Published on November 19th, 2012 | by EJC0
The 3.11 Japan Quake: Looking Back At News And Crowdsourcing On Media Coverage Map
What are the lessons learned from the experiences of the 3.11 Japan Earthquake? What were the roles of mainstream media and crowdsourcing in information sharing when this large-scale triple disaster hit the country?
On 11 March 2011, a magnitude 9.0 earthquake struck the east coast of Japan, causing over 16,000 deaths, 3,800 missing, and the complete destruction of over 118,600 buildings. Within just 9 hours of the initial 2:46 PM quake, the country experienced 1,599 aftershocks, including four of magnitude 7.0 or higher, 44 times at magnitude 6.0 level and 233 times at magnitude 5.0.
Japanese newsrooms are relatively well-equipped with natural disaster measurements due to the number and variety of catastrophic events that affect the country. For instance, NHK (Japan Broadcasting Corporation) has 460 robot cameras stationed across the country, recording footage of earthquakes as they happen. Major print and broadcast outlets own helicopters allowing them to fly into disaster-affected areas. At the same time, high 3G mobile and internet penetration rates in Japan have enabled widespread usage of social media networks like Twitter and Youtube during earthquakes.
Project 311: The Role of Media in the Great East Japan Earthquake
Organised by Google and Twitter Japan, the Big Data Workshop, also known as ‘Project 311’, took place during September and October 2012. Project 311 aimed to assess how information from news and social media circulated during/after the quake, and identify how it can be used for future disaster preparedness. This collaborative project gathered a number of key organisations such as the newspaper Asahi Shimbun, NHK, and Honda, which contributed the data sets that researchers and developers can work with to develop tools to achieve the project’s goal.
Mapping out News Coverage and Crowdsourced Information
As part of the activities of Big Data Workshop, Hidenori Watanabe, an associate professor at Tokyo Metropolitan University, launched a data visualisation project entitled ‘the ‘East Japan Earthquake Media Coverage Map’. Building on his background in developing the Hiroshima Archive and more recently the East Japan Earthquake Archive, Hidenori came up with the initiative, which aims to understand the reality of media coverage during the 3.11 Japan quake.
Introductory video of Hidenori’s past project ‘The East Japan Earthquake Archive’
“The media coverage gap can be detected by visualising the mainstream media’s news coverage after the quake on a map.” – Hidenori Watanabe
By using Google Earth, the Media Coverage Map combines different data layers varying from the data sets which are provided by NHK and JCC, to crowd-generated content such as geolocated tweets and WeatherNews Inc’s citizen reports.
By comparing multiple types of data, the map can potentially reveal what types of information on which locations were covered in the news. On the other hand, did the crowd report from the location wheres mainstream media did not cover? If so, what did they report?
The method of collecting data is an important factor to define the reliability of map. In the Media Coverage Map, three types of data layers caught our attention.
The map visualises textual data generated from NHK’s breaking news during the first 24 hours after the quake. Each piece of data contains the news anchor’s script, the types of information (e.g categorisation of updates into earthquake, tsunami, rescue, situation unknown etc), the exact times and the names of locations mentioned in the broadcast. On the Media Coverage Map, the red dots that spread throughout the country represent locations that were mentioned in NHK’s breaking news coverage.
One of the crowd-generated datasets used in the map is the geolocated tweets during the period of 11-18 March 2012. The list was compiled by a group of engineers at the Big Data Workshop, and each tweet has been marked with a green circle on the map. The ‘noise’, however, is not filtered out, which means that the list contains all kinds of geo-tagged tweets regardless of the content. “The role that geo-tagged tweets play here is where and how many people were tweeting during the crucial period after the quake, instead of what was being tweeted about”, says the map developer.
While the issue of ‘noise’ from geo-tagged tweets prevails, the quality of crowdsourced data being provided by WeatherNews Inc. greatly contributed to the mapping project, according to the associate professor. Its Gensai Report project, which means ‘report to minimise the damages caused by disaster’ in Japanese, gathered user-reported information sorted into twelve categories such as building wreckage, liquefaction, lifeline damage/recovery and cracks in the roads. The authenticity of crowd information is strengthened by the established network of experienced users who had contributed to previous WeatherNews’ crowdsourced projects. In short, the existing network of contributors who are prepared to effectively contribute to information sharing is a key to have a good crowdsourced data sets for better mapping.
The Power of the Crowd: Filling the Gap of Disaster News Coverage
So, what were the findings from this multi-layered map? Hidenori points out a number of cases where there was no news coverage but a lot of geotagged tweets and crowdsourced reports during the first week of the disaster. For example, while there was no news broadcasted about the city of Kamisu (Ibaragi prefecture) a number of geo-locatable tweets were found. In the below image, the tweet states, “Our factory is confronted with catastrophic liquefaction. It probably will not function even if the building did not collapse.”
Another example is the following case in which a citizen submitted a report to Gensai Report indicating the need of water, food and sanitary equipment in Ichinoseki city (Iwate prefecture). The individual reporting this had slept in a car because of cracks in the walls of his house. Again, there was no NHK coverage on the region for the entire week.
The following example highlights the devastating reality of the event. The large part of Kesennuma (Miyagi prefecture) was washed out by the tsunami, resulting 1,038 deaths and 259 residents still missing to date. A citizen reporter submitted a post around midnight on 11th March, writing “People are calling for help but because of the helicopter we cannot hear their voices. Please make them stop flying over here.”
There were a number of helicopters flying around that area for the first few nights, including some belonging to media outlets. One piece of footage from Asahi Shimbun shows that their helicopter was flying over Kesennnuma at around 1:30AM on 12 March.
Footage of Kesennuma which was shot by Asahi Shimbun at around 1:30AM on 12 March, 2011
Lessons-learned from Mapping Various Data on Media Coverage
While this mapping project is still in its beta stage, one valuable lesson learned so far is the significant power of the crowd, particularly during a large-scale disaster where media cannot cover the entire affected area. Having contributed significantly to Project 311, NHK is making use of the data sets to assess how their disaster coverage can be improved, and how they could make use of social media to gather story leads to fill the media coverage gap when a disaster of this magnitude occurs.
In the future, additional data layers such as panoramic images and articles from Asahi Shimbun that are already used in the East Japan Earthquake Archive will be integrated into the Media Coverage Map. Cooperating with key data providers and research institutes, the Media Coverage Map may reveal a more detailed reality of the experiences that major media outlets could not report during the 3.11 Japan quake.
Photo: wtnv studio