Media Data 4.5 million Geo‐coded Tweets Social Media Data Volume 5 1/10/2016 Social Media Data Sentiment Social Media Data User Network – Pre 6 1/10/2016 Social Media Data User Network – During Social Media Data User Network – Post 7 1/10/2016 Social Media Data Identifying and Codifying Pleading for Help Offering Help Organizing Help 8 1/10/2016 Software URL Software Codifier Module 9 1/10/2016 Software Emergency Manager Module Software Data Collection Module 10 1/10/2016 Software Example Image Software Example Image 11 1/10/2016 Software Example Image Software Example Image 12 1/10/2016 Software Example Image Future of Using Human Sensors for Obtaining Weather‐Related Information • ~150 million more Twitter users since Hurricane Sandy • Posting photographs has become much easier and more commonplace. Twitter hosts photos natively. • Advances in natural language processing and image recognition will increase the efficiency of the software • We do not condone capturing photographs of dangerous weather activity, but our software allows such photographs to be used for a greater good. 13 1/10/2016 Questions John F. Edwards, Ph.D. je@ssrc.msstate.edu 662.325.9726 32nd Conference on Environmental Information Processing Technologies New Orleans, Louisiana January 11, 2016 14