When disasters strike, media responses serve multiple functions. Media can focus the general public’s attention on a disruptive event, disseminate potentially lifesaving information to those in the disaster area, inform those not in the area of how they can help, provide information to the public on how they can better prepare for a potential similar disaster, and connect disaster outcomes to government policies or private-sector initiatives. However, not all media sources serve the same purpose, nor is every event covered to the same extent. What explains the variation between source coverage and content? How are people both in and out of disaster areas using the variety of news sources, and how does media usage affect society’s social construction of disaster risk?
If we assume that risk is not purely objective but embedded in the social structure of society in which media sources are actors (Stallings, 1990), then it is important to understand how media coverage shapes our perceptions of disaster response. Society’s construction of risk affects disaster perceptions, public policy prescriptions, private sector initiatives, and civil society solutions that contribute to resilience. The relationship between the characteristics of a disaster, resilience responses, and the extent of media coverage in an area of the literature with great potential for exploration. In this project, I incorporate data on multiple recent hurricane and wildfire disasters from traditional media sources like cable news and the New York Times and alternative media sources like Wikipedia and Flickr. The project is ongoing and is in many ways still in the data gathering and description phases, but initial findings show substantial differences between disasters and between media sources.
This project was initially inspired by articles in the Washington Post (Bump, 2018) and the Los Angeles Times (Battaglio, 2018) that use analysis of cable news closed-captioning data to show that cable news (particularly the big three cable channels of CNN, Fox News, and MSNBC) spend much more air time covering hurricanes like Irma and Harvey than wildfires like the Camp Fire in northern California. Bump argues that this is for several reasons. One, hurricanes can be forecasted so cable news has time to prepare and get people into place. Two, hurricanes cover a much larger landmass and as a result, are generally more dangerous for more people. Three, there is a bias towards flashy and familiar places. For example, when comparing wildfires, southern California wildfires in well-known places like Malibu receive more coverage than wildfires in places like Paradise, site of the Camp Fire. Battaglio adds to the analysis, arguing that proximity to wildfire is significantly more dangerous than proximity to hurricanes (and firefighters often strictly limit the movement of reporters), so reporters are much more willing to get close to the disaster areas. The analysis of cable news coverage allows for a comparison between traditional print news coverage like the New York Times and new information platforms like Wikipedia, Flickr, or YouTube.
I gathered data from different media sources using the platforms’ application programmer interfaces (API). Learning the API is a part of the process of engaging with each source, observing how the platforms structure their data differently, and determining how accessible the data is for public usage. The more open and accessible an API is, the easier it is for programmers and researchers to gather data or build applications. Both the private and public sector have APIs to scrape for research data.
From the New York Times, Wikipedia, Flickr, and YouTube APIs, I scraped data for five wildfires and five hurricanes: the Tubbs (2017), Camp (2018), Woolsey (2018), Kincade (2019), and Tick (2019) fires, and hurricanes Harvey (2017), Maria (2017), Irma (2017), Florence (2018), and Dorian (2019). These ten events were chosen because they are all within the last three years and there is a scale of destruction and intensity between them. I gathered the number of Times articles, Wikipedia page edits, and Flickr photos for these disasters. Unfortunately, Google’s YouTube API is very complex, and I have not yet been able to satisfactorily scrape the necessary data.
From an initial analysis of the New York Times data, the trends of cable news coverage are reinforced. In the Times coverage, hurricanes receive many more articles than wildfires, sometimes more than triple the number of articles in a single day (see figure 1 above). Note that for hurricanes, “day 0” is the day the hurricane forms in the Atlantic rather than landfall. Every hurricane has more content about it than any of the wildfires, including the Camp Fire of 2018, the most destructive fire in California history (CalFire, 2019). Hurricane Florence received the most single-day coverage at 31 items, but the least amount of coverage overall with 164 total pieces of coverage. Hurricane Harvey had 295 total pieces of coverage in the first month. Even Hurricane Dorian, which when compared to Irma, Florence, and Harvey in terms of damage and casualties was relatively small, received much more coverage than the average wildfire (182 total pieces of coverage). Table 1 below shows basic descriptive statistics of the hurricanes’ damage, casualties, and meteorological power.
All the hurricanes that affected the mainland United States, including Dorian which primarily damaged the Bahamas, had more coverage in the first 15 days than Hurricane Maria, though coverage of Maria had a long tail, resulting in the third most articles overall. See Table 1 below for descriptive statistics on the hurricanes. Puerto Rico was devastated by the hurricane at cost of approximately $90 billion, 65 direct casualties, and as many as 4,645 indirect deaths (Kishore, et al., 2018), many of which were caused by discontinued health care such as the inability to get medicine or damaged equipment. Hurricane Maria resulted in the second most direct casualties and damage next to Harvey, but only the third most coverage overall at 193 total pieces of content in the first month. Puerto Rico’s coverage gap can be explained by a few factors. Puerto Rico is a Spanish-speaking island off the mainland of the United States that is also not a state. Demographical, geographical, and legal factors combine to create a neglect by separating the island from typical hurricane coverage both physically and socially. Furthermore, many of the deaths caused by Hurricane Maria were indirect deaths in the weeks following the storm because of destroyed infrastructure. Hurricane Maria coverage may have a longer tail not captured in the first 30 days because many of the deaths occurred after the event, though because of the demographical and geographical factors above, I believe that hurricanes Harvey and Irma would still have longer, fatter tails past 30 days.
For the wildfires, the 2018 Camp Fire in Butte County, California received the most Times coverage (98 pieces), which is in line with the number of casualties, structures destroyed, and acres burned. See table 2 below for descriptive statistics on wildfire damage. The Tubbs Fire (2017) in Sonoma and Napa counties—which was the most destructive fire in California history for a year until the Camp Fire started—also received substantial coverage with 60 pieces of coverage. Additionally, the Woolsey Fire (2018) in Los Angeles and Ventura counties received slightly less coverage at 42 pieces. The Tick and Kincade fires (2019) were barely covered by the Times with 8 and 11 pieces of coverage, respectively. In all cases, wildfires were covered much less than hurricanes.
Wikipedia page edits present a similar hurricane bias, but the patterns of page edits present an interesting comparison to the Times coverage. Page edits are not the same thing as publishing an article or posting a video and they should not be directly compared, but Wikipedia is an important part of modern knowledge creation and dissemination. See figure 2 below for a graph of Wikipedia page edits. Wikipedia pages for hurricanes see significantly more edits than wildfires, but for some of the hurricane pages, a large majority of their edits appear before most of the Times coverage is published. For example, most of the edits (81%) on Hurricane Irma’s page in the first 30 days occurred within the first ten days, and 40% of edits were within the first five days. In comparison, 62% of the Times coverage was published within the first ten days and 21% within the first five days. The trend applies to Harvey, Maria, and Florence as well. Hurricane Dorian sees the most Wikipedia page edits and Times coverage within the same window of days seven to twelve.
Wildfire pages on Wikipedia (except for the Camp Fire) have much less overall activity than hurricanes. The Camp Fire received 1,022 edits over its first 30 days, which is lower than every hurricane except for Hurricane Florence, which has 909 edits. The average wildfire received 383 edits, while the average hurricane received 1,605 edits. If you exclude the Camp Fire, the average page received 178 edits, which shows what an outlier the Camp Fire was, likely because of how destructive and deadly the fire was.
One area of future research is to discover the citations being used by Wikipedia collaborators to see where the information that helps build the knowledge on the platform is coming from. Presumably, there would be much more information on hurricanes because of the extensive news coverage on cable news and in newspapers like the Times. However, as seen, many of the edits occurred before the hurricanes were covered by the Times. It is important to note that while there is a lot of editing activity, the edits could be relatively small and not necessarily adding substantial new information. One explanation for the boom in page edits is that breaking events like disasters see more contributors than a typical article which helps drive early collaboration (Keegan, 2013). Another explanation is that Wikipedia users are creating a collective memory of traumatic events (Ferron and Massa, 2011), though in the case of hurricanes, most of the edits occurred before the hurricane made landfall and caused significant damage. An additional future area of research and analysis is to analyze the content of the edits themselves in relation to the users making the edits.
The final media platform I used for this project is the photo collection site Flickr. Photographs on the website distinguish between when the photo was taken versus when the photo was uploaded to the website, and while I scraped data on both photo classifications, for the purpose of this post, I will be discussing when a photo was taken. Flickr is a good glimpse of media engagement during a disaster because it shows two related things that Wikipedia edits and Times coverage do not: people must be in the proximate area of a disaster in order to take a photo of it, and only people who actually use Flickr will be uploading photos to the platform. Therefore, we can get a rough estimate of how many people in different regions use a specific social media like Flickr. That is not to say that the data is perfect, as there may very well be some people who upload hurricane or fire photos from other internet sources, but those photos are likely a small minority.
Hurricanes Irma, Harvey, and Florence have the most photos uploaded to Flickr. Furthermore, the photos are generally taken in the window of hurricane landfall. See figure 3 below for a graph of Flickr photos taken. Hurricanes Maria and Dorian have many fewer photos on the platform. It may be that people in Puerto Rico do not use Flickr to the same degree as people in Florida or Texas. Hurricane Dorian was less destructive to the United States than the other hurricanes on the list, so people may not have had as many opportunities to take photos of the storm or damage. In terms of wildfires, the Woolsey Fire has the most photos with 2,957, though the highly destructive Camp Fire is close with 2,119 photos. The Tubbs, Tick, and Kincade fires only have 1,381 photos combined, 732 of which are from the Tubbs Fire and 444 from the Kincade Fire. Wildfire photos therefore do not clearly correlate with either destructiveness or length of the fire. A flaw with the Flickr data in its current state is that it is not normalized by the number of people affected by the disaster. In other words, Hurricane Harvey was a massive Texas hurricane that affected a great many people, so there may be more Harvey photos on Flickr simply because it affected more people. A similar phenomenon applies to the wildfire photos, where the Woolsey Fire may have affected more people than the Tubbs Fire because more people live in the Los Angeles and Ventura County areas than in the Napa and Sonoma County areas.
The way the media covers disasters shapes how people perceive danger and consider risks. Furthermore, how people are engaging with media and actively receiving information shapes how people think about disasters and make potentially life-changing decisions like evacuation. From this initial analysis, it is difficult to determine exactly what the factors are in disaster coverage across media platforms, but what is clear is that there is much more media activity for hurricanes than for wildfires. The differences between the platforms on the same disaster are intriguing because it does speak to the different purposes of the platforms, methods of public engagement, and types of information dissemination. The news media of primarily print platforms like the New York Times, imitates cable news even though it is not as visual and does not require video content. Flickr is a platform for self-expression, proximity is important for taking a photo, and people generally take photos around the time of hurricane landfall or at fire ignition. Finally, on Wikipedia, a large portion of activity occurs during the ignition and spread phase of a wildfire, and the formation phase of a hurricane.
Cable news, newspapers, Wikipedia, and Flickr are just some of the modern platforms where information is disseminated, and in future research I would like to include more platforms like YouTube or Twitter. Much of this project was spent learning the APIs of various platforms, and now that my code is written for platforms, the data scraping process can be scaled up and including dozens of disasters to really get a sense of larger patterns of media engagement. Furthermore, I would like to do more analysis of the content of the Times articles, YouTube videos, and Wikipedia “talk” pages, to get a better sense of the dialogue around the disasters.
Battaglio, Stephen. “Why California Fires Didn’t Dominate Cable News Like Hurricanes Did.” Los Angeles Times, 21 Nov. 2018, https://www.latimes.com/business/hollywood/la-fi-ct-fire-cable-news-coverage-20181121-story.html [Accessed October 14, 2019]
Blake, Eric S., and David A. Zelinsky. Hurricane Harvey. NOAA National Hurricane Center, 9 May 2018. https://www.nhc.noaa.gov/data/tcr/AL092017_Harvey.pdf.
Bump, Phillip. “Cable News Networks Spend Far More Time Talking About Hurricanes Than Wildfires.” Washington Post, 19 Nov. 2018. https://www.washingtonpost.com/politics/2018/11/19/cable-news-networks-spend-far-more-time-talking-about-hurricanes-than-wildfires/ [Accessed October 14, 2019]
CalFire. Top 20 Most Destructive California Wildfires. 8 Aug. 2019, https://www.fire.ca.gov/media/5511/top20_destruction.pdf. [Accessed January 8, 2020]
CalFire. CalFire Tubbs Fire (Central LNU Complex) General Information. https://www.fire.ca.gov/incidents/2017/10/8/tubbs-fire-central-lnu-complex/. [Accessed 8 Mar. 2018]
CalFire. Camp Fire General Information. https://www.fire.ca.gov/incidents/2018/11/8/camp-fire/. [Accessed 22 Jan. 2020]
CalFire. Kincade Fire General Information. https://www.fire.ca.gov/incidents/2019/10/23/kincade-fire/. [Accessed 22 Jan. 2020]
CalFire. Tick Fire General Information. https://www.fire.ca.gov/incidents/2019/10/24/tick-fire/. [Accessed 22 Jan. 2020]
CalFire. Woolsey Fire General Information. https://www.fire.ca.gov/incidents/2018/11/8/woolsey-fire/. [Accessed 22 Jan. 2020]
Cangialosi, Andrew, et al. Hurricane Irma. NOAA National Hurricane Center, 30 June 2018, https://www.nhc.noaa.gov/data/tcr/AL112017_Irma.pdf.
Ferron, Michela, and Paolo Massa. “WikiRevolutions: Wikipedia as a Lens for Studying the Real-Time Formation of Collective Memories of Revolutions.” International Journal of Communication, vol. 5, 2011.
Keegan, Brian, et al. “Hot Off the Wiki: Structures and Dynamics of Wikipedia’s Coverage of Breaking News Events.” American Behavioral Scientist, vol. 57, no. 5, May 2013, pp. 595–622.
Kishore, Nishant, et al. “Mortality in Puerto Rico After Hurricane Maria.” New England Journal of Medicine, vol. 379, no. 2, July 2018, pp. 162–70.
NOAA National Weather Service. Hurricane Dorian. NOAA’s National Weather Service, 6 Sept. 2019. https://www.weather.gov/mhx/Dorian2019.
Pasch, Richard J., et al. Hurricane Maria. NOAA National Hurricane Center, 14 Feb. 2019, https://www.nhc.noaa.gov/data/tcr/AL152017_Maria.pdf.
Stallings, Robert A. “Media Discourse and the Social Construction of Risk.” Social Problems, vol. 37, no. 1, Feb. 1990, pp. 80–95.