An Analysis of Web Coverage on the 2018 West Japan Heavy Rain Disaster

https://doi.org/10.20965/jdr.2019.p0531 An Analysis of Web Coverage on the 2018 West Japan Heavy Rain Disaster

Survey Report: An Analysis of Web Coverage on the 2018 West Japan Heavy Rain Disaster Shosuke Sato† and Fumihiko Imamura

International Research Institute of Disaster Science (IRIDeS), Tohoku University 468-1 Aoba Aramaki, Aoba-ku, Sendai, Miyagi 980-0845, Japan †Corresponding author, E-mail: [email protected] [Received November 6, 2018; accepted January 30, 2019]

600 This study analyzed quantitative big data from web 545 at maximum す᪥ᮏ㇦㞵䠄 ᖺ䠅 news on the West Japan Heavy Rain disaster for a two- (3rd day) West Japan Heavy2018 Rain (2018) month period. The retrieved information was com- 500 ⇃ᮏᆅ㟈䠄Kumamoto2016 Earthquakeᖺ䠅 (2016) ᪂₲┴୰㉺ᆅ㟈䠄Niigata-Chuetsu Earthquake2004ᖺ䠅 (2004) pared with previous natural disaster coverage. The re- 366 at maximum sults indicated the following. 1) For natural disasters 400 (8th day) that had occurred over the past 15 years, the “half-life 300 period for media exposure” (i.e., the period in which Half-life period of West the amount of news reporting halves) was approxi- 176 at maximum Japan Heavy Rain 200 (2nd day) (3rd time): 24th day Numberarticles of mately one week, while the half-life period of web me- Half-life period of Kumamoto (2nd time): 7th day dia exposure on the West Japan Heavy Rain disaster 100 was 24 days. Thus, the West Japan Heavy Rain disas- Half-life period of Chuetsu: 7th day ter appeared to be the most significant social concern 0 since the Great East Japan Earthquake. 2) The West 1 3 5 7 9 111315171921232527293133353739 Japan Heavy Rain disaster was large enough to affect ⅏ᐖⓎ⏕䛛䜙⤒㐣᪥ᩘ䠄⅏ᐖⓎ⏕᪥䜢Days after occurrence of disaster1᪥┠䛸䛧䛶䠅 (The disaster occurring day was counted as first day.) both the Chugoku and Shikoku Districts, but the available human support was comparable to the extent of Fig. 1. Time-series change concerning Yahoo! News ar- the human and material damages as well as the related ticles on the West Japan Heavy Rain disaster, Kumamoto amount of media coverage. No significant regional dif- Earthquake, and Niigata-Chuetsu Earthquake. ferences in the amount of media coverage or support were found. 2. Data Keywords: disaster information, web coverage, the 2018 West Japan Heavy Rain, media exposure This study focused on news coverage of the West Japan Heavy Rain disaster as released by Yahoo! News. The dataset was collected through an “OR” search using the 1. Introduction keywords “heavy rain (Japanese text: ‘Bg1+’)” and “heavy rainfall (Japanese text: ‘9k1+’)” to obtain news that was There has been significant web news coverage on every released from July 5, 2018 to September 6, 2018. Articles disaster since the Niigata-Chuetsu Earthquake in 2004. related to the West Japan Heavy Rain disaster were then The first author of this paper used these news sources selected for analysis. We then archived the data and de- to analyze social recognition of the covered disasters, veloped an analytical database consisting of release dates thereby providing a better understanding of these situa- and times, news media sources, headlines, and articles. tions as information sources [1, 2]. Web news provides useful linguistic data for understanding real-time events. This aids in analyzing and discussing social responses at 3. Time-Series Trend the time of their occurrence. This study analyzed and characterized big data from web news coverage of the Figure 1 shows a comparison of the number of arti- July 2017 heavy rain (referred to in this paper as the West cles on the West Japan Heavy Rain disaster, the Niigata- Japan Heavy Rain in 2018) for a two-month period after Chuetsu Earthquake in 2004, and the Kumamoto Earth- the disaster. quake in 2016 [3, 4]. All articles on the two earthquakes were collected from Yahoo! News. The number of articles on the Niigata-Chuetsu and Ku- mamoto earthquakes peaked between two and three days after the events (Fig. 1) [3]. This is because the public was unaware of these disasters immediately after, but gradu-

Journal of Disaster Research Vol.14 No.3, 2019 531

© Fuji Technology Press Ltd. Creative Commons CC BY-ND: This is an Open Access article distributed under the terms of the Creative Commons Attribution-NoDerivatives 4.0 International License (http://creativecommons.org/licenses/by-nd/4.0/). Sato,S.andImamura,F.

10,000 10,000 10,000 Number of completely Number of completely Number of victims destroyed and half destroyed destroyed houses Kurashiki houses and houses with Kurashiki inundation above floor level Kurashiki Hiroshima Hiroshima Hiroshima 1,000 Kure 1,000 Kure 1,000 Kure Uwajima Uwajima Uwajima OzuSeiyo Sakamachi SeiyoSakamachiMihara Seiyo OzuSakamachi OkayamaSoja Mihara Okayama Soja Mihara Soja Okayama MatsuyamaHigashihiroshima Higashihiroshima MatsuyamaHigashihiroshima TakahashiFukuyama Kumano FukuyamaKumanoTak ahas hi KumanoTak ahasFukuyama h i Fuchu Fuchu FuchuFuchu Onomichi Kaita Onomichi Kaita Ibara Ibara IbaraKaita 100KasaokaNiimi 100 Kasaoka 100 Niimi Kasaoka Imabari Ts uyYakage a maEdajima Iyo ImabariTsEdajima uy amaMiyoshiYakage Takehara Takehara Yahatahama Takehara Kagamino Shobara KagaminoKihoku Maniwa Kagamino KihokuMimasakuMatsuno Asakuchi Akitakata Akitakata Tamano AsakuchiAkitakataWakeManiwa MisakiKamijimaUchiko Number Number of articles Number of articles Number of articles HatsukaichiBizenOtake 10 10 10 KibichuoAinanSera Satosho IkataNishiawakura Osakikamijima Osakikamijima R=0.952 R=0.809 Akaiwa R=0.843 1 1 1 1101001101001,00010,000 1 10 100 1,000 10,000 Number of deaths and missing persons Number of completely destroyed houses Number of completely destroyed and half destroyed houses and houses with inundation above floor level

Fig. 2. Relationship between the damage scale and news coverage numbers for two months after the West Japan Heavy Rain disaster.

ally learned about them two or more days later [4]. Earthquake [4]. Figure 1 also shows the “half-life period for media This case study analyzed the West Japan Heavy Rain exposure of the disaster” [5]. This indicates the num- disaster to determine the relationship between the dam- ber of days subsequent to the disaster in which the num- age in each city, town, and village in the Hiroshima, ber of articles lowers by more than the half the “maxi- Okayama, and Ehime Prefectures as reported for a two- mum number of articles per day.” The half-life period month period after the heavy rain and the number of news thus indicates the persistence of social concern. The half- articles (i.e., the extent of media exposure) released in life period of media exposure was one week (i.e., seven that timeframe. In the surveyed areas of the Hiroshima, days) for both the Niigata-Chuetsu and Kumamoto Earth- Okayama, and Ehime Prefectures, the number of deaths quakes, while the half-life period of web media exposure and missing persons as of October 2018 was 114, 64, for the West Japan Heavy Rain disaster was 24 days. That and 29, respectively. These numbers were significantly is, it lasted about three times longer than coverage of ei- larger than those in other prefectures. The number of ther of the two earthquakes. After examining the half- news articles was then calculated. Here, an article was life period of media exposure, the author determined that counted as one if it contained the name of a city, town, no disaster coverage other than that concerning the Great or village (i.e., the name of an affected city, town, or vil- East Japan Earthquake had a half-life that exceeded one lage was presented in the article). Fig. 2 shows the re- week. Specifically, the half-life period for media exposure lationships between the number of news articles and the on the Great East Japan Earthquake lasted approximately number of deaths and missing persons (i.e., the number two months [5]. Thus, the West Japan Heavy Rain disas- of victims), the number of completely destroyed houses, ter attracted the second most significant amount of social the total number of completely and half-destroyed houses, concern in the last 15 years (Fig. 2). and houses with that were inundated with water above floor-level for each city, town, and village. These data were current as of September 6, 2018. Regression formula 4. Relationship Between the Damage Scale, Y = aX +b was created based on each scatterplot diagram Amount of News Coverage, and Amount of in Fig. 2, while the residual of the data from the straight Support line is shown in Fig. 3 for each city, town, and village. Two different damage scales were used. The first involved Sato et al. [5] showed that the extent of support from the amount of damage as calculated based on the num- areas other than those incurring damage significantly af- ber of deaths, number of completely destroyed houses, fected the amount of news coverage. For the Great East and other related factors. The second involved the rate Japan Earthquake, the amount of news coverage in each of damage as calculated based on the ratio of victims or area was highly correlated with the extent of available hu- the percentage of destroyed houses. In studies [4] and [6], man support. Even though there was significant human the extent of media exposure or the amount of available and material damage, some areas had relatively little hu- support was found to strongly correlate with the amount man support because of low media exposure. On the other of damage. We thus employed the amount of damage as hand, the amount of news coverage and extent of human an index for the damage scale. support concerning the Kumamoto Earthquake were pro- Figure 2 shows that a large number of news arti- portional to the scale of human and material damages in cles concerned Kurashiki City and Okayama City in the all cities, towns, and villages, while no significant discrep- Okayama Prefecture as well as Hiroshima City and Kure ancy was observed between the amount of news/support City in the Hiroshima Prefecture. These areas incurred R and the scale of damage found after the Great East Japan large-scale damages. Correlation coefficient was mea-

532 Journal of Disaster Research Vol.14 No.3, 2019 An Analysis of Web Coverage on the 2018 West Japan Heavy Rain Disaster

Number of articles gional differences in media exposure after the Great East -2,500-1,500 -5000 500 1,500 2,500 Japan Earthquake were also expected, but did not occur at significant levels. Hiroshima Higashihiroshima Figure 3 shows that the measured values exceeded Kumano Uwajima those estimated using the regression formula (X indicates Soja Kure Sakamachi the number of victims in Okayama City and Kurashiki Ok ayama Fukuyama City; the residuals are thus negative in these two cities, Kurashiki Ibara thereby indicating that they were exposed to significant Kasaoka Asakuchi media coverage). There was also significant media cov- Satosho Takahashi erage in Sendai City after the Great East Japan Earth- Niimi Kagamino quake and in Kumamoto City after the Kumamoto Earth- Tsuyama Tamano quake [4, 5]. Here, media exposure tended to be great- Akaiwa Maniwa est in designated cities and prefectural capitals. This ten- Wake Bizen Okayama Mimasaku dency was also observed after the West Japan Heavy Rain Yakage Pref. Nishiawakura disaster. Takehara City, Higashihiroshima City, and Saka Misaki Kibichuo Town in the Hiroshima Prefecture tended to receive less Setouchi Hayashima media coverage regarding the number of victims and scale Sboo Nagi of housing damage. Shinjo Takehara We then analyzed the relationship between the dam- Mihara Onomichi age scale and the extent of on-site human support. Fig. 4 Fuchu Hatsukaichi Akitakata shows the relationship between the damage scale provided Miyoshi Shobara in Fig. 2 and the cumulative number of volunteers who City, town, village town, City, Ot ak e Edajima arrived in each city, town, and village for a two-month Fuchu Kaita period subsequent the West Japan Heavy Rain disaster. Akiota Hiroshima Osakikamijima In addition, regression formula Y = aX + b was created Kitahiroshima Sera Pref. based on each scatterplot diagram shown in Fig. 4 and the Jinsek ik og en Seiyo residual of the data from the straight line shown in Fig. 5 Iyo EstimationṚ⪅࣭⾜᪉୙᫂ Oz u for each city, town, and village. Matsuyama residual of deaths Imabari ⪅ᩘ࡟ࡶ࡜࡙ࡃ Kihoku and missing Figure 4 indicates that a large amount of human sup- Uchiko persons᥎ᐃṧᕪ Yahatahama port (i.e., the number of volunteers) was available in Ainan Estimation residual of Kumakogen Kurashiki City in the Okayama Prefecture and Hiroshima completely඲ቯ࣭༙ቯ࣭ᗋ destroyed and Matsuno Ikata halfୖᲷᩘ࡟ࡶ࡜࡙ destroyed houses and City and Kure City in the Hiroshima Prefecture. These Tobe houses with inundation Ehime Shikokuchuo aboveࡃ᥎ᐃṧᕪ floor level areas experienced large-scale damage. Correlation coeffi- Kamijima Pref. Masaki cient R was measured between 0.700 and 0.894. Thus, no city, town, or village exhibited significant deviation con- Fig. 3. Residual calculated using a regression formula for cerning the data shown according to the straight line in the scale of damage and news coverage numbers. Fig. 2, which shows a different tendency from that observed after the Great East Japan Earthquake. Figures 4 and 5 show that a small number of volunteers were available in relation to the damage scale based sured between 0.809 and 0.952. The coefficient correla- on the number of victims in Kumano Town, Higashihi- tion between the number of news reports and the num- roshima City, Hiroshima City, and Akitakata City and for ber of completely destroyed houses was 0.809, while the the damage scale as measured by the number of com- correlation between the total number of completely and pletely and half-destroyed houses and houses that were half-destroyed houses and houses that were inundated inundated with water above floor-level in Okayama City, with water above floor-level had a larger coefficient value Fukuyama City, Miyoshi City, Kasaoka City, and Yakage of 0.843. This indicates that the number of news articles Town. On the other hand, a large number of volunteers was more significantly affected by the number of houses were available in both cases in Soja City, Kure City, and that were inundated with water than by the overall scale Saka Town and in relation to the scale of housing damage of housing damages. in Uwajima City. No city, town, or village largely devi- A previous study established a correlation coefficient ated from the straight line. of 0.789–0.972 for the Kumamoto Earthquake [4]. This We then analyzed the relationship between the dam- was almost the same as that for the West Japan Heavy age scale and monetary support. Previous studies ex- Rain disaster. The most severely damaged area resulting amined the relationship between the damage scale and from the Kumamoto Earthquake was within Kumamoto the extent of human support (e.g., the number of volun- Prefecture. However, the most severe damage resulting teers) [4, 5]. However, damaged areas also collected do- from the West Japan Heavy Rain disaster was not concen- nations and contributions, neither of which required the trated in the Hiroshima, Okayama, or Ehime Prefectures, movement of people. This study focused on data from but extended into the Chubu and Kyushu districts. Re- the hometown tax payment program, which could be sys-

Journal of Disaster Research Vol.14 No.3, 2019 533 Sato,S.andImamura,F.

100,000 100,000 100,000 Number of completely Kurashiki Kurashiki Number of completely Number of victims destroyed houses Kurashiki Kure Kure destroyed and half destroyed houses and houses with Kure Sakamachi Sakamachi Sakamachi Hiroshima Hiroshima inundation above floor level Hiroshima Soja Soja Soja Okayama MiharaUwajima Okayama UwajimaMihara Uwajima 10,000 OzuSeiyo 10,000 Seiyo Ozu MiharaSeiyo Ozu Okayama Higashihiroshima Higashihiroshima 10,000 Higashihiroshima

Takahashi Ta k a h a s h i Tak eh ara Takehara TakeharaTakahashi Onomichi Fuchu EdajimaOnomichi Edajima Kaita Matsuyama Kumano YakageMatsuyamaKaitaKumano Fuchu OnomichiKaitaYakage Fukuyama Fukuyama MatsuyamaKumano Fukuyama Imabari Imabari Imabari 1,000 1,000 Fuchu Fuchu 1,000 Fuchu Kasaoka Kasaoka Kasaoka Ibara Ibara Ibara Akitakata AkitakataKamijimaShobara Ikata KamijimaAkitakata Kihoku Number of volunteers Shobara Asakuchi SeraYahatahama KihokuSeraYahatahama

Number of volunteers Asakuchi Niimi Niimi Number of volunteers Niimi R=0.932 Osakikamijima R=0.700 Osakikamijima R=0.833 100 100 100 1101001 10 100 1,000 10,000 1 10 100 1,000 10,000 Number of deaths and missing persons Number of completely destroyed houses Number of completely destroyed and half destroyed houses and houses with inundation above floor level

Fig. 4. Relationship between the damage scale and human support (volunteers) for two months after the West Japan Heavy Rain disaster.

Number of people Choice’s Disaster Recovery Assistance website [7], which -14,000 -4,0000 6,000 16,000 is a portal site for the hometown tax payment designed to support disaster recovery. We were therefore able to look Hiroshima Higashihiroshima at the data. Kumano Uwajima Figure 6 Soja shows the relationship between the damage Kure Sakamachi scale shown in Fig. 4 and the amount of hometown taxes Ok ayama Fukuyama paid to each city, town, and village for a two-month pe- Kurashiki Ibara riod after the West Japan Heavy Rain disaster. In addi- Kasaoka A sak uchi tion, regression formula Y = aX +b was created based on Satosho Takahashi each scatterplot diagram in Fig. 6 and a residual of the Niimi Kagamino data from the straight line is shown in Fig. 7 for each city, Tsuyama Tamano town, and village. Akaiwa Maniwa Wake Figure 6 shows that large amounts of hometown taxes Bizen Okayama Mimasaku were paid to Kurashiki City in the Okayama Prefecture Yakage Pref. Nishiawakura and Kure City in the Hiroshima Prefecture, while Fig. 4 Misaki Kibichuo shows that these areas also received a large number of Setouchi Hayashima volunteers. This was because of the large damage scale. Sboo Nagi Correlation coefﬁcient R was measured between 0.727– Shinjo Takehara 0.832, and no city, town, or village largely deviated from Mihara Onomichi Fuchu the straight line. This differs from the case of the Great Hatsukaichi Akitakata East Japan Earthquake [5]. Miyoshi Shobara Figures 6 7 City, town, village and show that large amounts of hometown Ot ak e Edajima taxes were paid based on the damage scales in either case Fuchu Kaita in Kure City, Iyo City, Seiyo City, and Kurashiki City, Akiota Hiroshima Osakikamijima while the amounts were small for the damage scales in Kitahiroshima Sera Pref. Hiroshima City and Saka Town. On the other hand, no Jinsekikogen Seiyo Iyo city, town, or village largely deviated from the straight Oz u Estimation Matsuyama Ṛ⪅࣭⾜᪉୙᫂ line. Imabari residual of deaths Kihoku ⪅ᩘ࡟ࡶ࡜࡙ࡃ The analyses shown in Figs. 2–5 show that the large re- Uchiko and missing Yahatahama ᥎ᐃṧᕪ gional differences in media coverage as discussed in pre- Ainan persons Kumakogen Estimation residual of vious studies on the Great East Japan Earthquake were Matsuno ඲ቯ࣭༙ቯ࣭ᗋ Ikata completely destroyed and not found in the present case. This was also the case Tobe Ehime halfୖᲷᩘ࡟ࡶ࡜࡙ destroyed houses and Shikokuchuo houses with inundation during the Kumamoto Earthquake. A case study analy- Kamijima Pref. aboveࡃ᥎ᐃṧᕪ floor level Masaki sis of the Kumamoto Earthquake revealed the following two causes: Fig. 5. Residual calculated using a regression formula for the extent of human support and news coverage numbers. 1) Smaller areas were damaged by the Kumamoto Earth- quake than by the Great East Japan Earthquake, which not only damaged the Iwate, Miyagi, and Fukushima Prefectures, but also affected wide areas of the Kanto tematically collected from the surveyed areas at the time District. On the other hand, the Kumamoto Earthquake of writing. These data were acquired from the Furusato only caused deaths in the Kumamoto Prefecture. This

534 Journal of Disaster Research Vol.14 No.3, 2019 An Analysis of Web Coverage on the 2018 West Japan Heavy Rain Disaster

100,000 100,000 100,000 Number of victims Number of completely Number of completely destroyed and half destroyed destroyed houses houses and houses with Kurashiki Kurashiki inundation above floor level Kurashiki Kure Kure Kure

10,000 10,000 10,000

OzuSeiyo Iyo Seiyo Ozu Iyo Seiyo Ozu Takahashi Soja HigashihiroshimaHiroshima HigashihiroshimaTakahashiHiroshima Hiroshima Mihara SojaMihara HigashihiroshimaMiharaTak ahasSoja hi Onomichi Kumano KumanoOnomichi Kumano Takehara Tak ehar a OnomichiTa k e h a r a Sakamachi Sakamachi Sakamachi Uwajima Edajima Uwajima Edajima 1,000 Ibara 1,000 YakageIbara Uwajima Kaita 1,000 IbaraYakage Fuchu Fuchu Kaita Fuchu Kaita Kasaoka ManiwaFuchuKasaoka ManiwaFuchu KihokuFukuyamaOkayamaImabariMatsuyama MatsuyamaOkayamaImabari Matsuno MiyoshiKasaoka Okayama Satosho Akitakata Akitakata Fukuyama KihokuAkitakataImabariMatsuyama Fukuyama Shobara Shobara

Hometown Hometown tax paid (10,000 yen) Kagamino Kagamino Wake

Hometown Hometown tax paid (10,000 yen) KagaminoJinsekikogen

R=0.832 Hometown tax paid (10,000 yen) R=0.774 R=0.727 100 100 100 1101001 10 100 1,000 10,000 1 10 100 1,000 10,000 Number of deaths and missing persons Number of completely destroyed houses Number of completely destroyed and half destroyed houses and houses with inundation above floor level

Fig. 6. Relationship between the damage scale and hometown tax payment program for two months after the West Japan Heavy Rain disaster.

10,000 yen trated around the damaged areas. The case was differ- -8,000-4,000 00 4,000 8,00012,000 ent for the Great East Japan Earthquake.

Hiroshima 2) Media sources improved their skills and technolo- Higashihiroshima Kumano gies after the Great East Japan Earthquake. Lessons Uwajima Soja learned from the regional differences in media cover- Kure Sakamachi age after the Great East Japan Earthquake likely re- Ok ayama Fukuyama Kurashiki sulted in improved operations. Ibara Kasaoka Asakuchi The above causes are discussed from the viewpoint of Satosho Takahashi the current situation resulting from the West Japan Heavy Niimi Kagamino Tsuyama Rain disaster. Cause 2) (i.e., improved skills and tech- Tamano Akaiwa nologies for people involved in the media) was an impor- Maniwa Wake tant factor in the West Japan Heavy Rain disaster and the Bizen Okayama Mimasaku Kumamoto Earthquake. The West Japan Heavy Rain dis- Yakage Pref. Nishiawakura Misaki aster occurred only two years after the Kumamoto Earth- Kibichuo Setouchi quake. Thus, we believe that media technologies and ex- Hayashima Sboo periences were maintained. For cause 1) (i.e., the size of Nagi Shinjo Takehara the damaged area), we ﬁrst expected that different inﬂu- Mihara Onomichi ences existed during the West Japan Heavy Rain disaster Fuchu Hatsukaichi when compared to the Kumamoto Earthquake. The dam- Akitakata Miyoshi age caused by the West Japan Heavy Rain disaster was

City, town, village town, City, Shobara Ot ak e Edajima not limited to the above three prefectures, but extended Fuchu Kaita from the Chubu District to the Kyushu District. Still, there Akiota Hiroshima Osakikamijima were no significant regional differences in media coverage Kitahiroshima Sera Pref. Jinsekikogen or human support. This may be because the damages in Seiyo Iyo each of these areas were not as large as those caused by Oz u EstimationṚ⪅࣭⾜᪉୙᫂ Matsuyama the Great East Japan Earthquake. On the other hand, it is Imabari residual⪅ᩘ࡟ࡶ࡜࡙ࡃ of deaths Kihoku and missing notable that no regional differences were found in media Uchiko ᥎ᐃṧᕪ Yahatahama persons Ainan coverage or human support for a large-area disaster such Kumakogen Estimation residual of Matsuno completely඲ቯ࣭༙ቯ࣭ᗋ destroyed and as that experienced during the West Japan Heavy Rain. Ikata half destroyed houses and Tobe Ehime ୖᲷᩘ࡟ࡶ࡜࡙ To overview the above tendency, a cluster analysis was Shikokuchuo houses with inundation Kamijima Pref. aboveࡃ᥎ᐃṧᕪ floor level Masaki performed using the residuals shown in Figs. 3, 5,and 7 for each city, town, and village. The results are pre- Fig. 7. Residual calculated using a regression formula for sented in Fig. 8 (the Ward method with squared Euclid the amount of hometown taxes paid and news coverage num- distance was used). We used three sets of residuals ob- bers. tained from the regression formula. Here, x represents the number of deaths and missing persons, which provided the best fit for all analyses. Fig. 6 shows the results of each city, town, and village after classification into four could be because a relatively small area was damaged clusters with a distance threshold of 5 to 6. The names of by the Kumamoto Earthquake. In this case, media the 59 cities, towns, and villages used in Cluster 1 (top of resources would have easily been effectively concen- Fig. 8) were omitted because of their length.

Journal of Disaster Research Vol.14 No.3, 2019 535 Sato,S.andImamura,F.

59 cities, towns, and villages other than below Cluster 1:Almost average

Higashihiroshima (Hiroshima Pref.) Cluster 2: Few volunteers and low amount Kumano (Hiroshima Pref.) of hometown tax for the damage scale Hiroshima (Hiroshima Pref.)

Okayama (Okayama Pref.) Cluster 3: Many volunteers for the Saka (Hiroshima Pref.) damage scale Soja (Okayama Pref.) Cluster 4: Many volunteers and high amount Kure (Hiroshima Pref.) of hometown tax for the damage scale

Fig. 8. Cluster analysis results based on a regression residual of the number of articles concerning deaths and missing persons, volunteer numbers, and amount of hometown taxes paid.

The cluster features are described as follows: Saka Town and Kure City in Hiroshima Prefecture had “volunteer bus” and “volunteer ship” systems to transport Cluster 1: This cluster contained 59 average cities, volunteers [10] and Soja City in Okayama Prefecture had towns, and villages that were not classified a system designed to receive more than 1,000 volunteers into the other clusters. per day [11]. In addition, the Cabinet Office established the “Na- Cluster 2: This cluster contained cities, towns, and vil- tional Information Sharing Committee” after the West lages in which the numbers of volunteers and Japan Heavy Rain disaster. This was done in collaboration hometown taxes were both low in relation with the Japan Voluntary Organizations Active in Disaster to the damage scale (e.g., Higashihiroshima (JVOAD) and other organizations to control volunteer ac- City, Kumano City, and Hiroshima City in the ceptance practices over a wide area [12]. These measures Hiroshima Prefecture). compensated for regional differences. Namely, such dif- Cluster 3: This cluster contained cities, towns, and vil- ferences were not only corrected by mass media person- lages in which there were large numbers of nel who attempted to provide media coverage, but also volunteers in relation to the damage scale by supporters who took similar actions. This may be the (e.g., Okayama City in Okayama Prefecture reason for the absence of significant regional differences and Saka Town in Hiroshima Prefecture). after the West Japan Heavy Rain disaster. Large amounts of hometown taxes were paid to Soja City and Kure City Cluster 4: This cluster contained cities, towns, and vil- in Cluster 4. This was not only done because the cities lages in which there were large numbers of had their own tax payment accounts on their websites, but volunteers and hometown taxes in relation to also because three other cities called for hometown taxes the damage scale (e.g., Soja City in Okayama to be paid to each of the two cities [7]. Prefecture and Kure City in Hiroshima Pre- fecture). 5. Conclusions Low numbers of volunteers were present in the three cities classified into Cluster 2. This is because the progress rate This study analyzed and discussed the qualitative as- for volunteer activities had reached 99% as of Septem- pects of Yahoo! News articles released over a two-month ber 7 and the volunteer center was closed in Higashihi- period after the West Japan Heavy Rain disaster. Data roshima City [8]. Volunteers were called in for a short were compared with previous natural disasters. The re- period of time and their activities in Kumano Town were sults are summarized below. finished before August 26 [9]. Media coverage in Hi- roshima City was also not significantly higher than in or- 1) Natural disasters occurring over the last 15 years have dinary times because it is the prefectural capital. had “half-life media exposure periods” of approxi- There were large numbers of volunteers in relation to mately one week each. Here, the half-life period in- the damage scales in Clusters 3 and 4. This is because dicates the amount of time in which the amount of ini-

536 Journal of Disaster Research Vol.14 No.3, 2019 An Analysis of Web Coverage on the 2018 West Japan Heavy Rain Disaster

tial news coverage was halved. On the other hand, the its Associated Factors in the 2011 Great East Japan Earthquake Dis- half-life period of web media exposure for the West aster,” J. of Social Safety Science , No.19, pp. 93-103, 2013. [7] Furusato Choice’s Disaster Recovery Assistance, https://www. Japan Heavy Rain was 24 days. This indicates that the furusato-tax.jp/saigai/ [accessed October 30, 2018] disaster attracted the most significant amount of social [8] The Hiroshima City Council of Social Welfare, “Closing of Higashihiroshima City Sufferer Life Support Center,” https:// concern since the Great East Japan Earthquake. higashihiroshimashi-syakyo.jp/detail saigai.php?id=429&mode=1 2) The West Japan Heavy Rain disaster covered a wide [accessed October 30, 2018] [9] Kumano Town Council of Social Welfare, “Announcement from area, including both the Chugoku and Shikoku Dis- Kumano Town Disaster Relief Volunteer Center,” http://park11. tricts. However, the extent of available human support wakwak.com/∼kumano-shakyo/oshirase/index.html [accessed Oc- tober 30, 2018] was comparable to the extent of the human and mate- [10] Hiroshima Disaster Relief Volunteer Center, “Information of Hi- rial damages and the amount of related media cover- roshima Disaster Relief Volunteer,” https://www.facebook.com/ age. We found no significant regional differences in hiroshima.vc/ [accessed October 30, 2018] [11] The Sanyo Shimbun, “Soja City government accepts disaster relief the amounts of media coverage or support. volunteers – a thousand people in 14–16th, July –,” Press news: July 13, 2018, http://www.sanyonews.jp/article/750343/1/ [accessed Oc- This study on the West Japan Heavy Rain disaster and tober 30, 2018] [12] The Sankei Shimbun, “West Japan Heavy Rain: Disparity of previous studies on the Kumamoto Earthquake have indi- disaster relief volunteers,” Press news: July 13, 2018, https:// cated that improvements were made since the Great East www.sankei.com/west/news/180713/wst1807130118-n2.html [ac- Japan Earthquake. These improvements may be universal cessed October 30, 2018] and should be monitored in the future. This study’s analysis only focused on the scale of damage. Even with similar damage scales, cities, towns, and villages with smaller populations would likely be more significantly affected and thus require external support. A discussion of this issue not only requires focus on the damage scale, but also the damage rate and a related analysis of its relationship with support needs at that time. Deficiencies, excess support, and volunteer sources (i.e., whether they arrived from within or without the damaged areas) should also be studied. The big data examined in Name: this study were used to achieve an overview of the situa- Shosuke Sato tion. We thus did not examine the abovementioned essen- Affiliation: tial aspects. These issues should therefore be addressed in Associate Professor, International Research In- future studies. stitute of Disaster Science (IRIDeS), Tohoku University

Acknowledgements The data collection performed in this work was supported by Address: Kazumi Igarashi, who is a technician at the International Research 468-1 Aoba Aramaki, Aoba-ku, Sendai, Miyagi 980-0845, Japan Institute of Disaster Science (IRIDeS), Tohoku University. Brief Career: 2011 Ph.D. degree in Informatics, Kyoto University 2009- JSPS Research Fellows (DC2) References: 2011- Assistant Professor, Disaster Control Research Center (DCRC), Graduate School of Engineering, Tohoku University [1] S. Sato, H. Hayashi, K. Inoue, and T. Nishino, “Visualizing Choronological Behavior of Disaster Social Aspect Based on Web 2012- Assistant Professor, International Research Institute of Disaster News Articles on Disasters and Crises – Support of Creating Com- Science (IRIDeS), Tohoku University mon Operational Picture for National/Local Government Ofﬁcers Selected Publications: and Reseachers Related to Disaster Management through the Web • S. Sato et al., “Grasp of Disaster Situation and Support Need inside Publication of Keyword Extraction Results using TRENDREADER Affected Area with Social Sensing – An Analysis of Twitter Data before –,” J. of the Visualization Society, Vol.29, No.7, pp. 17-26, 2009. and after the 2011 Great East Japan Earthquake Disaster Occurring –,” J. [2] S. Sato, F. Imamura, and H. Hayashi, “Basic Analysis of the Web Disaster Res., Vol.11, No.2, pp. 198-206, 2016. News Corpus Broadcasted the 2011 Great East Japan Earthquake • S. Sato et al., “Online Information as Real-Time Big Data About Heavy Disaster,” J. of Social Safety Science, No.15, pp. 303-311, 2011. Rain Disaster and its Limitations: Case Study of Miyagi Prefecture, Japan, [3] S. Sato, H. Hayashi, N. Maki, and M. Inoguchi, “The Development During Typhoons 17 and 18 in 2015,” J. Disaster Res., Vol.12, No.2, of an Algorithm Using the TFIDF/TF Index to Extract Automati- cally the Set of Keywords of Corpus about Fields Related to Emer- pp. 335-346, 2017. gency Management: A Case Study Utilizing Web News Articles for • S. Sato et al., “Text Data Reduction Method to Grasp the Sequence of the 2004 Niigata-Ken-Chuetsu Earthquake Disaster,” J. of Social Disaster Situation: Case Study of Web News Analysis of the 2015 Safety Science, No.8, pp. 367-376, 2006. Typhoons 17 and 18,” J. Disaster Res., Vol.12, No.2, pp. 329-334, 2017. [4] S. Sato, F. Imamura, and M. Iwasaki, “An Analysis of Web Cover- Academic Societies & Scientiﬁc Organizations: age on the 2016 Kumamoto Earthquake Disaster,” J. Disaster Res., • Institute of Social Safety Science (ISSS) Vol.13, No.2, pp. 321-325, 2018. • Japan Society for Natural Disaster Science (JSNDS) [5] S. Sato, F. Imamura, and H. Hayashi, “Analyzing the 2011 Great • Japan Society of Civil Engineers (JSCE) East Japan Earthquake disaster based on Web News,” Analyzing • Japan Society for Disaster Information Studies (JASDIS) the 2011 Great East Japan Earthquake disaster 2 – Disaster Human, • Japan Society for Disaster Recovery and Revitalization (JSDRR) Society and Record –, Akashi Shoten, Part 3, Chapter 7, pp. 235- • 248, 2013. Institute of Electronics, Information and Communication Engineers (IEICE) [6] S. Sato, F. Imamura, and H. Hayashi, “An Analysis of Amount of Human Resource Support from Outside of the Affected Areas and

Journal of Disaster Research Vol.14 No.3, 2019 537 Sato,S.andImamura,F.

Name: Fumihiko Imamura

Afﬁliation: Professor and Director, International Research Institute of Disaster Science (IRIDeS), Tohoku University

Address: 468-1 Aoba Aramaki, Aoba-ku, Sendai, Miyagi 980-0845, Japan Brief Career: 1989 Dr.Eng. degree, Tohoku University 1989-1990 Research Associate, Tohoku University 1993-1995 Associate Professor, The School of Civil Engineering (SCE), Asian Institute of Technology 1995-2000 Associate Professor, Disaster Control Research Center, Tohoku University 1998-2000 Affiliated Faculty, Disaster Prevention Research Institute (DPRI), Kyoto University 2000-2012 Professor, Disaster Control Research Center, Tohoku University 2004-2006 Head of the Disaster Control Research Center 2012-2013 Deputy Director of IRIDeS, Tohoku University 2014-present Director of IRIDeS, Tohoku University Selected Publications: • F. Imamura, A. Suppasri, S. Sato, and K. Yamashita, “The Role of Tsunami Engineering in Building Resilient Communities and Issues to be Improved After the GEJE The 2011 Japan Earthquake and Tsunami: Reconstruction and Restoration,” Insights and Assessment after 5 Years, Springer, ISBN: 978-3-319-58691-5, 2017. • A. Muhari, I. Charvet, F. Tsuyoshi, A. Suppasri, and F. Imamura, “Assessment of tsunami hazards in ports and their impact on marine vessels derived from tsunami models and the observed damage data,” Natural Hazards, Doi: 10.1007/s11069-015-1772-0, 2015. Academic Societies & Scientific Organizations: • Japan Society of Civil Engineers (JSCE) • Japan Society for Natural Disaster Science (JSNDS) • American Geophysical Union (AGU) • Science Council of Japan, the Central Disaster Management Council in Japan, and the study group of the Reconstruction Design Council in Response to the Great East Japan Earthquake, Cabinet Office

538 Journal of Disaster Research Vol.14 No.3, 2019