Can machine learning and satellite imagery help improve humanitarian aid to Venezuelan migrants?
Forty-year-old Monica Acosta is a Venezuelan mother of three. She and her family live in a makeshift shelter along the highway in northern Colombia, close to the Venezuelan border. Constructed from scavenged wood and plastic sheets, Monica’s home offers scant shelter from the heat and rain. It also lacks electricity, running water, and toilets.
It’s been about a year since Monica and her family fled their home in Venezuela to settle in Colombia. Back home, Monica once made a modest living as a manicurist and hairstylist. Today, she survives in her new home by selling eggs, water, or collecting discarded plastic and aluminum to sell to recycling shops for a few dollars a day. Without passports or any other form of identification documents, neither Monica nor her children can legally work or attend school in Colombia.
Monica’s situation is common in her community. From just a few households in early 2018, has grown to 130 families as of May 2019. Of these families, 105 or 77% are Venezuelan migrants who recently arrived in Colombia. They arrived with no jobs, no money, and nowhere to go, they built new homes from whatever scrap materials they could find – cardboard boxes, discarded wood, plastic bags, blocks of caked dirt, even used car interiors. Each house accommodates multiple families. Without access to running water or sanitation, families spend as much as 20% of their meager earnings buying water and using the nearby garbage-strewn grass as a communal toilet.
Despite harsh conditions that await them in their new country, new migrant families continue to arrive in Colombia every day. "At least here, we can eat," says Monica. Grateful to have escaped worse conditions in Venezuela, Monica and her neighbors chose an apt name for their new home: Bendición de Dios. God’s Blessing.
The residents of Bendicion de Dios are just a fraction of the 3 million Venezuelans who have fled their country in recent years.
Imagine if all the money you had in the world – in the bank, hidden under your mattress, even the coins in your piggy bank – were one day worthless. This is what’s happening to people in Venezuela. Years of economic mismanagement has sent the country's economy into a downward spiral, which has accelerated in the past two or three years. In 2018, the country’s inflation rate was a staggering 80,000%. It’s estimated to reach as high as 10,000,000% in 2019. As of January 2019, the average Venezuelan family’s monthly income was barely enough to buy a pack of diapers.
Hunger, homelessness, power outages, disease, and lawlessness have forced people like Monica to flee their crumbling country. Neighboring Colombia has received the most migrants by far – some 1.2 million people to date, and counting. The influx has left government and humanitarian forces scrambling to provide these people with food, shelter, and livelihood.
To learn more about the Venezuelan crisis, check out these helpful resources:
- The Guardian: Why is Venezuela in crisis? (Video)
- Vox Media: The collapse of Venezuela, explained (Video)
- Vox Media: Why Colombia has taken in 1 million Venezuelan Migrants (Video)
- New York Times: Venezuela’s Collapse is the Worst Outside of War in Decades (Article)
- Bloomberg: Venezuela’s Cafe Con Leche Index - Tracking Hyperinflation One Cup of Coffee at a Time (Data Visualization)
- iMMAP: Tracking Venezuelan migration using Facebook marketing data
On a recent trip to Colombia for a statistics conference, our machine learning team was invited by the international non-profit iMMAP to learn more about the Venezuelan migrant situation. Specializing in data management for humanitarian situations, iMMAP gathers data about Venezuelan migrants in Colombia to guide organizations like UNICEF and World Food Program in deploying aid. iMMAP and UNICEF accompanied our team on a field mission in northern Colombia, where we visited three informal settlements of Venezuelan migrants.
Organizations like UNICEF and World Food Program are working to assist migrants in transit, by setting up water and sanitation facilities, emergency education camps, and community kitchens near the border. But aid efforts have been hampered by the lack of timely and reliable data about migrants in need, where they are, and the nature and size of their needs. “When this is lacking, it's very typical to have misspent resources, duplication of efforts, and a lack of reach to those who are truly most in need,” says Jeff Villaveces, the country head of iMMAP Colombia.
Currently, iMMAP supports humanitarian interventions by doing field surveys of settlements like Bendicion de Dios. But these surveys require a lot of manpower and time, and have limited coverage. The Colombian-Venezuelan border is over 2000 kilometers long, even longer than the US-Mexico border. With only seven formal immigration checkpoints along this border, countless migrants enter Colombia through unmapped roads or over mountain paths. This is why up to now, it is extremely difficult to keep track of new migrant settlements.
The residents of Bendicion de Dios said that they have not received any kind of assistance since arriving in their community. "There are entire communities of hundreds and even thousands of Venezuelan migrants who have never received so much as the smallest assistance. This is truly shocking and is in part due to the massive scale of this migration from Venezuela," says Jeff of iMMAP.
To help address the issue, our team at Thinking Machines is working with iMMAP to use satellite imagery to more rapidly detect migrant settlements along the Colombian-Venezuelan border. These communities are growing so rapidly and across such a vast area that it's impossible for field teams to locate them all. There are very likely many more informal settlements like Bendicion de Dios that humanitarian aid agencies don't yet know about.
Timelapse images from Google Earth Pro show how rapidly some of these communities have grown in the past three years.
In partnership with iMMAP, we plan to train a machine learning model to identify similar communities elsewhere in Colombia. If we can develop an accurate model, we can run it every quarter to produce quarterly maps of possible informal settlements across urban centers and areas along the Colombian-Venezuelan border. Organizations like iMMAP can then send field personnel to validate newly detected settlements. We also plan to explore ways of estimating the number of families and houses in each settlement, and their pace of growth over time.
"Many organizations will make use of this information if they have access. Consolidating a common baseline between responders will reduce duplication and improve the quality of humanitarian assistance in the emergency," says Jeff of iMMAP.
In Bendición de Dios, we told the community members that we were hoping to map their village. We took out our mobile phones and found our location on Google Maps. When we showed the satellite imagery to the community, their eyes brightened as they recognized their homes. "That's us!" they laughed as they pointed at their tiny rooftops on the smartphone screen.
Are you interested in helping Venezuelan migrants? Here are some ways that you can help:
- Contribute to our informal settlement detection project, which we will open-source soon
- Donate directly to these organizations that are assisting Venezuelan migrants.