Satellite Pictures, Data Help Scientists Predict Poverty In Hard To Reach Areas

In a major breakthrough, a new study has tried to map poverty in the hard-to-reach places around the globe by combining satellite images and machine learning. According to the study conducted by a team of researchers from Stanford University, the new poverty predicting technique could turn out to be helpful for policymakers and organizations to efficiently distribute their aid and funds, as well as evaluate and enact policies more effectively.

The researchers used three data sources including night light images, daytime images as well as survey data, to build an algorithm that predicts how rich or poor any given area is. Lead author Neal Jean and his research team build the algorithm using a two-step process called "transfer learning," reported The Verge.

The first step involved analyzing day and nighttime satellite pictures of five African countries including Uganda, Tanzania, Nigeria, Malawi, and Rwanda. The researchers claimed that areas that are brighter at night in satellite pictures are usually more developed. They also looked at daylight images of the same areas to capture features such as paved roads and metal roofs that are linked to economic development, reported Tech Times.

The next step was to feed additional information into the computer which was actual survey data from the Demographic Health Services and the World Bank Living Standards Measurement Study. The research team found that the method was efficient enough in mapping poverty distribution, and even excelled existing methods. In fact, the method was found to be 81 percent more effective at predicting poverty in an area under the poverty line as compared to a method using nighttime imagery alone. The researchers are claiming that as the new method is quite inexpensive and scalable, it could be applied to map poverty across the world. The research findings have been published in the journal Science.