One of the biggest problems in solving poverty worldwide is the scarcity of reliable data in developing countries.
Researchers have sought to address this by combining satellite imagery with artificial intelligence to identify impoverished areas from space, the BBC reports.
A team from Stanford University trained a computer system and surveyed information in five African countries. Researchers Neal Jean, Marshall Burke, and their colleagues say this method could go a long way in tracking and targeting poverty in specific countries.
Burke, an assistant professor of Earth system science at Stanford, says, “The World Bank, which keeps the poverty data, has for a long time considered anyone who is poor to be someone who lives on below $1 a day.”
He describes how traditional data-gathering on poverty is completed, specifically household surveys. “We use that data to construct our poverty measures,” he says.
Such surveys can be expensive, time-consuming, irregular and difficult to conduct in regions that are hard to reach or are undergoing armed conflict. This proved the need for an accurate, more efficient means of tracking household incomes in developing nations.
The researchers examined daylight satellite images that photographed features that could indicate different levels of economic health, such as paved roads and metal roofs. They then used a highly advanced computer model to classify the indicators in satellite images of Tanzania, Nigeria, Rwanda, Uganda and Malawi. Burke says,
If you give a computer enough data it can figure out what to look for. We trained a computer model to find things in imagery that are predictive of poverty.
The team then used images from countries with available survey data to confirm the satellite’s and the computer model’s findings. Burke and his team intend to continue the study, eventually scaling up their technique to cover the entire sub-Saharan Africa and later on, all developing countries.
Joshua Blumenstock, an assistant professor of development economics and data science at the University of California, Berkeley, says in an accompanying article to the study, this is an “exciting potential for adapting machine learning to fight poverty,” especially for social welfare programs. He was not involved in the research.
The study was published in the journal Science.
