When a severe earthquake struck Haiti in 2010, the world reacted swiftly. Supplies poured in to the country’s capital, Port-au-Prince. The problem? That’s not where the victims were. The population of Port-au-Prince fell by nearly 20% in the weeks after the quake struck.
Large numbers of displaced people are common in the aftermath of natural disasters. More than a million people were forced from their homes after the Haiti quake. To effectively deliver aid, international agencies need better information about their movements.
Since 2010, a Swedish operation has taken the innovative step of using mobile phone data to track the movement of people after a crisis. The data has the potential to transform the way aid groups respond to disaster — if they can figure out how to use it.
When a major aid operation starts, relief groups use a “needs assessment survey” to make sure the right kinds of supplies are brought to the country.
Those surveys need to be based on a representative sample of the people affected — and it’s difficult to make a sample representative without knowing how the population is distributed.
The trouble is that monitoring people’s movements in a post-crisis scenario is very difficult, according to Linus Bengtsson, executive director of Flowminder, the Swedish non-profit trying to help tackle this challenge.
“You might see lots of people on a road moving in a certain direction,” Bengtsson said. But that leaves many questions: where they end up, how many there are, how many people are moving in the opposite direction. “To have people placed out monitoring and compiling that — it’s absolutely impossible to do at scale.”
The main sources of information come from registration of people at international borders and camps. But most displaced people don’t cross borders or end up in camps, instead staying with family or friends or living in informal settlements.
Mind the flow
This problem, of unrepresentative and imprecise data, is what Flowminder aims to address. A one-off project in Haiti in 2010, when Bengtsson was a doctoral student in Stockholm, subsequently became a non-profit aiming to repeat its work in disaster zones across the world.
The core idea is simple: the movement of individual mobile phones can be tracked using cell towers. Comparing data from a post-crisis period with earlier records allows a detailed estimate of the movements caused by the disaster.
There are issues with this method. Some people are less likely to have phones — children, the elderly and the poor — which may make data from phone movements unrepresentative.
Tracking displacement is also challenging, according to Bengtsson, because a disaster can reduce movement as well as induce it — someone might normally move between two towns for business every day, but stop during a crisis.
Nonetheless, Flowminder’s estimates have proved impressively accurate. In Haiti, a UN-run survey seven months after the earthquake asked people if and where they had moved. The data Flowminder researchers had produced and shared with aid agencies closely matched the survey results.
Information from the government,distributed by the UN in the midst of the crisis, on the other hand, was “completely off”, Bengtsson said.
More reliable information makes it quicker and easier to find people in need and get them help. It can also help track when people are moving back to their homes — and when they aren’t, which can prompt closer investigation.
“They get a completely new picture of the situation,” Bengtsson said. Since 2010, Flowminder has worked with agencies including the World Food Programme and International Organization for Migration, supporting disaster response in Bangladesh, Nepal, and again in Haiti after the 2016 hurricane.
Beyond immediate operational help, the detailed information Flowminder produces can be used for policy development — in particular, preparing for future natural disasters.
Where urgent assistance should be directed can differ significantly depending on what time of the day or week a disaster hits. A busy business district might be empty at night but full at midday, for example, and mobile phone data can help identify those patterns. That dynamic population information, if prepared in advance, can provide a rapid guide to that question.
Patterns in older data, Bengtsson said, can also give a quite robust indication of where people are likely to go if disaster strikes. “People who had left Port-au-Prince at Christmas before the earthquake in Haiti very frequently went back to the same place,” he explained. “It makes a lot of sense — you don’t go into the forest and hide under a rock, you go to places where you have social support and family connections.”
Data uses and abuses
Flowminder’s work can’t rapidly be set up in entirely new contexts. That’s because accessing location data from mobile phone operators requires complex legal agreements to safeguard privacy concerns, which can take a long time to negotiate. “The largest part of the work is actually all these other considerations,” Bengtsson said.
Once Flowminder’s work is up and running, there’s one more challenge: getting aid agencies to take advantage of the data. “It’s really something that they’re not used to at all,” Bengtsson said. Relief workers operating under intense stress struggle to understand how they can integrate this new information source into their established ways of working.
That’s something Bengtsson and his colleagues are trying to change. “You need to have seen this data before, so you’re not seeing it for the first time when the disaster strikes,” he said. If it can achieve that level of familiarity, mobile phone data could revolutionise disaster response across the world. — Fergus Peace
(Picture credit: Flickr/UN Photo)