Is it possible to predict forced displacement?
When rain stops falling on some of the conflict-ridden regions of the Horn of Africa and the drought-stricken land becomes too barren to bear, people move.
They walk for days in search of water, in search of food, in search of new places to settle. But in a country like Somalia, where ethnic differences divide the territory and rival groups fight to control it, moving tends to increase violent conflict.
Escalated violence further exacerbates forced displacement, producing a cyclical phenomenon that leaves on-the-ground teams grappling with challenges that had been unforeseeable until now.
For the past year, UNHCR’s Innovation Service has been working to understand the intrinsic relationship between climate change, violent conflict, and forced displacement. Using supervised machine learning, they designed Jetson, an engine that is fed data and uses trained models to predict the displacement of persons in Somalia.
Jetson gives UNHCR and other humanitarian organizations the potential to become more proactive in their response efforts — a transformation that could significantly improve on-the-ground relief services and more importantly the lives of those who are forcibly displaced.
A fruitful request
Somalia has too often found itself at the volatile intersection of climate change, violent conflict and displacement.
Such was the case in 2011 when the country experienced, what researchers called, the worst famine in 25 years. In just six months, relentless violence compounded by severe drought forced more than 140,000 Somalis to cross the borders into neighboring Ethiopia, Kenya and Djibouti.
Dry trees and animal carcasses dotted the landscape along their route, but they kept their eyes on the horizon, waiting for the moment UNHCR tents would speckle their view. Many Somalis did not reach the camps on the other side of the border; their lives claimed by severe malnutrition and dehydration. Those who did often arrived in fragile health conditions.
During the peak of the crisis, the region of Dollo Ado in southeastern Ethiopia received some of the highest numbers of refugees with about 2,000 arrivals per day.
“Picture a concert of people arriving every day,” said Rebeca Moreno Jimenez, Data Scientist at UNHCR’s Innovation Service. “That is overwhelming for our teams on the ground.”
The Dollo Ado team was not fully equipped to assist such a large and unexpected influx of people. Their need for nutritional screening and protection support quickly depleted the resources they had been allocated for the year.
The unprecedented nature of these events left its mark on the team. They remember the bottlenecks in the registration system, the overcrowded camps, and the malnourished babies whose bones could be broken by a slight touch.
These memories have since turned into lessons for humanitarian preparedness.
In 2017, failed rains left Somalis teetering on the brink of famine. The land had once again become dry and unforgiving, but this time it yielded a fruitful request.
Fearful that displacement numbers would be as high as they were in 2011 and knowing that preparation would be essential to their response efforts, the team sent a request to the Innovation Service. They ventured to ask if it was possible to predict the number of arrivals to the region.
“That was a challenge question for me,” Moreno Jimenez said, “but it was the question that launched Project Jetson.”
Powered by open data
UNHCR began experimenting with predictive analytics in 2015 with the establishment of the Winter Cell — an interdisciplinary team tasked with addressing the challenges associated with the scale of population flows from Turkey through to Europe during an uncharacteristically cold winter.
To fulfill its mission, the team sought out value-based partnerships, including the World Meteorological Organization, the Met Office in the UK, academia, and other UN institutions. The collaborative partnership model granted them access to data, resources, and expertise that was not available in-house.
With the collaboration of these organizations in full swing and with the support of UN Global Pulse, a flagship innovation initiative of the UN’s Secretary-General on big data, the team was able to build a model that used meteorological data to predict population flow into Greece.
The humanitarian sector traditionally deals with the consequences of displacement, not with its specific origins. The Winter Cell’s work marked a shift in this dynamic by revealing the potential impact that proactive, data-driven decision-making could have on forcibly displaced individuals.
Although the Winter Cell was dismantled in June of 2016, the insights and partnerships gained during its short existence prepared the UNHCR’s Innovation team to respond to Dollo Ado’s request. Not only does Project Jetson build on the Winter Cell’s predictive model, but it also uses open data sharing and innovative approaches to achieve its objectives.
Forced displacement is a complex phenomenon. Mathematical evidence shows that there are certain factors, such as food insecurity and violent conflict, as well as the appeal of the destination region, in terms of prices, distance, and stability, that cause people to move. Certain combinations of these factors are more likely to make people move than others.
“If we can measure some the factors that make them flee, then we can predict their movements,” Moreno Jimenez said.
Jetson is thus an experiment to discover, understand, and measure the specific factors that cause or indicate the forced displacement of Somalis.
To find those critical variables, the team flew to Ethiopia and conducted interviews with UNHCR’ staff in Dollo Ado and refugees themselves. After months of desk and field research, ten variables including basic commodities market prices, rainfall, incidents of violent conflict, and historical population movement were identified as indicators of Somali displacement.
The Innovation Service team developed strategic partnerships with fourteen organizations to source seven year’s worth of data for the experiment. After processing the anonymized data using machine learning, they built an algorithm that recognizes the combination of factors that exacerbate displacement and makes predictions about population flow.
“Jetson has shown us that we can be doing so much more with data, especially when it’s openly shared,” she said.
The project is a testament to what the humanitarian sector can accomplish when it works together.
On the ground
Currently, Jetson can predict the displacement of persons in Somalia a month in advance, a breakthrough that can improve the efficiency and effectiveness of UNHCR’s services.
With one month’s notice, the Somalia team, including the protection cluster, and other relevant operational partners, could adjust resource allocation, better coordinate emergency preparedness and solicit support from other UN agencies or NGOs in neighboring regions. The insights gained from the project are also conducive to more informed scenario planning.
“It’s sometimes unbelievable to think that someone is actually using this theoretical craziness on the ground,” Moreno Jimenez said.
In recent months, she has been working with Martin Stobbs, Information Management Officer, and Andrea Bruhn, Durable Solutions Officer for Somalia, to crosscheck Jetson’s predictions with the actual number of arrivals on the ground. This data is verified by John Waweru, a Data Management Associate based in Nairobi, who sends the actual numbers of arrivals every month. The registration team in Dollo Ado, which is composed of seven people, also cross-checks the arrivals of refugees into the region.
As of June 2018, predictions were accurate for eleven of the country’s eighteen regions.
Looking to the future
The Innovation Service team hopes that Jetson will become a standard decision-making tool at UNHCR.
However, all refugee crises are different. The factors that indicate the forced displacement of Somalis will not necessarily indicate that of the Rohingya. In order to scale Project Jetson to other countries, a very detailed level of research is required to identify the displacement indicators specific to that region.
In the meantime, the Innovation Service team hopes to publish the project’s algorithms byproducts for public use and will encourage others to improve and adapt the technology to fit their needs. Jetson is a bright spot that the team hopes others in the humanitarian sector will look to that demonstrates the power of data sharing.
For projects like Jetson to continue to evolve, humanitarian organizations will have to create more safe spaces for experimentation and improve how the sector collects, uses, and shares data. If the team behind Jetson can begin to change these behaviors and mindsets, the proactive, data-driven humanitarian system they imagine, can be realized.
“Trying to predict the future is not impossible,” Moreno Jimenez said. “It can be done, and this is a new way of doing it.”
We’re always looking for great stories, ideas, and opinions on innovations that are led by or create impact for refugees. If you have one to share with us send us an email at innovation@unhcr.org.
If you’d like to repost this article on your website, please see our reposting policy.
