Heavy Rain Event in Japan

From LINKS Community Center
Jump to: navigation, search
Example case of a regional government affected by the disaster using a Twitter-based AI Disaster Risk Management Solution to confirm the damage and provide contact information for multiple rescues.
Flooding, Landslide
Kyushu and Chubu regions, Japan
Involved Organisations:
Oita Prefecture Officials
Publishing Organisation

Kobe University




Crowdsourcing, Social Media

  • Collecting and Analysing Information from SMCS
  • Ensuring Credible Information
  • Making Information Accessible
  • Mobilising Citizens
  • Targeting Communication
Disaster Management Phase



Overview: On July 5, 2020, record-breaking rainfall caused extensive damages in Kyushu and Chubu regions (southwest and central regions, Japan). Social media posts requesting rescue and safety confirmation were highlighted by people who had been isolated due to landslides and other damages. Oita Prefecture, one of the regions affected by the disaster, could reply to social media posts from the disaster-stricken areas, using a Twitter-based AI Disaster Risk Management Solution. This system collected messages from residents in risk situations during the disaster. Its use was effective for verifying damage situations and providing appropriate contact information, which was used for multiple rescues. (Ref. 1)


Using the AI solution, officials from Oita Prefecture found a post of a resident who had become isolated after a landslide blocked a road. In one notorious case, residents of a family in the area could not contact the authorities, because utility poles had fallen, and the landline and cellular phone services were out of service. While they managed to get an Internet connection, they posted on Twitter: "Mudslide right behind the evacuation center. Power outage, roads blocked, river flooding, we are isolated. It was too hard to evacuate with a one-and-a-half-year-old and a pregnant woman." (Ref. 2)

After Oita Prefecture officials used the AI solution to find the above message successfully, the government immediately used the official Twitter account to send a reply, giving the contact information (phone number) for the disaster headquarters. Then, residents who received the reply informed the prefecture of their family's condition and the current situation in the area. Afterward, the prefecture shared information with the local fire department. In response, fire department officials were dispatched to ensure the safety of those residents. (Ref. 2)

As subsequent reports, these residents said, according to local news, "At the time, I had a family member who was seven months pregnant. We also had young children and wanted someone to know what was happening. We wanted someone to find and help us in this difficult and stranded situation. That's how I felt when I posted this". Three months after the disaster, they gave birth safely. The mother said, "I am grateful to the prefectural government for catching our social media postings and connecting us to support." Also, the official from the Oita Prefecture, who used the AI solution to find the above tweet posted by these residents, said, "We in government do not have many opportunities to interact with residents. The social media was effective because it lets us get detailed information from people on the site". (Ref. 2)


Not all prefectures and municipalities have resources like the abovementioned AI solution. And even these smart solutions are the result of recent efforts. In the event of heavy rainfall that strike Kyushu region in 2017 (before the adoption of AI solutions), for example, people affected by the disaster can find a way to request rescue by Twitter using the hashtag #救助 (kyujyo, meaning "rescue" in Japanese) if they are unable to make a phone call. Therefore, when trying to find information from #救助, tweets unrelated to rescue requests were lined up, "burying" important information (Ref. 3).

After the event occurred in 2017, researchers from the International Research Institute of Disaster Science at Tohoku University (IRIDeS) analyzed 1,058 tweets in which #救助 was used and found that only 2% of the tweets were presumed to have come from the disaster area, while the rest were from outside the disaster area and were not urgent. In addition, when some news media tweeted articles calling for the use of this hashtag, they used #救助 as it is, which caused recipients to retweet the articles to spread them, thus contributing to the "burying" of information. The IRIDeS researchers concluded, "The challenge is to improve the social media manners of people outside the affected areas. Also, the news media should be aware of the magnitude of their influence." News media also emphasized that the hashtag #救助 is likely to be used as a "last recourse" for communication from disaster-stricken areas when telephone service is unavailable. Also, the Rescue-dedicated page on Twitter urges people not to spread the hashtag unnecessarily, but to call emergency services - 119 in Japan (Ref. 3).

Ref 1. Article from PR TIMES (Japanese), https://prtimes.jp/main/html/rd/p/000000072.000016808.html 21 Jan. 2021 (accessed 2023/03/28)

Ref 2. Article from NHK (Japanese), https://www.nhk.or.jp/ashitanavi/article/10326.html 2 Nov. 2021 (accessed 2023/03/28)

Ref 3. Article from Broadcasting Culture Research Institute, NHK (Japanese), https://www.nhk.or.jp/bunken/research/focus/f20170901_3.html , Sep. 2017 (accessed 20023/03/28)
What was the overall goal of the Use Case?
Crisis information collection, notification, visualization and forecasting
What limitations were identified?
Even with recent efforts based on Social Media and AI Disaster Risk Management Solutions, it is essential to note that some types of landslides may occur without any anomalies in prior observations and end up hitting residents suddenly. Under these circumstances, local community information loses its utility. All that remains is to rely on forecast information, which can contribute to estimating the risk and extent of damages resulting from landslides. Although such information should be interpreted carefully as it contains uncertainties, it should be as accurate as possible using the latest technology, based on information released by the meteorological agency and geological surveys.
Which social media and crowdsourcing technologies were used?
Which social media platforms were used?