Data journalism in disaster zones

How to report in emergencies

305, 8200, 922. At first sight, these numbers appear meaningless. But if we convert them to kilometres per hour, depth in metres, and barometric pressure, they represent the start of three major natural disasters. The speed of winds when Typhoon Haiyan hit the Filipino province of Eastern Samar. The depth of Nepal’s April 2015 Earthquake. And the intensity of Hurricane Andrew, when it made landfall in Florida 23 years prior.

Looking at these numbers another way, they also represent the most powerful storm to ever strike land, the worst natural disaster in Nepal’s recent history, and the United States’ sixth most intense hurricane.

Superlatives aside, the impact of these disasters raise significant challenges for all reporters -- data or not. In a rapidly changing reporting environment, how do you uphold journalistic standards of truth and accuracy? How do you find and verify data? What are the risks and best practices of data reporting, especially when your audience includes victims? And, perhaps even more crucially, how do you physically get through the weather to report?

To find out, we spoke to a global group of seasoned data journalists, all tasked with reporting in the midst of disaster:

  • Steve Doig: Professor at Arizona State University’s Cronkite School of Journalism, sharing a Pulitzer prize for reportage on Hurricane Andrew at the Miami Herald in 1993.
  • Arun Karki: Executive Director and Founder of the Center for Data Journalism Nepal, who reported on the April 2015 Nepal Earthquake.
  • John Maines: Database Editor for the Sun Sentinel in Florida, where he’s reported on the area’s various hurricanes and the 1996 Valujet airplane crash.
  • Joshua Mutisya: Data Journalist at Kenya’s Nation Newsplex, which has experience reporting on the nation’s droughts and food insecurity emergencies.
  • Cristen Tilley: Senior Journalist at the ABC News Interactive Digital Storytelling Team in Queensland, Australia -- an area frequented by floods and cyclones.
  • Norman Zafra: Journalist and documentary maker, who brings experience from multimedia reporting on the Typhoon Haiyan disaster in the Philippines.
Hurricane 63005 1280

How is reporting on a disaster different from other data stories?

Steve Doig, Hurricane Andrew: A key difference for a local reporter is that you and your family may be victims of the disaster yourselves. This means the added pressure of trying to take care of your own situation while at the same time doing your job of reporting on the disaster. In my case, Hurricane Andrew destroyed my house, so I and my wife and our two children had to live in a trailer in our driveway during the months of making the home livable again.

Cristen Tilley, cyclones and floods: If you're in the disaster, access to the basics like internet, electricity, food, and getting staff to work is hard enough. You need to prioritise those. When we were flooded out of our newsroom in the 2011 Brisbane floods, we had people doing shifts from their lounge rooms for a day or two while a makeshift newsroom was set up. Once we had the makeshift newsroom going, we had a core team to take care of the main news coverage, we deployed reporters to hotspots and then a smaller team worked on special coverage/data. If you're trying to work with data but not actually affected by the disaster, then it's all about finding reliable data in time.

5348468813 2523e1d91e z

Flooding of a Brisbane motorway during the floods illustrates the practical challenges facing reporters during a disaster. Credit: Martin Howard, CC BY 2.0.

Joshua Mutisya, Kenyan emergencies: Unlike other data stories, disaster stories are largely marred by a lack of reliable data to use. The authorities are also reluctant to provide information. Once a disaster occurs, little focus is given to the data-angle of the story as newsrooms are more interested in getting the right multimedia product to share with their audience. However, data journalism is different. It takes time to do a thorough analysis. Therefore, a worthy piece could be successfully completed days after the disaster has occurred.

Can you tell us more about the challenges around data reliability?

Norman Zafra, Typhoon Haiyan: The spread of unreliable information is a persistent feature of disaster events, especially when everyone is in a state of shock. Although natural disasters, such as typhoons, are often characterised by high predictability, there remains an absence of official data, given the enormity of the tasks shouldered by local officials. To overcome this, news organisations in the Philippines employ a dedicated research team that helps journalists coordinate, collect, gather, and verify information available from different sources. Access to this data is via an intranet.

Steve Doig, Hurricane Andrew: Anecdotal information after a disaster certainly can be unreliable -- witness the doctored ‘shark in the subway’ images from recent hurricanes. But much of the data that we would want to use for analysis -- damage surveys, property tax rolls, utility outages, and so on -- comes from official sources. The problem is that it can take weeks for authorities to gather a detailed house-by-house assessment of damage across a widespread disaster area. After Hurricane Andrew, the first damage assessment lists we got were from the Red Cross, but they were useless for analysis -- impressionistic and very general (stuff like ‘lots of roof damage along 125th Street’). But county inspectors began building a true house-level database with addresses and percentages of damage to roofs, windows, and other structural elements. That database, which came to us in pieces over several weeks, was crucial for our analysis.

Joshua Mutisya, Kenyan emergencies: The greatest challenge is the lack of reliable data. This could be largely attributed to over-reliance on singular sources for information, such as government agencies. To counter this, data journalists should have alternative ways of acquiring data from the ground. This implies media houses also investing in localised data collectors who can provide credible information in time.

What are some other ways that journalists can address unreliable and deficient data?

Cristen Tilley, cyclones and floods: For our coverage of Cyclone Debbie, we found a source of real-time wind speeds and visualised the current wind speeds in the cyclone zone. This was data from the weather bureau so knew it could be trusted. This is different from trying to provide information on a fast-moving, unpredictable bushfire. So the way you account for the risk involved in doing data work is to think about how it could affect the audience/reader, and don't go near coverage that could have adverse effects. One of the other problems we encountered with the cyclone coverage was that we didn't have time to scrape the data, so ended up updating by hand. This was time consuming but we didn't really have a choice.


The ABC’s visualisation of Cyclone Debbie’s wind speeds, using reliable information from the Australian Bureau of Meteorology.

Joshua Mutisya, Kenyan emergencies: Back in 2016 and parts of 2017, Newsplex carried out analysis about drought that had struck the country. To get a clearer picture, we used data from the National Drought Management Authority, which showed that warnings of an impending drought had been given way before. This provided a larger angle to the story instead of just focusing on figures from government about the estimated number of people affected.

Arun Karki, Nepal Earthquake: You can verify or fact-check the data to some extent using some online verification tools. Many hoaxes buzzed around on social media during April 2015 Nepal earthquake, and we were receiving so much information with discrepancies about the deaths and damages from local and individual sources. So we only quoted official sources -- they were slower, but credible.

In addition to data reliability, do disaster scenarios raise any other ethical concerns for data journalists? And how can these be addressed?

Arun Karki, Nepal Earthquake: Personal identifiers in data tables are very sensitive information -- and may pose risks to individuals or specific groups/communities if exposed or published. Two years after Nepal 2015 Earthquake, the Central Bureau of Statistics (CBS) of Nepal published a comprehensive dataset of housing damage, but to protect privacy, they didn’t share some private information of beneficiaries like the ‘geo-location tag’ of each household. Such sensitive information might be even more risky to women and other vulnerable communities.

‘Not fact-checked’ data could also be traumatising to audiences. However, after some time (a few hours or days -- it depends), online or social media crowdsourcing could be a good way to start verifying data, when authentic or real-time datasets are not available from trustworthy sources.

John Maines, Floridia disasters: Just be precise as possible, particularly in live television and radio newscasts. If you report, or interview someone who says “the whole town is leveled” or “the entire city is underwater”, is that really true? Better to report that the water has reached an XX feet depth in YY neighborhood, and is expected to continue rising until it reaches ZZ street. Audiences will still be traumatised, there is no way around that, but details are better than sweeping, non-specific statements. If you don’t know details, just say you’re working hard to find them.

Despite these challenges, how can data be used to improve the way that journalists report on disasters?

John Maines, Floridia disasters: You can use data to map out place that were not impacted by a disaster. We did this years ago. People in our poorer neighborhoods learned that they could get money from our Federal Emergency Management Agency (FEMA) by falsely reporting damage they did not have. People would remove stuff from their homes, spray it with a garden hose, and call FEMA. Hundreds of Millions of dollars in fraud, that taxpayers paid. We were a Pulitzer finalist for those stories, which we reported from several cities around the United States.


Sometimes disaster affected areas aren’t the story.

Steve Doig, Hurricane Andrew: Data can be used to track recovery, too. We used data from the municipal water/sewer utility to get dates when repaired homes were being reoccupied, and used that to do timeline maps showing where recovery was -- and wasn’t -- occuring.

Joshua Mutisya, Kenyan emergencies: One, in the wake of a disaster, journalists get quite emotive, and sometimes end up providing information that is exaggerated. Numbers don’t lie, and so with dependable sources, as data journalists, we don’t fail the audience by providing non-factual information. Two, using data we are able to provide different angles to a story. In the wake of the famine that hit Kenya in 2017, Newsplex covered different aspects; from the overall estimates of people affected, to the worst and least hit counties, to the invasion of destructive army worms in Kenyan farms, which threatened to worsen the situation.

Norman Zafra, Typhoon Haiyan: Data visualisation is a powerful form of contextual reporting and is able to bind together pieces of information collected from various sources. To improve the way journalists report on disasters, a centralised database provided by key government agencies and official sources is always useful. After Haiyan, for example, a database of aid funding was made available to the public. It was a rich source of information that allowed opportunities for journalists to interpret and innovate the presentation of complex figures.

2844108945 97f72085ec o 1

Steve Doig’s groundbreaking maps in the Miami Herald.

On data visualisation, what are the best ways to use it in disaster reporting?

Steve Doig, Hurricane Andrew: Maps. Our Andrew coverage was among the first to use GIS mapping to overlay the path of the hurricane winds onto a grid shaded by the percentage of homes that were damaged. But today’s GIS tools and property shapefiles make for fantastically better resolution of such maps, down to the house level, than was possible 25 years ago.

I’ll also recommend good aerial imagery. We acquired hundreds of detailed color pictures taken from an altitude of about 600 feet as part of a systematic damage inventory. The cover shot of our major damage analysis report was a single image showing three neighboring subdivisions with wildly differing levels of damage, attributable to varying construction standards. Another good use for a comprehensive collection of pre-disaster imagery is to create before-and-after slide shows.

2844893922 09b3642084 o 1

The report’s covershort, illustrating varying levels of damage between houses at the top, which were the most devastated, and those that survived towards the bottom.

Norman Zafra, Typhoon Haiyan: Make the data presentation sexy -- perhaps look for patterns, visually highlight the most relevant message and convey meaning that informs rather than complicates. Graphic designers should be part of the team too since designing is usually outside the skill set of most journalists. My suggestion is to tap the interactivity of web and mobile platforms to allow the audience to interact with data rather than treating the web as a repository of static information graphics. An interactive map is useful especially in the Philippines wherein the extent of devastation can only be realised when presented in a map. Personally, I find disaster data reporting extremely challenging -- too much complexity in the presentation can turn off the audience but oversimplification on the other hand can also be misleading.


A timeline interactive, produced by Norman to illustrate the extent of Typhoon Haiyan’s damage in the Philippines.

Since it is difficult to predict when a disaster might strike, what advice do you have for data journalists to prepare themselves?

John Maines, Floridia disasters: You need a storm/disaster kit. Don’t care if you are a data person. Don’t try to put it together when disaster is about to strike, that is stupid. I know that is what we are told even as regular people, but it is true. You need flashlights, a backup charger for your cell, water. Canned food. It is amazing to me how many people go to the supermarket in the hours before a storm hits. Just for water. Go online, order a few of those five gallon foldable plastic things campers use. Then stow them. Before the storm hits, fill them up. No trip to the market! Also, think about things you don’t normally think about. Like mosquito repellent. When Valujet crashed in 1996, I spent the night in the Florida Everglades getting eaten alive by mosquitos. Thank god in the middle of the night the American Red Cross came by with water and repellant. God bless them.

Also, news organisations tend to let batteries die on laptops, because they are plugged in most of the time. Check yours.

Steve Doig, Hurricane Andrew: As for data, gather what you can before disaster hits. Examples would include the property tax roll, which should have detail about ownership and value and age and type of construction for every structure in your area. Also, maybe a detailed historical database of the kinds of disasters which might hit your area, whether they’re hurricanes or earthquakes or wildfires or whatever. A couple of years before Andrew, I did a full-page graphic showing the history of hurricanes in Florida; immediately after Andrew, we dusted it off, added Andrew’s track and republished it.

Norman Zafra, Typhoon Haiyan: In the case of typhoon disasters, journalists must learn from previous data reportage to assess what worked and what didn’t work for the public’s understanding of the news. Evaluation is necessary. To prepare for it, there should be a way to routinise disaster data reporting (e.g. devising a design or content template for emergency data reporting, or a repository of previous data reports). For instance, we can learn from the principle of a ‘dark website’ in crisis communication. A dark website is a hidden (template) page that is activated only when there is a crisis. It’s prepared in case a crisis occurs without warning.

Cristen Tilley, cyclones and flood: We try to think about coverage before the storm season hits. Look at advances or developments in our technology, as well as in the data available, and see if there's anything that can be done differently. It's helpful to have a few go-to sources of reliable data if an unexpected disaster happens too, such as the weather bureau or other emergency authorities. Also, look back at your previous coverage and see if you can tweak anything so you're more prepared. For example, after the experience of updating the wind speed tracker by hand, I wrote a script to get that data automatically next time.

Hurricane 63005 1280

Any final thoughts?

John Maines, Florida disasters: Pace yourself. As reporters, our first instinct when disaster hits is to go to the newsroom. We learned years ago that this was a mistake. In fact, our editor in 1992 when Hurricane Andrew hit said the worst mistake he ever made is insisting that everyone come to the newsroom before the storm, and sit it out there. So, we had a bunch of people in the newsroom with no power, worried about their loved ones and if they survived. Not good. And they all got tired at the same time.

What we do now is have a red team and a blue team. The red team sits out the storm in the newsroom, arriving a day before it hits and bringing sleeping bags. The blue team, which I am part of because data work is delayed, comes in after the storm is gone.

Steve Doig, Hurricane Andrew: In a disaster, your first duty is to your family. After Andrew, it was a couple of days before I could even think about going into the newsroom. I had to make my heavily damaged house into something of a shelter for my wife and young children, including getting the toilets to operate with buckets of water and making sure exposed wiring would be safely out of the way when the electricity came back a couple of weeks later.

Arun Karki, Nepal Earthquake: Don’t hurry. Don’t break headlines unless the data is verified. It’s also good practice to keep all of your datasets (whatever comes in after the incident and from every last source) to use later for your follow-up reporting. Because every dataset could be useful for your future stories. For example, detailed datasets of house damage and housing grant beneficiaries were initially made public by local authorities in Nepal after 2015 earthquake. However, those granular or disintegrated datasets (very helpful to compare/correlate during reconstruction) could not be retrieved or accessed later. So, keep mining for data. Another tip: reach out to different government levels and sources. If data is not available at the national level, it could be at the sub-national, local, or periphery level. Moreover, if you’re not familiar with advanced data skills then find and collaborate with tech-savvy individuals or groups who can help you, from data gathering to visualisation.

Verificationhandbook 1
subscribe figure